MRS BULLETIN VOLUME 31 MAY

www.mrs.org/bulletin Spin-Polarized Current in Spin Valves and Magnetic Tunnel Junctions Introduction to Spin Valve and Magnetic Tunnel Junction Devices More than 70 years ago, it was realized that in simple ferromagnetic metals such as Fe, Co, and Ni, the current is carried by spin-polarized electrons. This phenomenon arises from a significant spin-dependent scattering of the majority ( up ) and minority ( down ) spin-polarized electrons. 1 Many of the magnetotransport properties of these elements and their alloys can be understood within a two-current model in which the electrical current consists of independent up and down spin currents. It took more than half a century, however, Stuart Parkin Abstract Spin-polarized currents can be generated by spin-dependent diffusive scattering in magnetic thin-film structures or by spin-dependent tunneling across ultrathin dielectrics sandwiched between magnetic electrodes. By manipulating the magnetic moments of the magnetic components of these spintronic materials, their resistance can be significantly changed, allowing the development of highly sensitive magnetic-field detectors or advanced magnetic memory storage elements. Whereas the magnetoresistance of useful devices based on spin-dependent diffusive scattering has hardly changed since its discovery nearly two decades ago, in the past five years there has been a remarkably rapid development in both the basic understanding of spindependent tunneling and the magnitude of useful tunnel magnetoresistance values. In particular, it is now evident that the magnitude of the spin polarization of tunneling currents in magnetic tunnel junctions not only is related to the spin-dependent electronic structure of the ferromagnetic electrodes but also is considerably influenced by the properties of the tunnel barrier and its interfaces with the magnetic electrodes. Whereas the maximum tunnel magnetoresistance of devices using amorphous alumina tunnel barriers and 3d transition-metal alloy ferromagnetic electrodes is about 70% at room temperature, using crystalline MgO tunnel barriers in otherwise the same structures gives tunnel magnetoresistance values of more than 350% at room temperature. Keywords: magnetic, memory, spintronic. before it was appreciated that these currents can be manipulated in inhomogeneous magnetic systems composed of magnetic and nonmagnetic regions so as to modify the flow of current in these systems and thereby their resistance. Examples include magnetic multilayers comprising alternating thin magnetic and nonmagnetic layers 2 4 such as Fe/Cr, Co/Ru, and Co/Cu, and granular magnetic alloys composed of immiscible magnetic and nonmagnetic metals such as Co and Cu. 5,6 These systems exhibit very large changes in resistance at room temperature in response to magnetic fields as the magnetization directions of neighboring magnetic layers or regions are changed. This phenomenon is often referred to as giant magnetoresistance (GMR). 7 14 The largest GMR effects are found in Co/Cu multilayers, 15 with changes in resistance exceeding 100% at room temperature. However, these large effects are found in multilayers in which the individual Co and Cu layers are ultrathin just two to three atomic layers thick because the origin of GMR arises largely from spindependent scattering not within the interior of the magnetic layers ( bulk scattering, as Mott had considered 1 ) but rather from the interfaces between the individual layers ( interface scattering). 16 The relative importance of interface as compared with bulk scattering was a topic of considerable debate in the early days of GMR, but now it is generally agreed that the contribution of spin-dependent scattering from a magnetic interface layer is perhaps 100 times as large as that from an interior magnetic layer. This is very important technologically because of the need to minimize and control the influence of long-range magnetic dipolar fields emanating from the magnetic component. GMR has found one extremely important application in the form of a highly sensitive magnetic recording read head for magnetic hard disk drives (see Reference 13). The GMR effect in magnetic multilayers is not suitable for sensors but can be engineered to create useful devices by several concepts of spin engineering. 13 In particular, one important concept is fixing the direction of the magnetic moment of individual magnetic layers in a thin-film device by the phenomenon of exchange bias. 17,18 Exchange bias is a magnetic exchange interaction at the interface between a ferromagnet and an antiferromagnet that can lead to a unidirectional exchange anisotropy in the ferromagnet. This interaction can be so large that the magnetic hysteresis loop of a ferromagnetic film can be shifted from zero field by an exchange-bias field, which can exceed 1 T or more in magnitude. A second very important concept is the use of oscillatory interlayer exchange coupling of 3d transition-metal ferromagnetic layers through intermediate nonmagnetic layers of the 3d, 4d, and 5d transition metals and the noble metals Cu, Ag, and Au. 3,19 All of these nonmagnetic metals exhibit a long-range indirect interlayer exchange coupling that oscillates between ferromagnetic and antiferromagnetic coupling. By using specific thicknesses of these nonmagnetic metals that give rise to antiferromagnetic coupling, artificial anti- MRS BULLETIN VOLUME 31 MAY 2006 389

ferromagnetic (AAF) thin-film structures can be created that have zero or small net magnetic moments and, consequently, significantly reduced magnetic dipolar fields. These concepts are illustrated in Figure 1. Figure 1d shows the basic structure of the most commonly used GMR read-head device today. The device, a spin valve, contains a reference magnetic electrode formed from an exchange-biased artificial antiferromagnet separated from the sensing layer by a thin Cu layer (yellow) 20 Å thick. The sensing layer is typically formed from a soft ferromagnetic alloy with the insertion of a very thin interface layer for enhanced interface scattering. The AAF layer universally contains an ultrathin ruthenium layer just 5 10 Å thick to provide the antiferromagnetic coupling layer. These devices act, in some ways, like a valve, where the flow of current through them is modified by applying a magnetic field. In the nearly two decades since the discovery of GMR and oscillatory interlayer coupling in transition-metal systems, the magnitude of the GMR signal exhibited by spin-valve structures has changed very little. The resistance of such structures is typically about 10 15% higher when the sensing and reference magnetic moments are antiparallel than when they are parallel to one another. Partly for this reason, interest has been renewed in the past decade in devices based not on spindependent diffusive scattering but rather on spin-dependent tunneling through an ultrathin dielectric layer forming a tunnel barrier. Such magnetic tunnel valve, or magnetic tunnel junction (MTJ), devices were first explored in the 1970s, but the tunneling magnetoresistance (TMR) was very small. 20,21 In the following 20 years, modest increases in TMR were obtained, with values of about 10% at room temperature achieved in 1995 in devices with Fe-Co electrodes. 22,23 Most work in this period of time used amorphous aluminum oxide tunnel barriers which had been extensively studied for Josephson junction superconducting devices. A method of forming the tunnel barrier in which a few monolayers of aluminum metal were first deposited and then were subsequently oxidized, either thermally using molecular oxygen or reactively using atomic oxygen, had been successful in creating high-quality tunnel barriers with very low leakage currents. 24,25 In a few years after 1995, the TMR of MTJs increased rapidly, with values of up to 60% for MTJs with CoFe electrodes and up to 70% for MTJs with amorphous ferromagnetic electrodes formed by incorporating small amounts (10 25 at.%) of glass-forming elements such as B, Zr, or Hf in the CoFe alloy. Interestingly, the magnitude of the TMR signal was found to be rather insensitive to the ferromagnetic electrode material. 26 The basic origin of TMR lies in the degree of spin polarization P of the tunneling current in the MTJ device. For many years, this was considered to be related to the fundamental electronic structure of the ferromagnetic electrodes and, in particular, to the degree of spin polarization of the density of states at the Fermi energy of the electrodes. Similar to the two-current model of transport in ferromagnetic metals, the tunneling current in an MTJ device may be considered as comprising independent majority spinpolarized and minority spin-polarized currents. Assuming that these currents are proportional to the corresponding spinpolarized density of states in the emitting and receiving magnetic electrodes, then the TMR can be defined as TMR P 1 P 2 (1 P 1 P 2 ), (1) where P 1 and P 2 are the spin polarizations of the density of states of the two magnetic electrodes. 20,27 In this simple model, the TMR is determined solely by the electronic structure of the magnetic electrodes and is insensitive to the tunnel barrier properties. It has become clear recently that this model is much too simpleminded, and that the properties of the tunnel barrier and its interface with the magnetic electrodes are extremely important in determining the degree of spin polarization of the tunnel current and, consequently, the tunneling magnetoresistance, as discussed in the following sections. Influence of Chemical Bonding on Spin-Polarized Tunneling It is well known in MIM (metal insulator metal) tunneling structures that the tunneling characteristics are strongly influenced by the electronic structure of the metal interface layers. 28 Similarly, in magnetic tunnel junctions, the interface electronic structure and the nature of the bonding between the ferromagnet and the insulator are clearly very important. Indeed, it was long postulated that the positive sign (majority spin electrons) of the polarization of the tunneling current from Co and Ni ferromagnetic electrodes reflects the predominant tunneling of the more delocalized conduction-band electrons with sp character rather than the more localized electrons with d character, even though the density of states near the Fermi energy of the d electrons is considerably higher. 29 Slight variations in the nature of the bonding between Fe and Co and oxygen in alumina tunnel barriers have been calculated to affect both the magnitude and sign of the polarization of the tunneling current. 30,31 Unfortunately, the role of oxygen bonding is difficult to probe experimentally because of the ease of formation of transition-metal oxides, many of which are nonferromagnetic or antiferromagnetic. An interesting system in which to explore the role of oxygen bonding on spindependent tunneling are transition-metal alloys formed from one or more of Fe, Ni, and Co diluted with nonmagnetic elements, in conjunction with tunnel barriers which do or do not contain oxygen. Figure 2 compiles results on alloys of Co-Pt and Co-V with tunnel barriers formed from Al 2 O 3 and AlN. 32 The figure includes measurements of the spin polarization of the tunneling current measured by superconducting tunneling spectroscopy in ferromagnet insulator superconductor (FIS) junctions formed by replacing one of the ferromagnetic electrodes in an MTJ with a superconducting layer of Al. Analysis of the voltage-dependence of the conductance of such structures at temperatures well below the superconducting transition temperature of Al ( 2 K) allows the direct determination of the tunnel current polarization. 27 The figure shows that for Al 2 O 3 barriers, the tunneling spin polarization (TSP) decreases rapidly when Co is diluted with V but hardly changes at all when Co is diluted with up to 50% Pt. When the Al 2 O 3 barrier is replaced with AlN, the results are quite different. For the case of V, the TSP now decreases more slowly as the Co is diluted with V, but for the case of Pt, the TSP decreases more quickly with increasing Pt concentration. These results can be rationalized based on the different strengths of the chemical bonds formed between the metal electrodes and O or N in the barrier, as illustrated schematically in Figure 3. Assuming that the tunneling process is one which is highly localized in nature and that the tunneling current may vary significantly between neighboring atomic sites in the metal interface layer, then the net tunneling current is composed of currents tunneling from individual Co and Pt or V atoms (see Figure 3). If we conjecture that these currents are directly related to the strength of the local chemical bonds formed at the interface with the tunnel barrier, then since Pt forms a much weaker bond with oxygen than does Co (see Table I), we conclude that the tunneling current from Co-Pt alloys for alumina tunnel barriers will be dominated by tunneling from the Co atoms. If we further assume that the current from the Co component is spin-polarized and that from Pt (or V) is not, then we can model the dependence of the spin polarization of the tunneling current on the Co-Pt 390 MRS BULLETIN VOLUME 31 MAY 2006

Figure 1. Spin-engineered magnetic devices. (a), (b) The easy axis of the free ferromagnetic layer in a magnetoresistive (MR) device is oriented based on the purpose for which it is engineered. Field sensor devices such as read heads rely on a free layer with an easy axis at right angles to the moment of the pinned layer. Impinging magnetic fields will rotate the moment away from this middle position and the sensor resistance changes. On the other hand, MR devices designed for use in memory applications will have a free-layer easy axis parallel to that of the pinned layer. (c) A very basic giant magnetoresistance/tunneling magnetoresistance (GMR/TMR) stack consisting of (1) a pinned ferromagnetic layer magnetically locked by exchange bias to the interfacial field of an antiferromagnetic (AF) layer and (2) a simple ferromagnetic free layer. The spin valve is such a stack using a conducting spacer layer between the ferromagnetic layers. (d) In this case, the pinned layer is in fact an element consisting of a pair of ferromagnetic layers antiferromagnetically coupled through a ruthenium spacer layer; the lower layer in this artificial antiferromagnet is pinned via exchange bias as in (c). This flux closure (light blue ellipses) increases the magnetic stability of the pinned layer and reduces coupling to the free layer. (e) Pinned element consisting of an AF-coupled pair of ferromagnetic layers acting as a single hard layer, with no exchange-bias layer to discourage rotation of the pinned element. (f) Both the pinned and free elements consist of AF-coupled pairs. (g) A double tunnel junction. All ferromagnetic elements consist of AF-coupled pairs. There are two pinned ferromagnets, both exchange-biased by AF layers. Spin-filtering occurs both as current tunnels from the first pinned layer to the free element and again as it tunnels from the free element to the second pinned element. (From Reference 13.) MRS BULLETIN VOLUME 31 MAY 2006 391

Figure 2. Tunneling spin polarization (TSP) at 0.25 K (red and green symbols) and saturation magnetization (blue symbols) at 5 K for (a) Co 1 x V x and (b) Co 1 x Pt x alloys as a function of the V and Pt atomic fractions for Al 2 O 3 and AlN tunnel barriers. Note that the TSP values for AlN, which are systematically lower than for Al 2 O 3, have been scaled by factors of 2 and 2.6, respectively, in (a) and (b). The dotted line in (b) is a fit to the spin polarization data for Al 2 O 3 barriers, assuming that the polarization of the electrons tunneling from Co sites is independent of Pt concentration. The fit yields a tunneling probability from Pt sites that is 3.8 times lower than that for Co sites. The dashed line in (b) is a similar fit to thetsp data for AlN barriers in which the tunneling probability from Pt is only 10% lower than from Co. (From Reference 32.) Figure 3. Schematic representation of the relative magnitude of the tunneling currents from the constituent elements in Co-Pt and Co-V alloys for tunneling for Al 2 O 3 and AlN tunnel barriers. and Co-V concentrations, as shown in Figure 2. For Co-Pt/Al 2 O 3, the experimental measurements can be fit with such a model if the probability of tunneling from Co is about 3.5 times higher than that from Pt. Perhaps coincidentally, this is nearly the same ratio as that of the strength of the Co-O to Pt-O bonds (see Table I), as inferred from the heat of formation of the corresponding oxygen bond per mole of metal atom. By contrast with Pt, V forms a much stronger bond with O than does Co, so the same model would predict a fast decrease of TSP as the Co is diluted with V since the (non spin-polarized) tunneling current from V would increase rapidly (faster than in proportion to the V concentration). This is as observed, as shown in Figure 2. As mentioned earlier in this section, when Al 2 O 3 is substituted by AlN, the dependence of TSP on Pt and V concentrations is quite different, but the results are consistent with a model in which the local tunneling probability is related to the local chemical bonding. The bonding of Co, Pt, and V with nitrogen is quite weak, so the TSP would then be diluted approximately in proportion to the Pt and V concentrations, as is found experimentally (see Figure 2). These results suggest that the magnitude of the tunneling spin polarization and, consequently, the tunneling magnetoresistance can be strongly modified by chemically engineering the bonding at the interfaces between the ferromagnetic electrodes and the tunnel barrier. Influence of Wave-Function Symmetry on Spin-Polarized Tunneling The influence of the chemical bonds formed at the ferromagnet/insulator interface on the magnitude of the tunneling current can be described in terms of a tunneling matrix element. The tunneling current is proportional to the density of states multiplied by the corresponding tunneling matrix element. Similarly, the tunneling matrix elements will also be strongly influenced by the symmetry of the wave functions of the conduction-band states in the ferromagnet. The electronic wave functions decay into the tunnel barrier evanescently, with a decay length that depends on the symmetry of the wave functions. Thus, states with a more delocalized character will decay less quickly into the tunnel barrier and so have a correspondingly larger tunneling matrix element. This means that if the majority and minority spin-polarized conduction-band states in the ferromagnet have significantly different symmetries, then these states will decay at different rates across the tunnel barrier, leading to an in- 392 MRS BULLETIN VOLUME 31 MAY 2006

TableI:Heat of Formation of Oxygen Bonds Formed with Co, Pt, and V Per Mole of Metal Atom. Oxide Nitride Co 237 kj mol CoN is only stable to 230 C, where it decomposes. 1 Pt 71 kj mol Pt forms a nitride only under extreme conditions, 2 but is more stable than PtO. 3 V 430 kj mol 1 K. Suzuki et al., J. Alloys Compd. 224 (1995) p. 232. 2 E. Gregoryanz et al., Nature Mater. 3 (2004) p. 294. 3 Friedman-Hill et al. J. Chem. Phys. 100 (1994) p. 6141. creased (or decreased) spin polarization of the tunneling current. Thus the tunnel barrier can act as a spin filter. This scenario was predicted for MTJs formed from singlecrystalline Fe/MgO/Fe sandwiches oriented in the (100) direction 33,34 where the Fe is bcc, the MgO is simple cubic, and the two lattices are rotated with respect to each other by 45 to allow for a nearly perfect epitaxial relationship. Following the theoretical predictions of very high TMR values in (100) Fe/MgO/Fe, a number of groups attempted to prepare single-crystalline epitaxial thin-film Fe/ MgO/Fe sandwiches, and although highly perfect structures were prepared, only modest TMR values were obtained for several years. 35 38 In 2001, our group used magnetron sputter-deposition techniques to successfully prepare highly (100)- textured, exchange-biased MTJs with MgO(100)-oriented tunnel barriers that exhibited very high TMR values exceeding 220%. 39,40 Nearly identical MTJ structures with amorphous alumina barriers exhibited much lower TMR values of 70%. These results appear to confirm the theoretical predictions of Butler et al. 33 With slight modifications of the structure reported in Reference 39, even higher TMR values are obtained, as shown in Figure 4. The resistance versus field curves in Figure 4 exhibit TMR values of more than 350% at room temperature and nearly 600% at helium temperatures. These structures are useful technologically both because they are formed at room temperature using simple sputter-deposition techniques and because they are prepared on amorphous substrates. In this case, the substrate was an amorphous SiO 2 layer formed on Si(100). A cross-section transmission electron micrograph of a typical MTJ is shown in Figure 5. Underlayers of TaN/Ta are deposited first on the SiO 2 substrate, forming a template on which grows a highly (100)- textured fcc antiferromagnetic exchangebias layer of Ir 76 Mn 24. Subsequently, a reference ferromagnetic electrode of Co 70 Fe 30 is deposited that is bcc and grows Figure 4. Resistance versus field curves at (a) 2.6 K and (b) 290 K for an exchanged-biased magnetic tunnel junction with a highly (100)-textured MgO tunnel barrier. Figure 5. Transmission electron micrographs of a magnetic tunnel junction showing a highly oriented (100) MgO tunnel barrier. (a) Low-magnification image showing the growth of ultrasmooth underlayers formed from TaN, Ta, IrMn, and CoFe, each readily distinguishable, which form a template for the growth of the (100)-oriented MgO tunnel barrier (lightest layer). (b), (c) High-resolution images along the [110] zone axes, showing atomically resolved lattice planes with (100) planes perpendicular to the growth direction. The (100) planes in the grain in the center of (c) are rotated by about 15. (Based on Reference 39.) Note that in (b) the amorphous CoFeB layer that forms the upper part of the top ferromagnetic electrode can be seen. During annealing, the B diffuses from this layer to the upper part of the device, leaving CoFe, which crystallizes, increasing the extent of the crystalline CoFe interface layer. In structures without any CoFe interface layer above the MgO layer, the diffusion of B away from the MgO layer nevertheless results in the formation of a crystalline interface layer, within the experimental accuracy of electron energy loss spectroscopy, from CoFe. MRS BULLETIN VOLUME 31 MAY 2006 393

(100)-oriented. The MgO barrier is formed by first depositing a thin layer of Mg, 4 6 Å thick, followed by the reactive sputter-deposition of Mg in an Ar-O 2 plasma ( 2% oxygen) to form MgO. The Mg underlayer is used to prevent oxidation of the underlying ferromagnetic electrode, but this layer is converted to MgO by the reactive oxygen introduced into the sputter chamber during the deposition of the MgO layer. Finally, a counter ferromagnetic electrode is formed from a nominally amorphous alloy of [Co 70 Fe 30 ] 80 B 20. The TMR of this structure, as deposited, is modest ( 60 80% at room temperature), but on annealing at temperatures of up to 450 C, the TMR is significantly increased to values of more than 350% at room temperature (see Figure 4). Conclusions Spin-dependent tunneling junctions have a history dating back more than 30 years, but it is only in the past decade that magnetic tunnel junctions with significant values of tunneling magnetoresistance at room temperature have been fabricated. A detailed understanding of the relationship of the tunneling magnetoresistance and the corresponding spin polarization of the tunneling current to the ferromagnetic and tunnel barrier materials forming the magnetic tunnel junction is developing. It is clear that it is not simply the electronic structure of the ferromagnetic electrode that determines the magnitude of the spindependent tunneling. Rather, the spin polarization of the tunneling current can be strongly modified by tunneling matrix elements that themselves depend on chemical bonding at the ferromagnet/tunnel barrier interface and also on the symmetry of the electrode s conduction-band wave functions. This means that magnetic metals which are only weakly magnetic can give rise, by suitable wave-function or chemicalbond engineering, to highly spin-polarized currents. Conversely, strongly magnetic metals may give only very weakly spinpolarized tunnel currents. An important conclusion is that the tunneling spin polarization is not related in any simple way to the magnetization of the magnetic electrodes. For example, rare-earth transitionmetal ferrimagnetic electrodes with no net magnetic moment can give rise to highly spin-polarized tunneling current. 41 Magnetic tunnel junctions have a promising future, both as highly sensitive field sensors and as magnetic memory storage elements. 13 By contrast with metallic spinvalve sensors, whose magnetoresistance is limited to 10 20% at room temperature, there is no theoretical limit to the tunneling magnetoresistance of magnetic tunnel junctions. While the very high tunneling magnetoresistance of 350% at room temperature observed with crystalline MgO tunnel barriers is very attractive for sensing and memory applications, it seems likely that new materials with even higher magnetoresistance values will be found in the future, which could have even wider technological applications. Acknowledgments I am deeply indebted to my co-workers who have significantly contributed to this work, including especially Christian Kaiser, Xin Jiang, Alex Panchula, Roger Wang, Hyunsoo Yang, See-Hun Yang, Mahesh Samant, Brian Hughes, Kevin Roche, Andrew Kellock, and Phil Rice. Financial support from DARPA, DMEA, and NEDO is also gratefully acknowledged. References 1 N.F. Mott and H. Jones, Theory of the Properties of Metals and Alloys (Oxford University Press, 1936). 2. M.N. Baibich, J.M. Broto, A. Fert, F.N. van Dau, F. Petroff, P. Etienne, G. Creuzet, A. Friederich, and J. Chazelas, Phys. Rev. Lett. 61 (1988) p. 2472. 3. S.S.P. Parkin, N. More, and K.P. Roche, Phys. Rev. Lett. 64 (1990) p. 2304. 4. S.S.P. Parkin, R. Bhadra, and K.P. Roche, Phys. Rev. Lett. 66 (1991) p. 2152. 5. A.E. Berkowitz, J.R. Mitchell, M.J. Carey, A.P. Young, S. Zhang, F.E. Spada, F.T. Parker, A. Hutten, and G. Thomas, Phys. Rev. Lett. 68 (1992) p. 3745. 6. C.L. Chien, in Annu. Rev. Mater. Sci., Vol. 25, edited by B.W. Wessels (1995) p. 129. 7. S.S.P. Parkin, in Ultrathin Magnetic Structures, Vol. II, edited by B. Heinrich and J.A.C. Bland (Springer, Berlin, 1994) p. 148. 8. A. Fert and P. Bruno, in Ultrathin Magnetic Structures, Vol. II, edited by B. Heinrich and J.A.C. Bland (Springer, Berlin, 1994) p. 82. 9. S.S.P. Parkin, in Annu. Rev. Mater. Sci., Vol. 25, edited by B.W. Wessels (1995) p. 357. 10. M.A.M. Gijs and G.E.W. Bauer, Adv. Phys. 46 (1997) p. 285. 11. J. Mathon, Contemp. Phys. 32 (1991) p. 143. 12. P.M. Levy, in Solid State Physics, Vol. 47, edited by H. Ehrenreich and D. Turnbull (Academic Press, New York, 1994) p. 367. 13. S.S.P. Parkin, X. Jiang, C. Kaiser, A. Panchula, K. Roche, and M. Samant, Proc. IEEE 91 (2003) p. 661. 14. A. Barthélémy, A. Fert, J.-P. Contoura, M. Bowen, V. Crosa, J.M.D. Teresa, A. Hamzica, J.C. Fainib, J.M. George, J. Grollier, F. Montaigne, F. Pailloux, F. Petroff, and C. Vouille, J. Magn. Magn. Mater. 242 (2002) p. 68. 15. S.S.P. Parkin, Z.G. Li, and D.J. Smith, Appl. Phys. Lett. 58 (1991) p. 2710. 16. S.S.P. Parkin, Phys. Rev. Lett. 71 (1993) p. 1641. 17. W.H. Meiklejohn and C.P. Bean, Phys. Rev. 102 (1956) p. 1413. 18. J. Nogués and I.K. Schuller, J. Magn. Magn. Mater. 192 (1999) p. 203. 19. S.S.P. Parkin, Phys. Rev. Lett. 67 (1991) p. 3598. 20. M. Julliere, Phys. Lett. 54A (1975) p. 225. 21. The tunneling magnetoresistance is defined as (R AP -R P )/R P, where R AP -R P are the resistances of a magnetic tunnel junction in its highest- and lowest-resistance state. For a simple MTJ comprising a sandwich of two ferromagnetic layers separated by a single tunnel barrier, these states correspond to antiparallel and parallel orientations of the moments, respectively. 22. J.S. Moodera, L.R. Kinder, T.M. Wong, and R. Meservey, Phys. Rev. Lett. 74 (1995) p. 3273. 23. T. Miyazaki and N. Tezuka, J. Magn. Magn. Mater. 139 (1995) p. L231. 24. J.M. Rowell, M. Gurvitch, and J. Geerk, Phys. Rev. B 24 (1981) p. 2278. 25. W.H. Mallison, R.E. Miller, and A.W. Kleinsasser, IEEE Trans. Appl. Supercond. 5 (1995) p. 2330. 26. S.S.P. Parkin, K.P. Roche, M.G. Samant, P.M. Rice, R.B. Beyers, R.E. Scheuerlein, E.J. O Sullivan, S.L. Brown, J. Bucchigano, D.W. Abraham, Y. Lu, M. Rooks, P.L. Trouilloud, R.A. Wanner, and W. J. Gallagher, J. Appl. Phys. 85 (1999) p. 5828. 27. R. Meservey and P.M. Tedrow, Phys. Rep. 238 (1994) p. 173. 28. E.L. Wolf, Principles of Electron Tunneling Spectroscopy (Oxford University Press, New York, 1989). 29. M.B. Stearns, J. Magn. Magn. Mater. 5 (1977) p. 167. 30. E.Y. Tsymbal, I.I. Oleinik, and D.G. Pettifor, J. Appl. Phys. 87 (2000) p. 5230. 31. I.I. Oleinik, E.Y. Tsymbal, and D.G. Pettifor, Phys. Rev. B 62 (2000) p. 3952. 32. C. Kaiser, S. van Dijken, S.-H. Yang, H. Yang, and S.S.P. Parkin, Phys. Rev. Lett. 94 247203 (2005). 33. W.H. Butler, X.-G. Zhang, T.C. Schulthess, and J.M. MacLaren, Phys. Rev. B 63 054416 (2001). 34. J. Mathon and A. Umerski, Phys. Rev. B 63 220403 (2001). 35. M. Klaua, D. Ullmann, J. Barthel, W. Wulfhekel, J. Kirschner, R. Urban, T.L. Monchesky, A. Enders, J.F. Cochran, and B. Heinrich, Phys. Rev. B 64 134411 (2001). 36. M. Bowen, V. Cros, F. Petroff, A. Fert, C.M. Boubeta, J.L. Costa-Krämer, J.V. Anguita, A. Cebollada, F. Briones, J.M. de Teresa, L. Morellón, M.R. Ibarra, F. Güell, F. Peiró, and A. Cornet, Appl. Phys. Lett. 79 (2001) p. 1655. 37. J. Faure-Vincent, C. Tiusan, E. Jouguelet, F. Canet, M. Sajieddine, C. Bellouard, E. Popova, M. Hehn, F. Montaigne, and A. Schuhl, Appl. Phys. Lett. 82 (2003) p. 4507. 38. S. Mitani, T. Moriyama, and K. Takanashi, J. Appl. Phys. 93 (2003) p. 8041. 39. S.S.P. Parkin, C. Kaiser, A. Panchula, P. Rice, B. Hughes, M. Samant, and S.-H. Yang, Nature Mater. 3 (2004) p. 862. 40. W.H. Butler and A. Gupta, Nature Mater. 3 (2004) p. 845. 41. C. Kaiser, A.F. Panchula, and S.S.P. Parkin, Phys. Rev. Lett. 95 04720 (2005). 394 www.mrs.org/bulletin MRS BULLETIN VOLUME 31 MAY 2006