# Fileset

[cigs_tmr.pdf](https://mdr.nims.go.jp/filesets/ff422efb-f2fb-4bef-9721-f7f31a6afbe5/download)

## Creator

[Keisuke Masuda](https://orcid.org/0000-0002-6884-6390), [Yoshio Miura](https://orcid.org/0000-0002-5605-5452)

## Rights

© 2017 The Japan Society of Applied Physics
<br>
This is an author-created, un-copyedited version of an article accepted for publication/published
in Japanese Journal of Applied Physics. IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or any version derived from it. The Version of Record is available online at https://doi.org/10.7567/jjap.56.020306.[In Copyright](http://rightsstatements.org/vocab/InC/1.0/)

## Other metadata

[First-principles study on magnetic tunneling junctions with semiconducting CuInSe2and CuGaSe2barriers](https://mdr.nims.go.jp/datasets/8f709ef5-aaff-49f2-946a-319bcaf4f53c)

## Fulltext

Japanese Journal of Applied Physics RAPID COMMUNICATIONFirst-principles study on magnetic tunneling junctions with semiconducting CuInSe 2 andCuGaSe2 barriersKeisuke Masuda1 and Yoshio Miura1,2,3,4 ∗1Research Center for Magnetic and Spintronic Materials, National Institute for Materials Science (NIMS), Tsukuba, Ibaraki 305-0047,Japan2Kyoto Institute of Technology, Electrical Engineering and Electronics, Kyoto 606-8585, Japan3Center for Materials Research by Information Integration, National Institute for Materials Science (NIMS), Tsukuba, Ibaraki 305-0047,Japan4Center for Spintronics Research Network (CSRN), Graduate School of Engineering Science, Osaka University, Toyonaka, Osaka560-8531, JapanWe theoretically investigate two magnetic tunneling junctions (MTJs) with different semiconductor barriers, CuInSe2 (CIS) andCuGaSe2 (CGS), sandwiched between Fe electrodes. We find that ∆1 wave functions provide dominant contributions to spin-dependent tunneling transport in both CIS- and CGS-based MTJs. We also find that the CGS-based MTJ has a much highermagnetoresistive (MR) ratio than the CIS-based MTJ, which indicates that a higher MR ratio is expected for a higher Ga concentra-tion x in the recently reported CuIn1−xGaxSe2-based MTJs. Furthermore, we show that the CIS- and CGS-based MTJs have muchsmaller resistance-area products (RA) than the conventional MgO-based MTJs.Magnetoresistive (MR) devices with high MR ratios andsmall resistance-area products (RA) are required for realizingread sensors of ultrahigh-density hard disk drives and Gbit-class spin transfer torque magnetoresistive random accessmemories (STT-MRAMs). Various attempts have been madeto reduce theRA of MgO-based magnetic tunneling junc-tions (MTJs)1,2) to less than 1Ω µm2 while keeping high tun-neling magnetoresistance (TMR) ratios. Elaborate techniquesto deposit ultrathin MgO barriers (∼1 nm) have been estab-lished,3–5) enabling the reduction inRA to ∼1Ω µm2 whilekeeping the high MR ratio of around 200% at room temper-ature. On the other hand, the use of half-metallic Co-basedHeusler alloys as ferromagnetic (FM) electrodes increasedthe MR ratio in small-RA current-perpendicular-to-plane gi-ant magnetoresistive (CPP-GMR) devices.6–11) The highestMR ratio reported so far is 82% at room temperature forthe Co2FeGa0.5Ge0.5/Ag/Co2FeGa0.5Ge0.5 system withRAof∼40 mΩ µm2.11) However, thisRA value is too small to ob-tain sufficiently high voltage output under the current for readsensor applications. Most recently, Kasaiet al. reported highMR ratios of 40% at room temperature and 100% at 8 K inthe MTJ using the compound semiconductor CuIn0.8Ga0.2Se2(CIGS) with the chalcopyrite crystal structure as a barrier incombination with Co2FeGa0.5Ge0.5 FM layers.12) This is thefirst observation of high MR output for MTJs with a com-pound semiconductor barrier. Since the band gap of the CIGSis much smaller than that of the insulator MgO, a smallerRA is expected. Actually, sufficiently smallRA values rang-ing from 0.3 to 3Ω µm2 were observed in the CIGS-basedMTJs. Moreover, these MTJs are expected to have high con-trollability and high breakdown voltage because their barrierthicknesses (∼2 nm) are two times larger than those of theabove-mentioned MgO-based MTJs with comparably smallRA. Such semiconductor barriers open up another path to-wards realizing both smallRAand high MR output.From the theoretical point of view, several previous stud-ies have focused on the transport properties of the MTJs∗E-mail: MIURA.Yoshio@nims.go.jpwith semiconductor barriers. MacLarenet al.13) studied MTJsconsisting of Fe electrodes and a ZnSe semiconductor bar-rier within the first-principles calculations using the layerKorringa−Kohn−Rostoker approach. They showed that thesystem has spin-dependent tunneling transport properties, inwhich ∆1 bands provide dominant contributions. From theirdata, the MR ratio was estimated to be∼500% for the bar-rier of 50 atomic units (∼2.6 nm). In another theoretical work,Autèset al.14) discussed the MTJ composed of a GaAs bar-rier sandwiched between Fe electrodes. They calculated spin-dependent conductance and predicted a maximum MR ratioof nearly 400% for around 10 atomic layers of GaAs. More-over, they considered the spin-orbit interaction and found thatthe effect of such interaction is significant for sufficientlythick barriers (≫20 atomic layers,∼2.8 nm). Although thesetheoretical approaches predicted high MR ratios, such a no-table output has not been experimentally observed in theZnSe- and GaAs-based MTJs; only a low MR ratio (<2%)has been reported in the GaAs-based MTJs.15,16)On the otherhand, no theoretical studies have focused on CIGS-basedMTJs in spite of the recent report of high MR output. To un-derstand the origin of such a high MR ratio in small-RATMRdevices, a theoretical understanding of the transport proper-ties of FM/CIGS/FM MTJs is essential.In this work, we study transport properties of two MTJswith different barriers, CuInSe2 (CIS) and CuGaSe2 (CGS),which are the terminal compounds of a CuIn1−xGaxSe2 mixedcrystal. Since the band gap of CuIn1−xGaxSe2 continuouslyincreases asx increases, CIGS is located between CIS andCGS not only chemically but also physically. Therefore, thepresent study of the CIS and CGS terminal compounds is ade-quate for obtaining sufficient information on the CIGS-basedMTJ. As electrodes, we adopt ferromagnetic bcc Fe with awell-known band structure. Since thea-axis length of CIS(CGS) is almost twice as large as that of bcc Fe, the latticemismatch between them is expected to be small. Note that wedo not consider the effect of the spin-orbit interaction in thiswork, because we focus on a thin barrier of∼2 nm, in which1Jpn. J. Appl. Phys. RAPID COMMUNICATIONFe FeFeFe FeFeFe FeSeFe FeFeFe FeFeFe FeSeSeSeSeSeSeSeSeSeSeSeSeSeSeSeSeSeCuCuCuCuCuCuCuCuCuCu Cu CuXXX XXXXXX XXXX = In, GaxzFig. 1. (Color online) Schematic of the supercell used in this study.the effect of the spin-orbit interaction is sufficiently small asshown in a related study on an MTJ with a GaAs barrier.14)We prepared supercells of Fe/CIS/Fe and Fe/CGS/Fe (Fig.1), each of which includes 2 unit cells (=17 layers) of CISor CGS and 1 unit cell (=3 layers) of Fe on both sides of thebarrier. This barrier length is comparable to 2 nm estimatedin experiments on the CIGS-based MTJ.12) We fixed thea-axis length to 0.5782 nm in the case of CIS17) and 0.5614 nmin the case of CGS.18) As termination layers, we selected Selayers on the basis of the TEM observation results in the ex-perimental work.12) We next optimized the positions of atomsin the supercells using the density functional theory withinthe generalized gradient approximation implemented in theViennaab-initio simulation program (VASP).19,20) In this op-timization, we used a 10× 10× 1 k-point mesh and assumedthat the spins of all Fe atoms align parallel to each other.As a result of the calculation, the distance between Fe andSe layers in the CIS-based (CGS-based) supercell is deter-mined to be∼0.167 nm (∼0.144 nm) in the left boundary andas∼0.168 nm (∼0.151 nm) in the right boundary. Such a dif-ference in distance between the left and right boundaries isdue to the lack of inversion symmetry along thec-axis in CISand CGS.To discuss the transport properties of the CIS- and CGS-based MTJs, we consider the quantum open system composedof the above-mentioned supercell attached to the left and rightsemi-infinite electrodes of Fe atoms. The conductance wascalculated with the aid of the quantum code ESPRESSO.21)In the present work, the Coulomb repulsionU for the Cu 3dstates in the barriers was considered to investigate the changein MR ratio upon changing the amplitude of the band gap sys-tematically.22,23)First, we obtained the wave functions in eachregion of the quantum open system by means of the densityfunctional theory and the generalized gradient approximation.The number ofk points was taken to be 10× 10 × 1, andMethfessel-Paxton smearing with the broadening parameter0.01 Ry was used. The cutoff energies for the wave functionand charge density were set to be 30 and 300 Ry, respectively.Since our system has a two-dimensional periodicity in thexyplane, the scattering states can be classified by an in-planewave vectork∥ = (kx, ky). For each fixedk∥ and spin index,we solved the scattering equations derived under the conditionthat the wave function and its derivative in the supercell areconnected to those in the electrodes.24,25) In this process, wecan also obtain the complex band structures, which are use-ful for understanding the transport properties of the system.Conductance is calculated by substituting the amplitudes ofthe scattering wave functions into the Landauer formula.We first set the Coulomb repulsionU to 5 eV and focused1.0-1.00.5-0.50.0(a) CIS (b) CGSFig. 2. (Color online) Real and complex band structures of (a) CIS and(b) CGS withU = 5 eV atk∥ = (0,0) along the out-of-plane wave vectorkz.on the difference between the CIS- and CGS-based MTJs.Figures 2(a) and 2(b) show the real and complex band struc-tures of the CIS and CGS, respectively, atk∥ = (0,0) alongthe out-of-plane wave vectorkz. The band gaps are estimatedas ECISg ≃ 0.17 eV for CIS andECGSg ≃ 0.41 eV for CGS.Although these values are smaller than the experimental ob-servations (ECISg ≃ 1.0 eV andECGSg ≃ 1.7 eV), the magnituderelation (ECISg < ECGSg ) is the same for the present calculationsand experiments. We see that the complex band with the∆1components has the smallest imaginary partκmin = |Im(kz)|minaround the Fermi level in both CIS and CGS. This means thatthe propagating state with the∆1 components in the electrodecouples to the evanescent state (κmin) in the barrier and pro-vides the largest contribution to the tunneling conductance.26)In Fig. 3, we show the in-plane wave vectork∥ = (kx, ky)dependence of the conductances at the Fermi energy for var-ious situations in the CIS- and CGS-based MTJs. The uppertwo panels, Figs. 3(a) and 3(d), show the majority-spin con-ductances of the CIS- and CGS-based MTJs with the paral-lel magnetization of the electrodes, in which the sharp peaksaroundk∥ = (0,0) are seen for both MTJs. As shown inFigs. 2(a) and 2(b), since the∆1 evanescent state is the dom-inant conducting channel atk∥ = (0,0) in the barriers, thesepeaks can be considered as evidence of the tunneling trans-port by the∆1 wave functions. Figures 3(b) and 3(e) showthe minority-spin conductances of the CIS- and CGS-basedMTJs with the parallel magnetization of the electrodes. Com-pared with the majority-spin cases, conductances have muchsmaller values and are distributed over a wide region of thek∥ Brillouin zone for both CIS- and CGS-based MTJs. Thelower two panels, Figs. 3(c) and 3(f), show the conductancesof the CIS- and CGS-based MTJs for the majority-spin statesof the left electrodes in the case of the antiparallel magnetiza-tion. Although the∆1 wave function in the left electrode de-cays more rapidly than in the case of parallel magnetization,it still has a small amplitude in the right electrode after pass-ing through the barrier, which is the origin of the small con-ductances aroundk∥ = (0,0). Note that thek∥ dependencesin Figs. 3(c) and 3(f) break the fourfold rotational symmetry,2Jpn. J. Appl. Phys. RAPID COMMUNICATION0.30.20.10.00.80.60.40.20.00.080.060.040.020.00(a) CIS, majority-spin, parallel(b) CIS, minority-spin, parallel(c) CIS, majority-spin, antiparallel0.80.60.40.20.0(d) CGS, majority-spin, parallel0.0080.0060.0040.0020.000(e) CGS, minority-spin, parallel0.080.060.040.020.00(f) CGS, majority-spin, antiparallel000000000000Fig. 3. (Color online) In-plane wave vectork∥ = (kx, ky) dependence ofthe conductances at the Fermi energy for various cases in CIS- andCGS-based MTJs withU = 5 eV. (See the text for details.) The unit of thecolor bar isG0 = e2/h in all panels.which might be due to the twofold symmetry of CIS and CGSin the xy plane. Although not shown here, the minority-spinconductances in the case of antiparallel magnetization havek∥ dependences that are identical to those obtained by rotat-ing conductance distributions in Figs. 3(c) and 3(f) by 180◦ inthek∥ plane.In this work, we adopt the usual optimistic definition ofthe MR ratio: MR ratio [%]= 100× (TP − TAP)/TAP, whereTP (TAP) is the sum of the majority- and minority-spin con-ductances averaged over thek∥ Brillouin zone in the case ofparallel (antiparallel) magnetization. We obtained 62.3 and300.9% MR ratios for the CIS- and CGS-based MTJs, respec-tively. The difference is mainly due to a large difference inconductance in the case of antiparallel magnetization [Figs.3(c) and 3(f)]. We also estimatedRA values from the con-ductances in the case of the parallel magnetizationTP, where0.408 and 0.680Ω µm2 were obtained for the CIS- and CGS-based MTJs, respectively. In the present study, the in-planelattice constant of Fe/CIS(CGS)/Fe supercells is twice as longas that of bcc Fe (see Fig. 1). Since CIS and CGS have quitesmall but finite displacements in Se atoms from fractional po-sitions,27) we need to use a larger supercell as compared withother semiconductors (Si, GaAs or ZnSe). In such a largersupercell, the band folding of Fe can reduce MR ratios, asdiscussed for Fe/MgAl2O4/Fe(001) MTJs.28,29) However, inthe present case, the folded band hardly affects MR ratios ofthe CIS- and CGS-based MTJs. Actually, we confirmed thatthe folded minority-spin band of Fe crossing the Fermi levelprovides no contribution to the conductance in both MTJs be- 0 50 100 150 200 250 300 350 400 450 0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9  1.0  1.1CISCGSU = 0 eVU = 0 eV5 eV10 eV5 eV10 eVCGS ( U = 0, 5, 10 eV,  tCGS = 2.64 nm )MgOtMgO = 1.24 nm1.64 nm2.05 nm2.45 nm 2.86 nm 3.26 nmCIS ( U = 0, 5, 10 eV,  tCIS = 2.76 nm )(a)(b)Fig. 4. (Color online) MR ratios andRAvalues of (a) CIS- andCGS-based MTJs and (b) CIS-, CGS-, and MgO-based MTJs on adouble-logarithmic scale. In the panel (b), barrier thickness (tCIS, tCGS, ortMgO) is defined as the distance between two Fe layers closest to the barrier.cause of the very small displacements of Se atoms.Let us further discuss the relationship between the bandgaps and MR ratios by changing the Coulomb repulsionUfor the Cu 3d states in the barriers. By increasing the repul-sion U from 0 to 10 eV, the band gap of bulk CIS (CGS) isincreased from 0.044 (0.114) to 0.468 (0.691) eV. In Fig. 4(a),we show the MR ratios andRAvalues of the CIS- and CGS-based MTJs withU = 0,5, and 10 eV. As the repulsionUbecomes larger, the MR ratio andRA increase in both MTJs.It is also found that for a fixed repulsionU, the CGS-basedMTJ has a higher MR ratio and a largerRAthan the CIS-basedMTJ. From these, we can conclude, at least for the CIS- andCGS-based MTJs, that a larger gap system has a higher MRratio and a largerRA.Figure 4(b) shows the MR ratios andRAvalues of the CIS-, CGS-, and MgO-based MTJs. The MR ratio andRAof theMgO-based MTJ are calculated in the same manner as thoseof the CIS- and CGS-based MTJs. For the barrier thickness of2.6−2.8 nm, theRAvalues of the CIS- and CGS-based MTJsare nearly six orders of magnitude smaller than those of theMgO-based MTJs, which is a great advantage of the CIS-and CGS-based MTJs for certain device applications wherea smallRA is needed, e.g., read sensors. Even if we reducethe thickness of the MgO to 1.24 nm,RA is still larger thanthose of the CIS and CGS systems. On the other hand, theCIS- and CGS-based MTJs have the possibility of achiev-3Jpn. J. Appl. Phys. RAPID COMMUNICATIONing even smallerRA values by reducing their barrier thick-nesses. In that case, establishing a method to keep the MR ra-tio high, e.g., the use of highly spin-polarized ferromagneticelectrodes, is essential. In fact, such an example has recentlybeen reported by Kasaiet al.12)Finally, let us discuss the reason why only low MRratios have been observed in the GaAs- and ZnSe-basedMTJs,15,16,30) as opposed to the CIGS-based MTJ. To thisend, we calculated MR ratios of Fe/GaAs/Fe and Fe/ZnSe/FeMTJs using supercells with 17 barrier layers and 3 Fe layerson both sides of the barrier. The termination layers of GaAsand ZnSe were determined to be As and Se layers, respec-tively, from the energy minimization of the supercells. Allconditions in the transport calculations were the same as thosefor the CIS- and CGS-based MTJs. In the GaAs-based MTJ,we obtained an MR ratio of 12%, which is more than oneorder smaller than the value reported in the previous theoret-ical work.14) Such a difference comes from the sensitivity ofthe MR ratio in the GaAs-based MTJ to the position of theFermi level.14) In general, the Fermi level of magnetic junc-tions strongly depends on the in-plane lattice constant, the in-terfacial distance, and other calculation conditions, leading todifferent positions of the Fermi level in each calculation. Inour case, the Fermi level is located near the interfacial reso-nance states of Fe, which yields a low MR ratio. On the otherhand, in Ref. 14, the Fermi level was set to the middle of theband gap in the barrier, where the effect of the interfacial reso-nance states is small and a high MR ratio is obtained. As men-tioned in the introduction, experiments on Fe/GaAs/Fe MTJshave revealed low MR ratios (<2%),15,16) which can be ex-plained as an effect of the interfacial resonance states. In thecase of the ZnSe-based MTJ, we obtained a high MR ratio of292%, which is consistent with the previous theoretical esti-mation.13) We also found that the interfacial resonance statesdo not provide a significant contribution to the MR ratio inthis system. Therefore, a low MR ratio (∼10%) observed inexperiments30) would be due to experimental imperfections.Actually, the existence of a large film roughness (∼0.9 nm) isconfirmed in Ref. 30.In summary, we studied transport properties of MTJs withsemiconductor barriers, Fe/CIS/Fe and Fe/CGS/Fe. Our first-principles-based calculations showed that∆1 wave functionsdominate the tunneling transport in both MTJs. The theoreti-cal transport calculations predicted MR ratios of around 50%for the CIS-based MTJ and around 300% for the CGS-basedone, which means that a higher MR ratio is expected for ahigher Ga concentrationx in CuIn1−xGaxSe2-based MTJs. Wefurther discussed the relationship between the band gap andMR ratio by changing the Coulomb repulsion in the CIS andCGS barriers. We found that a larger band gap in the bar-rier gives a higher MR ratio. Through the comparison withthe MgO-based MTJ, we confirmed that the CIS- and CGS-based MTJs have much smallerRA values than the MgO-based MTJ, which is consistent with the experimental resultsof the CuIn0.8Ga0.2Se2-based MTJ.12)Acknowledgment The authors are grateful to K. Hono, S. Kasai, and K.Mukaiyama for useful discussions and critical comments. This work wasin part supported by Grant-in-Aids for Scientific Research (S) (Grant No.16H06332) and (B) (Grant No. 16H03852) from the Ministry of Education,Culture, Sports, Science and Technology, Japan, by NIMS MI2I, and also bythe ImPACT Program of Council for Science, Technology and Innovation,Japan.1) S. S. P. Parkin, C. Kaiser, A. Panchula, P. M. Rice, B. Hughes, M.Samant, and S.-H. Yang, Nat. Mater.3, 862 (2004).2) S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Nat.Mater.3, 868 (2004).3) Y. S. Choi, Y. Nagamine, K. Tsunekawa, H. Maehara, D. D.Djayaprawira, S. Yuasa, and K. Ando, Appl. Phys. Lett.90, 012505(2007).4) S. Isogami, M. Tsunoda, K. Komagaki, K. Sunaga, Y. Uehara, M. Sato,T. Miyajima, and M. Takahashi, Appl. Phys. Lett.93, 192109 (2008).5) H. Maehara, K. Nishimura, Y. Nagamine, K. Tsunekawa, T. Seki, H.Kubota, A. Fukushima, K. Yakushiji, K. Ando, and S. Yuasa, Appl.Phys. Express4, 033002 (2011).6) K. Yakushiji, K. Saito, S. Mitani, K. Takanashi, Y. K. Takahshi, and K.Hono, Appl. Phys. Lett.88, 222504 (2006).7) K. Nikolaev, P. Anderson, P. Kolbo, D. Dimitroc, S. Xue, X. Peng, T.Pokhil, H. Chho, and Y. Chen, J. Appl. Phys.103, 07F533 (2008).8) T. Furubayashi, K. Kodama, H. Sukegawa, Y. K. Takahashi, K.Inomata, and K. Hono, Appl. Phys. Lett.93, 122507 (2008).9) J. Sato, M. Oogane, H. Naganuma, and Y. Ando, Appl. Phys. Express4, 113005 (2011).10) S. Li, Y. K. Takahashi, T. Furubayashi, and K. Hono, Appl. Phys. Lett.103, 042405 (2013).11) J. W. Jung, Y. Sakuraba, T. T. Sasaki, Y. Miura, and K. Hono, Appl.Phys. Lett.108, 102408 (2016).12) S. Kasai, Y. K. Takahashi, P.-H. Cheng, Ikhtiar, T. Ohkubo, K.Kondou, Y. Otani, S. Mitani, and K. Hono, Appl. Phys. Lett.109,032409 (2016).13) J. M. MacLaren, X. G. Zhang, W. H. Butler, and X. Wang, Phys. Rev.B 59, 5470 (1999)14) G. Autès, J. Mathon, and A. Umerski, Phys. Rev. B82, 115212 (2010).15) S. Kreuzer, J. Moser, W. Wegscheider, and D. Weiss, Appl. Phys. Lett.80, 4582 (2002).16) J. Moser, M. Zenger, C. Gerl, D. Schuh, R. Meier, P. Chen, G.Bayreuther, W. Wegscheider, D. Weiss, C.-H. Lai, R.-T. Huang, M.Kosuth, and H. Ebert, Appl. Phys. Lett.89, 162106 (2006).17) J. Parkes, R. D. Tomlinson, and M. J. Hampshire, J. Appl. Crystallogr.6, 414 (1973).18) H. W. Spiess, U. Haeberlen, G. Brandt, A. Räuber, and J. Schneider,Phys. Status Solidi B62, 183 (1974).19) G. Kresse and J. Furthm̈uller, Phys. Rev. B54, 11169 (1996).20) G. Kresse and D. Joubert, Phys. Rev. B59, 1758 (1999).21) S. Baroni, A. Dal Corso, S. de Gironcoli, and P. Giannozzi, QuantumESPRESSO package. This is open-source computer codes forelectronic-structure calculations on the basis of the density functionaltheory, plane waves, and pseudopotentials. For more information, seehttp://www.pwscf.org.22) V. I. Anisimov, J. Zaanen, and O. K. Andersen, Phys. Rev. B44, 943(1991).23) G. Sclauzero and A. Dal Corso, Phys. Rev. B87, 085108 (2013).24) H. J. Choi and J. Ihm, Phys. Rev. B59, 2267 (1999).25) A. Smogunov, A. Dal Corso, and E. Tosatti, Phys. Rev. B70, 045417(2004).26) W. H. Butler, X.-G. Zhang, T. C. Schulthess, and J. M. MacLaren,Phys. Rev. B63, 054416 (2001).27) J. E. Jaffe and A. Zunger, Phys. Rev. B29, 1882 (1984).28) Y. Miura, S. Muramoto, K. Abe, and M. Shirai, Phys. Rev. B86,024426 (2012).29) H. Sukegawa, Y. Miura, S. Muramoto, S. Mitani, T. Niizeki, T.Ohkubo, K. Abe, M. Shirai, K. Inomata, and K. Hono, Phys. Rev. B86, 184401 (2012).30) X. Jiang, A. F. Panchula, and S. S. P. Parkin, Appl. Phys. Lett.83,5244 (2003).4