aa r X i v : . [ c ond - m a t . m t r l - s c i ] N ov Large enhancement in the generation efficiency of pure spin currents in Geusing Heusler-compound Co FeSi electrodes
K. Kasahara, Y. Fujita, S. Yamada, K. Sawano, M. Miyao, and K. Hamaya ∗ Department of Electronics, Kyushu University, 744 Motooka, Fukuoka 819-0395, Japan and Advanced Research Laboratories, Tokyo City University, 8-15-1 Todoroki, Tokyo 158-0082, Japan (Dated: March 5, 2018)We show nonlocal spin transport in n -Ge based lateral spin-valve (LSV) devices with highly or-dered Co FeSi/ n + -Ge Schottky tunnel contacts. Clear spin-valve signals and Hanle-effect curves aredemonstrated at low temperatures, indicating generation, manipulation, and detection of pure spincurrents in n -Ge. The obtained spin generation efficiency of ∼ PACS numbers:
Semiconductor based spintronic devices have been pro-posed in a scalable solid state framework.[1–4] For thecompatibility with the existing electronic devices onsemiconductor platforms, all electrical means of gener-ation, transport and detection of spin polarized carriersthrough semiconductor channels such as GaAs,[5] Si,[6]Ge,[7] and so forth, are important technologies. To re-alize low power consumption for actual applications, theuse of Schottky-tunnel contacts is more effective thanthat of insulating tunnel-barrier contacts because of lowparasitic resistance in nano-scaled devices.[5, 8–10]Recently, since high electron and hole mobilities arerequired as next-generation channel materials, Ge basedspintronics compatible with Si-LSIs has also been ex-plored by many groups.[7, 11–14, 16] So far, by usingnonlocal spin-valve (NLSV) measurements in Ge basedlaterally configured devices, generation and detection ofpure spin currents were observed up to ∼
200 K.[7, 15, 17]Unfortunately, above 100 K, there has not been spinmanipulation via Hanle-type spin precession in nonlo-cal four-terminal geometries. Despite the devices withFe/MgO tunnel-barrier contacts,[7, 15] the Hanle-effectcurves in nonlocal four-terminal geometries were able tobe seen only at less than 50 K. In general, even at lowtemperatures, the spin lifetime of the n -Ge channels canbe estimated frequently to be less than 1 ns,[7, 15] whichis significantly shorter than that of n -GaAs[5] and n + -Si.[18] The short spin lifetime can lead to short spin dif-fusion length in the channel region. This means that, inlateral device structures, the demonstration of the purespin current transport in n -Ge is more difficult than thatin n -GaAs and n + -Si experimentally.Until now, highly ordered Co FeSi − x Al x Heusler com-pounds have been explored as the spin injector and de-tector to improve the spin generation efficiency in theGaAs channels[19, 20] although n -GaAs channels can ∗ E-mail: [email protected] not be compatible with Si-LSIs, where Co FeSi − x Al x compounds have relatively large spin polarization likea half-metallic material.[21–24] Since we have estab-lished the growth technique for highly ordered Co FeSi( x = 0) on Ge by using low-temperature molecularbeam epitaxy,[25] we can combine the Co FeSi elec-trodes with spin injection/detection techniques reportedso far.[14, 17] If large spin accumulation is generated inthe n -Ge channel by using the highly ordered Co FeSi,the lateral transport with the spin manipulation, i.e.,Hanle-effect curve, can be observed at higher tempera-tures.In this study, using L -ordered Co FeSi/ n + -GeSchottky-tunnel contacts, we demonstrate lateral trans-port of pure spin currents detected by four-terminal non-local Hanle effects in n -Ge up to 225 K. The spin gen-eration efficiency is markedly enhanced, which is abouttwo orders of magnitude larger than that of the previousreport by Zhou et al .[7] The use of Co based Heusler com-pounds will be a candidate for metallic source and drainin Ge based spintronic applications. We also discuss thespin related behavior with temperature evolution.As schematically shown in Fig. 1(a), we fabri-cated n -Ge based lateral spin-valve devices (LSVs) withCo FeSi/ n + -Ge Schottky tunnel contacts. The follow-ings are fabrication processes. First, we formed a phos-phorous (P)-doped n -Ge(111) channel (P + ∼ cm − )with a thickness of ∼
50 nm on non-doped Ge(111) sub-strates ( ρ = ∼
40 Ωcm) by using an ion implantationtechnique and post annealing at 700 ◦ C. Then, an n + -Ge(111) layer consisting of Sb δ -doped layer and a 5-nm-thick Ge epitaxial layer was grown by molecular beamepitaxy (MBE) at 400 ◦ C,[26] where the doping den-sity of Sb was 2 × cm − . After the fabricationof the n + -Ge(111) layer, a 10-nm-thick Co FeSi epitax-ial layer was grown on top of the n + -Ge(111) layer byroom-temperature MBE.[27] In Fig. 1(b), we show ahigh-resolution cross-sectional transmission electron mi-crograph (TEM) of the formed Co FeSi/ n + -Ge inter-face. The heterojunction is atomically flat, leading to Co FeSi n + -Ge(111) AuSb n -Ge channel Co FeSi (10 nm) δ -doped Sb 100 μ m Co (20 nm) non-doped Ge 0.3 μ m 0.5 μ m V NL + - I
50 nm (a)(b) z yx (c) L = 1.0 μ m
000 111022111 n + -Ge(111) (c)
000 400422022 311 Co FeSi
FIG. 1: (Color online) (a) Schematic diagram of a lateral four-terminal device with Co FeSi/ n + -Ge contacts. (b) A cross-sectional TEM image of the Co FeSi/ n + -Ge interface fabri-cated by a room-temperature MBE technique. (c) Nanobeamelectron diffraction patterns for each Co FeSi and n + -Ge. Thezone axis is parallel to the [011] direction. the reduction in the presence of interface states.[28, 29]We also observe < > and < > superlattice reflec-tions in the nanobeam electron diffraction patterns ofthe Co FeSi layer [Fig. 1(c)], resulting from the presenceof L -ordered structures. Thus, the Co FeSi Heusler-compound electrodes were high quality as shown in ourprevious work.[25] In Fig. 1(c), the quality of the n + -Ge layer is guaranteed. In order to align the magneticmoments in the in-plane direction for the Hanle-effectmeasurements, a polycrystalline Co layer with a thick-ness of ∼
20 nm was deposited on the Co FeSi layer byusing electron beam evaporation.Conventional processes with electron-beam lithogra-phy, Ar + ion milling, and reactive ion etching were usedto fabricate four-terminal LSVs. The size of each con-tact is presented in Fig. 1(a) (0.3 × µ m and 0.5 × µ m ), and the center-to-center distance ( L ) be-tween the Co FeSi/ n + -Ge contacts was 1.0 µ m. Thecurrent-voltage ( I - V ) characteristics of the Co FeSi/ n + -Ge junctions showed almost no rectifying behavior atroom temperature, indicating tunneling conduction ofelectrons through the Co FeSi/ n + -Ge interface. Al-though we confirmed no change in the forward-bias cur-rent with temperature variation, small decreases in thereverse-bias current were observed with decreasing tem-perature. Nearly same features have already been seen -60-40-200204060 ∆ R N L ( m Ω ) -400 0 400 B z (Oe)-100-80-60-40-200 ∆ R N L ( m Ω ) -1000 -500 0 500 1000 B y (Oe) ∆ R NL (a)(b) FIG. 2: (Color online) (a) Nonlocal magnetoresistance curveat 150 K. (b) Hanle-effect curves at 150 K for parallel andanti-parallel magnetization configuration. The | ∆ R NL | valueis defined from the Hanle curves. in our previous work for the n + -Ge(111) layer formedby Sb δ -doping.[17] In order to avoid the change in theinterface resistance with changing temperature, we con-centrate on measurements in the forward-bias conditionsfor the Co FeSi/ n + -Ge contacts. In this study, the resis-tance area product ( RA ) of the Co FeSi/ n + -Ge interfacewas nearly constant of ∼ Ω µ m . In order to preciselyunderstand spin-related phenomena, we also fabricateda micro fabricated Hall-bar device with the same n -Gechannel and AuSb ohmic contacts. The electron carrierdensity ( n ) of the fabricated channel was experimentallyestimated from electrical Hall-effect measurements. Theestimated values were n ∼ × cm − at 300 Kand n ∼ × cm − at 150 K, respectively. As aresult, the electron mobility ( µ e ) of this channel is ∼ V − s − at 300 K and ∼
665 cm V − s − at 150 K.Figure 2(a) shows a nonlocal magnetoresistance(∆ R NL = ∆ V NL I ) measured at I = +1.0 mA at 150 K,where the positive sign of I ( I >
0) means that the elec-trons are extracted from the Ge channel into Co FeSi,i.e., spin extraction condition, through the Schottky tun-nel barrier. By applying in-plane magnetic fields ( B y ),a large spin-valve signal ( ∼
100 mΩ) can be seen evenat 150 K. This spin-valve feature is attributed to thechange in the magnetization direction of two differentCo FeSi contacts used here between nearly parallel andanti-parallel configurations. Note that the magnitude ofthe nonlocal signal is almost compatible with that ob-served at 4 K for a device (Device B in Ref. [7]) withFe/MgO tunnel contacts and L = 1.0 µ m in the previouswork reported by Zhou et al .[7] This feature implies that,even at a higher temperature, we demonstrate the spinaccumulation in the n -Ge channel equivalent to that byZhou et al .[7] Using this nonlocal four-terminal geometry,we also applied out-of-plane magnetic field ( B Z ) underparallel and anti-parallel magnetic configurations for theCo FeSi electrodes and recorded ∆ R NL as a function of B Z . As a result, clear Hanle-type spin precession curves,which are evidence for the generation, manipulation, anddetection of pure spin currents in the n -Ge channel, canbe seen in Fig. 2(b). At further higher temperatures,such Hanle-effect curves were observable for several de-vices.By fitting the Hanle-effect curves with the one-dimensional spin drift diffusion model,[5, 30] the spin life-time ( τ s ) in the channel region can roughly be extracted.The model used for our device is,∆ R NL ( B Z ) ∝ ± Z ∞ √ πDt exp (cid:20) − L Dt (cid:21) cos( ω L t ) exp( − tτ s ) dt, (1)where ± is the sign depending on the magnetization con-figuration (parallel or anti-parallel), D is the diffusionconstant of the Ge channel, ω L = gµ B B Z / ~ is the Lamorfrequency, g is the electron g -factor ( g = 1.56),[31] µ B isthe Bohr magneton. The representative fitting resultsare shown in the solid curves in Fig. 2(b). At 150 K,the τ s and D values for the n -Ge channel used are esti-mated to be ∼
420 ps and ∼ s − , respectively.The obtained D value is consistent with experimentallyestimated value of 8.6 cm s − at 150 K from Einsteinrelation, D = k B Tq µ e , where k B is Boltzmann’s constant.We will comment on τ s later.From the | ∆ R NL | shown in Fig. 2(b), we also esti-mate the spin generation efficiency of this device with theCo FeSi/ n + -Ge contacts. In general, | ∆ R NL | detected bynonlocal four-terminal geometry can be expressed as, | ∆ R NL | = P gen P det ρ N λ N S exp( − Lλ N ) , (2)where P gen and P det are spin polarization which are gen-erated and detected, respectively, at the ferromagneticelectrodes. Thus, ( P gen × P det ) / can roughly be re-garded as the spin generation efficiency. ρ N and λ N are the resistivity and the spin diffusion length of thenonmagnetic channel. In this study, ρ Ge and λ Ge at150 K are 28.2 mΩcm and 0.59 µ m, respectively, where λ Ge = √ Dτ s . S is the cross section of the Ge channel ( ∼ µ m ) and we used L = 1.0 µ m. Using these parame-ters, we can obtain ( P gen × P det ) / of ∼ ∼
12 %).Despite a higher temperature of 150 K the spin gener-ation efficiency is about two orders of magnitude largerthan that for an Fe/MgO/Ge device (Device B at 4 Kin Ref. [7]). It should be noted that the order of thisefficiency is consistent with that for a GaAs based device -1000 -500 0 500 1000 -400 0 40010040200140 160 180 200 220 240 260 2801001000 B z (Oe) B y (Oe)
100 m Ω τ S ( p s ) Temperature (K) Temperature (K) ∆ R N L ( m Ω ) ∆ R N L ( m Ω ) (a) (b)(c)(d)
50 m Ω ∆ R N L ( m Ω )
140 160 180 200 220 240 260 280 ρ G e
160 200 240 280 Temp. (K) ( m Ω c m )
175 K250 K275 K225 K 175 K250 K225 K
FIG. 3: (Color online) Temperature dependent (a) NLSV sig-nals, (b) Hanle curves, (c) ∆ R NL , and (d) τ s . The inset of (c)shows ρ Ge as a function of temperature for the used channel. with L -ordered Co FeSi electrodes ( ∼ L -ordered Co FeSi electrodes enable us tomarkedly enhance the generation of spin accumulation inthe channel. Considering this study for Ge and a previ-ous one for GaAs,[19] the Heusler-compound electrodescan open a possible way for highly efficient spin genera-tion even in semiconductor channels.[8, 10, 14, 17, 19, 20]Next, we discuss spin signals with temperature evolu-tion for this device. Since the temperature dependent RA can affect the spin signals, shown in our previouswork,[17] we simply focus on the data which can be mea-sured in the condition of a constant RA ( ∼ Ω µ m )above 150 K. As shown in Figs. 3(a) and 3(b), nonlocalspin signals and Hanle curves in the parallel magneticconfiguration of the Co FeSi electrodes are gradually de-creased with increasing temperature. In this experiment,the clear Hanle curves can be observed up to 225 K al-though spin-valve like nonlocal signal can be seen at 275K. Unfortunately, both features could not be verified atroom temperature. Figure 3(c) shows ∆ R NL , which isroughly extracted from NLSV signals, as a function oftemperature. In order to understand ∆ R NL versus tem-perature, we also estimated τ s from the fitting of theHanle curves with increasing temperature, where D val-ues were estimated from Einstein relation and experimen-tal transport measurements of carriers in micro Hall-bardevices. Interestingly, Fig. 3(d) reveals that there is noclear dependence with temperature evolution, which islargely different from the features in high-temperatureregime in the previous works.[7, 15] As a whole, the τ s values ( ∼
380 ps τ s ∼
480 ps ) extracted hereare markedly shorter than theoretically expected intrinsicones ( τ s ∼
10 ns).[32] These features of τ s with temper-ature evolution cannot be explained by the Elliot-Yafetspin relaxation mechanism.[7, 33–36] We should also rec-ognize the large influence of the spin-flip scattering dueto the heavily doped Sb near the Co FeSi/ n + -Ge inter-face on the spin generation and detection,[15] giving riseto the much shorter τ s values. Taking this special situ-ation for our devices into account, we infer that the τ s value with temperature evolution is limited by the ex-trinsic factors rather than conventional spin relaxationmechanism.[17]According to Eq. (2), we find that the magnitude of∆ R NL depends not only on λ N (= √ Dτ s ) but also on ρ N .Thus, we should reconsider the change in ρ Ge with tem-perature evolution. The inset of Fig. 3(c) shows the tem-perature dependence of ρ Ge for the n -Ge channel used inthis study. Since the electron density of the used Gechannel is reduced with decreasing temperature, ρ Ge de-creases monotonically. As expected, this feature is close to the temperature dependent ∆ R NL in the main panelof Fig. 3(c). From these considerations, we can judgethat the contribution of ρ Ge to ∆ R NL is more dominantthan that of the spin relaxation in the Ge channel in thisstudy.We finally comment on the short λ Ge of ∼ µ m atlow temperatures. As described before, our Ge channelswere fabricated by a conventional ion implantation tech-nique and then, we used some etching processes to fab-ricate devices for transport measurements. Thus, a lackof optimization of these processes gave rise to damagedchannels with relatively poor mobility ( ∼
300 cm V − s − at 300 K) compared to bulk wafers (more than ∼ V − s − ).[38–40] As a result, the D values of ∼ s − are relatively low at low temperatures, leadingto short λ Ge . Even though we detect the lateral trans-port of pure spin currents in n -Ge at higher temperaturesby using Heusler-compound Co FeSi electrodes becauseof the large enhancement in the spin generation efficiency,a marked improvement of λ Ge might also be required forachieving Ge spintronics. Now, optimization of the fab-rication processes for lateral structures is underway forobtaining long λ Ge in Ge channels for actual device ap-plications.In summary, we showed large enhancement in spin gen-eration efficiency in n -Ge by using L -ordered Co FeSiHeusler-compound electrodes. Thanks to this technolog-ical development, we demonstrated the lateral transportof pure spin currents in n -Ge up to 225 K. For actual Gespintronic applications, the improvement of the devicefabrication processes is also required.This work was partly supported by Industrial Technol-ogy Research Grant Program from NEDO and Grant-in-Aid for Scientific Research (A) (No. 25246020) from TheJapan Society for the Promotion of Science (JSPS). S.Y.acknowledges JSPS Research Fellowships for Young Sci-entists. [1] S. Datta and B. Das, Appl. Phys. Lett. , 665 (1990).[2] S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J.M. Daughton, S. von Moln´ar, M. L. Roukes, A. Y.Chtchelkanova and D. M. Treger, Science , 1488(2001).[3] I. ˇZuti´c, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. , 323 (2004).[4] H. Dery, P. Dalal, L. Cywi´nski, and L. J. Sham, Nature(London) , 573 (2007).[5] X. Lou, C. Adelmann, S. A. Crooker, E. S. Garlid, J.Zhang, S. M. Reddy, S. D. Flexner, C. J. Palmstrøm,and P. A. Crowell, Nat. Phys. , 197 (2007).[6] I. Appelbaum, B. Huang, and D. J. Monsma, Nature , 295 (2007).[7] Y. Zhou, W. Han, L.-T. Chang, F. Xiu, M. Wang, M.Oehme, I. A. Fischer, J. Schulze, R. K. Kawakami, andK. L. Wang, Phys. Rev. B , 125323 (2011).[8] K. Hamaya, Y. Ando, T. Sadoh, and M. Miyao, Jpn. J.Appl. Phys. , 010101 (2011). [9] K. Hamaya, Y. Ando, K. Masaki, Y. Maeda, Y. Fujita,S. Yamada, K. Sawano, and M. Miyao, J. Appl. Phys. , 17C501 (2013).[10] R. Farshchi and M. Ramsteiner, J. Appl. Phys. ,191101 (2013).[11] K.-R. Jeon, B.-C. Min, Y.-H. Jo, H.-S. Lee, Il-J. Shin,C.-Y. Park, S.-Y. Park, and S.-C. Shin, Phys. Rev. B ,165315 (2011).[12] H. Saito, S. Watanabe, Y. Mineno, S. Sharma, R. Jansen,S. Yuasa, and K. Ando, Solid State Commun. , 1159(2011).[13] A. Jain, J.-C. Rojas-Sanchez, M. Cubukcu, J. Peiro, J.C. Le Breton, E. Prestat, C. Vergnaud, L. Louahadj, C.Portemont, C. Ducruet, V. Baltz, A. Barski, P. Bayle-Guillemaud, L. Vila, J.-P. Attan´e, E. Augendre, G. Des-fonds, S. Gambarelli, H. Jaffr´es, J.-M. George, and M.Jamet, Phys. Rev. Lett. , 106603 (2012).[14] K. Kasahara, Y. Baba, K. Yamane, Y. Ando, S. Yamada,Y. Hoshi, K. Sawano, M. Miyao, and K. Hamaya, J. Appl. Phys. , 07C503 (2012).[15] L.-T. Chang, W. Han, Y. Zhou, J. Tang, I. A. Fischer,M. Oehme, J. Schulze, R. K. Kawakami, and K. L. Wang,Semicond. Sci. Technol. , 015018 (2013).[16] M. Koike, E. Shikoh, Y. Ando, T. Shinjo, S. Yamada,K. Hamaya, and M. Shiraishi, Appl. Phys. Exp. ,023001(2013).[17] K. Hamaya, Y. Baba, G. Takemoto, K. Kasahara, S. Ya-mada, K. Sawano, and M. Miyao, J. Appl. Phys. ,183713 (2013).[18] M. Shiraishi, Y. Honda, E. Shikoh, Y. Suzuki, T. Shinjo,T. Sasaki, T. Oikawa, K. Noguchi, and T. Suzuki, Phys.Rev. B , 241204(R) (2011).[19] P. Bruski, Y. Manzke, R. Farshchi, O. Brandt, J. Her-fort, and M. Ramsteiner, Appl. Phys. Lett. , 052406(2013).[20] T. Saito, N. Tezuka, M. Matsuura and S. Sugimoto, Appl.Phys. Lett. , 122401 (2013).[21] K. Inomata, N. Ikeda, N. Tezuka, R. Goto, S. Sugimoto,M. Wojcik, and E. Jedryka, Sci. Technol. Adv. Mater. ,014101 (2008).[22] H. Sukegawa, W. Wang, R. Shan, T. Nakatani, K. Ino-mata, and K. Hono, Phys. Rev. B , 184418 (2009).[23] T. Kimura, N. Hashimoto, S. Yamada, M. Miyao, K.Hamaya, NPG asia materials , e9 (2012).[24] K. Hamaya, N. Hashimoto, S. Oki, S. Yamada, M. Miyao,and T. Kimura, Phys. Rev. B , 100404(R) (2012).[25] K. Kasahara, K. Yamamoto, S. Yamada, T. Murakami,K. Hamaya, K. Mibu, and M. Miyao, J. Appl. Phys. ,09B105 (2010).[26] K. Sawano, Y. Hoshi, K. Kasahara, K. Yamane, K.Hamaya, M. Miyao, and Y. Shiraki, Appl. Phys. Lett. , 162108 (2010).[27] S. Yamada, J. Sagar, S. Honda, L. Lari, G. Takemoto, H. Itoh, A. Hirohata, K. Mibu, M. Miyao, and K. Hamaya,Phys. Rev. B , 174406 (2012).[28] K. Yamane, K. Hamaya, Y. Ando, Y. Enomoto, K. Ya-mamoto, T. Sadoh, and M. Miyao, Appl. Phys. Lett. ,162104 (2010).[29] K. Kasahara, S. Yamada, K. Sawano, M. Miyao, and K.Hamaya, Phys. Rev. B , 205301 (2011).[30] F. J. Jedema, H. B. Heersche, A. T. Filip, J. J. A. Basel-mans, and B. J. van Wees, Nature , 713 (2002).[31] R. Vrijen, E. Yablonovitch, K. Wang, H. W. Jiang, A. Ba-landin, V. Roychowdhury, T. Mor, and, D. DiVincenzo,Phys. Rev. A , 012306 (2000).[32] P. Li, Y. Song, and H. Dery, Phys. Rev. B , 085202(2012).[33] R. J. Elliott, Phy. Rev. , 266 (1954).[34] Y. Yafet, in Solid State Physics , edited by F. Seitz andD. Turnbull (Academic, New York, 1963).[35] F. Pezzoli, F. Bottegoni, D. Trivedi, F. Ciccacci, A. Gior-gioni, P. Li, S. Cecchi, E. Grilli, Y. Song, M. Guzzi H.Dery, and G. Isella, Phys. Rev. Lett. , 156603 (2012).[36] C. Guite and V. Venkataraman, Appl. Phys. Lett. ,252404 (2012).[37] Y. Ando, Y. Maeda, K. Kasahara, S. Yamada, K. Masaki,Y. Hoshi, K. Sawano, K. Izunome, A. Sakai, M. Miyao,and K. Hamaya, Appl. Phys. Lett. , 132511 (2011).[38] S. Nakaharai, T. Tezuka, N. Sugiyama, Y. Moriyama andS. Takagi, Appl. Phys. Lett. , 3516 (2003).[39] C. O. Chui, K. Gopalakrishnan, P. B. Griffin, J. D. Plum-mer, and K. C. Saraswat, Appl. Phys. Lett. , 3275(2003).[40] A. Satta, T. Janssens, T. Clarysse, E. Simoen, M. Meuris,A. Benedetti, I. Hoflijk, B. De Jaeger, C. Demeurisse andW. Vandervorst, J. Vac. Sci. Technol. B24