Electrical spin injection and detection of spin precession in room temperature bulk GaN lateral spin valves

Transcription

1 Electrical spin injection and detection of spin precession in room temperature bulk GaN lateral spin valves Aniruddha Bhattacharya a), Md Zunaid Baten a) and Pallab Bhattacharya a) a) Center for Photonic and Multiscale Nanomaterials Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI 48109, USA ABSTRACT We report the measurement of diffusive electronic spin transport characteristics in an epitaxial wurtzite GaN lateral spin valve at room temperature. Hanle spin precession and non-local spin accumulation measurements have been performed with the spin valves fabricated with FeCo/MgO spin contacts. Electron spin relaxation length and spin-flip lifetime of 176 nm and 37 ps, respectively, are derived from analysis of results obtained from four-terminal Hanle spin precession measurements at 300 K. The role of dislocations and defects in bulk GaN has also been examined in the context of electronic spin relaxation dynamics.

2 The III-nitride family of wide bandgap semiconductors, such as GaN and their alloys (InGaN, AlGaN, InAlN), are important for diverse applications ranging from solid-state lighting, ultra-violet and deep ultra-violet coherent light sources and detectors and high-power and highfrequency electronics [1, 2]. GaN is also deemed to be suitable for room temperature spintronics applications owing to its intrinsically weak spin-orbit coupling (SOC) [3-4]. Values of electron spin relaxation lifetimes in bulk n-doped wurtzite GaN, ranging from 35 to 50 ps, have been reported from room temperature time-resolved Kerr-rotation and time-resolved Faraday rotation spectroscopy measurements [5, 6]. Spin diffusion parameters in relatively defect-free single wurtzite GaN nanowires (NWs) have also been determined from four-terminal Hanle spin precession measurements [7], which yielded a maximum spin lifetime ~ 100 ps and a spin diffusion length of ~ 260 nm at room temperature for samples having an unintentional background n-type doping concentration of ~ 1 x cm -3. Spin relaxation in GaN is generally explained by invoking the D yakonov-perel (DP) spin scattering mechanism [8, 9].The DP mechanism describes spin relaxation due to both precessional dephasing (as result of Rashba and Dresselhaus spin-orbit coupling) and scattering, where the orientation of spin precession changes randomly by momentum scattering. The four-terminal Hanle spin precession measurement [10, 11], which is generally regarded as the most convincing proof of electrical spin injection, transport and detection, is yet to be demonstrated in bulk GaN. In the present study we have made measurements on bulk GaN lateral spin valves with FeCo/MgO ferromagnetic contacts. The spin diffusion length and spin lifetime are independently estimated from channel-length dependent non-local magnetoresistance (MR) measurements [12] and four-terminal Hanle spin precession measurements respectively. The spin-flip lifetime derived from four-terminal Hanle spin precession measurement is 2

3 estimated to be ~37 ps and the corresponding spin diffusion length is ~176 nm. Similar results are obtained from channel-length dependent non-local MR measurements. These values of the spin transport parameters, which agree well with most other previously published results, are discussed in the context of large defect density in the material and related Elliott-Yafet (EY) spin scattering [13, 14]. The lateral ferromagnet-semiconductor spin valves were fabricated upon n (Si)-doped wurtzite c-plane GaN (0001) grown by metal-organic chemical vapor deposition (MOCVD). The nominal thickness of the epitaxial GaN layer is ~ 5 µm and the doping concentration is ~ 1x10 18 cm -2. The defect density in these substrates have been measured to be ~ 6.44x10 8 cm -2 from etchpit dislocation density measurements. Device fabrication is initiated by removal of the native oxide with HCl: H 2 O solution followed by definition of three contact regions by standard UV lithography. Next, 1 nm MgO tunnel barrier and 80 nm FeCo spin injector are deposited by electron beam evaporation to form the ferromagnetic tunnel contacts. Prior to this deposition step, the thickness of the MgO and FeCo layers were calibrated employing spectroscopic ellipsometry. Finally, the middle contact is bifurcated into two contacts of different widths, by focused ion-beam etching, to create the transport channel. We believe this processing step does not introduce active defects. Several ferromagnet/semiconductor/ferromagnet (FM/SC/FM) lateral spin valves were fabricated with varying channel lengths ranging from 250 nm to 450 nm. The channel width in all the spin valves is 400 µm. A schematic of the device, along with the four terminal non-local MR measurement scheme, is illustrated in Fig. 1. Contacts I and IV have identical dimensions (400 µm X 500 µm) and are positioned 150 µm( > several diffusion lengths(~ 140 nm ) )away from the spin detection (II) and spin injection (III) ferromagnetic electrodes, which ensures that the average spin polarization at contacts I and IV is essentially 3

4 zero [17]. Contacts II and III have dimensions of 400 µm X 800 µm and 400 µm X 400 µm, respectively. Magneto-optic Kerr effect (MOKE) measurements were performed at temperatures ranging from 20K to 300K to ascertain the magnetization characteristics of the FeCo films by sweeping an applied magnetizing field in-plane, parallel to the film surface. Magnetization switching characterized by hysteresis of the Kerr signal is observed at all temperatures and a coercive field of 109 Oe is measured at room temperature. Figure 2(a) shows the measured twoterminal I-V characteristics of the lateral spin-valve devices at room temperature. The current through the semiconductor channel is found to vary non-linearly with the applied bias. The measured conductance as a function of bias is shown in the inset of the figure, together with a parabolic fit to the data. Further, the weak insulator-like dependence of normalized zero-bias resistance (ZBR), R 0 (T)/R 0 (300 K) on temperature shown in Fig. 2(b) is strong evidence of single-step tunneling injection of the spin polarized carriers into the semiconductor. Non-local magnetoresistance and spin accumulation measurements were made on the GaN-based devices. The measurements were performed, using standard four-probe ac lock-in techniques, at 300 K in a magneto-optical cryostat. Samples of varying channel lengths were mounted between the poles of an electromagnet such that the applied external magnetizing field is aligned to the plane of spin transport. The MR measurements were performed in the non-local spin valve geometry to ensure that spurious magnetostatic effects such as anisotropic magnetoresistance (AMR) and local Hall effects [10,15,16,17], which may potentially resemble MR behavior arising from true spin injection, are eliminated. It has been shown that local measurements are also highly susceptible to spurious magnetic effects originating from impurityassisted tunneling magnetoresistance [18].Problems with interpretation of origin of spin-valve 4

5 signals in local measurements in general, and three-terminal measurements in particular, have been discussed in the literature [18-24]. Results from non-local measurements performed at T=300 K for the spin valve with channel length L=250 nm are shown in Fig. 3(a). The magnetoresistance, defined as, R( ) R( ) MR peaks at H ~ 292 Oe in both directions of the R( ) magnetic field. A peak magnetoresistance of % is recorded for L=250 nm at an applied bias of 1V. Figure 3(b) shows a plot of the peak magnetoresistance as a function of L. From the value of peak magnetoresistance for L 0, a maximum spin polarization ( s 1 ) of 7.9 % 1 at the injector ferromagnetic contact is derived. The value of the peak magnetoresistance as L 0, can be expressed as 2 R R ( 1) MR. Here, R 4 R TB is defined as the spin R TB selectivity of the tunnel barrier. Analysis of the channel-length dependent data yields a spin-flip lifetime ~ 23 ps and a corresponding spin diffusion length of ~ 140 nm at 300 K. To conclusively establish electrical spin injection into the channel, four-terminal Hanle spin precession measurements were performed with the same lateral spin valves at room temperature. We observe a gradual change in the measured non-local voltage due to Larmor precession of the diffusing electron spins and consequent spin dephasing in an applied transverse magnetic field B z. The Larmor frequency g B / L B z, where g is the Landé g-factor, µ B is the Bohr magneton, and ћ is reduced Plank s constant.the measurements were made by magnetizing contacts II and III (Fig. 1(a)) either in parallel or antiparallel directions and recording the nonlocal voltage developed across contacts III and IV while sweeping an applied transverse magnetic field perpendicular to the plane of the semiconductor channel and passing a 5

6 constant current through contacts I and II. The data shown in Fig. 4 (a) is a clear demonstration of the Hanle effect, where the top and bottom branches correspond to parallel and antiparallel magnetization of contacts II and III. The spin-flip lifetime, τ sf, was deduced by analyzing the data for a channel length of L=250 nm (solid lines in Fig. 4) with the following equation [11] V I NL inject 2 1 y2 y1vd t t exp cos( Lt)exp dtdy1dy 2 (1) 4 y2, y1, t 4 Dt Dt sf where D is the electron diffusion constant, τ sf is the spin-flip lifetime, v d is the drift velocity which is assumed to be zero for electron spin transport mediated entirely by diffusion [11], and + (-) signs indicate parallel (antiparallel) magnetization state of the FM electrodes. From Hall mobility measurements on a bulk GaN-on-sapphire sample we obtain a majority carrier mobility µ ~ 261 cm 2 /V-s and a majority carrier concentration n ~ 1x10 18 cm -3. From the Einstein s 1 S( E) F( E) de D q 0 relation F S( E) de E 0, where S is the density of states, F is Fermi-Dirac distribution, q is electronic charge and D is the diffusion constant, we derive a value of D=8.44 cm 2 /s. Using a g-factor of 1.94 for GaN [6], we obtain a value of spin-flip lifetime τ sf 37 ps, which translates to a spin diffusion length of λ sf 176 nm at T=300 K using the relation λ sf = (D τ sf ) 1/2. This value is in good agreement with the spin diffusion length derived from the non-local magnetoresistance measurements, within the limits of errors incurred during measurements and analysis of data. A value ~ 37 ps for the spin relaxation time in GaN is in excellent agreement with previously reported values obtained from time-resolved Faraday rotation [5] and time-resolved Kerr rotation measurements [6] performed on GaN samples having similar n-doping levels. The 6

7 latter results were explained by invoking the D yakonov-perel spin scattering mechanism. A common feature in the samples used in the previous studies and the present one is that the defect density in the active region is ~ 10 8 cm -2. While differences in doping levels have been accounted for, the large defect density has not been taken into account. It may be remembered that a significantly larger value of τ spin ~ 100 ps was measured at room temperature by Hanle spin precession and non-local spin transport measurements on a single GaN nanowire in which the density of defects is ~ cm -2.Further, although the spin-orbit coupling in bulk GaN is appreciably weaker than that of GaAs (the valence-band splitting in GaN is ~ 20 mev whereas in GaAs it is ~340 mev),we observe that room temperature spin-flip lifetimes in bulk GaN (~ 40 ps) is smaller than that of bulk GaAs (~ 200 ps) in which the dislocation density is almost negligible. It is therefore apparent that defects in GaN do play a significant role in determining spin transport characteristics. Spin-polarized electrons trapped in defects encounter further precessional dephasing owing to local fields, such as the hyperfine field, due to the stronger interaction of the trapped and consequently confined electrons with the nuclear system. Electron spin dephasing mediated by hyperfine interaction with the nuclear spin bath is an important effect for confined or localized electrons which is true for these electrons trapped in the defects [25]. We also believe another major dephasing channel is the strong exchange interaction between electrons trapped in adjacent defect states. It has been recently demonstrated experimentally and theoretically described how inelastic scattering between conduction and shallow donor impurity-bound electrons in unintentionally doped n-type Si(100) leads to dephasing of both spins by mutual spin exchange [26].We have also plotted the spin-flip lifetime measured in the present study and the reported spin-flip lifetimes as a function of dislocation density in Fig. 4(b). The n-doping level in all the samples is ~ cm -3. It is observed that the 7

8 spin lifetime decreases significantly with the increase of dislocation density, which may be potentially attributed to larger Elliott-Yafet spin scattering rates in these materials with larger defect densities. Detailed analysis of the data with this scattering, including spin dephasing brought about by the defects, in these wurtzite crystals is beyond the scope of this Letter. In conclusion, we have demonstrated electrical spin injection, transport, and detection in epitaxial GaN-based lateral spin-valves with FeCo/MgO tunnel contacts at room temperature. We have verified diffusive electronic spin transport in GaN from channel length-dependent nonlocal MR measurements and four-terminal Hanle spin precession measurements performed in the same devices. The spin diffusion length and spin lifetime obtained from both measurements are in good agreement. We have also examined the dependence of the spin relaxation lifetime on the dislocation density in the framework of Elliot-Yafet spin relaxation dynamics. The calculated spin lifetimes are in good agreement with values measured in this study and those reported in the literature. 8

9 Acknowledgement This work is supported by the National Science Foundation under the MRSEC program (Grant DMR ). The devices were fabricated at the Robert H. Lurie Nanofabrication Facility, a member of the National Nanotechnology Infrastructure Network funded by the National Science Foundation. 9

10 References 1. T. Frost, A. Banerjee, K. Sun, S. L. Chuang, and P. Bhattacharya, IEEE J. Quantum Electron 49(11), 923 (2013). 2. S. Nakamura, Rev. Mod. Phys. 87, 1139 (2015). 3. J. H. Bub, J. Rudolph, F. Natali, F. S. D. Hagele, Appl. Phys. Lett. 95, (2009). 4. A. Banerjee, F. Dog an, J. Heo, A. Manchon, W. Guo, and P. Bhattacharya, Nano Lett. 11, 5396 (2011). 5. B. Beschoten, E. Johnston-Halperin, D. K. Young, M. Poggio, J. E. Grimaldi, S. Keller, S. P. DenBaars, U. K. Mishra, E. L. Hu, and D. D. Awschalom, Phys. Rev. B 63, (R) (2001). 6. J. H. Bub, J. Rudolph, F. Natali, F. Semond, and D. Hagele, Phys. Rev. B 81, (2010). 7. H. Kum, J. Heo, S. Jahangir, A. Banerjee, W. Guo, and P. Bhattacharya, Appl. Phys. Lett. 100, (2012). 8. M.I. Dyakonov and V.I. Perel, Sov. Phys. JETP 33, 1053 (1971). 9. M.I. Dyakonov and V.I. Perel, Sov. Phys. Solid State 13, 3023 (1972). 10. F. J. Jedema, A. T. Filip and B. J. van Wees, Nature 410, (2001). 11. X. Lou, C. Adelmann, S. A. Crooker, E. S. Garlid, J. Zhang, K. S. M. Reddy, S. D. Flexner, C. J. Palmstrom, and P. A. Crowell, Nat. Phys. 3, (2007). 12. D. Saha, M. Holub, P. Bhattacharya, and Y. C. Liao, Appl. Phys. Lett. 89, (2006). 13. Y. Yafet Solid State Physics, edited by F. Seitz and D. Turnbull (Academic, New York, 1963), Vol R. J. Elliott, Phys. Rev. 96, 266 (1954). 10

11 15. M. Johnson and R. H. Silsbee, Phys. Rev. Lett. 55, 1790 (1985). 16. F. J. Jedema, H. B. Heersche, A. T. Filip, J. J. A. Baselmans, and B. J. van Wees, Nature (London) 416, 713 (2002). 17. O. M. J. van t Erve, A. T. Hanbicki, M. Holub, C. H. Li, C. Awo-Affouda, P. E. Thompson, and B. T. Jonker, Appl. Phys. Lett. 91, (2007). 18. O. Txoperena, Y. Song, L. Qing, M. Gobbi, L. E. Hueso, H. Dery, and F. Casanova, Phys. Rev. Lett. 113, (2014). 19. O. Txoperena, M. Gobbi, A. Bedoya-Pinto, F. Golmar, X. Sun, L. E. Hueso, and F. Casanova, Appl. Phys. Lett. 102, (2013). 20. H. N. Tinkey, P. Li, and I. Appelbaum, Appl. Phys. Lett. 104, (2014). 21. A. G. Swartz, S. Harashima, Y. Xie, D. Lu, B. Kim, C. Bell, Y. Hikita, and H. Y. Hwang, Appl. Phys. Lett. 105, (2014). 22. Y. Song and H. Dery, Phys. Rev. Lett. 113, (2014). 23. I. Appelbaum, H. N. Tinkey, and P. Li, Phys. Rev. B 90, (2014). 24. H. Inoue, A. G. Swartz, N. J. Harmon, T. Tachikawa, Y. Hikita, M. E. Flatté, and H. Y. Hwang, Phys. Rev. X 5, (2015). 25. Ł. Cywin ski, W. M. Witzel, and S. Das Sarma, Phys. Rev. Lett. 102, (2009). 26. L. Qing, J. Li, I. Appelbaum, and H. Dery, Phys. Rev. B 91,241405(R) (2015). 11

12 Figure Captions Figure 1 (color online). Schematic illustration of the four-terminal non-local measurement scheme (not drawn to scale). Figure 2 (color online) (a) Two-terminal I-V characteristics of the tunneling FeCo/MgO/GaN contact recorded at 300 K. The inset shows differential conductance as a function of applied biasing voltage at 300 K; (b) normalized zero-bias resistance as a function of temperature. Figure 3 (color online) (a) Magnetoresistance as a function of applied magnetic field for a 250 nm channel length lateral spin valve measured at 300 K. The black and red lines indicate increasing and decreasing magnetic field sweeps, respectively; (b) peak magnetoresistance as a function of channel length measured at 300 K. Figure 4 (color online) (a) Four-terminal Hanle spin precession data measured for a 250 nm channel length lateral spin valve at 300 K; (b) measured spin relaxation time τ spin as a function of dislocation density. The dashed line is a guide to the eye. 12

13 Fig. 1

14 80 T = 300 K Current (A) di/dv (A/V) Experimental Parabolic Analysis Bias (V) Bias (V) (a) R 0 (T)/R 0 (300K) R 0 (300K)=1.5 M T (K) (b) Fig. 2

15 Magnetoresistance (%) T=300 K H (Oe) (a) Peak Magnetoresistance (%) sf = nm D=8.44 cm 2 /s sf ~ ps T=300 K L channel (nm) (b) Fig. 3

16 0.1 T = 300 K V 4T (mv) sf ~ 374 ps D = 8.44 cm 2 /s sf ~ 1769 nm L channel = 250 nm B z (T) (a) spin (ps) Present study Ref. 5 Ref T= 300 K n2d x 10 8 (cm -2 ) (b) Fig. 4