Spin Transfer Torque in a Graphene Lateral Spin Valve Assisted by an

Letter pubs.acs.org/NanoLett

Spin Transfer Torque in a Graphene Lateral Spin Valve Assisted by an External Magnetic Field Chia-Ching Lin, Ashish Verma Penumatcha, Yunfei Gao, Vinh Quang Diep, Joerg Appenzeller, and Zhihong Chen* School of Electrical and Computer Engineering and Birck Nanotechnology Center, Purdue University, West Lafayette, Indiana 47907, United States S Supporting Information *

ABSTRACT: Spin-based devices are widely discussed for post-complementary metal−oxide−semiconductor (CMOS) applications. A number of spin device ideas propose using spin current to carry information coherently through a spin channel and transfering it to an output magnet by spin transfer torque. Graphene is an ideal channel material in this context due to its long spin diffusion length, gate-tunable carrier density, and high carrier mobility. However, spin transfer torque has not been demonstrated in graphene or any other semiconductor material as of yet. Here, we report the first experimental measurement of spin transfer torque in graphene lateral nonlocal spin valve devices. Assisted by an external magnetic field, the magnetization reversal of the ferromagnetic receiving magnet is induced by pure spin diffusion currents from the input magnet. The magnetization switching is reversible between parallel and antiparallel configurations, depending on the polarity of the applied charged current. The presented results are an important step toward developing graphene-based spin logic and understanding spin-transfer torque in systems with tunneling barriers. KEYWORDS: Graphene, spin transfer torque, nonlocal spin valve, all spin logic ue to its potential for low-power operation, “All Spin Logic”1 using pure spin currents to communicate information is considered an attractive approach to continue the downscaling of post-complementary metal−oxide−semiconductor (CMOS). Information in terms of magnetization gets transferred and imprinted onto nanomagnets by means of spin transfer torque.2,3 While separation of the charge and spin currents has been experimentally realized in lateral nonlocal spin valve (LNLSV) structures,4,5 it is essential to demonstrate that spin currents can be unambiguously utilized to switch the magnetization of nanomagnets. Yang et al.6 first showed that, when a sufficiently large spin current diffuses through a Cu channel in a LNLSV structure, the magnetization of the receiving magnet can be altered by spin transfer torque. This result has been confirmed by Zou and Ji,7 while the same has not been achieved in any other channel material although comparable or even longer spin diffusion lengths have been found in those. Expectations of low power and high performance operation lead to a detailed exploration of graphene for both electronic and spintronic applications. Graphene has been proposed as an ideal channel material.8 It exhibits a very long spin diffusion length even at room temperature9−16 due to its weak spin− orbit coupling. Moreover, both the carrier concentration in the graphene channel and contact resistance at the graphene/ ferromagnetic interface can be modulated by a vertical gate field,17 which offers device designers more flexibility. Most

D

© XXXX American Chemical Society

importantly, the high carrier mobility and high current carrying capability in graphene are essential to achieve strong spin transfer torque. Although large spin valve signals have been observed in graphene devices at room temperature, it had not been clear whether spin angular momentum from electrons in graphene can be successfully transferred onto a detector magnet. Here, we demonstrate for the first time magnetic field assisted, reversible magnetization switching through spin transfer in graphene LNLSV devices. These findings are an important step toward developing a graphene based spin logic. A magnetic field dependent critical current is observed and analyzed. In addition to the traditional spin transfer by a polarized spin current that can affect the magnetization of the detector magnet, the so-called “Slonczewski” spin torque18 that had been observed before in Cu devices,6 we suggest an extra “effectivefield” like spin torque19,20 to account for the observed switching of magnetization in graphene devices. For our study we have fabricated LNLSV structures on high quality peeled multilayer graphene (seven layers in the presented device) using permalloy (Py) as the ferromagnetic contacts (contact 2 and 3), as shown in Figure 1. Different from conventional LNLSV devices where the thickness of the Py Received: July 10, 2013 Revised: October 15, 2013

A

dx.doi.org/10.1021/nl402547m | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Figure 1. Graphene LNLSV device structure. (a) Schematic structure and (b) scanning electron micrograph of a graphene LNLSV device used for spin torque experiments. The graphene channel length between the Py electrodes is 300 nm, and the channel width is 1.1 μm. The size of the Py injector is 300 nm (W) × 25 nm (H), and the detector is 100 nm (W) × 5 nm (H). Spin currents are separated from the charge current loop and diffuse to the detector side. Magnetization configurations of the Py injector and detector can be identified by electrochemical potential difference between the two voltage electrodes.

Figure 2. Nonlocal spin valve measurements. (a) Nonlocal spin valve signal at high carrier density regime (VG = +40 V) as a function of external magnetic field along the easy axis for the device shown in Figure 1. (b) The coercive fields of Py nanomagnets for widths of 300 nm and 100 nm from three different samples with Py thicknesses of 40 nm, 25 nm, and 10 nm. The coercive field decreases with Py thickness. The coercive field of the 25 nm thick injector is identified as ∼15 mT, and the extrapolated data point (black dot) indicates a smaller coercive field of ∼5 mT for the 5 nm thick detector. The respective magnet responsible for each sharp switching in (a) can be identified accordingly.

graphene to be ∼4 μm for our devices with 5−7 layer graphene.11 As expected, the sharp transitions found at the fields of ±5 mT and ±15 mT correspond to the magnetization switching of the Py nanomagnets. However, since the coercive field of a magnet strongly depends on its dimension, it is necessary to identify in our device structure at which fields the injector and detector switch. From measurements on other samples where the two magnets have the same thickness but different widths, the coercive fields are found to be in general larger for narrower magnets and decrease with thickness, as shown in Figure 2b. This trend is in agreement with theoretical predictions.25,26 Note that a narrow magnet can switch at a smaller coercive field than a wider one if its thickness is substantially smaller. Given the designed dimensions of the injector [300 nm (W) × 25 nm (H)] and detector [100 nm (W) × 5 nm (H)], we found that different from conventional NLSV structures our detector switches at ± 5 mwhile ±15 mT corresponds to the injector switching. Knowing the coercive fields of the magnets, we now study the magnetization reversal by spin transfer torque in the presence of an external magnetic field along the easy axis. The device is preset to a P state at which the magnetization of both the Py injector and the detector are aligned in the negative field direction. The external field is then scanned from negative to positive and stops at +4 mT, right before the coercive field (+5 mT) at which the detector will switch (Path 1 in Figure 2a). A

injector and detector is typically kept the same, scaling down the thickness of the detector to 5 nm is found to be critical for easy switching19,21 while keeping the injector side as thick as 25 nm to ensure fully polarized spin injection. To achieve sufficient spin injection from the injector to the graphene channel and to reduce spin-flip scattering at the interface,10,15,22−24 a tunneling barrier formed from a thin film of oxidized Al (0.6 nm) has been inserted between the Py and the graphene channel. A gate voltage (VG) of +40 V is applied to the Si substrate based on our previous finding that the optimal spin diffusion length in graphene is achieved at higher carrier concentrations in multiple layer graphene with 5−7 layers.11 Contact resistance per unit width and graphene sheet resistance are gate voltage dependent as well. They are measured to be ∼1.5 kΩ·μm and ∼0.3 kΩ/□, respectively, at VG = +40 V. All measurements are performed at 77 K, using standard ac lock-in technique. The measured nonlocal spin valve (NLSV) resistance, Rnonlocal, is plotted in Figure 2a as a function of the external magnetic field applied along the easy axis (x-axis shown in Figure 1) of the Py electrodes. The nonlocal spin valve signal, the resistance difference between the parallel (P) and antiparallel (AP) states, is obtained as ∼0.6 Ω for this device. Our early work employing multiple devices with various channel lengths allowed extraction of the spin polarization at the contact interface to be ∼4% and spin diffusion length of B

dx.doi.org/10.1021/nl402547m | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

Figure 3. Spin transfer torque measurements. (a) In the preset parallel state (P), the magnetization of the detector switches to the opposite direction by a positive current pulse of +4.5 mA assisted by Bext = +4 mT. (The positive current is defined by the current flow from contact 2 through channel to contact 1 in the device shown in Figure 1.) (b) In a subsequent scan, Rnonlocal changes at Bext = +15 mT due to the field-driven switching of the magnetization in the injector. (c) In the preset antiparallel state (AP), the magnetization of the detector switches to align with the injector by a negative current pulse of −4.5 mA assisted by Bext = −4 mT. (d) In the following scan, no Rnonlocal change is observed since the magnetizations of the two magnets are already in parallel.

spin transfer torque, no further Rnonlocal change is observed in the negative field scan (Figure 3d). In our LNLSV graphene devices, the charge current flows between contacts 2 and 1 to induce the magnetization reversal in the detector magnet (contact 3) through spin diffusion. With the assistance of a magnetic field of +4 mT, as the injected dc current is increased, Rnonlocal sharply decreases at Idc = +4.5 mAindicating a switching of the Py nanomagnets from the P to the AP state. Similarly, assisted by a magnetic field of −4 mT, Rnonlocal increases at Idc = −4.5 mA, showing an AP to P switching. Since the polarity of the current pulse determines the injected spins are either aligned (negative current) or opposite (positive current) to the injector, a positive current is always required if a P to AP switching is desired, while a negative current leads to AP to P switching. The complete plot of Rnonlocal as a function of applied dc current is presented in Figure 4, as the first experimental demonstration of magnetization reversal by pure spin current in a nonlocal graphene channel. Current-induced heating and Oersted field from the nonlocal charge flow have also been ruled out for the presented magnetization reversal in this study. Under the same preset P condition in which the two magnets are aligned with the negative field direction, we find the same current pulse, but with a negative polarity cannot result in the P to AP switching (see Figure S2 in the Supporting Information). This distinct current direction dependence provides the evidence that current-induced heating is not responsible for the observed Rnonlocal change. In another case, both magnets are preset to be aligned to the positive magnetic field. No P to AP switching is observed in a backward scan during which a negative current pulse is applied (see Figure S3 in the Supporting Information).

current pulse of +4.5 mA with duration of 5 μs is then applied between contact 2 and 1. The positive current extracts spins from the graphene channel through the injector, which is aligned to the negative field direction, and leaves electrons with opposite spins accumulated right underneath the injector. Given the long spin diffusion lengths in graphene, a large number of electrons with this “opposite spin” is transferred to the detector and switches its magnetization to the opposite direction (positive field direction). The preset P state is now transformed into an AP state at a magnetic field smaller than the coercive field, and a clear Rnonlocal change is observed as shown in Figure 3a. Next, we continue to scan the external magnetic field from +4 to +20 mT, as shown in Figure 3b. A change in resistance is observed at +15 mT at which the magnetization of the injector is reversed and no longer appears at +5 mTa clear evidence that the magnetization of the detector had already been reversed by the current pulse. The AP to P state switching is also demonstrated. By scanning the magnetic field from a large negative field to ∼+11 mT, we preset the device to an AP statethe magnetization of the injector is aligned with the negative field direction, while the detector is oppositely aligned. The magnetic field is then reversed to scan back to −4 mT, a field that is not yet able to switch the detector magnet (Path 2 in Figure 2a). A negative current pulse of −4.5 mA is applied to inject spins which are aligned to the negative field into the graphene channel and transfer the spin angular momentum to the detector to reverse its magnetization to be aligned with the injector. An increase of Rnonlocal is observed in Figure 3c upon current injection, indicating a successful magnetization reversal from the AP to the P state. Again, since the P state has been already enabled by C

dx.doi.org/10.1021/nl402547m | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

magnetic multilayer systems.19 Through this exchange force, both spin torque and an effective field, are induced by the spin current and both contribute to current driven magnetization reversal. The torque contribution, named “Slonczewski” term, hsl, is consistent with the spin current hs in the second term of eq (1), while the effective field, named “field-like” term, hfl, can be added to the first term of eq (1), as (1 + h + hfl). Although induced by spin current, hfl acts as a field-driven switching of the magnetization in the easy-plane,19,20 without an easy-plane anisotropy energy involved. The switching threshold now becomes |h + hfl| = 1. Since Hk and Hext have similar values in our experiment, the magnetization reversal induced by hfl is expected to show a strong dependence on the external field as compared to a pure Slonczewski spin torque. Note that in magnetic tunnel junctions (MTJs), which also have tunneling interfaces in their device structures, the contribution of the field-like term is significant in experimental demonstrations27,28 as well which is consistent with our results. In conclusion, the first experimental measurement of spin torque in graphene devices is presented. To explain our experimental data consistently, we have suggested that spin torque switching in graphene LNLSV devices with tunneling barriers relies on the contributions from both, the Slonczewski term and an appreciable field-like term.

Figure 4. Reversible magnetization switching as a function of the injected current. The decrease of Rnonlocal at Idc = +4.5 mA assisted by Bext = +4 mT indicates the change from the P state to the AP state. The increase of Rnonlocal at Idc = −4.5 mA assisted by Bext = −4 mT implies the change from the AP state to the P state.

Note that, in the spin torque effect described above, a positive current is always needed for P to AP switchingno matter whether the P state is aligned with the positive or negative field. On the other hand, in a current induced Oersted field effect, a preset P state aligned to the negative magnetic field requires a positive current to switch to an AP state, while a P state aligned to the positive field needs a negative current to switchwhich is not observed in our experiment. Therefore, the possibility of magnetization switching through Oersted field generated by nonlocal charge currents is also excluded. The detail experiments and discussions are included in the Supporting Information. The successful magnetization reversal in graphene devices is found to be dependent on both the magnitude of the injected current and the external magnetic field. At the critical charge current of |4.5 mA|, no magnetization reversal is observed below a magnetic field of |4 mT|. To examine spin current induced magnetic precession dynamics, we first adopt Sun’s analytical model to explain spin torque by evaluating the average energy variation rate in the detector magnet:18 1 K

dU dτ

⎤ ⎡ ⎛ 1 ⎞ = −2(1 + h)⎢α⎜1 + h + hp⎟ + hs⎥θ02 ⎦ ⎣ ⎝ 2 ⎠

■

ASSOCIATED CONTENT

S Supporting Information *

Experiments performed to exclude switching induced by heating and Oersted field; mathematical model of the “Fieldlike” term and the “Slonczewski” term. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

(1)

where U is the potential energy, τ is a natural time unit, K = (1/ 2)MHk with M being the magnetization and Hk being the uniaxial anisotropy field, h = Hext/Hk is the normalized external magnetic field, hp = 4πMs/Hk is the normalized easy-plane anisotropy field and Ms is the saturation magnetization, hs is the effective spin current normalized with the uniaxial anisotropy energy MsHk/2, and α is the Gilbert damping coefficient. In eq (1), in the low field regime, the instability occurs when hs exceeds a critical value, hs < −(1 + h + (1/2)hp)α. Since 4πMs of the detector is typically >1 Tesla, which is much larger than Hk ∼ |5 mT| and Hext = |4 mT| from the experiments, it is obvious that the easy-plane anisotropy field, hp, dominates the spin torque. Hence, the impact of the external magnetic field is negligible, which is different from our experimental results. The first term, (1 + h), from eq (1) is treated as pure magnetic switching with a switching threshold of |h| = 1, completely independent of the spin current. To understand our experimental observation of spin-torque switching in the presence of a dc current together with a small external magnetic field, we have to include an exchange force between the perpendicular spin current and background magnetization, as proposed by Zhang et al. for vertical

■

ACKNOWLEDGMENTS

■

REFERENCES

We would like to acknowledge our very fruitful conversations with Prof. Supriyo Datta and Dr. Dmitri Nikonov. We also gratefully acknowledge the support of this work by Nanoelectronics Research Initiative (NRI) through the Institute for Nanoelectronics Discovery and Exploration (INDEX) center.

(1) Behin-Aein, B.; Datta, D.; Salahuddin, S.; Datta, S. Proposal for an all-spin-logic device with built-in memory. Nat. Nanotechnol. 2010, 5, 266−270. (2) Slonczewski, J. C. Current-driven excitation of magnetic multilayers. J. Magn. Magn. Mater. 1996, 159, L1−L7. (3) Berger, L. Emission of spin waves by a magnetic multilayer traversed by a current. Phys. Rev. B 1996, 54, 9353−9358. (4) Johnson, M.; Silebee, R. H. Coupling of electronic charge and spin at a ferromagnetic-paramagnetic metal interface. Phys. Rev. B 1988, 37, 5312−5325. (5) Jedema, F. J.; Heersche, H. B.; Filip, A. T.; Baselmans, J. J. A.; van Wees, B. J. Electrical detection of spin precession in a metallic mesoscopic spin valve. Nature 2002, 416, 713−716. (6) Yang, T.; Kimura, T.; Otani, Y. Giant spin-accumulation signal and pure spin-current-induced reversible magnetization switching. Nat. Phys. 2008, 4, 851−854. D

dx.doi.org/10.1021/nl402547m | Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

Letter

(7) Zou, H.; Ji, Y. Temperature evolution of spin-transfer switching in nonlocal spin valves with dipolar coupling. J. Magn. Magn. Mater. 2008, 323, 2448−2452. (8) Geim, A. K.; Novoselov, K. S. The rise of graphene. Nat. Mater. 2007, 6, 183−191. (9) Hill, E. W.; Geim, A. K.; Novoselov, K.; Schedin, F.; Blake, P. Graphene spin valve devices. IEEE Trans. Magn. 2006, 42, 2694−2696. (10) Tombros, N.; Jozsa, C.; Popinciuc, M.; Jonkman, H. T.; van Wees, B. J. Electronic spin transport and spin precession in single grapheme layers at room temperature. Nature 2007, 448, 571−574. (11) Gao, Y.; Kubo, Y.; Lin, C. C.; Chen, Z.; Appenzeller, J. Optimized spin relaxation length in few layer graphene at room temperature. IEEE Electron Devices Meeting (IEDM) 2012, San Francisco, CA, Dec 10−13, 2012; 10.1109/IEDM.2012.6478978. (12) Cho, S.; Chen, Y. F.; Fuhrer, M. S. Gate-tunable graphene spin valve. Appl. Phys. Lett. 2007, 91, 123105. (13) Ohishi, M.; et al. Spin injection into a graphene thin film at room temperature. Jpn. J. Appl. Phys. 2007, 46, L605−L607. (14) Han, W.; et al. Electrical detection of spin precession in single layer graphene spin valves with transparent contacts. Appl. Phys. Lett. 2009, 94, 222109. (15) Han, W.; et al. Tunneling Spin Injection into Single Layer Graphene. Phys. Rev. Lett. 2010, 105, 167202. (16) Yang, T. Y.; et al. Observation of long spin-relaxation times in bilayer graphene at room temperature. Phys. Rev. Lett. 2011, 107, 047206. (17) Knoch, J.; Chen, Z.; Appenzeller, J. Properties of metalgraphene contacts. IEEE Trans. Nanotechnol. 2012, 11, 513. (18) Sun, J. Z. Spin-current interaction with monodomain magnetic body: A model study. Phys. Rev. B 2000, 62, 570−578. (19) Zhang, S.; Levy, P. M.; Fert, A. Mechanisms of spin-polarized current-driven magnetization switching. Phys. Rev. Lett. 2002, 88, 236601. (20) Ralph, D. C.; Stiles, M. D. Spin transfer torques. J. Magn. Magn. Mater. 2008, 320, 1190−1216. (21) Zhang, J.; Levy, P. M.; Zhang, S.; Antropov, V. Identification of transverse spin currents in noncollinear magnetic structures. Phys. Rev. Lett. 2004, 93, 256602. (22) Schmidt, G.; Ferrand, D.; Molenkamp, L. W. Fundamental obstacle for electrical spin injection from a ferromagnetic metal into a diffusive semiconductor. Phys. Rev. B 2000, 62, R4790−R4793. (23) Rashba, E. I. Theory of electrical spin injection: tunnel contacts as a solution of the conductivity mismatch problem. Phys. Rev. B 2000, 62, R16267−R16270. (24) Fert, A.; Jaffrès, H. Conditions for efficient spin injection from a ferromagnetic metal into a semiconductor. Phys. Rev. B 2001, 64, 184420. (25) Osborn, J. A. Demagnetizing factors of the general ellipsoid. Phys. Rev. 1945, 67, 351−357. (26) Beleggia, M.; Graef, M. D.; Millev, Y. T. The equivalent ellipsoid of a magnetized body. J. Phys. D.: Appl. Phys. 2006, 39, 891−899. (27) Sankey, J. C.; et al. Measurement of the spin-transfer-torque vector in magnetic tunnel junctions. Nat. Phys. 2008, 4, 67−71. (28) Kubota, H.; et al. Quantitative measurement of voltage dependence of spin-transfer torque in MgO-based magnetic tunnel junctions. Nat. Phys. 2008, 4, 37−41.

E

dx.doi.org/10.1021/nl402547m | Nano Lett. XXXX, XXX, XXX−XXX