Large inverse magnetoresistance of CrO2/Co ... - APS Link Manager

RAPID COMMUNICATIONS

PHYSICAL REVIEW B 69, 220413(R) (2004)

Large inverse magnetoresistance of CrO2 / Co junctions with an artificial barrier J. S. Parker, P. G. Ivanov, D. M. Lind, and P. Xiong Center for Materials Research and Technology and Physics Department, Florida State University, Tallahassee, Florida 32306, USA

Y. Xin National High Magnetic Field Laboratory, Tallahassee, Florida 32310, USA (Received 16 December 2003; revised manuscript received 10 March 2004; published 30 June 2004) We report on magnetotransport measurements on magnetic junctions consisting of Co and half-metallic CrO2 as the electrodes. The insulating barrier in between is a CrOx-AlOx layer created via a chemical modification of the native CrO2 surface, followed by the deposition and oxidation of a thin Al layer. The junctions exhibit a hysteretic low-field magnetoresistance with switching closely matching that of the magnetization of the CrO2 and Co layers. The magnetoresistance is inverse in sign, with a maximum value of −24% at 5 K, implying a negative spin polarization for Co in such structures. The magnetoresistance shows strong temperature and bias dependence, diminishing quickly with increasing temperature and bias voltage. DOI: 10.1103/PhysRevB.69.220413

PACS number(s): 85.75.Dd

Devices in which both the charge and spin of electrons play important roles have generated substantial interest in recent years. One such device consists of two ferromagnetic (FM) layers separated by an insulating barrier, typically referred to as a magnetic tunnel junction (MTJ). The two electrodes of such structures are mostly made of transition-metal ferromagnets. Various MTJ’s have exhibited large tunneling magnetoresistance (TMR) effects at low fields:1 the junction resistance changes abruptly and substantially when the magnetization of the two ferromagnetic electrodes switches from parallel to antiparallel. Such effect promises applications for MTJ’s as sensitive magnetic-field sensors and in nonvolatile magnetic random access memory.2 The junction magnetoresistance (JMR) of an ideal MTJ depends only on the electronic density of states (DOS) and spin polarization, P, of the electrodes described by a model proposed by Jullière,3 and the magnitude and sign of the JMR are related to P of the electrodes through a simple equation. Recently, it was shown4 that substantial JMR can be obtained even in the case of hopping transport through the barrier instead of direct tunneling. The JMR was understood with spin conserving hopping through localized states in the barrier and was used to infer the magnitude and sign of P of the electrode. The maximum magnetoresistance (MR) reported to date of metalbased MTJ’s at room temperature is found to be around 40%,5,6 consistent with typical P values 共30% – 50% 兲 of the common ferromagnetic metals. From a materials standpoint, an obvious next step to enhance the JMR is to fabricate MTJ’s from materials with higher spin polarization. One class of materials that has attracted particular interest is the so-called half metals. In a half metal the two spin species have different DOS: the Fermi level lies within one spin band, while the other spin band has a gap, thus the itinerant charge carriers are 100% spin polarized. An MTJ with all half-metallic electrodes 共P1 = P2 = 1兲 would produce an infinite JMR and the junction would exhibit true on-off behavior. One material that offers the potential for such applications is the binary oxide CrO2, which has been shown to be half metallic, by point-contact Andreev reflection (PCAR) measurements, which yielded a spin polarization for CrO2 as high as 97%.7,8 However, junctions made of CrO2 and Co 0163-1829/2004/69(22)/220413(4)/$22.50

with the native surface oxide Cr2O3 as the insulating barrier have shown MR much below what is expected from the spin polarizations of the electrodes.9,10 This is consistent with the PCAR measurements8 that showed a precipitous decline in the measured P of CrO2 with increasing Cr2O3 barrier thickness. Recently, we have developed a method to chemically modify the CrO2 surface and create an artificial barrier.11 Planar junctions with superconducting counterelectrodes (Pb and Al) demonstrated that the barriers thus created preserve the full spin polarization of CrO2 at the interface. In particular, Meservey-Tedrow-type12 measurements on CrO2 / I / Al junctions clearly demonstrated near 100% spin polarization at and across the artificial barrier.11 We have applied this Br-etching technique to the fabrication of CrO2 MTJs. In this paper, we report on the fabrication and magnetoelectronic characterization of CrO2 / I / Co junctions separated by an artificial insulating barrier. The junctions used in this study were fabricated as follows: First, single crystal CrO2 films were grown on TiO2 (100 or 110) substrates employing a chemical vapor deposition (CVD) technique described in detail elsewhere.13,14 The films were patterned into stripes, ⬃300 ␮m wide, by photolithography and wet chemical etching. The surface of the CrO2 was then modified using a Br/ methanol etch (⬃22% Br by volume) for 2 min before the deposition of a thin layer of Al via electron-beam evaporation in an UHV system. The Al layer was then oxidized in situ under controlled oxygen pressure (⬃2.5 mTorr for 20 min). The chamber was then evacuated and a 100-nm-thick Co counterelectrode, ⬃300 ␮m wide, was deposited through a shadow mask in a cross-stripe geometry. The cross stripes were carefully aligned along the magnetocrystalline easy axis (c axis) of the CrO2. Crosssectional transmission electron microscopy of the samples shows a well-defined barrier with two distinct layers (Fig. 1). There is an oxygen-deficient crystalline CrOx layer, followed by an amorphous AlOx layer, capped by the Co electrode. The junction resistance was measured with the ac lock-in technique. For MR measurement, the field is applied in the plane of the films and along the magnetocrystalline easy axis of CrO2 and the length of the Co stripe. After the electrical measurements, magnetic measurements were performed on

69 220413-1

©2004 The American Physical Society



PARKER et al.

FIG. 1. Cross-sectional TEM image of a CrO2 / CrOx-AlOx / Co junction fabricated with the process described in the text. The upper image shows a well-defined insulating barrier between the CrO2 and Co electrodes. The lower image shows two distinct layers within the barrier.

the same samples in a similar setup with a SQUID magnetometer. Figure 2 shows the JMR as a function of the applied field, and the corresponding magnetization curves, at 5 K for a CrO2 / I / Co junction fabricated with the above procedure. The junction resistance showed sharp switching behavior, corresponding closely with the magnetic-moment reversal of the two electrodes, indicating this is a spin-dependent effect. The junction exhibited an inverse MR: its resistance decreases to a lower value in the antiparallel configuration. In this particular junction the AlOx thickness, ␦, was approximately 4.5 nm, which resulted in a maximum MR of ⬃24%. This represents a significant improvement over junctions with native oxide on CrO2 as the barriers.9,10 When ␦ devi-

FIG. 2. Resistivity and magnetization as a function of field for the same CrO2 / CrOx-AlOx / Co junction, taken at T = 5 K.

ates away from this optimum thickness, the JMR slowly decreases. In junctions where the Br-etching step was omitted and the Al was deposited directly onto an untreated CrO2 surface, the junction resistance increased by a factor of 100– 1000, while the JMR was significantly diminished to less than 5% at 5 K. Most likely, when Al is deposited on a freshly grown CrO2 film, it reacts with CrO2, which results in an oxygen-depleted CrOx layer plus an AlOx layer, forming a thicker barrier. The Br etching probably leaves behind a thin layer of oxygen-deficient CrOx layer on CrO2, which minimizes the reaction between Al and CrO2. Clearly, this process not only gave us a high degree of control over the barrier thickness,11 but also resulted in a higher quality barrier. Below we discuss possible explanations for the magnitude and sign of the JMR observed. We have studied more than a dozen junctions of various AlOx thicknesses, and found that the JMR peaked at ␦ ⬃ 4.5 nm. Such a barrier thickness is too large for direct tunneling. Most likely, the conduction across the junction is via hopping through localized states in the barrier. This scenario is supported by an examination of the quality of such insulating barriers via superconducting tunneling using CrO2 / I / Pb junctions. In junctions where the insulating layer 共I兲 was formed by Br etching only, we observed the Pb superconducting DOS and phonon structures, which indicate elastic tunneling across the barrier. In contrast, in CrO2 / I / Pb junctions where I includes the oxidized Al, there was a significant inelastic broadening of the conductance spectrum, consistent with hopping transport through the barrier. Although an inelastic process, some types of hopping conduction have been shown to be capable of preserving spin polarization and sustaining large MR in magnetic junctions.4 In our junctions, there probably exist both spin conserving and spin-flipping hopping channels: The fact that we do observe a significant JMR in our samples implies that there must be a spin-polarized conduction channel, possibly due to hopping through nonmagnetic defects; on the other hand, the JMR is still much lower than what is expected theoretically, indicating the presence of localized states that are magnetic in nature and depolarizing the spins. These states may have resulted from Cr (oxide) impurities in the AlOx, which we will discuss in detail. In general, larger MR was observed for samples grown on TiO2 (100) than for those on (110). This is likely due to the difference in coercivity of the CrO2 layer on the two substrates. The (110) CrO2 films have a Hc ⬃ 120 Oe, which is closer to the Hc ⬃ 90 Oe of the Co than the (100) CrO2 共Hc ⬃ 50 Oe兲. Thus the junctions grown on the (100) substrates allow the moments of the FM electrodes to be closer to a true antiparallel alignment and exhibit a higher MR. It is possible that even in junctions with (100) CrO2, a true antiparallel state was not realized and the JMR could be higher than what was observed. As for the sign of the JMR, although inverse MR has been observed in CrO2/native barrier/Co junctions,9,10 its observation in our junctions is quite surprising. The negative sign of the JMR implies that both majority and minority spinpolarized carriers contribute to the transport across the junction. Because of the thick barrier, we consider the two electrode/insulator interfaces independently. From the

220413-2



LARGE INVERSE MAGNETORESISTANCE OF CrO2 / Co…

FIG. 3. The bias dependence of (a) the junction conductance in parallel and antiparallel configurations, and (b) the magnitude of the JMR for a CrO2 / CrOx-AlOx / Co at T = 5 K.

Meservey-Tedrow-type spin-polarization measurements on CrO2 / CrOx-AlOx / Al junctions,11 we know that CrO2 has P ⬃ + 100% (majority spin polarized) in such a structure. Therefore, one needs to assume that the minority spins come from states in the Co layer or at the Co-insulator interface. Although a negative P in Co would be in agreement with band-structure calculations, which indicate that Co has minority spin-polarized d states,15 Meservey and Tedrow12 consistently observed majority spin polarization for Co in their Al/ AlOx / Co junctions. More recently, several groups16,17 have convincingly demonstrated the importance of the interface in determining the sign of the JMR in MTJ’s. In particular, De Teresa et al.17 studied a series of La0.7Sr0.3MnO3 / I / Co junctions with different insulating barriers. An inverse (negative) JMR was observed when SrTiO3 or Ce0.69La0.31O1.845 was used as the insulating barrier. However, when SrTiO3 / Al2O3 or Al2O3 was used as the barrier, a normal (positive) JMR was seen. The authors attributed this to different interfacial bonding between the barrier and the Co electrode, producing different spin polarization for Co at the interface. A negative MR has also been observed in spin valves18 with LSMO and Co electrodes separated by an organic semiconductor 共Alq3兲. The inverse MR in these structures is again related to minority polarized states in the Co electrode. There is no clear understanding as to why the Co minority d states are favored over the majority s states in this case. However, this demonstrates that the inverse MR is not an effect limited to tunnel junctions. Now, at first glance, it would appear that the negative MR in our CrO2 / CrOx -AlOx / Co junctions is inconsistent with the results from DeTeresa et al.17 in junctions with an AlOx / Co interface where a positive MR was observed. We believe this discrepancy may be due to defect states in the barrier resulting from Cr inclusions in the AlOx layer, possibly formed during the oxidation of the Al layer. This conjecture is supported by several experimental observations. There exist remarkable similarities in the bias dependence of the junction conductance between our junctions and the Al/ Al2O3-Cr2O3 / Pb junctions studied by Kirtley et al.19 An asymmetric, almost linear background conductance over a large bias range was observed in all of our junctions. An example is shown in Fig. 3(a). Such a bias dependence is distinctively different from the parabolic conductance characteristic of a typical tunnel junction. It has been shown that a linear conductance can be created in a conventional Al/ I / Pb junction when Cr was intentionally doped into the Al oxide barrier.20 The authors attributed the appearance of

the linear background to inelastic tunneling through a broad continuum of localized states in the barrier created by the Cr inclusion. The presence of the linear conductance in our junctions led us to believe that there existed similar Cr inclusions in the Al oxide barriers. These localized states may be magnetic in nature and depolarizing the spins. This could be the origin for the JMR with lower than theoretically expected magnitude and inverse sign. The exact mechanism of the sign inversion due to Cr inclusion in the barrier is not clear. A study of MTJs with conventional ferromagnetic metal electrodes and an AlOx barrier with various degrees of Cr doping may provide convincing evidence for the above scenario and help elucidate the mechanism. The presence of spin-depolarizing hopping sites in the barrier could also explain the fast diminishing JMR with increasing bias, as shown in Fig. 3(b). This bias dependence is much stronger than that of transition-metal junctions with clean Al2O3 barriers.21 The diffusion of Cr into the AlOx most likely happened after the Al was deposited and during its oxidation in oxygen, since the CrO2 / I / Al junctions (which were absent of this oxidation step) exhibited nearly 100% spin-polarized transport across the barrier.11 We have also examined the temperature dependence of the junction resistance and MR. A set of typical results are plotted in Figs. 4(a) and 4(b), respectively. Unlike most transition-metal MTJ’s whose JMR generally have weak temperature dependence, our junctions exhibited a rapid decay of the JMR as a function of temperature, for this particular sample from −24% at 5 K to less than −1% at room temperature. In the same temperature range the junction resistance increases by more than a factor of 6 from 300 to 5 K. The strong increase of the junction conductance with increasing temperature is consistent with increasing thermally assisted hopping conduction through the barrier. The presence of a spin-independent hopping channel would lead to a rapid decay of the MR with increasing temperature, as suggested by Shang et al.22 Another possible reason for the rapid decline of the JMR would be a decrease in the spin polarization of one or both of the electrodes. While surface spin-wave excitations do cause a small variation of the spin polarization much below the Curie temperature, the effect is far from sufficient to cause such a large change in JMR. Therefore, in this scenario, the temperature dependence of the JMR would imply a decrease of the spin polarization of CrO2 with increasing T. The transport measurements on CrO2 thin films reveal an activated behavior for the resistivity at low temperatures,23 which becomes a T2 dependence

220413-3



PARKER et al.

FIG. 4. Temperature dependence of (a) the magnetoresistance and (b) the junction resistance of a CrO2 / CrOx-AlOx / Co Junction.

above a characteristic temperature, ⌬ ⬃ 80 K. The appearance of the power-law dependence is indicative of the onset of electron-magnon scattering. Furthermore, an anomalous Hall effect and a negative MR emerge above ⌬ indicating the appearance of magnetic scattering. Although these observations are suggestive of a possible loss of spin polarization above ⌬ ⬃ 80 K for CrO2, by the same token they also imply that the P below this temperature should be relatively constant. In contrast, the JMR of our junctions showed a strong T dependence all the way down to 5 K. This is probable evidence that, at least at low temperatures, the T dependence of the JMR in our junctions results primarily from the barrier rather than the decline of P of the electrodes. In summary, we have fabricated magnetic junctions with epitaxial CrO2 and polycrystalline Co electrodes separated by an artificial insulating barrier. A low-field inverse MR was observed with a maximum value of −24% at 5 K, a signifi-

1 J.

S. Moodera, L. R. Kinder, T. M. Wong, and R. Meservey, Phys. Rev. Lett. 74, 3273 (1995). 2 S. Tehrani, J. M. Slaughter, M. Deherrera, B. N. Engel, N. D. Rizzo, J. Salter, M. Durlam, R. W. Dave, J. Janesky, and B. Butcher et al., Proc. IEEE 91, 703 (2003). 3 M. Jullière, Phys. Lett. 54A, 225 (1975). 4 G. Hu and Y. Suzuki, Phys. Rev. Lett. 89, 276601 (2002). 5 S. Tehrani, B. Engel, J. M. Slaughter, E. Chen, M. DeHerrera, M. Durlam, P. Naji, R. Whig, J. Janesky, and J. Calder, IEEE Trans. Magn. 36, 2752 (2000). 6 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, and D. W. Abraham, et al., J. Appl. Phys. 85, 5828 (1999). 7 R. J. Soulen, J. M. Byers, M. S. Osofsky, B. Nadgorny, T. Ambrose, S. F. Cheng, P. R. Broussard, C. T. Tanaka, J. Nowak, and J. S. Moodera et al., Science 282, 85 (1998). 8 Y. Ji, G. J. Strijkers, F. Y. Yang, C. L. Chien, J. M. Byers, A. Anguelouch, G. Xiao, and A. Gupta, Phys. Rev. Lett. 86, 5585 (2001). 9 A. Barry, J. M. D. Coey, and M. Viret, J. Phys.: Condens. Matter 12, L173 (2000). 10 A. Gupta, X. W. Li, and G. Xiao, Appl. Phys. Lett. 78, 1894 (2001). 11 J. S. Parker, S. M. Watts, P. G. Ivanov, and P. Xiong, Phys. Rev.

cant improvement over similar junctions with a native oxide barrier. However, this value is still much below the theoretical expectation. We attribute the magnitude, the sign, and the strong bias and temperature dependence of the JMR to the inclusion of Cr in the barriers. Even higher JMR values should be possible with further improvement over the barriers. The authors would like to acknowledge S. von Molnár and P. Schlottmann for useful discussions and S. Watts for technical assistance. This work was supported by (DARPA) Defence Advanced Research Project Agency through the ONR under Contract Nos. N00014-99-1-1094 and MDA972-02-1-002. The TEM imaging presented in this paper was carried out using the microscope facilities at NHMFL, which is supported by the National Science Foundation under cooperative agreement DMR-0084173, and the State of Florida.

Lett. 88, 196601 (2002). Meservey and P. M. Tedrow, Phys. Rep. 283, 173 (1994). 13 P. G. Ivanov, S. M. Watts, and D. M. Lind, J. Appl. Phys. 89, 1035 (2001). 14 P. A. Stampe, R. J. Kennedy, S. M. Watts, and S. von Molnár, J. Appl. Phys. 89, 7696 (2001). 15 V. L. Moruzzi, J. F. Janek, and A. R. Williams, Calculated Electronic Properties of Metals (Pergamon, New York, 1978). 16 M. Sharma, S. X. Wang, and J. H. Nickel, Phys. Rev. Lett. 82, 616 (1999). 17 J. M. DeTeresa, A. Barthélémy, A. Fert, J. P. Contour, R. Lyonnet, F. Montaigne, P. Seneor, and A. Vaurès, Phys. Rev. Lett. 82, 4288 (1999). 18 Z. H. Xiong, D. Wu, Z. V. Vardeny, and J. Shi, Nature (London) 427, 821 (2004). 19 J. R. Kirtley, S. Washburn, and D. J. Scalapino, Sov. J. Low Temp. Phys. 18, 384 (1992). 20 J. R. Kirtley, S. Washburn, and D. J. Scalapino, Phys. Rev. B 45, 336 (1992). 21 A. C. Marley and S. S. P. Parkin, J. Appl. Phys. 81, 5526 (1997). 22 C. H. Shang, J. Nowak, R. Jansen, and J. S. Moodera, Phys. Rev. B 58, R2917 (1998). 23 S. M. Watts, S. Wirth, S. von Molnár, A. Barry, and J. M. D. Coey, Phys. Rev. B 61, 9621 (2000). 12 R.

220413-4