Perpendicular,magnetic,anisotropy,induced, by Rashba ... · negative bias voltages enhanced the perpendicular anisotropy, in contrast with the odd function dependence previously observed

Perpendicular magnetic anisotropy induced by Rashba spin-‐orbit interaction

In collaboration with S. E. Barnes (U. Miami),S. Maekawa (ASRC, JAEA)

家田淳一Jun’ichi Ieda

Advanced Science Research Center (ASRC),Japan Atomic Energy Agency (JAEA),

Tokai, Japan

New Perspectives in Spintronics and Mesoscopic Physics ISSP Kashiwa, U Tokyo2015 June 16

Outline• Introduction– Rashba effects in Spintronics

• Model– Rashba mechanism of magnetic anisotropy

• Electric-‐Field Effect– Experiments and predictions

• Summary

Rashba Effect

E. I. Rashba: Fiz. Teverd. Tela (Leningrad) 2 (1960) 1224;Y. A. Bychkov and E. I. Rashba: J. Phys., C 17 (1984) 6039.

http://news.harvard.edu/

physics of spin-‐orbit coupling penetrated into numerous branches of condensed matter physics …. I guess that our paper became successful because it was one of the first, and timely, steps of this journey.

by E. I. Rashba from the editor blog of JETP Letters

HR =

αR

p ⋅ σ × z( )

Inversion symmetry breaking -‐> E(k,↑) ≠ E(-‐k,↑)Time reversal symmetry -‐> E(k,↑) ≠ E(k,↓)

Surface Rashba Splitting

• Au (111)

~3.5meV ~ 35T!

ε k,σ =

2k2

2m−σαR k =

2

2mk −σ k0( )2 − ER

k2 = kx

2 + ky2

ER =

2

2mk02 = mαR

2

22= m2

eηSO

⎛⎝⎜

⎞⎠⎟2

E2

structure. Using the concept of surface alloying, theAg(111) (Z ! 47) surface layer was doped with the heavyelement Bi (Z ! 83). The 2D band structure of theBi=Ag"111# surface alloy was investigated by ARPES [atthe EPFL in Lausanne, Switzerland, as well as at the Syn-chrotron Radiation Center (SRC) in Wisconsin, U.S.A.]and exhibits a spin splitting of unprecedented magnitude.We have shown elsewhere [15] that the large spin-orbitsplitting opens up new opportunities for STM investigationof SO split states. Here we focus on the electronic structureof the surface alloy, and provide a theoretical explanationfor the effect by means of first-principles electronic struc-ture calculations. The sizable spin splitting is explained bythe in-plane potential gradient.

The Ag(111) surface was cleaned in ultrahigh vacuumusing successive sputtering and annealing cycles. The dep-osition of 1=3 of a monolayer of Bi atoms results in a long-range ordered "

!!!3p$

!!!3p#R30% substitutional surface alloy

[see Fig. 1(a)]. The Bi atoms protrude slightly out of thesurface due to size mismatch [Fig. 1(b)] [15]. A similarscenario is observed for the Sb=Ag"111# surface alloy [16].

The experimental electronic structure of Bi=Ag"111#[Fig. 1(c)] consists of one pair of bands that showsstrong photoemission intensity (dark) at the maxima(&0:135 eV), which becomes weaker (lighter) with in-creasing binding energy [17]. The maxima are symmetri-cally shifted from the !" point by 0:13 #A&1; the effectivemass m? is &0:35me. A second pair with much lessintensity crosses the Fermi level at positions indicated bythe four arrows in Fig. 1(c). The separation between thesebands is 0:12 #A&1.

The above findings—a symmetric offset of band ex-trema [Bi=Ag"111#: maxima; Au(111): minima] in con-junction with a crossing at !"—clearly indicate a spin-split

surface band structure due to spin-orbit interaction, as inthe RB model and for Au(111). However, the splitting isconsiderably larger than those reported for other systems[9–12,18]. Different splittings can be compared on a wavenumber scale, where k0 describes the shift of the band ex-tremum away from !" as well as on an energy scale, whereER ! @2k2

0=2m? is called Rashba energy [see Fig. 1(c)]. Ifthere were no spin splitting the dispersion would reduce toa spin-degenerate band with the maximum at the highsymmetry point. The Rashba parameter !R ! @2k0=m? isthe coupling constant in the spin-orbit Hamiltonian[Eq. (2)]. A selection of systems is given in Table I, wherethe Rashba parameter has been computed from the experi-mentally accessible quantities k0 and m?.

In the following we argue that the RB model, as given byEq. (2), cannot explain our findings for a surface alloy. Itwas shown that the Rashba parameter !R of the L-gapsurface state of a AgxAu1&x alloy depends linearly on theconcentration x, as in the virtual crystal approximation [6].Hence, one is tempted to conclude that !R is essentiallydetermined by the number of heavy atoms (here Au, Z !79) probed by the surface state. If one were to apply thesame idea to Bi=Ag"111#, with Bi being the heavy element(Z ! 83), one would expect !R to be about an order ofmagnitude smaller (0.3 to 0:4 eV #A) than what was foundexperimentally (see Table I). The potential gradient alongthe z direction is expected almost independent of thespecific surface, and a strong atomic SO interaction alonecannot account for our findings. Hence, there must be anadditional mechanism responsible for the observed giantsplitting.

The electronic structure of the Bi=Ag"111# surface alloywas calculated from first principles within the frameworkof the local spin-density approximation to density-functional theory (DFT). A relativistic multiple-scatteringapproach (layer Korringa-Kohn-Rostoker method), inwhich spin-orbit coupling is fully accounted for (Diracequation), was used to compute spin-, layer-, and atom-resolved Bloch spectral densities, hence allowing one tocharacterize the electronic structure in much detail. Itssuitability for investigating RB systems was alreadyproven for the L-gap surface state of Au(111) [19].Calculations for the clean Ag(111) surface agree wellwith experiment. In the theory, the SO interaction can bescaled between the fully relativistic (Dirac equation) andthe scalar-relativistic case [20,21]. In the latter case (notshown), the spin splitting vanishes. This clearly proves thatSO coupling is the base of the observed effect.

A first comparison of experiment and theory is done byconsidering I"E; kx; ky# slices, where I is either the experi-mental photoemission intensity or the theoretical Blochspectral density of the surface layer (cf. Ref. [15]). I isshown in Fig. 2 as gray scale for fixed ky (along !"- !K) andvarying both energy E and kx (along !M- !"- !M). For ky ! 0[Fig. 2(a)], both experiment (left-hand panels) and theory(right-hand panels) exhibit the characteristic dispersion of

FIG. 1 (color online). (a) Schematic top view of the "!!!3p$!!!

3p#R30% Bi=Ag"111# surface alloy [Bi, light gray (orange); Ag,

dark gray (blue)]. (b) Side view of the schematic, illustrating theoutward relaxation of Bi in the surface layer. (c) Experimentalband structure obtained by ARPES. The abscissa is the wavevector kx along the !M- !"- !M line in the vicinity of the center of thesurface alloy Brillouin zone ( !", i.e., kk ! 0). The ordinate givesthe energy below the Fermi level (0 eV).

PRL 98, 186807 (2007) P H Y S I C A L R E V I E W L E T T E R S week ending4 MAY 2007

186807-2

Huge “Magnetic Field”

• Bi on Ag (111)

Ast et al. PRL (2007).

~200meV ~ 2000T!

Rashba Splitting

ε !k,σ = "

2

2mk# −σ k0( )2

− ER

Energy gain~ 3.5meV

for Au(111)~ 35T!

ER =

mαR2

2!2

Rashba splitting band:

Rashba energy:

Spins are in-‐plane

µBBR =αR −ky ,kx ,0( )

Rashba in SpintronicsDomain wall motion

[Miron: Nat. Mat. (2011)]Spin filtering

[Kohda: Nat. Com. (2012)]Spin-‐charge conversion

[J. Sanchez: Nat. Com. (2013)]

• Non-‐equilibrium state: current, magnetization dynamics…

Aim• Perpendicular Magnetic Anisotropy (PMA)

è Indispensable properties for MRAM– Simple bit structure for efficient magnetization flips– High thermal stability (vs. superparamagnetism)– High density memory bit

• Gate control of PMAè Electric control of magnetism

We propose Rashba-‐induced PMA.

Question

How does the in-‐plane Rashba field lead to PMA?

AnswerCooperation with the exchange interaction.

Ultrathin ferromagnetic film

Rashba Field: BR

B R R , ,0y xk kB

exchange

Rashba field

Reinforcement

Model

• Ferromagnet with Rashba SOI:

Magnetization (Order parameter)

Rashba parameter: αR

Rashba Field: BR

H =

2

2mkx2 + ky

2( )− J0Sm ⋅ σ +αR kyσ x − kxσ y( )

R SO zE

ˆ sin cos ,sin sin ,cosm

B R R , ,0y xk kB

Ultrathin ferromagnetic film

αR & αR2 terms

• “Total” field (J0S in the y-‐z plane).

µB B tot = J0S + µBBRy sinθ( )2

+ µBBRy cosθ( )2

+ µBBRx( )2⎡

⎣⎢⎤⎦⎥

1/2

! J0S +αRkx sinθ + 12αR

2

J0Skx

2 cos2θ + ky2( ).

V J0S

BRy

BRx

y

z

x BRycosV

BRysinV

Fermi sea shift &(partial) Rashba splitting

exchange reinforcement

Out-‐of-‐plane configuration

Rashba Field

quantization axis

Energy gain due to reinforced exchange splitting (no energy gain from Rashba splitting)

J0SJS

RB

In-‐plane configurationJ0S

JS

RxB

RyB

Rashba fields go into exchange (BRx) and Rashba (BRy) splittings.

Rashba-‐PMA

ε k,σ = 2

2mkx −σ k0 sinθ( )2 + ky2⎡

⎣⎤⎦ − ER sin

2θ

−σ J0S( )2 +αR2 kx

2 cos2θ + ky2( )⎡

⎣⎤⎦1/2

Magnetic Anisotropy Energy:

2MA R

0

1 co2 sTJ S

E E

ER = mαR

2

22 ~ 35T (Au)

T =

2

2mkx2

↑− kx

2↓( )

Averaging over occupied states…

Rashba splitting gain→ in-‐planeMA

exchange splitting gain→ PMA

Two interfaces interaction

• I/F/I tri-‐layer with gatingMagnetic Anisotropy Energy:

EMA = ER 1− 2TJ0S

⎛⎝⎜

⎞⎠⎟cos2θ

ER =m αR

top +αRbottom( )2

2!2

= me2

2!2 ηSOtop E0

top + Ebias( ) +ηSObottom −E0

bottom + Ebias( )⎡⎣

⎤⎦

2

~2me2ηSO

2

!2 Ebias( )2

MgO/FeB/MgO

Voltage-Induced Magnetic Anisotropy Changes in an Ultrathin FeB Layer

Sandwiched between Two MgO Layers

Takayuki Nozaki1;2!, Kay Yakushiji1;2, Shingo Tamaru1, Masaki Sekine1, Rie Matsumoto1;2,Makoto Konoto1;2, Hitoshi Kubota1;2, Akio Fukushima1;2, and Shinji Yuasa1;2

1National Institute of Advanced Industrial Science and Technology, Spintronics Research Center, Tsukuba, Ibaraki 305-8568, Japan2CREST, JST, Kawaguchi, Saitama 332-0012, Japan

E-mail: [email protected]

Received May 7, 2013; accepted June 10, 2013; published online June 24, 2013

We investigated the effect of voltage on the perpendicular magnetic anisotropy in an ultrathin Fe80B20 layer sandwiched between two MgO barrierlayers with different thicknesses, in which the bias voltage is predominantly applied to one of the MgO layers. The application of both positive and

negative bias voltages enhanced the perpendicular anisotropy, in contrast with the odd function dependence previously observed in a MgO/

ferromagnetic metal/non-magnetic metal structure. Moreover, for positive bias voltages, the large anisotropy change slope of 108 fJ/Vm was

demonstrated. These results indicate that the MgO-sandwich structure has high potential to extend the availability of the voltage effect.# 2013 The Japan Society of Applied Physics

In spintronics, direct voltage control of the magneticproperties of thin ferromagnetic films is currently ofhigh interest as the ultimate technology for low-power

spin manipulation.1–12) In particular, voltage control of themagnetic anisotropy, which has been observed for a junctionin an insulator/ultrathin ferromagnetic metal/non-magneticmetal structure, is one of the most promising phenom-ena.1–10) In principle, the electric field cannot penetrate intothe metal due to the screening effect of the free electrons.However, if the ferromagnetic metal layer is thin enough(e.g., on the order of a few monolayers), accumulated chargeat the interface can influence the electronic occupation stateof 3d orbitals and the band structure of the ultrathinferromagnetic layer, leading to a change in the perpendicularmagnetic anisotropy through the spin–orbit interaction.13–16)

This technique has advantages for practical applications. Forexample, we can employ practical 3d transition ferromag-netic metals such as Fe, Co, Ni, and their alloys simply byreducing their thickness to a few monolayers, and conse-quently, they can be easily introduced into magnetoresis-tance (MR) devices such as MgO-based magnetic tunneljunctions.17–20)

An experimental demonstration of voltage controlledmagnetic anisotropy was first reported in a thin FePt filmusing a liquid electrolyte, resulting in a change in coercivityby 4.5%.1) We have successfully demonstrated this phenom-enon in solid state structures, such as Au/ultrathin Fe(Co)/MgO/polyimide/indium tin oxide (ITO) junctions2,3) andMgO-based magnetic tunnel junctions,17,18) leading to therealization of dynamic magnetization switching21,22) andferromagnetic resonance excitation23,24) solely driven by theapplication of a voltage.

Recently, the strong perpendicular anisotropy in anFeCoB layer sandwiched between MgO layers has createdwidespread interest in the realization of an ultrahigh-density spin-transfer-torque magnetoresistive random accessmemory (STT-MRM) based on perpendicularly magnetizedMTJs.25,26) Voltage-control of the perpendicular magneticanisotropy in a MgO/FeCoB/MgO structure is thereforean interesting subject for future voltage-driven spintronicdevices with perpendicular magnetization.

In this study, we investigated the variation in the voltage-induced anisotropy in an ultrathin Fe80B20 layer sandwichedbetween two MgO layers with different thicknesses and,

surprisingly, observed increases in the perpendicular mag-netic anisotropy for both positive and negative biases, incontrast with the odd function dependence previously ob-served in a MgO/ferromagnetic metal/non-magnetic metalstructure.

Figure 1(a) shows a schematic diagram of the magnetictunnel junction stack, consisting of a Cr buffer/Fe (50 nm)/MgO (2.5 nm)/Fe80B20 (1.5 nm)/MgO (1.5 nm)/Ru (10 nm),deposited on a single crystal MgO(001) substrate usinga combination of molecular beam epitaxy (MBE) andmagnetron sputtering techniques. Here, the perpendicularanisotropy of the 1.5-nm-thick Fe80B20 layer is controlledby the voltage. In this study, we used Fe80B20 instead of anFeCoB alloy because the Fe80B20 layer shows the largestperpendicular anisotropy energy among the (Fe1"xCox)80B20

alloys.25) This large anisotropy energy is suitable for theexperiments in this study. A high-quality Fe(001) electrodewas deposited on the Cr buffer at room temperature (RT) byelectron-beam evaporation in an ultrahigh vacuum (UHV)MBE chamber (base pressure: 1:0# 10"8 Pa), followed byannealing at 300 $C. After cooling the substrate to 70 $C, a2.5-nm-thick MgO(001) layer was deposited on top of thebottom electrode and annealed at 300 $C under UHV. Then,the sample was moved to a sputtering deposition chamberthrough a UHV transfer chamber (base pressure: 0:9# 10"7

Pa), and a 1.5-nm-thick Fe80B20 layer was deposited at RT.

Ru cap

Fe80B20 (1.5 nm)

MgO (2.5 nm)

Fe (50 nm)

Seed/Buffer layers

MgO (001) substrate

MgO (1.5 nm)

(a)

-8000 -6000 -4000 -2000 0 2000 4000 6000 800012

14

16

Res

ista

nce

(kΩ

)

Magnetic field (Oe)

(b)

Fig. 1. (a) Schematic illustration of the magnetic tunnel junction with theFe/MgO/Fe80B20/MgO structure. (b) TMR curve measured under the in-plane magnetic fields. Inset drawings show the magnetization configurationof the bottom Fe and top Fe80B20 layers.

Applied Physics Express 6 (2013) 073005

073005-1 # 2013 The Japan Society of Applied Physics

http://dx.doi.org/10.7567/APEX.6.073005

Voltage-Induced Magnetic Anisotropy Changes in an Ultrathin FeB Layer

Sandwiched between Two MgO Layers

Takayuki Nozaki1;2!, Kay Yakushiji1;2, Shingo Tamaru1, Masaki Sekine1, Rie Matsumoto1;2,Makoto Konoto1;2, Hitoshi Kubota1;2, Akio Fukushima1;2, and Shinji Yuasa1;2

1National Institute of Advanced Industrial Science and Technology, Spintronics Research Center, Tsukuba, Ibaraki 305-8568, Japan2CREST, JST, Kawaguchi, Saitama 332-0012, Japan

E-mail: [email protected]

Received May 7, 2013; accepted June 10, 2013; published online June 24, 2013

We investigated the effect of voltage on the perpendicular magnetic anisotropy in an ultrathin Fe80B20 layer sandwiched between two MgO barrierlayers with different thicknesses, in which the bias voltage is predominantly applied to one of the MgO layers. The application of both positive and

negative bias voltages enhanced the perpendicular anisotropy, in contrast with the odd function dependence previously observed in a MgO/

ferromagnetic metal/non-magnetic metal structure. Moreover, for positive bias voltages, the large anisotropy change slope of 108 fJ/Vm was

demonstrated. These results indicate that the MgO-sandwich structure has high potential to extend the availability of the voltage effect.# 2013 The Japan Society of Applied Physics

In spintronics, direct voltage control of the magneticproperties of thin ferromagnetic films is currently ofhigh interest as the ultimate technology for low-power

spin manipulation.1–12) In particular, voltage control of themagnetic anisotropy, which has been observed for a junctionin an insulator/ultrathin ferromagnetic metal/non-magneticmetal structure, is one of the most promising phenom-ena.1–10) In principle, the electric field cannot penetrate intothe metal due to the screening effect of the free electrons.However, if the ferromagnetic metal layer is thin enough(e.g., on the order of a few monolayers), accumulated chargeat the interface can influence the electronic occupation stateof 3d orbitals and the band structure of the ultrathinferromagnetic layer, leading to a change in the perpendicularmagnetic anisotropy through the spin–orbit interaction.13–16)

This technique has advantages for practical applications. Forexample, we can employ practical 3d transition ferromag-netic metals such as Fe, Co, Ni, and their alloys simply byreducing their thickness to a few monolayers, and conse-quently, they can be easily introduced into magnetoresis-tance (MR) devices such as MgO-based magnetic tunneljunctions.17–20)

An experimental demonstration of voltage controlledmagnetic anisotropy was first reported in a thin FePt filmusing a liquid electrolyte, resulting in a change in coercivityby 4.5%.1) We have successfully demonstrated this phenom-enon in solid state structures, such as Au/ultrathin Fe(Co)/MgO/polyimide/indium tin oxide (ITO) junctions2,3) andMgO-based magnetic tunnel junctions,17,18) leading to therealization of dynamic magnetization switching21,22) andferromagnetic resonance excitation23,24) solely driven by theapplication of a voltage.

Recently, the strong perpendicular anisotropy in anFeCoB layer sandwiched between MgO layers has createdwidespread interest in the realization of an ultrahigh-density spin-transfer-torque magnetoresistive random accessmemory (STT-MRM) based on perpendicularly magnetizedMTJs.25,26) Voltage-control of the perpendicular magneticanisotropy in a MgO/FeCoB/MgO structure is thereforean interesting subject for future voltage-driven spintronicdevices with perpendicular magnetization.

In this study, we investigated the variation in the voltage-induced anisotropy in an ultrathin Fe80B20 layer sandwichedbetween two MgO layers with different thicknesses and,

surprisingly, observed increases in the perpendicular mag-netic anisotropy for both positive and negative biases, incontrast with the odd function dependence previously ob-served in a MgO/ferromagnetic metal/non-magnetic metalstructure.

Figure 1(a) shows a schematic diagram of the magnetictunnel junction stack, consisting of a Cr buffer/Fe (50 nm)/MgO (2.5 nm)/Fe80B20 (1.5 nm)/MgO (1.5 nm)/Ru (10 nm),deposited on a single crystal MgO(001) substrate usinga combination of molecular beam epitaxy (MBE) andmagnetron sputtering techniques. Here, the perpendicularanisotropy of the 1.5-nm-thick Fe80B20 layer is controlledby the voltage. In this study, we used Fe80B20 instead of anFeCoB alloy because the Fe80B20 layer shows the largestperpendicular anisotropy energy among the (Fe1"xCox)80B20

alloys.25) This large anisotropy energy is suitable for theexperiments in this study. A high-quality Fe(001) electrodewas deposited on the Cr buffer at room temperature (RT) byelectron-beam evaporation in an ultrahigh vacuum (UHV)MBE chamber (base pressure: 1:0# 10"8 Pa), followed byannealing at 300 $C. After cooling the substrate to 70 $C, a2.5-nm-thick MgO(001) layer was deposited on top of thebottom electrode and annealed at 300 $C under UHV. Then,the sample was moved to a sputtering deposition chamberthrough a UHV transfer chamber (base pressure: 0:9# 10"7

Pa), and a 1.5-nm-thick Fe80B20 layer was deposited at RT.

Ru cap

Fe80B20 (1.5 nm)

MgO (2.5 nm)

Fe (50 nm)

Seed/Buffer layers

MgO (001) substrate

MgO (1.5 nm)

(a)

-8000 -6000 -4000 -2000 0 2000 4000 6000 800012

14

16

Res

ista

nce

(kΩ

)

Magnetic field (Oe)

(b)

Fig. 1. (a) Schematic illustration of the magnetic tunnel junction with theFe/MgO/Fe80B20/MgO structure. (b) TMR curve measured under the in-plane magnetic fields. Inset drawings show the magnetization configurationof the bottom Fe and top Fe80B20 layers.

Applied Physics Express 6 (2013) 073005


http://dx.doi.org/10.7567/APEX.6.073005

about 3 times that observed in the Au/FeCo/MgO/Fejunction (37 fJ/Vm).17) For negative bias voltages, the slopeis estimated to be about 24 fJ/Vm. In the previous studyusing the Au/Fe/MgO junction,2) charge depletion at theFe/MgO interface caused an increase in the perpendicularanisotropy. This tendency is consistent with our results forpositive bias voltages. However, a similar increase in theperpendicular anisotropy with voltage was observed evenfor negative bias voltages, which is an electron-chargingcondition.

Although the origin of the even function voltage de-pendence is not clear at the present stage, one helpful cluehas been reported by Shimabukuro et al.27) According totheir calculation, the symmetric MgO/Fe/MgO structureexhibits a complete even function dependence on the electricfield application. On the other hand, in the Fe/MgO structurewith no metal layer adjacent to the Fe layer, large anisotropychange is obtained in the positive electric-field direction,while almost no change is induced in the negative direction.If the MgO thicknesses in the MgO/Fe/MgO structureare largely different, the intermediate feature between theMgO/Fe/MgO and Fe/MgO structures is expected, whichappears to qualitatively agree with our experimental results.These discussions may indicate that the topside MgO barrierwith the lower RA value also plays an important role in thevoltage effect. To clarify the mechanism, further systematicinvestigations such as the dependences on ferromagneticlayer and MgO layer thickness are required in the futurework.

From the viewpoint of device applications, the observedeven function voltage dependence of the perpendicularanisotropy is expected not only to extend the applications ofMTJs but also to help realize novel spintronic devices.

In summary, we investigated the effect of bias voltage onthe perpendicular anisotropy of an ultrathin Fe80B20 layersandwiched between MgO layers with different thicknesses.

A highly effective anisotropy change was induced bypositive bias voltages with a slope of 108 fJ/Vm. In addition,an even function voltage dependence of the perpendicularanisotropy was observed, in contrast with the odd functionvoltage dependence observed previously for a single MgObarrier structure. These results indicate that the MgO-sandwich structure has the potential to offer a new route tocontrolling the voltage effect.

Acknowledgments We thank H. Imamura, H. Maehara, and K. Nakamurafor fruitful discussion and E. Usuda for assistance with the experiments.

1) M. Weisheit, S. Fahler, A. Marty, Y. Souche, C. Poinsignon, and D.Givord: Science 315 (2007) 349.

2) T. Maruyama, Y. Shiota, T. Nozaki, K. Ohta, N. Toda, M. Mizuguchi,A. A. Tulapurkar, T. Shinjo, M. Shiraishi, S. Mizukami, Y. Ando, and Y.Suzuki: Nat. Nanotechnol. 4 (2009) 158.

3) Y. Shiota, T. Maruyama, T. Nozaki, T. Shinjo, M. Shiraishi, and Y.Suzuki: Appl. Phys. Express 2 (2009) 063001.

4) M. Endo, S. Kanai, S. Ikeda, F. Matsukura, and H. Ohno: Appl. Phys. Lett.96 (2010) 212503.

5) M. Zhernenkov, M. R. Fitzsimmons, J. Chlistunoff, J. Majewski, I. Tudosa,and E. E. Fullerton: Phys. Rev. B 82 (2010) 024420.

6) T. Seki, M. Kohda, J. Nitta, and K. Takanashi: Appl. Phys. Lett. 98 (2011)212505.

7) F. Bonell, S. Murakami, Y. Shiota, T. Nozaki, T. Shinjo, and Y. Suzuki:Appl. Phys. Lett. 98 (2011) 232510.

8) U. Bauer, M. Przybylski, J. Kirschner, and G. S. D. Beach: Nano Lett. 12(2012) 1437.

9) K. Kita, D. W. Abraham, M. J. Gajek, and D. C. Worledge: J. Appl. Phys.112 (2012) 033919.

10) N. Tournerie, A. P. Engelhardt, F. Maroun, and P. Allongue: Phys. Rev. B86 (2012) 104434.

11) L. Gerhard, T. K. Yamada, T. Balashov, A. F. Takacs, R. J. H. Wesselink,M. Dane, M. Fechner, S. Ostanin, A. Ernst, I. Mertig, and W. Wulfhekel:Nat. Nanotechnol. 5 (2010) 792.

12) D. Chiba, S. Fukami, K. Shimamura, N. Ishiwata, K. Kobayashi, and T.Ono: Nat. Mater. 10 (2011) 853.

13) C.-G. Duan, J. P. Velev, R. F. Sabirianov, Z. Zhu, J. Chu, S. S. Jaswal, andE. Y. Tsymbal: Phys. Rev. Lett. 101 (2008) 137201.

14) K. Nakamura, R. Shimabukuro, Y. Fujiwara, T. Akiyama, T. Ito, and A. J.Freeman: Phys. Rev. Lett. 102 (2009) 187201.

15) M. Tsujikawa and T. Oda: Phys. Rev. Lett. 102 (2009) 247203.16) K. H. He, J. S. Chen, and Y. P. Feng: Appl. Phys. Lett. 99 (2011) 072503.17) T. Nozaki, Y. Shiota, M. Shiraishi, T. Shinjo, and Y. Suzuki: Appl. Phys.

Lett. 96 (2010) 022506.18) Y. Shiota, S. Murakami, F. Bonell, T. Nozaki, T. Shinjo, and Y. Suzuki:

Appl. Phys. Express 4 (2011) 043005.19) W. Skowronski, P. Wisniowski, T. Stobiecki, S. Cardoso, P. P. Freitas, and

S. van Dijken: Appl. Phys. Lett. 101 (2012) 192401.20) W.-G. Wang, M. Li, S. Hageman, and C. L. Chien: Nat. Mater. 11 (2012)

64.21) Y. Shiota, T. Nozaki, F. Bonell, S. Murakami, T. Shinjo, and Y. Suzuki:

Nat. Mater. 11 (2012) 39.22) S. Kanai, M. Yamanouchi, S. Ikeda, Y. Nakatani, F. Matsukura, and H.

Ohno: Appl. Phys. Lett. 101 (2012) 122403.23) T. Nozaki, Y. Shiota, S. Miwa, S. Murakami, F. Bonell, S. Ishibashi, H.

Kubota, K. Yakushiji, T. Saruya, A. Fukushima, S. Yuasa, T. Shinjo, andY. Suzuki: Nat. Phys. 8 (2012) 491.

24) J. Zhu, J. A. Katine, G. E. Rowlands, Y.-J. Chen, Z. Duan, J. G. Alzate, P.Upadhyaya, J. Langer, P. K. Amiri, K. L. Wang, and I. N. Krivorotov:Phys. Rev. Lett. 108 (2012) 197203.

25) H. Kubota, S. Ishibashi, T. Saruya, T. Nozaki, A. Fukushima, K. Yakushiji,K. Ando, Y. Suzuki, and S. Yuasa: J. Appl. Phys. 111 (2012) 07C723.

26) H. Sato, M. Yamanouchi, S. Ikeda, S. Fukami, F. Matsukura, and H. Ohno:Appl. Phys. Lett. 101 (2012) 022414.

27) R. Shimabukuro, K. Nakamura, T. Akiyama, and T. Ito: Physica E 42(2010) 1014.

0 2000 4000 60000.0

0.2

0.4

0.6

0.8

1.0

Min

-pla

ne/ M

S (a.

u.)

H (Oe)

-2000-1500-1000 -500 0 500 1000 1500 2000300

320

340

360

380

400

Bias voltage (mV)

Epe

rp· t

FeB (µ

J/m

)2

Fig. 3. Bias voltage dependence of the estimated Eperp ! tFeB. Inset figureshows an example of the estimated magnetization curve of the Fe80B20 layercalculated from the normalized TMR curve obtained under the bias voltageof 30mV.

T. Nozaki et al.Appl. Phys. Express 6 (2013) 073005


Magne

tic Anisotrop

yEne

rgy

2

MA0

SO2

22 212

biasTJ

E EeS

mh

I/F/N tri-‐layer

• Depending on N-‐layer…

Rashba fields cancel

Rashba fields add

gating experiments distinguish the cases.

MgO/Co16Fe64B20/Ta(Ru)

the relative angle of the two ferromagnetic layers, hr, and theconductance, G, with the following equation:29

M==;norm:ðHÞ ¼ cos hr ¼GðHÞ $ G90%

G0% $ G90%; (1)

where G90% and G0% are the conductance in the 90% and paral-lel configurations, respectively. Here, we employed thisdetermination, in the case, that the magnetization of refer-ence layer is aligned to the direction of external magneticfield. Because the stray dipole field from the top thickCoFeB layer is estimated to be less than 0.01 kOe, which isnegligible compared to the external magnetic field, G90% isobtained at zero field, and G0% is obtained at 3 kOe.

Figures 2(a) and 2(b) show the normalized magnetiza-tion curves for voltages of 6700 mV. With an applied volt-age, clear changes in the saturation field are observed in bothMTJs. To evaluate the voltage-induced PMA change in aquantitative way, the PMA energy per unit volume of thefilm, Eperp, is calculated from the normalized magnetizationcurves, assuming that the saturation magnetization is inde-pendent of the applied voltage

Eperp ¼ l0MS

ð1

0

HdM==;norm:; (2)

where l0, MS, and H are the permeability of free space, thesaturation magnetization, and the external magnetic field,

respectively. We use a value for the saturation magnetizationl0MS¼ 1.78 T for Ta-MTJ and l0MS¼ 1.98 T for Ru-MTJ,as measured by vibrating sample magnetometry in separatesamples. Figures 2(c) and 2(d) present the applied bias volt-age dependence of EperptCoFeB, where tCoFeB is the thicknessof the thin CoFeB layer. Linear changes in the anisotropyenergy are observed in both MTJs, and the change of the sur-face anisotropy energy for an electric field of 1 V/nm is eval-uated to be $33 fJ/(Vm) for Ta-MTJ and þ18 fJ/(Vm) forRu-MTJ. Interestingly, these two values have opposite signs.In Ta-MTJ, a negative voltage (positively charged thinCoFeB layer) increases the PMA. This tendency agrees withother works with TajCoFeBjMgO junctions.11,16,20,21 On theother hand, a positive voltage in Ru-MTJ (negativelycharged thin CoFeB layer) increases the PMA. This tendencyhas not yet been observed in the stacks with CoFeBjMgOjunctions.

Clear evidence of these opposite signs is observed in therf voltage-induced ferromagnetic resonance spectra. An rfvoltage is applied to the MTJ junctions through the rf port ofa bias tee. When the ferromagnetic resonance is excited bythe voltage effect, the product of the oscillating resistanceand the small tunneling current produce a dc output voltage.This output dc voltage is measured through the dc port of thebias tee. The applied rf power is 10 lW ($20 dBm), and theexternal magnetic field is tilted by 55% with respect to thefilm plane with its in-plane component parallel to the longaxis of pillar. This tilted angle for the external magnetic fieldis chosen to maximize the voltage torque.19

Figures 3(a) and 3(b) show the magnetic field depend-ence of the homodyne detection signals. All spectra exhibita clear resonance peak with a dispersion structure (anti-symmetric peak). Because the rf voltage-induced anisot-ropy change can be considered as an rf field appliedperpendicular to the film plane, it can excite ferromagneticresonance. As discussed in Ref. 19, the shape of resonancespectra depends on the magnetization configuration andsign/magnitude of voltage effect. In particular, a dispersionstructure is expected in our case because the in-planecomponents of the two magnetizations are parallel (/1¼ 0in Eq. (S2) of Ref. 19). Moreover, the sign of the peakdepends on the sign of the anisotropy change (@U/@V inEq. (S2) of Ref. 19).

The obtained resonant frequencies as a function of exter-nal magnetic field fit well with the modified Kittel formula(not shown here), and the effective demagnetization fields inthe perpendicular direction are evaluated to be about 0.11kOe for Ta-MTJ and 0.075 kOe for Ru-MTJ. The most sig-nificant difference between the two MTJs is the oppositesign of their resonance peaks. As described above, the signof voltage effect between Ta-MTJ and Ru-MTJ is opposite,and this is clearly observed in the homodyne detection spec-tra as a sign change of the peaks. From the peak intensity,we evaluate the magnitude of the voltage effect according toEq. (1) in Ref. 19. For an external magnetic field of 3.0 kOe,the slope of the effective demagnetization field, @H?/@V, isevaluated to be 0.14 kOe/V for Ta-MTJ and $0.081 kOe/Vfor Ru-MTJ. These values correspond to the surface mag-netic anisotropy energy change for an applied electric fieldof $29 fJ/(Vm) for Ta-MTJ and þ18 fJ/(Vm) for Ru-MTJ,

FIG. 1. Tunnel magnetoresistance as a function of the in-plane magneticfield for (a) Ta-MTJ and (b) Ru-MTJ. The resistance-area product (RA) isplotted for the designed junction size area of 4' 1 lm2. The insets show thelayer structures of the magnetic tunnel junctions. The thicknesses of eachlayer in nanometers are given in parentheses.

082410-2 Shiota et al. Appl. Phys. Lett. 103, 082410 (2013)

This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:133.53.133.149 On: Tue, 03 Dec 2013 14:46:08

the relative angle of the two ferromagnetic layers, hr, and theconductance, G, with the following equation:29

M==;norm:ðHÞ ¼ cos hr ¼GðHÞ $ G90%

G0% $ G90%; (1)

where G90% and G0% are the conductance in the 90% and paral-lel configurations, respectively. Here, we employed thisdetermination, in the case, that the magnetization of refer-ence layer is aligned to the direction of external magneticfield. Because the stray dipole field from the top thickCoFeB layer is estimated to be less than 0.01 kOe, which isnegligible compared to the external magnetic field, G90% isobtained at zero field, and G0% is obtained at 3 kOe.

Figures 2(a) and 2(b) show the normalized magnetiza-tion curves for voltages of 6700 mV. With an applied volt-age, clear changes in the saturation field are observed in bothMTJs. To evaluate the voltage-induced PMA change in aquantitative way, the PMA energy per unit volume of thefilm, Eperp, is calculated from the normalized magnetizationcurves, assuming that the saturation magnetization is inde-pendent of the applied voltage

Eperp ¼ l0MS

ð1

0

HdM==;norm:; (2)

where l0, MS, and H are the permeability of free space, thesaturation magnetization, and the external magnetic field,

respectively. We use a value for the saturation magnetizationl0MS¼ 1.78 T for Ta-MTJ and l0MS¼ 1.98 T for Ru-MTJ,as measured by vibrating sample magnetometry in separatesamples. Figures 2(c) and 2(d) present the applied bias volt-age dependence of EperptCoFeB, where tCoFeB is the thicknessof the thin CoFeB layer. Linear changes in the anisotropyenergy are observed in both MTJs, and the change of the sur-face anisotropy energy for an electric field of 1 V/nm is eval-uated to be $33 fJ/(Vm) for Ta-MTJ and þ18 fJ/(Vm) forRu-MTJ. Interestingly, these two values have opposite signs.In Ta-MTJ, a negative voltage (positively charged thinCoFeB layer) increases the PMA. This tendency agrees withother works with TajCoFeBjMgO junctions.11,16,20,21 On theother hand, a positive voltage in Ru-MTJ (negativelycharged thin CoFeB layer) increases the PMA. This tendencyhas not yet been observed in the stacks with CoFeBjMgOjunctions.

Clear evidence of these opposite signs is observed in therf voltage-induced ferromagnetic resonance spectra. An rfvoltage is applied to the MTJ junctions through the rf port ofa bias tee. When the ferromagnetic resonance is excited bythe voltage effect, the product of the oscillating resistanceand the small tunneling current produce a dc output voltage.This output dc voltage is measured through the dc port of thebias tee. The applied rf power is 10 lW ($20 dBm), and theexternal magnetic field is tilted by 55% with respect to thefilm plane with its in-plane component parallel to the longaxis of pillar. This tilted angle for the external magnetic fieldis chosen to maximize the voltage torque.19

Figures 3(a) and 3(b) show the magnetic field depend-ence of the homodyne detection signals. All spectra exhibita clear resonance peak with a dispersion structure (anti-symmetric peak). Because the rf voltage-induced anisot-ropy change can be considered as an rf field appliedperpendicular to the film plane, it can excite ferromagneticresonance. As discussed in Ref. 19, the shape of resonancespectra depends on the magnetization configuration andsign/magnitude of voltage effect. In particular, a dispersionstructure is expected in our case because the in-planecomponents of the two magnetizations are parallel (/1¼ 0in Eq. (S2) of Ref. 19). Moreover, the sign of the peakdepends on the sign of the anisotropy change (@U/@V inEq. (S2) of Ref. 19).

The obtained resonant frequencies as a function of exter-nal magnetic field fit well with the modified Kittel formula(not shown here), and the effective demagnetization fields inthe perpendicular direction are evaluated to be about 0.11kOe for Ta-MTJ and 0.075 kOe for Ru-MTJ. The most sig-nificant difference between the two MTJs is the oppositesign of their resonance peaks. As described above, the signof voltage effect between Ta-MTJ and Ru-MTJ is opposite,and this is clearly observed in the homodyne detection spec-tra as a sign change of the peaks. From the peak intensity,we evaluate the magnitude of the voltage effect according toEq. (1) in Ref. 19. For an external magnetic field of 3.0 kOe,the slope of the effective demagnetization field, @H?/@V, isevaluated to be 0.14 kOe/V for Ta-MTJ and $0.081 kOe/Vfor Ru-MTJ. These values correspond to the surface mag-netic anisotropy energy change for an applied electric fieldof $29 fJ/(Vm) for Ta-MTJ and þ18 fJ/(Vm) for Ru-MTJ,

FIG. 1. Tunnel magnetoresistance as a function of the in-plane magneticfield for (a) Ta-MTJ and (b) Ru-MTJ. The resistance-area product (RA) isplotted for the designed junction size area of 4' 1 lm2. The insets show thelayer structures of the magnetic tunnel junctions. The thicknesses of eachlayer in nanometers are given in parentheses.



which are similar to the magnitude of voltage effect esti-mated from the static magnetoresistance measurements.

Here, we discuss the origin of the observed opposite signfor the voltage effects in Ta-MTJ and Ru-MTJ. According to

first-principles calculations, the PMA is induced by Fe-Ohybridization at the FejMgO interface. Therefore, it isconsidered that the voltage effect originates from theCoFeBjMgO interface, not from the underlayerjCoFeB inter-face. Surprisingly, our experimental results show oppositesigns for the voltage effect depending on the underlayermaterial. A possible reason could be a difference in the spin-orbit strength in the underlayer. A second possible reason isthe difference in crystallinity because of the different under-layer lattices. The bulk lattice structures of Ta and Ru are bccand hcp, respectively. It is known that an fcc underlayer candisturb the crystallization of bcc FeCoB.30 Moreover, thecrystallization generally results from the diffusion of boronand its incorporation in the underlayer. Ta is particularlyeffective for absorbing boron,31 but Ru may be less efficient.A third possible reason could be a difference in the Fe-Ohybridization at FejMgO interface. It has been shown theoret-ically that the modification of the magnetic anisotropy by theapplied voltage is affected by the oxygen concentration in theinterfacial FeO layer.25 More generally, our result may becompared to PtjFe and PdjFe systems,24,26 which also showan opposite sensitivity to the voltage. This was explained bythe differences in the electronic structures around the Fermilevel. A similar effect probably occurs in our films, but firstprinciples calculations would be necessary to evidence it.Therefore, the reason for the origin of the effect of underlayeris still an open question.

In summary, we investigated the voltage effect on thePMA in CoFeBjMgO junctions with a Ta or Ru underlayer.In both the bias voltage dependence of the magnetoresistancecurve and homodyne detection measurements, we observedthe opposite voltage effects depending on the underlayer ofthe CoFeBjMgO junction. We hope that the present resultswill motivate theoretical studies in order to clarify the

FIG. 2. Normalized in-plane magnet-ization curves determined from thetunnel magnetoresistance for variousapplied voltages and bias-voltage de-pendence of the magnetic anisotropyenergy per area times the thickness ofCoFeB, EperptCoFeB, for (a) and (b) Ta-MTJ and (c) and (d) Ru-MTJ.Opposite directions for the voltage-induced magnetic anisotropy changewere observed for different underlayermaterials.

FIG. 3. rf-voltage-induced ferromagnetic resonance spectra for (a) Ta-MTJand (b) Ru-MTJ. The magnetic field was applied at an angle tilted 55! withrespect to the film plane.



which are similar to the magnitude of voltage effect esti-mated from the static magnetoresistance measurements.

Here, we discuss the origin of the observed opposite signfor the voltage effects in Ta-MTJ and Ru-MTJ. According to

first-principles calculations, the PMA is induced by Fe-Ohybridization at the FejMgO interface. Therefore, it isconsidered that the voltage effect originates from theCoFeBjMgO interface, not from the underlayerjCoFeB inter-face. Surprisingly, our experimental results show oppositesigns for the voltage effect depending on the underlayermaterial. A possible reason could be a difference in the spin-orbit strength in the underlayer. A second possible reason isthe difference in crystallinity because of the different under-layer lattices. The bulk lattice structures of Ta and Ru are bccand hcp, respectively. It is known that an fcc underlayer candisturb the crystallization of bcc FeCoB.30 Moreover, thecrystallization generally results from the diffusion of boronand its incorporation in the underlayer. Ta is particularlyeffective for absorbing boron,31 but Ru may be less efficient.A third possible reason could be a difference in the Fe-Ohybridization at FejMgO interface. It has been shown theoret-ically that the modification of the magnetic anisotropy by theapplied voltage is affected by the oxygen concentration in theinterfacial FeO layer.25 More generally, our result may becompared to PtjFe and PdjFe systems,24,26 which also showan opposite sensitivity to the voltage. This was explained bythe differences in the electronic structures around the Fermilevel. A similar effect probably occurs in our films, but firstprinciples calculations would be necessary to evidence it.Therefore, the reason for the origin of the effect of underlayeris still an open question.

In summary, we investigated the voltage effect on thePMA in CoFeBjMgO junctions with a Ta or Ru underlayer.In both the bias voltage dependence of the magnetoresistancecurve and homodyne detection measurements, we observedthe opposite voltage effects depending on the underlayer ofthe CoFeBjMgO junction. We hope that the present resultswill motivate theoretical studies in order to clarify the

FIG. 2. Normalized in-plane magnet-ization curves determined from thetunnel magnetoresistance for variousapplied voltages and bias-voltage de-pendence of the magnetic anisotropyenergy per area times the thickness ofCoFeB, EperptCoFeB, for (a) and (b) Ta-MTJ and (c) and (d) Ru-MTJ.Opposite directions for the voltage-induced magnetic anisotropy changewere observed for different underlayermaterials.

FIG. 3. rf-voltage-induced ferromagnetic resonance spectra for (a) Ta-MTJand (b) Ru-MTJ. The magnetic field was applied at an angle tilted 55! withrespect to the film plane.



Opposite signs of voltage-induced perpendicular magnetic anisotropychange in CoFeBjMgO junctions with different underlayers

Yoichi Shiota,1,2 Fr!ed!eric Bonell,1,2 Shinji Miwa,1,2 Norikazu Mizuochi,1 Teruya Shinjo,1

and Yoshishige Suzuki1,2,a)

1Graduate School of Engineering Science, Osaka University, 1-3 Machikaneyama, Toyonaka, Osaka 560-8531, Japan2CREST, Japan Science Technology, 4-1-8 Honcho, Kawaguchi, Saitama 332-0012, Japan

(Received 3 June 2013; accepted 7 August 2013; published online 23 August 2013)

We report a voltage-induced perpendicular magnetic anisotropy (PMA) change in sputter-depositedTajCoFeBjMgO and RujCoFeBjMgO junctions. The PMA change is quantitatively evaluated bythe field dependence of the tunneling magnetoresistance for various bias voltages. We find thatboth the sign and amplitude of the voltage effect depend on the underlayer, Ta or Ru, belowthe CoFeB layer. The rf voltage-induced ferromagnetic resonance spectra also support the underlayer-material-dependent direction of the voltage torque. The present study shows that the underlayer is oneof the key parameters for controlling the voltage effect. VC 2013 AIP Publishing LLC.[http://dx.doi.org/10.1063/1.4819199]

Voltage control of magnetic properties has been investi-gated in magnetic semiconductors,1–3 multiferroics,4,5 orhybrids with piezoelectric materials and ferromagnetic materi-als.6,7 Recently, a voltage-induced perpendicular magnetic ani-sotropy (PMA) change in ferromagnetic metals has been ofgreat interest for magnetization control in spintronics devices,because it enables room temperature operation with low-powerconsumption and coherent manipulation of the magnetization.Even though the electric field is screened at the metal/insulatorinterface, a significant modification of the surface magnetic an-isotropy by applying a voltage has been experimentally8–22 andtheoretically23–27 observed, provided that the films are thinenough. Concerning the sign of the voltage effect, the PMAincreases as electrons are depleted from the ferromagnetic layerin PtjFePtjelectrolyte,8 AujFe(FeCo)jMgO,9,10 AgjFejMgO,18

and TajCoFeBjMgO,11,16,20–22 whereas it decreases inPdjFePdjelectrolyte,8 PdjFePdjMgO,12 and AujFePtjMgO.13

Because the PMA is highly sensitive to the interfaces (hybrid-ization, oxidation, interdiffusion, roughness, etc.), both interfa-ces with the metal underlayer and dielectric should play animportant role in the voltage effect for PMA. So far, effortshave focused on using high capacitance ion gels17 or solid-stateinsulators with high dielectric constant,21 but little is knownabout the role of the underlayer.

In this letter, we investigate the voltage-induced PMAchange in CoFeBjMgO junctions with a Ta or Ru underlayerand find that the sign of the voltage-induced PMA is affectedby the underlayer material. This is confirmed by both staticmagnetoresistance and rf voltage-induced ferromagnetic res-onance measurements.

The sample structures are schematically illustrated inthe insets of Fig. 1. Magnetic tunnel junctions (MTJs) weredeposited on surface-oxidized Si substrates using magnetronsputtering (Canon ANELVA, E-880S-M) with the followingstructures: bufferjTa (5 nm)jCo16Fe64B20 (1.4 nm)jMgO(2.0 nm)jthick Co16Fe64B20jcap and bufferjTa (5 nm)j

Ru (5 nm)jCo16Fe64B20 (1.4 nm)jMgO (2.0 nm)jthick Co16

Fe64B20jcap (hereafter called Ta-MTJ and Ru-MTJ, respec-tively). The MTJ films were patterned into pillar-shapedjunctions (4! 1 lm2) by electron-beam lithography and Arion milling. Since the films with small saturation field areeasy to observe the voltage-induced magnetic anisotropychange, the annealing temperature of 300 "C for 1.0 h(Ta-MTJ) or 200 "C for 0.5 h (Ru-MTJ) was employed tomake the CoFeB layers with small saturation field. Note that,according to Refs. 11 and 22, the anneal condition does notinfluence the sign of the voltage effect in TajCoFeBjMgOjunctions. Further, a top thick CoFeB layer, which hasin-plane magnetic easy axis, was used as a reference layer.Tunneling magnetoresistance (TMR) measurements werecarried out using a conventional two-terminal technique withmagnetic fields applied along the long axis of the junctions.In addition, rf voltage-induced ferromagnetic resonancemeasurements, as previously introduced in Ref. 19, were car-ried out by employing the homodyne detection technique.28

In our definition, a positive bias voltage accumulated elec-trons at the thin CoFeBjMgO interface. All measurementswere carried out at room temperature.

Figures 1(a) and 1(b) show the TMR curve for Ta-MTJand Ru-MTJ measured at a bias voltage of 50 mV. As themagnetic easy axis of thin CoFeB layer in both MTJs is per-pendicular to the film plane, typical hard-axis TMR curvesare observed. The resistance-area product in the parallelmagnetization configuration is 15.6 kXlm2 for Ta-MTJ and65.4 kXlm2 for Ru-MTJ. From the surface roughness mea-surement by using an atomic force microscope, the Taunderlayer is more flatness than Ru underlayer; the root-mean-square of the surface roughness was 0.31 nm and0.50 nm in Ta and Ru underlayer, respectively. The largerroughness of underlayer may result in the poorer crystalliza-tion of CoFeBjMgO structure, and then Ru-MTJ has higherRA value than Ta-MTJ.

We determine the normalized in-plane component of theCoFeB magnetization, M//,norm, from the relation betweena)[email protected]

0003-6951/2013/103(8)/082410/4/$30.00 VC 2013 AIP Publishing LLC103, 082410-1

APPLIED PHYSICS LETTERS 103, 082410 (2013)


Relative work function: ϕTa-ϕFe=-0.25 eV <0 ϕRu-ϕFe=0.21 eV >0

Summary• PMAmechanism due to Rashba effect• Two energy gain processes: – Reinforced exchange splitting -‐> perp. MA– Residual Rashba splitting -‐> in-‐plane MA

• Magnetic anisotropy energy scales with Rashba energy ER and can be changed by applied field E.

Ref.: Scientific Reports 4, 4105 (2014).

Documents

Perpendicular,magnetic,anisotropy,induced, by Rashba ... · negative bias voltages enhanced the perpendicular anisotropy, in contrast with the odd function dependence previously observed