Engineering spin-orbit torque in CoPt multilayers with perpendicular magnetic anisotropy

Engineering spin-orbit torque in Co/Pt multilayers with perpendicular magnetic anisotropyKuo-Feng Huang, Ding-Shuo Wang, Hsiu-Hau Lin, and Chih-Huang LaiCitation: Applied Physics Letters 107, 232407 (2015); doi: 10.1063/1.4937443View online: /10.1063/1.4937443View Table of Contents: /content/aip/journal/apl/107/23?ver=pdfcovPublished by the AIP PublishingArticles you may be interested inSpin-orbit torque opposing the Oersted torque in ultrathin Co/Pt bilayersAppl. Phys. Lett. 104, 062401 (2014); 10.1063/1.4864399Interlayer exchange coupled composite free layer for CoFeB/MgO based perpendicular magnetic tunnel junctions J. Appl. Phys. 114, 203901 (2013); 10.1063/1.4833252Effect of annealing on the magnetic tunnel junction with Co/Pt perpendicular anisotropy ferromagnetic multilayers J. Appl. Phys. 107, 09C711 (2010); 10.1063/1.3358249The intermixing induced perpendicular magnetic anisotropy in ultrathin Co/Pt multilayersJ. Appl. Phys. 105, 07C316 (2009); 10.1063/1.3070639Magnetization reversal and field annealing effects in perpendicular exchange-biased Co ∕ Pt multilayers and spin valves with perpendicular magnetizationJ. Appl. Phys. 99, 083901 (2006); 10.1063/1.2180527Engineering spin-orbit torque in Co/Pt multilayers with perpendicular magnetic anisotropyKuo-Feng Huang,1Ding-Shuo Wang,1Hsiu-Hau Lin,2and Chih-Huang Lai 1,a)1Department of Materials Science and Engineering,National Tsing Hua University,Hsinchu 30013,Taiwan 2Department of Physics,National Tsing Hua University,Hsinchu 30013,Taiwan(Received 2October 2015;accepted 26November 2015;published online 10December 2015)To address thermal stability issues for spintronic devices with a reduced size,we investigate spin-orbit torque in Co/Pt multilayers with strong perpendicular magnetic anisotropy.Note that the spin-orbit torque arises from the global imbalance of the spin currents from the top and bottom interfaces for each Co layer.By inserting Ta or Cu layers to strengthen the top-down asymmetry,the spin-orbit torque efficiency can be greatly modified without compromised perpendicular mag-netic anisotropy.Above all,the efficiency builds up as the number of layers increases,realizing robust thermal stability and high spin-orbit-torque efficiency simultaneously in the multilayersstructure.VC 2015AIP Publishing LLC .[/10.1063/1.4937443]Manipulating magnetization by electrical means,suchas spin-transfer torque 1–4or electric-field control of magne-tism,5–7is an important task for realizing spintronic devices.Recently,the so-called spin-orbit-torque (SOT)has attracted intense attention because it provides an efficient way for manipulating magnetization and offers a route for spintronic device designs.The spin-orbit torque arises mainly from two distinct origins:the spin current generated by spin Hall effect 8or the effective magnetic field induced by Rashba interaction.9By simply applying a charge current through the asymmetrical structure composed of a ferromagnet (FM)layer and adjacent heavy metal layers with strong spin-orbit coupling,the spin-orbit torque can be generated in the FM and drives the magnetization reversal.10In consequence,SOT has been proposed as an alternative writing mechanism for the next generation magnetoresistive random access memory (MRAM),8,11especially to manipulate the FM layer with perpendicular magnetic anisotropy (PMA)for better thermal stability.12–14Nevertheless,most of the existing works are restricted to the single FM layer around 1nm in thickness for preserving the PMA,where the relatively poor thermal stability due to the limited volume of FM layer is a big concern when scaling down to several tens of nano-meters.In this work,we would like to evaluate the SOT in the Co/Pt multilayers (MLs),because Co/Pt MLs possess PMA with tunable volume of the FM layer.Furthermore,the combinations of Co/Pt MLs and the interfacial CoFeB layers show a large magnetoresistance (over 100%)in MgO-based magnetic tunnel junctions.15,16Although a high SOT has been reported on the Co/Pd MLs recently,14no systematic works have been done for explaining the presence of SOT in the multilayers,where the spin currents from heavy metal may counteract each other and no Rashba interaction can be seen without structure inversion asymmetry.By conducting the lock-in technique,17–19the dependence of SOT efficien-cies of the Co/Pt MLs on different repeating layer number (N )is investigated.Besides,Cu and Ta are inserted into theCo/Pt MLs for modifying the interfacial conditions to clarify the origin of SOT in multilayers system and to tune the strength of SOT efficiency.We found that the magnitude of SOT in Co/Pt MLs strongly depends on the interfacial condi-tions.In addition,as the N increases,the FM volume and SOT efficiency increase accordingly,showing great potential for improving the thermal stability and the writability.By DC magnetron sputtering,we first prepared the Co/Pt MLs,sub./Ta(2.5nm)/[Pt(2nm)/Co(0.9nm)]N /Pt(2nm)/Ta(2.5nm),with varied repeating layer numbers (N ¼1,2,4,and 8)and the Pt/Co/Pt tri-layers with varied Pt thicknesses,sub./Ta(2.5nm)/Pt(x nm)/Co(0.9nm)/Pt(x nm)/Ta(2.5nm),on thermally oxidized Si (100)substrates.Furthermore,0.5nm Cu or 0.15nm Ta was inserted into some of the Co/Pt MLs on the top of each Co layers,denoted as Co/Cu/Pt and Co/Ta/Pt MLs,respectively.By a vibrating sample magne-tometer (VSM),we confirmed that all of those film structures show perpendicular magnetic anisotropy and extracted both the saturation magnetization-thickness product (M S *t ,acquired by normalizing the total magnetization by its area)and the magnetic anisotropy field (H K )for calculating the corresponding magnetic anisotropy energy (K eff ¼M S H K /2)from the in-plane and out-of-plane hysteresis loops.Then,these multilayers were patterned into Hall-bar (5l m in width for both the wire and bars,as shown in Figure 1(a))by photolithography and reactive-ion etching.On all of the Hall-bars,we further fabricated the Ta (10nm)/Pt (100nm)electrodes by photolithography,DC magnetron sputtering,and lift-off process sequentially to conduct the following measurements.To study the SOT quantitatively,the lock-in measurement with planar Hall effect correction 19was con-ducted for acquiring the in-phase first harmonic (V x )and the out-of-phase second harmonic (V 2x )Hall voltages.We defined the longitudinal direction as the direction along the applied current and transverse direction as the direction transverse to the current,as shown in Figure 1(a).The repre-sentative V x and V 2x curves are shown in Figure 1(b),revealing the resultant V x and V 2x with the sweeping exter-nal magnetic field along either the longitudinal (H L )or the transverse (H T )direction.The effective field alonga)Author to whom correspondence should be addressed.Electronic mail:chlai@.tw0003-6951/2015/107(23)/232407/5/$30.00VC 2015AIP Publishing LLC 107,232407-1APPLIED PHYSICS LETTERS 107,232407(2015)longitudinal and transverse directions (D H L and D H T ,respec-tively)can therefore be quantitatively evaluated from the measured V x and V 2x based on the formula derived by Hayashi et al.19D H L T ðÞ¼À2B L T ðÞ62n B T L ðÞ1À4n 2;(1)where the B L(T)defined as @V 2x @H =@2V x@H 2j H ==L ðT Þ,which is the ra-tio between the slope of V 2x and the curvature of V x along longitudinal (transverse)direction,the n defined as D R PHE /D R AHE representing the ratio of planar Hall and anomalous Hall resistances,and the 6sign corresponds to the direction of initial magnetization (6M )along out-of-plane axis.The D R AHE was obtained from the V x of þM and ÀM saturated states under zero external fields and its corresponding applied current;the D R PHE was evaluated from the ampli-tude of the in-plane angular dependence (Sin 2/)of the V x under an in-plane saturation field ($1.5T).By adopting Equation (1),we can obtain the D H L and D H T under different charge current density (J e )with a planar Hall effect correc-tion.To calculate the charge current density,J e ,we first measured the resistivity of a single layer Ta and full structure (Ta/[Co/Pt]N /Ta)MLs.We treated [Co/Pt]N as one unit and subtracted the current shutting in the Ta layers based on thethickness and the measured resistivity.Finally,we divided the current flowing through Co/Pt MLs by its cross-sectional area to get J e .We first evaluated the SOT in the [Co (0.9nm)/Pt (2nm)]8MLs by adopting the lock-in measurement 19to mea-sure the D H L and D H T with varied J e .The J e dependence of D H L(T)is shown in Figure 1(c).The magnitude of both D H L and D H T increases linearly with J e ,where the longitudinal and transverse spin-orbit-torque efficiencies (b L(T)),defined as D H L(T)/J e ,can therefore be extracted from the slope.The b L and b T of the [Co/Pt]8sample (K eff $4.6Â106erg/cm 3,M s *t $6.7Â10À4emu/cm 2)are 62and 24Oe cm 2/107A,respectively,showing a comparable b L(T)and K eff to the cases with single FM layer,12,20but with largely enhanced M S *t ,that is,improved thermal stability.16Besides,because the b L is 2.6times larger than the b T ,the damping-like torque arisen by the spin current from the giant spin Hall effect of Pt layers should be the main source of the SOT in the Co/Pt MLs.18,21Note that the symmetric structure of the Co/Pt MLs implies null SOT due to the cancellation of the spin currents from dif-ferent Pt layers.To address the unexpected SOT observed in Co/Pt MLs,we focus on the longitudinal SOT efficiency b L first.To simplify the system,we started with a series of tri-layers,Pt (x nm)/Co (0.9nm)/Pt (x nm),the basic repeating units comprising Co/Pt MLs,where the thicknesses of the top and bottom Pt are equal,denoted as a symmetric tri-layer.Due to the equal thickness of the top and bottom Pt,the bulk contribution of the spin current from the Pt layers due to spin Hall effect is expected to be the same magnitude but with opposite spin directions,cancelling each other with no b L .12,22However,as shown in Figure 1(d),we still observed significant b L (about 21–33Oe cm 2/107A)in the symmetric tri-layers,indicating the spin currents from the Pt layers are not completely cancelled.We speculate that the difference between top and bottom interfaces is the key for the presence of b L in the symmetric tri-layers.The different interfacial structures can enormously change the transport of spin current,for example,modifying the ratio of spin back-flow 23,24or changing the degree of spin-flip scattering.25Considering the interfaces of Co/Pt MLs,the different inter-facial structures between the Pt/Co and the Co/Pt were reported in the sputtered Co/Pt MLs;26the bottom Pt/Co shows chemically sharp interfaces but the top Co/Pt suffers from the intermixing during sputtering,which may be the reason for the presence of b L .Based on the direction of b L ,which is identical to the reported b L of the Pt/Co/Pt tri-layers with thicker top Pt layer,27the unbalanced spin current may come from the top Pt,suggesting a higher transmission ratio of spin current through the top Co/Pt interface.Besides,the b L also shows thickness dependence.If the difference between top and bottom interfaces is not a strong function of Pt thickness,the b L increase with increasing Pt thickness in symmetric Pt/Co/Pt tri-layers may mainly result from the Pt bulk contribution,that is,larger thickness generates more spin current.22On the other hand,the interface difference might be varied by increasing Pt thickness due to more inter-mixing,which might also contribute to b L increase.However,this effect should be confined to the certain Pt thickness,thicker than which should not significantlyalterFIG.1.(a)Sketch of the lock-in measurement setup for measuring the spin-orbit-torque effective fields in both of the longitudinal (D H L )and the trans-verse (D H T )directions.(b)Representative in-phase first harmonic,V x (left two figures),and the out-of-phase second harmonic,V 2x (right),Hall voltage curves obtained with the sweeping external magnetic field along either the longitudinal,H L (top two figures),or the transverse,H T (bottom),direction for the Ta(2.5nm)/[Co(0.9nm)/Pt(2nm)]8/Ta(2.5nm)multilayers under 1.15Â107A/cm 2.(c)D H L and D H T of the Ta(2.5nm)/[Co(0.9nm)/Pt(2nm)]8/Ta(2.5nm)multilayers after planar Hall effect correction with different charge current density (J e ).The longitudinal (b L )and transverse (b T )spin-orbit-torque efficiencies defined as D H L(T)/J e of (d)the symmetric tri-layers film structures and (e)the tri-layers with thicker varied top Pt layer,with positive/negative initial magnetization states (þ/ÀM ).the interface.Furthermore,as shown in Figure 1(e),with increasing top Pt thickness (bottom keeps 2nm),denoted as asymmetric tri-layers,we found a monotonic increase of the magnitude of b L without sign change.The b L increase can be attributed to the enhancement of spin current density from the thicker top Pt layer,leading to an effective net spin cur-rent.Notice that the SOT generated from the bottom Ta layer should be a minor effect.Conductivity of Ta is much lower than Pt,so the current density in the bottom Ta layer is less than 20%of that through bottom Pt layer.Besides,the 2nm Pt may also screen part of the spin current generated by the Ta.Since the sign of b L is all identical for the symmetric and asymmetric tri-layers,we confirmed the effective net spin current in the symmetric tri-layers,which leads to the unex-pected b L ,is also from the top Pt.To verify the b L modification due to the discrepancy between top and bottom interfaces in multilayers,we pre-pared a series of Co/Pt MLs with different repeating layer numbers (N ¼1,2,4,and 8)and inserted Cu (0.5nm)or Ta (0.15nm)in some of the Co/Pt MLs at the top interfaces of each Co in the multilayers for altering the properties of top interfaces but keeping the bottom intact.For these multi-layers,we found that all the M S *t increases linearly with N ,as shown in Figure 2(a).The M S *t reveals slight reduction when the Cu or Ta was inserted,implying the top interfaces are modified by the insertions.Because of the different impacts on the Co layer during Cu and Ta sputtering (due to the different atomic mass)and the different chemical affinity with Co (immiscible for Cu;alloying for Ta),Ta insertion may create magnetic dead layer,28but Cu insertion can pre-vent the interdiffusion at Co/Pt interfaces.29In addition,insertion of Cu and Ta may both reduce proximity-induced moment in Pt.30,31Accordingly,both effects can contribute to the reduction of M S *t with Ta insertion,but the reduction of proximity-induced moment may be the main reason for Cu insertion.On the other hand,although the K eff shows a slight increase with increasing N or with the extra Cu inser-tion,the K eff remains in the same level with good PMA,as shown in Figure 2(b),indicating that the bottom interfaces (the origin of the PMA)29are preserved.Figures 3(a)–3(c)show that the measured b L and b T of Co/Pt,Co/Cu/Pt,and Co/Ta/Pt MLs with increasing N ,respectively.We found that the b L was enhanced with the extra Ta insertions,but degraded with the Cu insertions.Besides,the sign of b L remains the same for all samples,indicating the effective net spin current generated by the top Pt layers dominates.Note that the insertion of Cu at Co/Pt interfaces has been reportedto degrade the transmission of spin current through the inter-face either due to the enhanced spin-flip scattering 25or the suppressed spin mixing conductance.24,32,33As a result,the decreased b L may be attributed to the reduction of spin cur-rent transmission through the top interfaces,which lowers the disparity between top and bottom interfaces.On the con-trary,regarding the Ta insertions,possessing a giant spin Hall angle (h SH $À0.12)8but with the opposite sign to Pt,we observed the b L enhancement from 62Oe cm 2/107A for the [Co/Pt]8MLs to 88Oe cm 2/107A for the [Co/Ta/Pt]8MLs,shown in Figure 3(c).Similar Pt/Co/Ta heterostruc-tures have been reported by Woo et al.,34where the enhance-ments of spin-orbit torques in Pt/Co/Ta heterostructures are mainly attributed to the opposite sign of h SH for the bottom Pt and the top Ta.The Ta thickness they used is around 0.5–4nm,and thus,Ta may reveal its bulk properties of spin Hall effect to offer a spin current.However,in our case,if we consider the bulk contribution of Ta with its opposite sign of h SH to that of Pt,the combination of the Ta insertions and the top Pt is expected to counteract each other,and thus,lowers the value or even reverses the sign of b L ,which is contradictory to our observations.Besides,the 0.15nm Ta insertion in our [Co/Ta/Pt]N MLs is still in the region for forming magnetic dead layer,35as observed and discussed in Figure 2(a),where the bulk spin Hall effect may be limited or disappeared.Therefore,we suggest that the interfaces modification by forming a magnetic dead layer is the key factor to enhance the b L by the 0.15nm Ta insertions.The randomness in the magnetic dead layer may alter the spin-mixing conductance at the interfaces 23,36and leads to a more transparent top interface with larger discrepancy between the top and bottom interfaces based on the increased b L .Regarding the case in pure Co/Pt MLs,we speculate that the intermixing of the top Co/Pt interfaces may also alter the spin-mixing conductance as the magnetic dead layer intro-duced by Ta insertion layers does,which gains the transpar-ency of spin current,leading to a discrepancy between the chemical sharp bottom and the top interface with severe intermixing.Consequently,we believe that the b L are sensi-tive to the discrepancy between top and bottom interfaces in the multilayers system and can be tuned by modifying the interfaces.In addition,we also found that the b L enhanced with increasing N in all MLs with different insertion materials,asFIG.2.Variations with increasing repeating layer numbers,N ,of (a)satura-tion magnetization-thickness products (M S *t )and (b)effective magnetic ani-sotropy energy (K eff )for [Pt(2nm)/Co (0.9nm)]N ,[Pt(2nm)/Cu (0.5nm)/Co (0.9nm)]N ,and [Pt(2nm)/Ta (0.15nm)/Co (0.9nm)]NMLs.FIG.3.Longitudinal (b L )and transverse (b T )spin-orbit-torque efficiencies of (a)[Pt(2nm)/Co (0.9nm)]N multilayers,(b)[Pt(2nm)/Cu (0.5nm)/Co (0.9nm)]N multilayers,and (c)[Pt(2nm)/Ta (0.15nm)/Co (0.9nm)]N multi-layers with different repeating layer numbers,N ,and positive/negative initial magnetization states (þ/ÀM ).also shown in Figures3(a)–3(c).The enhanced b L indicates that the effective net spin current absorbed by each Co layer increases with N,where the total Pt thickness also increases even though the Pt layers are separated by the0.9nm Co layers.That is,the adjacent Pt layers may couple to each other across the Co layer to enhance the driving force of the spin current.To understand this N dependence of b L in Co/Pt MLs,we need to consider the spin diffusion length of Co (l sf,Co)and Pt(l sf,Pt)first,where Co possesses a relatively long l sf,Co(>30nm)37compared to the Pt($4nm).In the Co/Pt MLs,the spin accumulation at the interfaces due to spin Hall effect is the main driving force for the spin current flowing into each Co layers.When the FM layer(d F)is thick enough(d F)l sf,Co),where the spin accumulation at the interfaces can be totally relaxed or the spin current can be totally dissipated in the FM layer,the spin accumulations of the adjacent Pt layers are independent to each other;there-fore,the spin accumulation and spin current may depend only on the thickness of each Pt layer and the interface con-ditions mentioned above.However,when the FM layer is thin(d F<l sf,Co),the spin accumulation cannot be totally relaxed within such a thin FM layers and the spin current from one side may affect the one from the other side,leading to the coupling of the spin accumulation between Pt layers. The coupling of the Pt layers across Co layers enhances the spin accumulation at the interfaces,effective to the increase of the Pt thickness,and thus leads to the b L enhancement.Regarding the relatively small b T,it can be explained by the lack of structure inversion asymmetry in the all metallic Co/Pt structure.38Besides,we found that b T of all samples scales with its corresponding b L,which is induced by spin Hall effect.That is,the varied b T of different Co/Pt MLs or tri-layers might also be originated from spin Hall effect, which is also found in other material systems,13,18where the reflection of the spin current may contribute to the b T.21 Compared to the b L(T)reported in Co/Pd MLs14that shows b T>b L,our Co/Pt MLs reveal b L>b T,indicating the spin-orbit torque is dominated by different sources.21In the Co/ Pd MLs,the large b T was attributed to the presence of a Rashba-like spin-orbit splitting due to different lattice distor-tion between top and bottom interfaces.However,in our Co/ Pt tri-layers and multilayers,we always observed a relative small b T,which is coherent to the reported Pt/Co/Pt tri-layers showing no significant b T.38That is,the interfaces may be very different between Co/Pt and Co/Pd,which needs more experiments to clarify.In terms of b L,our value (68Oe cm2/107A for N¼8)is on the same order as that reported in Co/Pd MLs(117Oe cm2/107A for N¼20).14 The difference of b L can arise from several factors:Co thick-ness,different h SH in Pd and Pt,and interface conditions var-ied by process as discussed in the Co/Pt MLs.In summary,in Co/Pt MLs,because of the combination of the spin accumulation of the coupled Pt layers across the Co layer and the discrepancy between top and bottom inter-faces,the b L increases with increasing N and can be tuned by modifying its interfaces without sacrificing the perpendicular magnetic anisotropy.The Ta insertion on the top of each Co layer enhances the discrepancy between top and bottom interfaces,leading to larger b L compared to the pure Co/Pt MLs.On the other hand,Cu insertion reduces the interface discrepancy with smaller b L.Consequently,we believe that the multilayers setup is of great potential for real applica-tions because the enhancing SOT efficiency can be achieved without compromise of thermal stability.This work was partially supported by the Ministry of Science and Technology of Republic of China under Grant Nos.NSC102-2221-E-007-043-MY2and MOST104-2221-E-007-047-MY2.1A.Brataas,A.D.Kent,and H.Ohno,Nat.Mater.11,372(2012).2D.C.Ralph and M.D.Stiles,J.Magn.Magn.Mater.320,1190(2008).3S.Mangin,D.Ravelosona,J.A.Katine,M.J.Carey,B.D.Terris,and E.E.Fullerton,Nat.Mater.5,210(2006).4H.Meng and J.-P.Wang,Appl.Phys.Lett.88,172506(2006).5W.G.Wang,M.Li,S.Hageman,and C.L.Chien,Nat.Mater.11,64 (2012).6Y.Shiota,T.Nozaki,F.Bonell,S.Murakami,T.Shinjo,and Y.Suzuki, Nat.Mater.11,39(2012).7D.Chiba,S.Fukami,K.Shimamura,N.Ishiwata,K.Kobayashi,and T. Ono,Nat.Mater.10,853(2011).8L.Liu,C.F.Pai,Y.Li,H.W.Tseng,D.C.Ralph,and R.A.Buhrman, Science336,555(2012).9I.M.Miron,K.Garello,G.Gaudin,P.J.Zermatten,M.V.Costache,S. Auffret,S.Bandiera,B.Rodmacq,A.Schuhl,and P.Gambardella,Nature 476,189(2011).10A.Hoffmann,IEEE Trans.Magn.49,5172(2013).11M.Cubukcu,O.Boulle,M.Drouard,K.Garello,C.Onur Avci,I.Mihai Miron,nger,B.Ocker,P.Gambardella,and G.Gaudin,Appl.Phys. Lett.104,042406(2014).12L.Liu,O.J.Lee,T.J.Gudmundsen,D.C.Ralph,and R.A.Buhrman, Phys.Rev.Lett.109,096602(2012).13C.-F.Pai,M.-H.Nguyen,C.Belvin,L.H.Vilela-Le~a o,D.C.Ralph,and R.A.Buhrman,Appl.Phys.Lett.104,082407(2014).14M.Jamali,K.Narayanapillai,X.Qiu,L.M.Loong,A.Manchon,and H. Yang,Phys.Rev.Lett.111,246602(2013).15K.Yakushiji,T.Saruya,H.Kubota,A.Fukushima,T.Nagahama,S. Yuasa,and K.Ando,Appl.Phys.Lett.97,232508(2010).16S.Ishikawa,H.Sato,M.Yamanouchi,S.Ikeda,S.Fukami,F.Matsukura, and H.Ohno,J.Appl.Phys.115,17C719(2014).17U.H.Pi,K.Won Kim,J.Y.Bae,S.C.Lee,Y.J.Cho,K.S.Kim,and S. Seo,Appl.Phys.Lett.97,162507(2010).18J.Kim,J.Sinha,M.Hayashi,M.Yamanouchi,S.Fukami,T.Suzuki,S. Mitani,and H.Ohno,Nat.Mater.12,240(2013).19M.Hayashi,J.Kim,M.Yamanouchi,and H.Ohno,Phys.Rev.B89, 144425(2014).20K.Garello,I.M.Miron,C.O.Avci,F.Freimuth,Y.Mokrousov,S. Blugel,S.Auffret,O.Boulle,G.Gaudin,and P.Gambardella,Nat. Nanotechnol.8,587(2013).21P.M.Haney,H.-W.Lee,K.-J.Lee,A.Manchon,and M.D.Stiles,Phys. Rev.B87,174411(2013).22L.Liu,T.Moriyama,D.C.Ralph,and R.A.Buhrman,Phys.Rev.Lett. 106,036601(2011).23C.-F.Pai,Y.Ou,L.H.Vilela-Le~a o,D.C.Ralph,and R.A.Buhrman, Phys.Rev.B92,064426(2015).24W.Zhang,W.Han,X.Jiang,S.-H.Yang,and S.S.P.Parkin,Nat.Phys. 11,496(2015).25J.C.Rojas-Sanchez,N.Reyren,czkowski,W.Savero,J.P.Attane, C.Deranlot,M.Jamet,J.M.George,L.Vila,and H.Jaffres,Phys.Rev. Lett.112,106602(2014).26S.Bandiera,R.C.Sousa,B.Rodmacq,and B.Dieny,Appl.Phys.Lett. 100,142410(2012).27C.Hin Sim,J.Cheng Huang,M.Tran,and K.Eason,Appl.Phys.Lett. 104,012408(2014).28J.Sinha,M.Hayashi,A.J.Kellock,S.Fukami,M.Yamanouchi,H.Sato, S.Ikeda,S.Mitani,S.-H.Yang,S.S.P.Parkin,and H.Ohno,Appl.Phys. Lett.102,242405(2013).29S.Bandiera,R.R.Sousa,B.B.Rodmacq,and B.Dieny,IEEE Magn.Lett. 2,3000504(2011).30C.J.Lin,G.L.Gorman,C.H.Lee,R.F.C.Farrow,E.E.Marinero,H.V. Do,H.Notarys,and C.J.Chien,J.Magn.Magn.Mater.93,194(1991). 31K.S.Ryu,S.H.Yang,L.Thomas,and S.S.Parkin,mun.5, 3910(2014).32P.Deorani and H.Yang,Appl.Phys.Lett.103,232408(2013).33C.Du,H.Wang,F.Yang,and P.C.Hammel,Phys.Rev.Appl.1,044004 (2014).34S.Woo,M.Mann,A.J.Tan,L.Caretta,and G.S.D.Beach,Appl.Phys. Lett.105,212404(2014).35Y.-H.Wang,W.-C.Chen,S.-Y.Yang,K.-H.Shen,C.Park,M.-J.Kao, and M.-J.Tsai,J.Appl.Phys.99,08M307(2006).36S.Emori,T.Nan,T.M.Oxholm,C.T.Boone,J.G.Jones,B.M.Howe, G.J.Brown,D.E.Budil,and N.X.Sun,Appl.Phys.Lett.106,022406 (2015).37L.Piraux,S.Dubois,A.Fert,and L.Belliard,Eur.Phys.J.B4,413 (1998).38I.M.Miron,G.Gaudin,S.Auffret,B.Rodmacq,A.Schuhl,S.Pizzini,J. Vogel,and P.Gambardella,Nat.Mater.9,230(2010).。