Longitudinal Spin Seebeck Effect in Silver Strip on CoFe Film. and K. Y. Wang, 1,b) Semiconductors, Chinese Academy of Sciences, Beijing , China

Transcription

1 Longitudinal pin eebeck Effect in ilver trip on CoFe Fil Y. heng, 1,2 M. Y. Yang, 1 Y. Cao, 3 K. M. Cai, 1 G. N. Wei, 1 G. H. Yu, 3 B. Zhang, 1 X. Q. Ma, 2,a) and K. Y. Wang, 1,b) 1 tate Ke Laborator of uperlattices and Microstructures, Institute of eiconductors, Chinese Acade of ciences, Beijing , China 2 Departent of Phsics, Universit of cience and Technolog Beijing, Beijing , China 3 Departeent of aterials phsics, Universit of cience and Technolog Beijing, Beijing , China We report the experiental observation of the spin eebeck effect (E) in Ag/CoFe noble etal/agnetic etal bilaers with a longitudinal structure. Theral voltages jointl generated b the anoalous Nernst effect (ANE) and the E were detected across the Ag/CoFe/Cu strip with a perpendicular theral gradient. To effectivel separate the E and the ANE part of the theral voltages, we copared the experiental results between the Ag/CoFe/Cu strip and Cu/CoFe/Cu strip, where two saples processed with the heating power instead of the teperature difference through the thin CoFe fil. The respective contributions of the ANE and E to theral voltage were deterined, and the have the ratio of 4:1. The spin current injected through CoFe/Ag interface is calculated to be 1.76 A/W. There are three essential technologies in developing spintronics: the generation, detection and anipulation of spin currents. 1 A pure spin current can be realized b spin Hall effect 2,3, spin puping 4,5 and spin eebeck effect 6. Pure spin current is beneficial for spintronic operation with uch reduced energ dissipation. The spin eebeck effect (E), which includes the spin current generation and detection, has attracted a lot of attentions since its first discover in perallo in This effect consists of the generation of a spin a) b) Electronic ail: xqa@sas.ustb.edu.cn. Electronic ail: kwang@sei.ac.cn. 1

2 current as a result of an applied theral gradient in a ferroagnet (FM), and a voltage detected b a noble etal (NM) 3 or agnetic etal 8 which has a strong spin orbital coupling attached on FM b eans of the inverse pin Hall Effect (IHE). It offers applications of waste energ recover or spin current generation, and a new point of view to discover the interpla between heat and spin currents in aterials. Materials including agnetic etals, 1 seiconductors, 9 and insulators 10 have been used to investigate the E in two structures, the transverse (TE) 11,12,13 with a teperature gradient applied in the saple plane and the longitudinal E (LE) 14,15,16 with a teperature gradient out of the saple plane, respectivel. The longitudinal configuration is a ore siple structure for easuring the E, and is good for high densit integration. But the LE configuration is ainl used to investigate insulators other than the ferroagnetic etals, which are widel eploed in spintronic applications such as the tunnel junction 17, spin valve 18 and eor devices 19. When the FM is agnetic etal, the voltage of the E and anoalous Nernst effect (ANE) 20 in longitudinal structures are indistinguishable. The E j HE due to E and ANE are collinear, where E HE, j,, E T with owning the sae direction with E ANE, T, are the electric field fro the HE, the spin current across FM/NM interface, the spin polarization of FM, the electric field of ANE, the teperature gradient of FM, and the agnetization of FM. Considering the ANE coes fro the FM itself, the general wa to distinguish these two effects is coparing two sae FM laers with one adjoining to a NM laer and the other not, under the sae teperature gradient. 21 However, the teperature gradient of the FM laer for LE structure is hardl to be easured precisel, leading to the inaccurac of E signal analsis fro the experient. In this letter, we provide a schee to separate the ANE and E in Ag/CoFe/Cu saple unabiguousl. We use the ferroagnetic etal CoFe thin fil as the spin injector, and noral etal Ag laer as the spin detector, and the ANE and E are entangled with each other. In addition, we choose Cu/CoFe/Cu structure for coparison, in which the Ag laer 2

3 was replaced b a Cu laer. We chose Cu for its week spin Hall effect and leading no E signal, thus onl ANE exists. The theral signals of two saples were easured, and copared after noralized b their heating power. The heating power was used to represent the teperature gradient which couldn t be precisel easured directl, because the theral power can transfor into the teperature gradient according to saple's theral conductivit and geoetr. We finall separate these two effects, and obtain the value of ANE, E current per Watt. Using the obtained E results, the spin current injected into FM/NM interface was also obtained. Two saples were prepared for the present experients following the procedures below: Ag(3 n)/cofe(5 n)/cu(2 n) (saple 1) and Cu(3 n)/cofe(5 n)/cu(2 n) (saple 2) ultilaers were deposited b agnetron sputtering on a therall oxidized i substrate (0.5 in thickness) and patterned into a rectangle structure with the diension b Ar-ion bea etching. The E was easured using the so called longitudinal configuration [see Fig. 1(a)]. The two saples were placed between two Cu plates as the hot source and cold source. To reduce the theral contact and relieve the non-unifor heating between the Cu blocks and the saple, theral grease was used between the Cu plate and the saple. A piece of insulating cotton was placed on the top of the hot side to ake sure that the heat flux ainl flowed through the saple, as shown in the insert of Fig. 1(c). Two Pt-100 theroeters were placed on the two Cu plates to onitor the teperature. The theral voltage was easured through two probes attached to both ends of the saple using a Keithle 2182A digital nanovolteter. All easureents were perfored under vacuu (saller than bar) in a cover of aluinu, in order to iniize theral conduction through air, convection and disturb of electro-agnetic induction fro electroagnetic wave. It would take s until the teperature difference between the hot and the cold end becoes constant, and the teperature of both cold and hot sides keep slightl increasing parallell for longer tie, as shown in Fig. 1(b). 3

4 FIG. 1. (a) tructure of both saple 1 and saple 2, the red part eans the hot side with diension of 3 1 2, the blue part is the cold side and the gre part is Ag(Cu)/CoFe/Cu ultilaers with diension of The teperature gradient is out of plane, and the theral voltage is easured along the length of FM laer. is the angle of in-plane agnetic field and the length direction. (b) The tie-dependence of teperature difference T between hot and cold sides of Cu plates; the insert shows the tie-dependence of teperature on either hot or cold side, indicating that the teperature of both sides keep slightl increasing parallell when T is constant. As it is not necessar to take ever easureent until T coes to zero, T doesn t begin with zero as the tie does. Heating power P dependence of T (c) and V (d) of saple 1 and their linear fitting curves are shown in the end. Figure 1(c) shows the linear relation between the teperature difference of the two Cu plates ( T ) and the heating power (P). In vacuu condition, P could be taken as the theral power passing through the saple, as the heating power would transport ainl through the saple to the cold side, and onl ver sall part will be dissipated b air and the insulating cotton. The relation between P and TFM is P= FM x TFM / d, where FM is the theral conductivit of the FM laer, d is the thickness of the FM laer. The accurate TFM can be 4

5 obtained if FM is known. The FM of the two saples are equivalent because of the sae agnetic aterial. Thus if the sae heating power is applied to saple 1 and 2, the teperature gradient of the CoFe laer in these two saples will be the sae. Figure 1(d) shows that the theral voltages V [ V ( H ) V ( H )]/ 2 in saple 1 linearl increase with increasing the heating power, consisting with the linear V Trelation in previous stud 7,9,21,22, where H is the saturated agnetic field applied in x direction. Besides the teperature gradient, the E is strongl dependent on the saturation agnetization ( M ) of the FM laer. In order to exclude the contribution fro the change of M during increasing of the heating power, we easured the agnetic hsteresis loops using Magneto optical Kerr effect when sweeping the agnetic field fro -450 to 450 Oe in x direction at different teperature ranging fro 25 to 78 C, as shown in the insert of Fig. 2. Figure 2 shows that the saturation agnetization is independent to the teperature in saple 1, indicating that M is hardl influenced b the teperature in the range of easureent and a constant agnetization can be taken. Thus the power, and we can use the heating power instead of when coparing the theral voltages of the two saples. FM is barel changed during the increase of heating T to noralize the theral voltage FIG. 2. Teperature dependence of the saturation agnetization of CoFe fil. The insert shows the hsteresis loop at both the axiu (red inverted triangle) and iniu (blue square) teperature in easureent. 5

6 Figure 3(a) shows the theral voltage hsteresis loops with the agnetic field applied along different in-plane orientations for saple 1, in which the classical eebeck effect coponent was subtracted b oving the center of the loop to the origin point. Because the classical eebeck effect coing fro the non-unifor distribution of the teperature in x- plane, was unchanged during sweeping the agnetic field. As a result of the existence of theral gradient in x- plane, the Planar Nernst effect (PNE) due to the scattering of electrons, which were driven b heat flux in the saple plane, would contribute to the theral voltage. To analze the PNE, the angular dependence of was shown in Fig. 3(b), where V / P R, in which = 1 and = 2 describe aple 1 and 2 respectivel. shares the sae dependence with the theral voltage. V PNE can be characterized b a sin(2 ) dependence, while V ANE and V E are both sin( ) dependence. It is therefore possible to separate the V PNE and the other two ters b quantitativel analzing the V relation. In order to exclude the contribution of the agnetization change during the rotation of the agnetic field, the angular dependence of V was easured at agnetic field well above the coercive field. The the saple is full agnetized. The angle dependence of well fit to sine function, thus no PHE signal is detectable. curve is obtained b averaging the V values when for both the saple 1 and 2 are FIG. 3. (a) Hsteresis of theral voltages V across aple 1 at fixed heating power with agnetic field applied along in-plane different orientations with angles: 90 (red square), 50 (green circle), 10 (blue triangle) and 330 (orange hexagon). (b) Angular dependence of in aple 1 (blue square) and 2 (red circle), and solid lines are result 6

7 calculated b B Asin( ). (c) Angular dependence of 1, E in saple 1. We proceed to separate the contribution of the ANE and E to the theral voltage. Considering the expression fro the electron transport theor: J = σ E-α i J stands for the electron current, E is the electric field, gradient in k-direction, and the coefficients ij σ and i ij j ik k T, where T k is the applied theral ik α are the eleents of the conductivit and heating power tensor, respectivel. We consider the current in direction, which is an open circuit. The sources of the current can be divided into two groups: one is theral gradient, and the other is theral-induced electric field. The conductive current, resulting fro the electric field due to the theral effects, which force electrons to ove directionall along the saple and accuulate on the edge of the saple, can be calculated b the voltage easured along direction. The total currents of both sources are zero. We obtain the following expressions under our experiental condition: V J = = J J, (1) J 1 1, con 1, ANE 1, E R1xz V = = J 2 2, con 2, ANE R 2xz, (2) where J 1, con and J 2, con stands for the conductive current of saple 1 and 2 in direction, R 1 and R 2 are the resistance of the heating area of saple 1 and 2, J 1, con and J 2, con are the conductive current of saple 1 and 2. 1, ANE J and J 2, ANE should be equal when the FM fil of the two saples has the sae teperature gradient. Assuing there is not an spin Hall effect in Cu, there is onl ANE in saple 2, while both ANE and E exist in saple 1. ubtracting the results of saple 2 fro saple 1, the J 1, E can be obtained accordingl. To copare the theral-induced current in two saples under the sae heating power, we defined the entioned above. This factor represents the value of the theral induced current per Watt. Figure 3(b) shows the angular dependence of 7 for both saples. The best fitting of the both curves follow sine relation, which give the aplitude

8 na/w and na/w. The obtained 1 is uch larger than that of 2, where the botto etal laer silver in saple 1 has stronger spin orbital coupling than copper in saple 2. Thus the extra part of in saple 1 is attributed to the E, which is shown in Fig. 3(c). The axiu value of the E ( E 0.02 na/w) can be achieved with the saturation agnetization direction along x axis. We can obtain the ratio of these two effects to be V ANE : V E =1 : 4, indicating that the E coponent is 3 ties larger than the ANE. Considering the resistance value of 9.08 KΩ of saple 1, the voltage of E is 0.18 μv/w. Fro the obtained E, the abilit of the spin current generation at the Ag/CoFe interface in saple 1 can be calculated based on the expression 23 as follows : j tnm NM tfm FM EIHE (3) tnm H D tanh( ) 2 D where t NM ( t FM ), NM ( FM ), D, H and E IHE is the thickness of the NM (FM) laer, electron conductivit of NM (FM) laer, the spin diffusion length, the spin hall angle and the electric field due to IHE of the NM laer. As the D of silver is 700 n, which is uch larger than t NM (3 n), the tanh( tnm / 2 D) could be approxiated to t /2, and the ter of D can be eliinated, which eans spin current decreasing in the Ag strip could be ignored. o equation (3) can be siplified to j j1, E / H, and the H of silver is , 24 and we can calculate the theral-induced spin current per Watt in FM laer long z direction to be I j x 1.76 A/W, where x is the area spin current across. NM D In conclusion, we have observed the spin eebeck effect in Ag/CoFe bilaers with longitudinal structure at roo teperature. Noralizing the theral voltage b heating power, thus we can directl copare the theral voltages between the Ag/CoFe/Cu and Cu/CoFe/Cu saples. The contribution of E and ANE in Ag/CoFe/Cu saple is obtained to be E 0.02 na/w and ANE na/w, respectivel. The spin current injected into the silver laer is I 1.76 A/W, indicating the heating power of 1 Watt could generate a spin current of 1.76 A in this structure. The ethod used in this paper is not onl 8

9 effective to separate the ANE and E in FM(etal)/NM structure, but also convenient to obtain the abilit of spin current generation. This work was supported b 973 progra No. 2014CB643903, NFC Grant Nos , and K. Uchida, T. Ota, K. Harii, K. Ando, H. Nakaaa, and E. aitoh, J. Appl. Phs. 107, 09A951 (2010). 2.O. Valenzuela and M. Tinkha, Nature 442, 176 (2006). 3 E. aitoh, M. Ueda, H. Miajia, and G. Tatara, Appl. Phs. Lett. 88, 1 (2006). 4 K. Ando, Y. Kajiwara, K. asage, K. Uchida, and E. aitoh, IEEE Trans. Magn. 46, 3694 (2010). 5 H. Nakaaa, K. Ando, K. Harii, Y. Fujikawa, Y. Kajiwara, T. Yoshino, and E. aitoh, J. Phs. Conf. er. 266, 4 (2011). 6 J.-C. Le Breton,. hara, H. aito,. Yuasa, and R. Jansen, Nature 475, 82 (2011). 7 K. Uchida,. Takahashi, K. Harii, J. Ieda, W. Koshibae, K. Ando,. Maekawa, and E. aitoh, Nature 455, 778 (2008). 8 B.F. Miao,.Y. Huang, D. Qu, and C.L. Chien, Phs. Rev. Lett. 111, 1 (2013). 9 C.M. Jaworski, J. Yang,. Mack, D.D. Awschalo, J.P. Hereans, and R.C. Mers, Nat. Mater. 9, 898 (2010). 10 K. Uchida, J. Xiao, H. Adachi, J. Ohe,. Takahashi, J. Ieda, T. Ota, Y. Kajiwara, H. Uezawa, H. Kawai, G.E.W. Bauer,. Maekawa, and E. aitoh, Nat. Mater. 9, 894 (2010). 9

10 11 M. chid,. richandan, D. Meier, T. Kuschel, J.-M. chalhorst, M. Vogel, G. Reiss, C. trunk, and C.H. Back, Phs. Rev. Lett. 111, (2013). 12 M. chid,. richandan, D. Meier, T. Kuschel, M. Vogel, G. Reiss, C. trunk, and C.H. Back, 2 (2013). 13.H. Wang, L.K. Zou, J.W. Cai, B.G. hen, and J.R. un, Phs. Rev. B - Condens. Matter Mater. Phs. 88, 1 (2013). 14 T. Kikkawa, K. Uchida,. Daion, Y. hioi, H. Adachi, Z. Qiu, D. Hou, X.-F. Jin,. Maekawa, and E. aitoh, Phs. Rev. B 88, (2013). 15 K. Uchida, T. Nonaka, T. Kikkawa, Y. Kajiwara, and E. aitoh, Phs. Rev. B - Condens. Matter Mater. Phs. 87, 1 (2013). 16 T. Kikkawa, K. Uchida, Y. hioi, Z. Qiu, D. Hou, D. Tian, H. Nakaaa, X.F. Jin, and E. aitoh, Phs. Rev. Lett. 110, 1 (2013). 17..P. Parkin, C. Kaiser, A. Panchula, P.M. Rice, B. Hughes, M. aant, and.-h. Yang, Nat. Mater. 3, 862 (2004). 18 Z.H. Xiong, D. Wu, Z.V. Varden, and J. hi, Nature 427, 821 (2004). 19..P. Parkin, M. Haashi, and L. Thoas, cience 320, 190 (2008). 20 T. Miasato, N. Abe, T. Fujii, a. Asaitsu,. Onoda, Y. Onose, N. Nagaosa, and Y. Tokura, Phs. Rev. Lett. 99, 1 (2007). 21 R. Raos, T. Kikkawa, K. Uchida, H. Adachi, I. Lucas, M.H. Aguirre, P. Algarabel, L. Morellón,. Maekawa, E. aitoh, and M.R. Ibarra, Appl. Phs. Lett. 102, 7 (2013). 22 K.I. Uchida, H. Adachi, T. Ota, H. Nakaaa,. Maekawa, and E. aitoh, Appl. Phs. Lett. 97, (2010). 10

11 23 E. hikoh, K. Ando, K. Kubo, E. aitoh, T. hinjo, and M. hiraishi, Phs. Rev. Lett. 110, 1 (2013). 24 H.L. Wang, C.H. Du, Y. Pu, R. Adur, P.C. Hael, and F.Y. Yang, Phs. Rev. Lett. 112, 1 (2014). 11