Materials Science

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

commentary

A window on the future of spintronics


Hideo Ohno Despite low transition temperatures, ferromagnetism in diluted magnetic semiconductors has been essential in exploring new ideas and concepts in spintronics, some of which have been successfully transferred to metallic ferromagnets.

n the past couple of decades, a great deal of effort has been put into the investigation of the mechanisms behind the ferromagnetism of diluted magnetic semiconductors and oxides (DMSs and DMOs), in the hope of finding ways to raise the Curie temperature, TC, to above room temperature, as described in the article by Tomasz Dietl in this issue1. The outcome of this ongoing quest will eventually determine whether DMSs and DMOs can be used in practical applications. However, the ferromagnetism discovered in these materials, in particular III-V-based DMSs, has already played a critical role in exploring new physics, ideas and concepts in spintronics that have not readily been accessible in metal ferromagnets.

electrical manipulation of magnetism

Ferromagnetism in prototypical Mn-doped III-V semiconductors such as (In,Mn)As


a
Paramagnetic state +

and (Ga,Mn)As, in which Mn substitutes for a small percentage of the group III element (Ga or In), is stabilized by exchange interaction between charge carriers and localized spins. The carrier concentration in these ferromagnetic semiconductors is approximately equal to the Mn concentration and thus is low, of the order of 1025 to 1026 m3, as opposed to 1029 m3 in metals. It is thus possible to control a considerable portion of carriers by external electric fields using metalinsulatorsemiconductor or pn junction structures involving a thin ferromagnetic semiconductor layer, to the extent that the field significantly alters the stability of the ferromagnetic phase and other magnetic properties (Fig. 1). Modulation of TC in a ferromagnet by an electric field was first observed2 back in 2000 in a metalinsulator semiconductor structure with a thin film of (In,Mn)As; this was one of the earliest
50

22.5 K

25 RHall ()

0 E (x108 V m 1) 0 +1.6 1.6 0 0.5 0 0H (mT) 0.5 1.0

Ferromagnetic state

+ 25 l~ 1 nm

50 1.0

Figure 1 | Electric control of ferromagnetic phase in ferromagnetic semiconductors. a, Schematic diagram of the effect. The thickness, l, of the controlled region is of the order of 1 nm from the interface. b, Magnetization curves as obtained through a measure of the Hall resistivity in (In,Mn) As at different gate voltages, in the order 0, +1.6, 1.6 and back to 0 (108 V m1), showing the transition from paramagnetic to ferromagnetic states. Panel b reprinted with permission from ref. 2, 2000 NPG.
952

realizations of electrical manipulation of collective phenomena. Subsequently, electrical manipulation of the coercive field (HC) was also shown to be possible3, once again in (In,Mn)As. The ability to modify HC implies that the applied electric field changes the magnetic anisotropy. This can be understood in the following way: the presence of spontaneous magnetization along with the exchange interaction splits carrier states according to the spinorbit interaction1. With electric fields, the carrier concentration can be tuned to populate carriers differently among the states, resulting in the change in magnetic anisotropy. This electrical manipulation of the magnetic anisotropy can be used to control the direction of magnetization (Fig. 2)4, because the energy minima of the potential landscape set by the magnetic anisotropy determine the direction of magnetization. This series of experiments points to the possibility of magnetization reversal solely by means of an electric field through control of carrier concentration4, without resorting to using a magnetic field as in hard disk drives, spin-current in spin-torque transfer non-volatile random access memories, strain-induced magnetic anisotropy or multiferroics. A number of research groups started working on metal ferromagnets to explore the existence of any electric field effect. Conventional wisdom was that the carrier concentration that can be modulated per area is 1018 m2 at most (limited by dielectric breakdown of 109 V m1; this sets the upper bound unless one resorts to exotic high-permittivity materials), which is only a fraction of the electrons residing in one atomic plane (1019 m2), and thus there would be virtually no measurable effect. In contrast, coercivity modulation was observed in an experiment carried out with FePt or FePd using electrolyte to

nature materials | VOL 9 | DECEMBER 2010 | www.nature.com/naturematerials

2010 Macmillan Publishers Limited. All rights reserved

commentary
produce a high electric field of several times 108 V m1 (ref. 5), and a clear modulation of magnetocrystalline anisotropy was seen in an ultrathin iron layer (a few monolayers thick) sandwiched between MgO (insulator) and gold6, followed by an electrical magnetocrystalline anisotropy modulation in CoFe using femtosecond electron beams7. It was also shown that CoFeB (0.82 nm) sandwiched between MgO and Ta, a heterostructure known to produce giant tunnel magnetoresistance and high-performance magnetic tunnel junctions812 if there is a ferromagnetic counterelectrode on top of MgO, produced a significant modulation of anisotropy of the order of 30 J m2 per 109 V m1 (ref. 13). It is important to note that these experiments on ferromagnetic metals were done at room temperature. The electric field effect, as in the case of ferromagnetic semiconductors, can be understood as originating from a change in the carrier occupancy of d-orbitals with different spinorbit interaction at the interface, and hence contribute differently to magnetic anisotropy. Theoretical studies have also emerged1416. As our understanding deepens, the possibility of electrical magnetization reversal in magnetic nanostructures at room temperature is becoming more and more realistic. For example, a possible way to obtain it would be to design the magnetocrystalline anisotropy in such a way that it produces a ratchet-type motion on cycling the electric field. Another possibility is that by using the precessional motion brought about when the anisotropy is abruptly changed by an electric pulse, it should be possible to observe the reversal of magnetization4,6,17. Electrical reversal, when realized, should offer an extremely powerful and efficient way of manipulating magnetization. The energy needed to switch magnetization by spin-transfer torque in a state-of-the-art magnetic tunnel junction is of the order of 0.1 pJ. But for electrical reversal it should be possible to reduce this energy by two orders of magnitude, down to just 1 fJ. It is also interesting to see whether reversal of magnetization by magnetic field can be assisted by electric fields in material systems that are relevant to current storage and non-volatile memory applications. Such electric-field-assisted reversal has been shown in a ferromagnetic semiconductor structure at low temperatures3, and has subsequently been seen in ultrathin Fe films18. An initial attempt along this direction has been made to see the effect in a setting closer to hard disk drive applications19.

2K 5 0 5 10 0H = 0 4 2 0 2 E (x108 V m1) 4 (100) axis

Ferromagnet

Metal Insulator

Figure 2 | With electric fields, the carrier concentration can be tuned to populate carriers differently among the spinorbit states, resulting in a change in the magnetic anisotropy, which in turn changes the direction of magnetization m indicated by an angle (left panel). The right panel shows experimental results from (Ga,Mn)As. Reprinted with permission from ref. 4, 2008 NPG.

a new universality class

Universality manifests itself in the critical exponent of the power law that describes phenomena in complex systems with competing forces. The phenomena, no matter how different in appearance (ranging from earthquakes to the stock market), belong to the same universality class when they are described by power-law curves exhibiting the same exponent.

Understanding creep phenomena is essential for the development of devices based on domain walls.
Magnetic domain-wall motion in a disordered pinning background driven by a low external magnetic field below threshold value for depinning the domain wall also known as domain-wall creep is a classic example of phenomena described by a power law 20. Domain-wall manipulation by spin-transfer torque exerted from spin-polarized current has recently gathered attention, fuelled by suggestions that based on domain-wall motion it will be possible to build the next generation of mass storage devices with the concept known as racetrack memory 21 and a fast three-terminal non-volatile memory cell22. Domain walls are known to have an intrinsic threshold current density for spin-torque depinning given by jC = 2eKw/(hP), where K is the hard axis anisotropy, w the domain-wall width and P the spin polarization of current 23. The jC of in-plane ferromagnetic metal nanowires is beyond the experimentally accessible level. The reason one sees domain-wall motion in this case is because of the presence of the field-like torque produced by the so-called -term (also

called non-adiabatic torque; the direction of the torque, as its name suggests, is the same as the magnetic-field torque and is orthogonal to the spin-transfer torque)24. In contrast, ferromagnetic semiconductor systems with perpendicular anisotropy make it possible to go beyond jC, owing to its low magnetization (100 times lower than that of the metal counterparts), which reduces K accordingly. Whether one induces the same type of creep by spintransfer torques as by magnetic fields is a non-trivial and intriguing issue. This has made it interesting to look into currentinduced and magnetic-field-induced creep in ferromagnetic semiconductors. It turns out that at a current density below jC, we do observe domain-wall creep in (Ga,Mn) As (ref. 25). Comparison with the motion induced by magnetic fields in the same sample revealed that the two creeps, by spin-polarized currents and by external magnetic fields, are fundamentally different, belonging to different universality classes. In fact, the universality class to which the spin-polarized current-induced creep belongs was not previously known. The difference stems from the fact that magnetic field and spin transfer produce two different torques orthogonal to each other on localized spins in the domain wall (Fig. 3); until the introduction of spin-transfer torque and ferromagnetic semiconductors, there was no means of producing torque orthogonal to that produced by magnetic field. Recent experiments on metal ferromagnets using Pt/Co/Pt nanowires showed that the spin-polarized currentinduced creep in this case belongs to the same universality class as that driven by the magnetic field27. The difference is presumably due to the relative contribution of spin-transfer torque to the field-like torque; in (Ga,Mn)As, in contrast to metals, spin-transfer torque dominates. Whether
953

nature materials | VOL 9 | DECEMBER 2010 | www.nature.com/naturematerials

2010 Macmillan Publishers Limited. All rights reserved

commentary
Current-induced creep Current Domain wall Field-induced creep

Magnetic eld
Figure 3 | Magnetic domain-wall creep is a motion of the domain wall induced by spin-polarized current or by magnetic field through a disordered potential landscape. Creep induced by spin-polarized current in (Ga,Mn)As belongs to a different universality class from magnetic-field induced creep, characterized by a different exponent. This can be understood in terms of the direction of torque acting on spins in the domain wall. Yellow arrows in the figure indicate magnetic spins; green, spin torque induced by field; purple, spin torque induced by current; and black, the direction of domain wall creep.

the current-induced creep belonging to the (Ga,Mn)As universality class can be observed in ferromagnetic metals remains an open question at this stage. In any case, understanding creep phenomena is essential for the development of devices based on domain walls.

Ferromagnetic III-V semiconductors and their variants continue to function as a test bench for new and exploratory physics and ideas.
anisotropic tunnelling
The anisotropy of spinorbit split (Ga,Mn)As states manifests itself in tunnelling resistance in a very different way from the ordinary tunnel magnetoresistance. Conventional tunnel magnetoresistance requires two ferromagnetic layers separated by a thin insulator, through which electrons tunnel, and the resistance of this twoterminal device depends on the relative orientation of the magnetizations of the two ferromagnetic layers. Instead, a nonmagnetic metal/thin insulator/(Ga,Mn)As structure, with only one ferromagnetic layer in it, has been shown to have a tunnel resistance that depends on the orientation of the (Ga,Mn)As magnetization; the resistance of the structure at 4.2 K is 3% higher when magnetization lies along the [100] direction of (Ga,Mn)As than when it lies along [010] (ref. 28). This tunnelling anisotropic magnetoresistance is due to the fact that the tunnelling density of states of the spinorbit split (Ga,Mn)As bands differ with the direction of magnetization. It should be generic to any ferromagnetic material that shows anisotropic tunnelling
954

density of states, and therefore one expects to see this in ferromagnetic metal structures as well. Indeed, tunnel anisotropic magnetoresistance has been observed in a Fe/GaAs/Au tunnel junction29, with the resistance varying by a magnitude 0.4% at 4.2 K. This new tunnelling magnetoresistance may be of future use, once the effect is seen at room temperature and is made comparable to existing magnetoresistances such as giant magnetoresistance (>10%) and tunnel magnetoresistance (>100%).

Hideo Ohno is at the Center for Spintronics Integrated Systems and the Laboratory for Nanoelectronics and Spintronics, Research Institute of Electrical Communication, Tohoku University, 2-1-1, Katahira, Aoba-ku, Sendai 980-8577, Japan. e-mail: ohno@riec.tohoku.ac.jp References
1. Dietl, T. Nature Mater. 9, 965974 (2010). 2. Ohno, H. et al. Nature 408, 944946 (2000). 3. Chiba, D., Yamanouchi, M., Matsukura, F. & Ohno, H. Science 301, 943945 (2003). 4. Chiba, D. et al. Nature 455, 515518 (2008). 5. Weisheit, M. et al. Science 315, 349351 (2007). 6. Maruyama, T. et al. Nature Nanotech. 4, 158161 (2009). 7. Gamble, S. J. et al. Phys. Rev. Lett. 102, 217201 (2009). 8. Butler, W. H., Zhang, X. G., Schulthess, T. C. & MacLaren, J. M. Phys. Rev. B 63, 054416 (2001). 9. Mathon, J. & Umerski, A. Phys. Rev. B 63, 220403 (2001). 10. Parkin, S. S. P. et al. Nature Mater. 3, 862867 (2004). 11. Yuasa, S., Nagahama, T., Fukushima, A., Suzuki, Y. & Ando, K. Nature Mater. 3, 868871 (2004). 12. Ikeda, S. et al. Nature Mater. 9, 721724 (2010). 13. Endo, M., Kanai, S., Ikeda, S., Matsukura, F. & Ohno, H. Appl. Phys. Lett. 96, 212503 (2010). 14. Duan, C.-G. et al. Phys. Rev. Lett. 101, 137204 (2008). 15. Nakamura, K. et al. Phys. Rev. Lett. 102, 187201 (2009). 16. Tsujikawa, M. & Oda, T. Phys. Rev. Lett. 102, 247203 (2009). 17. Chiba, D., Nakatani, Y., Matsukura, F. & Ohno, H. Appl. Phys. Lett. 96, 192506 (2010). 18. Shiota, Y. et al. Appl. Phys. Exp. 2, 063001 (2009). 19. Zhou, T. J. et al. Appl. Phys. Lett. 96, 012506 (2010). 20. Giamarchi, T., Kolton, A. B. & Rosso, A. Lect. Notes Phys. 688, 91 (2006). 21. Parkin, S. S. P., Hayashi, M. & Thomas, L. Science 320, 190194 (2008). 22. Suzuki, T., Fukami, S., Nagahara, K., Ohshima, N. & Ishiwata, N. IEEE Trans. Magn. 45, 37763779 (2009). 23. Tatara, G. & Kohno, H. Phys. Rev. Lett. 92, 086601 (2004). 24. Ralph, D. C. & Stiles, M. D. J. Magn. Magn. Mater. 320, 11901216 (2008). 25. Yamanouchi, M. et al. Science 317, 17261729 (2007). 26. Moore, T. A. et al. Appl. Phys. Lett. 93, 262504 (2008). 27. Lee, J. C. et al. Preprint at http://arXiv.org/abs/1006.1216 (2010). 28. Gould, C. et al. Phys. Rev. Lett. 93, 117203 (2004). 29. Moser, J. et al. Phys. Rev. Lett. 99, 056601 (2007). 30. Jaworski, C. M. et al. Preprint at http://arXiv.org/abs/ 1007.1364 (2010). 31. Liu, Q., Liu, C. X., Xu, C., Qi, X. L. & Zhang, S. C. Phys. Rev. Lett. 102, 156603 (2009). 32. Liu, C. Hughes, T. L., Qi, X. L., Wang, K. & Zhang, S. C. Phys. Rev. Lett. 100, 236601 (2008).

Prospects

Ferromagnetic III-V semiconductors and their variants continue to function as a test bench for new and exploratory physics and ideas. They can be used as ferromagnets that have very unique ferromagnetic properties (low magnetization and high-spin polarization among them), a recent example being observation of the spin-Seebeck effect in (Ga,Mn)As (ref. 30), a test of the generality of an effect in which thermal gradient produces spin polarization. Epitaxial integration of ferromagnetism and semiconductor heterostructures also offers opportunities for new ideas as well as physics. An interesting possibility is to explore a broken-gap semiconductor heterostructure system, InAsGaSb, to create an interfacial topological insulator 31, where one might be able to find yet another way of stabilizing ferromagnetic order in semiconductors32. These few examples demonstrate that there is ample room for the exploration of new physics as well as future technologies in the ferromagnetic semiconductor arena, even without knowing when the Curie temperature will reach levels that will allow the use of these materials in commercial devices.

Acknowledgements

I thank Fumihiro Matsukura, Michihiko Yamanouchi and Tomasz Dietl for discussions.

nature materials | VOL 9 | DECEMBER 2010 | www.nature.com/naturematerials

2010 Macmillan Publishers Limited. All rights reserved

You might also like