Journal of Magnetism and Magnetic Materials 240 (2002) 251­253 Asymmetric magnetization reversal on exchange biased CoO/Co bilayers F. Radua,b,*, M. Etzkorna, T. Schmittea, R. Siebrechta,c, A. Schreyera, K. Westerholta, H. Zabela a Institut f .ur Experimentalphysik, Festk.orperphysik, Ruhr-Universit.at Bochum, D 44780 Bochum, Germany b Departamentul de Fizica Experimentala, Institutul National de Fizica si Inginerie Nucleara, P.O. BOX MG-6, 76900, Magurele-Bucuresti, Romania c Institute Laue-Langevin, 38042, Grenoble cedex 9, France Abstract We study magnetic hysteresis loops after field cooling of a CoO/Co bilayer by MOKE and polarized neutron reflectivity. The neutron scattering reveals that the first magnetization reversal after field cooling is dominated by domain wall movement, whereas all subsequent reversals proceed essentially by rotation of the magnetization. In addition, off-specular diffuse scattering indicates that the first magnetization reversal induces an irreversible change of the domain state in the antiferromagnet. r 2002 Elsevier Science B.V. All rights reserved. Keywords: Exchange bias; Magnetization reversal; Neutron reflectivity Since the discovery of the exchange bias (EB) effect describe properly the asymmetry of the magnetization [1,2], mucheffort has been devoted to a basic under- reversal and the relaxation processes. standing of the exchange interaction across ferro- Here we present magnetization and neutron reflectiv- magnetic/antiferromagnetic (F/AF) interfaces [3,4]. ity studies of a Co/CoO bilayer grown by RF-sputtering Extensive data has been collected on the exchange bias methods. The sample is a Co (E200 (A) layer deposited field, HEB; and the coercivity fields, Hc; from a large on a Ti(2000 (A)/Cu(1000 (A)/Al2O3 template. The CoO number of bilayer systems [4]. The experimental results (30 (A) layer is formed on top of the Co layer by reflect the following characteristics: (1) HEB and Hc oxidation in air. The sample was characterized by X-ray increase as the system is cooled in an applied magnetic diffraction at the HASYLAB, by AFM, MOKE, and by field below the blocking temperature of the AF layer; (2) polarized neutron reflectometry (PNR). The sample is the magnetization reversal might be different for the polycrystalline witha strong (1 1 1) texture growthalong ascending and descending part of the hysteresis loop, as the growth direction. The surface roughness measured was first pointed out in Ref. [5]; (3) time relaxation by AFM is about 3 (A, which has been confirmed by X- effects of HEB and Hc indicates that a stable magnetic ray reflectivity measurements. state is reached only at very low temperatures. Several Fig. 1a shows the magnetic hysteresis loop measured theoretical models have been proposed for describing by MOKE at T ¼ 50 K, after cooling in an applied field possible mechanisms of the EB effect [6­11]. So far none of +2000 Oe. Upon descending the field for the first of them is able to explain satisfactorily all macroscopic time to negative values, an abrupt magnetization characteristics of EB systems. One of the problems is to reversal is observed at the coercivity field Hc1: Ascending again to positive field values, the magnetization curve at H *Corresponding author. Fax: +49-234-3214-173. c2 is more rounded. In subsequent cycles the magne- E-mail address: radu@spin.ep4.ruhr-uni-bochum.de tization curves at Hc1 and Hc2 are of about the same (F. Radu). shape characterized by HEB=30 Oe and DHc ¼ 200 Oe. 0304-8853/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 0 1 ) 0 0 8 1 5 - 0 252 F. Radu et al. / Journal of Magnetism and Magnetic Materials 240 (2002) 251­253 0.50 0.48 [a. u.] 0.46 Hc1 Training Curve 0.44 0.42 err Rotation 0.40 (a) K 0.38 Hc2 -1200 -900 -600 -300 0 300 600 900 1200 Magnetic Field [Oe] 90 90 80 80 70 I++ 70 I++ I-- I-- 60 60 I+- I+- 50 I-+ 50 I-+ 40 (c) 40 30 30 20 (b) 20 10 10 Spin-dependent Intensities [cts/sec] -1000 -900 -800 -700 -600 -500 Spin-dependent Intensities [cts/sec] -1500 -1000 -500 0 500 1000 1500 H [Oe] H [Oe] 10 15 8 FWHM = 9 Oe 14 ] 6 13 12 (e) 4 11 R +- [cts/sec] I+- [cts/sec 10 2 (d) 9 0 8 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 -1100 -1000 -900 -800 -700 -600 [deg] H [Oe] Fig. 1. (a) MOKE hysteresis loop of a CoO/Co bilayer after field cooling to 50 K in an external field of 2000 Oe. The black dots denote the first hysteresis loop, the open circles the second loop. Any further loops are not significantly different from the second. (b) and (c) Neutron hysteresis loops from the same sample but at 10 K. I þ þ; I ; I þ and I þ are non-spin flip and spin-flip intensities as a function of external magnetic field. They are measured at special scattering vector values of the reflectivity curves (see text); (d) Off- specular diffuse spin-flip scattering taken at Hc1 ¼ 750 Oe (full dots) and in saturation at 1400 Oe (open dots). (e) Specular spin-flip at Hc1 enlarged from panel (b) for better recognition. From these measurements, we conclude that the first at the resonance peak near the critical edge. While the magnetization reversal of the virgin sample after field magnetic field is swept like in a conventional MOKE set cooling is conspicuously different from any subsequent up, the sensitivity to any neutron spin-flip processes `trained' reversals. This difference also becomes obvious from spin-canting, magnetic roughness, domain walls, by a strong thermal relaxation process, which is or rotation is enhanced (about 10 to 40 times) due to the observable only in the virgin state of the sample at sample design as a neutron resonator, in analogy to a constant field value HoHc1; i.e. just before the sharp Fabry Perot interferometer [13,14]. magnetization reversal [12]. In Fig. 1b and c the non-flip intensities I þ þ and MOKE is a fast method for determining hysteresis I are plotted as a function of external field. loops but cannot reveal the spin configuration at the The normalized intensity difference DI=I ¼ ððI þ þÞ interface. Therefore we measured, in addition, neutron ðI ÞÞ=ððI þ þÞ þ ðI ÞÞ is proportional to the mag- hysteresis loops (NHL) of the same sample at T ¼ 10 K netization component parallel to the neutron polariza- using the ADAM reflectometer at the ILL. Fig. 1b and c tion axis and proportional to the magnetization curve show corresponding neutron results. The NHL method as determined, for instance, by MOKE. The points will be detailed in a forthcoming paper [12]. Briefly, the where the two curves I þ þ and I intersect (and, non-spin-flip intensities (I þ þ and I ) are measured more precisely, where the spin-flip intensities I þ and at the wave vector transfer Q corresponding to the I þ reachmaximum) are defined as the coercive fields inflection point of the non-polarized neutron reflectivity Hc1 and Hc2: The sharp intensity change at Hc1 and the (near the critical edge for total external reflection), while more rounded change at Hc2 reflects the corresponding the spin-flip intensities (I þ and I þ) are measured parts of the hysteresis loops in the MOKE measure- F. Radu et al. / Journal of Magnetism and Magnetic Materials 240 (2002) 251­253 253 ments. Note that the neutron data are taken at 10 K as would be spin misalignment at the interface or magnetic compared to 50 K for the MOKE measurements, which roughness [17]. At the present stage, we cannot explains the different HEB and Hcin panels (a) and (b-e) distinguish whether the off-specular signal is a char- of Fig. 1. acteristic of the interface only or of the interface plus the The different shapes of the I þ þ and I intensities whole antiferromagnet layer. However, the question at Hc1 and Hc2 are also reflected in the different spin-flip may be answered by measuring samples withincreasing intensities. We first discuss the specular spin-flip antiferromagnetic layer thickness. An increase of the off- intensities shown by triangles in panels (b) and (c). specular signal will then verify the formation of The magnetization reversal at Hc2 exhibits strong spin- antiferromagnetic domains by the increase of the flip intensities I+ and I +. This is always observed domain wall lengths. for round or `trained' hysteresis loops and is character- In conclusion, we have shown that polarized neutron istic for a magnetization reversal via (domain) rotation. scattering results give deep insight into the origin of the Magnetization reversal by rotation provides a large striking difference between the first magnetization magnetization component perpendicular to the polar- reversal at Hc1 and all subsequent reversal characteristic ization axis, giving rise to neutron spin-flip. Vice versa, for CoO/Co bilayers withvery thin CoO layer. the rather low spin-flip intensities (I þ and I þ), The results suggest that the field cooling forces the which are observed during the first magnetization thin AF-layer into a single domain state with the reversal of the virgin sample at Hc1 are indicative of a sublattice magnetization direction essentially parallel domain wall movement. The step like intensity change at (or antiparalel) to the Co magnetization direction. This Hc1 ¼ 750 Oe from high to low, which is enlarged in metastable original state characterized by very large panel (e) for better recognition, is followed by a steady exchange bias field HE is destroyed upon the first decrease as the system approaches saturation. From this magnetization reversal and transformed into a stable figure, it is quite obvious that the enhanced spin-flip multidomain state witha muchlower HE: scattering takes place in a very narrow field range of not more than 15 Oe [FWHM=9 Oe]. This could be easily This work was supported by the Sonderforschungsber- attributed to the domain walls in the ferromagnet during eich 491 of the Deutsche Forschungsgemeinschaft. The the magnetization reversal. neutron reflectivity measurements were carried out at The striking differences noticed for the specular spin- the CRG-ADAM reflectometer of the ILL, which is flip scattering are also expressed in the off-specular supported by the BMBF under grant 03ZAE8BO. diffuse scattering. While the diffuse spin-flip intensity at Hc2 (not shown here) is rather low in intensity and symmetrically centered around the specular peak, this is References not the case at Hc1; as seen in panel (d). The off-specular diffuse intensities (corrected for efficiency and footprint [1] W. Meiklejohn, C.P. Bean, Phys. Rev. 102 (1956) 1413. contributions) taken at Hc1 ¼ 750 Oe and in saturation [2] W. Meiklejohn, C.P. Bean, Phys. Rev. 105 (1957) 904. at H ¼ 1400 Oe are rather strong. Therefore, we infer [3] A.E. Berkowitz, K. Takano, JMM 200 (1999). that the dominant part of the off-specular spin-flip [4] J. Nogues, I.K. Schuller, JMM 192 (1999). scattering is due to the antiferromagnetic domains since [5] M.R. Fitzsimmons, P. Yashar, C. Leighton, I.K. Schuller, the ferromagnetic Co layer is already in saturation. J. Nogues, C.F. Majkrzak, J.A. Dura, Phys. Rev. Lett. 84 The specular spin-flip intensity, both near H (2000) 3986. c1 and in saturation, indicates the presence of domain walls at the [6] E. Fulcomer, S.H. Charap, J. Appl. Phys. 43 (1972) 4190. interface [9]. The off-specular spin-flip signal, as was [7] A.P. Malozemoff, Phys. Rev. B 35 (1987) 3679. [8] N.C. Koon, Phys. Rev. Lett. 78 (1997) 4865. pointed out by experimental and theoretical studies [9] M.D. Stiles, R.D. McMichael, Phys. Rev. B 59 (1999) 3722. [15,16], arises from magnetic domains smaller than the [10] P. Milt!enyi, M. Gierlings, J. Keller, B. Beschoten, lateral coherence length (micron size range) of the G. G.untherodt, U. Nowak, K.D. Usadel, Phys. Rev. Lett. neutron beam. A Zeeman splitting in the external field 84 (2000) 4224. takes place as well, but is usually not strong enoughto [11] R.L. Stamps, J. Phys. D 33 (2000). explain the off-specular intensity [14]. To strengthen this [12] F. Radu, M. Etzkorn, R. Siebrecht, V. Leiner, A. Schreyer, conclusion, we have recorded the I þ rocking curve at K. Westerholt, H. Zabel, in preparation. H ¼ 200 Oe, after first magnetization reversal, and not [13] F. Radu, V.K. Ignatovich, Physica B 267 (1999). detected any significant shift of the specular and off- [14] V.L. Aksenov, Yu.V. Nikitenko, F. Radu, Yu.M. Glede- specular pattern. Thus, our data suggests that during the nov, P.V. Sedyshev, Physica B 276­278 (2000) 916. [15] S.G.E. te Velthuis, A. Berger, G.P. Felcher, B.K. Hill, first field reversal at Hc1 the CoO layer brakes into E. Dan Dahlberg, J. Appl. Phys. 87 (2000) 5046. antiferromagnetic domains. The weak spin-flip signal is [16] B.P. Toperverg, Physica B 297 (2001) 160. then due to AF domain walls and uncompensated spins. [17] H. Zabel, R. Siebrecht, A. Schreyer, Physica B 276­278 Another interpretation for the off-specular reflectivity (2000) 17­21.