Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 Preparation and magnetic properties of the CoO/Co bilayer B. Raquet , R. Mamy *, J.C. Ousset , N. Ne gre , M. Goiran , C. Guerret-PieŽcourt Laboratoire de Physique de la Matie%re Condense&e de Toulouse, INSA. SNCMP, Universite Paul sabatier, Complexe Scientifique de Rangueil, 31077 Toulouse Cedex 4, France LAAS-CNRS, 7 av. du Colonel Roche, 31077 Toulouse Cedex, France Received 28 July 1997; received in revised form 7 November 1997 Abstract The results of the preparation, characterization and magnetic properties of a CoO/Co interface are discussed. This interface was realized by ultraviolet irradiation of a 4 nm Co layer deposited with a gold buffer layer on a GaAs(1 1 1) substrate. The Co oxide analyzed by XPS and AFM revealed a continuous 1.7 nm thick layer with a rms roughness of 0.6 nm and mostly CoO at the interface. Longitudinal magneto-optical Kerr loops evidence a strong initial in-plane anisotropy which disappears on oxidation. Values of the unidirectional exchange coupling at the Co/CoO interface were obtained: 0.1 erg/cm from SQUID measurement at 5 K after field cooling and 0.6 erg/cm with magneto-optical Kerr effect in polar configuration from modelization of the M(H) curve. This discrepancy is discussed in terms of the difference of sensing the interfacial coupling in the two cases due to atomic steps. 1998 Elsevier Science B.V. All rights reserved. PACS: 75-60; 75-70; 81-15G Keywords: Exchange coupling; Anisotropy - uniaxial; Anisotropy - unidirectional; Thin films 1. Introduction tion [5]. The best interface arises from (1 1 1) ori- entation with alternating ferromagnetic sheets of We present a study of the preparation and char- CoO [6]. However, actual insulating layers suffer acterization of the antiferro/ferromagnetic CoO/Co from lack of homogeneity, roughness and crystal- interface. The knowledge of the insulating-mag- linity. To get a better quality of the oxide layer, we netic nature of an ultra-thin CoO layer as well as propose an ultra-violet (UV) oxidation method the magnetic coupling (exchange anisotropy) may (UVOX) that seems to be in our case quite satisfac- be very useful to realize spin-dependent tunneling tory. Co was deposited on a gold buffer layer on devices [1,2] and spin-valve structures with pinned a GaAs(1 1 1) substrate. The CoO/Co interface was layers [3,4]. Their use in magnetoresistive random realized by air UVOX. The chemical analysis con- access memories (MRAM) is also called for ques- firms the CoO nature of the interfacial oxide. An obvious proof of the antiferromagnetic nature of * Corresponding author. Tel.: #33 5 61 55 61 85; e-mail: the oxide is given by the shift of the magnetic loops mamy@insa-tlse.fr. which appears below the NeŽel temperature of the 0304-8853/98/$19.00 1998 Elsevier Science B.V. All rights reserved. PII S 0 3 0 4 - 8 8 5 3 ( 9 7 ) 0 1 1 0 5 - 0 42 B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 oxide layer and gives by the way an evaluation of Fig. 1a-Fig. 1c show three typical M-H curves the interfacial exchange anisotropy. Finally, this illustrating the magnetic anisotropy in the plane of exchange energy is also extracted from the magnet- the layer. The angle ization measurement of the Co/CoO bilayer along & defines the direction of the magnetic field with respect to the [1 1 0] crystallo- the normal direction of the film. graphic directional axis of the GaAs(1 1 1). The MOKE intensity is normalized to the Kerr rotation obtained in the saturated state. We first observe 2. Experiment unambiguously a strong unexpected in-plane uni- axial anisotropy with the easy-axis along a well- The GaAs surface was obtained by homoepitaxy defined crystallographic direction [1 1 0] of the from a nominal (1 1 1)GaAs wafer in a MBE appar- GaAs(1 1 1) substrate. The easy-axis loop (Fig. 1a) atus [7] and furnished covered with a passivation presents an approximate square-loop revealing a As cap in the metal deposition chamber where it small angular distribution of the easy direction of was reacted by heating so as to obtain a (2;2) around $15°. However, the lack of hysteresis for surface reconstruction. A 20 nm gold and 4 nm Co an applied magnetic field perpendicular to the layers were sequentially deposited at room temper- easy-axis suggests the high film quality and the ature by thermal evaporation [8]. We chose to uniaxial nature of the in-plane anisotropy, which oxidize the Co layer by UVOX (ozone generation) dominates over higher-order anisotropies. which has already proven efficient in semiconduc- The in-plane uniaxial anisotropy constant K tor technology [9]. The chemical analysis was per-  is easily estimated using a coherent rotation model formed using electron spectroscopy for chemical [11-13] in a restricted range of analysis (ESCA) with a commercial apparatus & and H values. We reasonably assume that the magnetization from vacuum generators (VG) using non-mono- behavior is dominated by a reversible coherent chromatized Mg K radiation. The atomic force rotation for a decreasing magnetic field from the microscopy (AFM) images were obtained from saturation field H a Nanoscope III apparatus from Digital Instru-  to 0 with the angle & between the field and the easy-axis inferior to 40°. For ments Inc. The magnetic hysteresis loops were the Co (4 nm) layer discussed in this paper, we find measured at room temperature with a magneto- that it is sufficient to include only the uniaxial optical Kerr effect magnetometer and at low tem- anisotropy term K perature (5 K), with a SQUID magnetometer.  and the Zeeman energy to express the total magnetic energy Another magneto-optical Kerr effect apparatus [10] with a pulsed high magnetic field was used in E"K sin( )!HM cos( ! &), (1) polar configuration down to 5 K. where and & are the angles defining, respectively, the orientation of the magnetization and the applied field with the easy axis. The shape aniso- 3. Uniaxial in-plane anisotropy tropy energy is large enough to favor in-plane magnetization, we therefore assume the reversal We report the study of the in-plane magnetic process confined to the film plane. Besides, we switching behavior of the Co (4 nm) film performed neglect any magneto-crystalline contribution in the by longitudinal magneto-optical Kerr effect. The (0 0 0 1) HCP plane of the cobalt compared to the magnetization measurements were realized ex situ strength of the observed uniaxial anisotropy. and at room temperature just after the preparation. The magnetization curves are obtained by min- Our follow-up of the time-dependent oxidation of imizing Eq. (1) with respect to and for a fixed the Co layer in air ensures the negligible alteration orientation of the applied field. Fig. 1d shows two of the uncovered Co layer two days after the expo- examples of the calculated magnetization as solid sure in air. The lack of oxidation at the beginning of lines in good agreement with the experimental the exposure in air reveals the high crystallographic measurements corresponding to quality of the uncovered magnetic layer. &"24.5 and 37°. The only one parameter K is adjusted in order to B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 43 Fig. 1. (a)-(d). Longitudinal magneto-optical Kerr loops for different values of the angle & between the direction of the magnetic field in the plane of the (1 1 1) surface and the [1 1 0] crystallographic direction of the GaAs substrate. The fitting with a coherent rotation model is shown in solid lines in (a) and (d). obtain the best fit. Whatever the value of &, we the persistence of a stepped topography along the deduce the same value of K"3.6;10 erg/cm. same axis. We therefore attribute the strong It is noticeable that such a well-defined aniso- uniaxial in-plane anisotropy of the Co layer to tropy is absolutely neither owing to a magneto- a shape memory of the substrate, which disturbs crystalline direction, contrary to experimental the Au/Co interface on an atomic scale. Previous results observed in Co(BCC)/GaAs(0 0 1) [14] or studies dealing with the origin of such an unex- Fe/Ag(100) [15] and not induced by a step bunch- pected uniaxial in-plane anisotropy provide insight ing phenomena on a vicinal surface. The uniaxial into the contribution of atomic steps of the substra- anisotropy constant deduced from our measure- te to the magnetic anisotropy [15-17]. A lateral ments is at least 40% greater than the in-plane pseudo-periodic variation of the interfacial topo- uniaxial anisotropy estimated in other Co thin graphy may induce, among other things, a shape layers deposited on nominal surfaces [14]. anisotropy of the Co islands within the film or an Before deposition of Au(1 1 1) at room temper- oriented residual strain in the magnetic layer. ature, the GaAs(1 1 1) surface is analyzed by AFM. A microstructural study by transmission electron Some regular atomic steps at the surface are ob- microscopy is in progress to clarify the effects of the served along the [1 1 0] crystallographic direction atomic steps on the growth of the Co layer. of the substrate (Fig. 2). By scanning the Au(20 nm) Let us remark the large discrepancy between the deposited on GaAs(111), the AFM image reveals theoretical magnetization curve (Fig. 1a, solid line) 44 B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 Fig. 2. AFM topography (10 m;10 m) of the GaAs(1 1 1) surface showing the atomic steps along the [0!1 1] direction. and the experimental loop when the field is applied along the easy axis ( &+0°). The switching behav- ior is most likely to deviate from the coherent rotation model. A dynamical study of the magneti- zation and relaxation experiments along the [1 1 0] substrate direction presented evidence that the Fig. 3. (a) Evolution of the longitudinal magneto-optical Kerr loops performed at room temperature for the 4 nm Co layer magnetization reversal is largely dominated by the as a function of the number of days exposed to air. The domain-wall motion process [18]. attenuation of the Kerr intensity is shown in the inset. (b) Comparison of the effect of a 15 min UV exposure with respect to 120 d in air on longitudinal magneto-optical 4. Characteristics of the oxidization process Kerr loops. We followed the evolution of the oxidization The slow air oxidation process can be brought process of a sample left in air over weeks as com- together with other studies of kinetics of ultra-thin pared to another sample submitted to an UV ir- Co layers [19] which give a thickness-dependent radiation during 15 min. In the air oxidation case, process: very rapid oxidation (however limited the Kerr intensity and the uniaxial anisotropy regu- to 2.5 nm of oxide) for thickness '5 nm and larly decrease while the hysteresis loops broaden very long time constants for lower thickness. As (Fig. 3a and inset), in the UV oxidation case the the Kerr measurements plotted in Fig. 3 are per- uniaxial anisotropy completely disappears and the formed at room temperature, the Kerr intensity is hysteresis loops broaden even more (Fig. 3b). The mainly proportional to the Co thickness. The solid AFM image of the UV oxidized surface (Fig. 4) line in inset of Fig. 3a represents an exponential shows a continuous oxide layer with a root mean fit of the air oxidation process as a function of square (rms) roughness of 0.6 nm while oxidation in time. The decay rate of oxidation is estimated air gives oxide patches. So UV oxidation turns out to 22 d which is in good agreement with the pre- as a more rapid and suitable method. vious works [19]. B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 45 Fig. 4. AFM topography (1 m;1 m) of the CoO surface after UV irradiation (15 min), the measured RMS roughness is 0.6 nm. 5. Analysis of the oxide layer ESCA is a powerful technique to determine the Fig. 5. XPS spectra of (a) the Co oxidation number and the thickness of the oxide  and (b) the O core levels. from the analysis of the core state shifts (namely Co and O here). Moreover, the different spin surface region [20]. In fact the angular variation of configuration of Co'' (para and high-spin config- the O uration) and Co''' (dia and low-spin configuration)  peak shows that the intensity of the peak I decreases towards grazing angle and consequently ions permits the identification of CoO which pres- arises from underneath the oxide layer, i.e. near the ents a strong satellite shake-up structure on the Co interface. Finally, starting from a 4 nm Co thick- 2p states which is considerably weaker in the ness, the UV oxidization turns 1.7 nm into oxide CoO case [20,21]. We present in Fig. 5 the and the interfacial part is pure CoO. Co and Co region for the UV oxidized sample which shows the satellite structure (786 eV) of the Co peak (780.5 eV) and which identify 6. Unidirectional exchange anisotropy unambiguously CoO. Each Co2p peak has unresol- ved oxidized and non-oxidized component 2 eV The antiferromagnetic behavior of the oxide wide which gives after line-shape analysis following overlayer is borne out by measuring hysteresis classical methods [22] a 1.7 m oxide thickness. loops in the temperature range from 5 to 300 K, by Useful information can also be extracted from the SQUID magnetometer. O peak which exhibits two components at 529.5 Fig. 6a shows two hysteresis loops for the oxi- (I) and 531 eV (II). Although the interpretation of dized film performed at 5 K, to which the sample these components in the Co oxide case is disputed was cooled from 300 K in a demagnetized state and [23] it seems that the less intense component I, is below which a magnetic field of 50 kOe was applied due to CoO and the other to non-stoichiometric parallel to the film plane. 46 B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 cooled from 300 K in a positive magnetic field of 50 kOe and performed at two distinct temperatures 5 and 100 K. The centered hysteresis loop obtained at 100 K reveals that the NeŽel temperature of the CoO/Co layer is below this temperature. It is in agreement with our estimated CoO thickness by XPS and the known NeŽel temperature dependence on the CoO thickness [25]. The exchange energy per unit area of the inter- face J can be explicitly calculated as the field offset H has been shown to follow the relation [19] J H  " . (2) Mt! The thickness of the non-oxidized Co underlayer t! is roughly estimated considering the absolute value of the saturation magnetization measured by SQUID. We deduce approximately a 2 nm Co layer thickness. This value is consistent with the thickness of the oxide layer estimated by XPS. Inserting M&1.833 T, the Co bulk value of the saturated magnetization in Eq. (2), we obtain, J "0.12 erg/cm. This exchange constant is in the same order of magnitude of those defined in pre- vious works in CoO/Co bilayers [6,19,26,27]. Fig. 6. (a) Comparison of SQUID hysteresis loops for CoO/Co However, one usually notices a large scatter of cooled without and under an applied field. The field-cooled loop the experimental exchange coupling values as is shifted by an offset field of !420 Oe. (b) Field-cooled the strength of the exchange AF/F drastically SQUID loop taken above (100 K) and below (5 K) the NeŽel depends on the structural quality of the CoO/Co temperature of CoO. interface. Therefore, we propose another experiment to analyze the effect of the interface roughness on the Under the last condition, an unidirectional ex- exchange coupling Co/CoO per unit of area. In change anisotropy is manifested by a negative field a polar configuration, we measure at 5 K the per- offset H of !420 Oe, defined as pendicular magnetization of the Co layer in an high H pulsed magnetic field perpendicular to the layer.  "("H "!"H ")/2, The results given in Fig. 7 show the effect of where H and H are the coercive fields corre- the oxide layer on the magnetization behavior sponding, respectively, to the left and right sides of of the Co along the [0 0 0 1] axis. As the field the loop. No shift is observed on the hysteresis loop needed to rotate the CoO magnetic moments in when the sample is cooled in the absence of an the bulk is around 380 T, we consider that in our applied field. It is well known that the shifted mag- range of applied field, the CoO magnetic mo- netization is induced by the antiferromagnetic/fer- ments keep blocked in the plane of the layer. Thus, romagnetic exchange coupling at the CoO/Co assuming a coherent rotation process along the interface [24]. The high magnetic field of 50 kOe is hard axis, the total energy can be simply expressed applied during the cooling to obtain a single anti- as follows: ferromagnetic domain in the CoO. Fig. 6b com- pares the hysteresis loops for the oxidized film E"K sin( )!MH cos( )!J /t! sin( ), (3) B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 47 steps. Such a magnetic arrangement drastically de- creases the effect of the exchange coupling mea- sured in the longitudinal configuration of the applied magnetic field. 7. Conclusions The ultra-violet oxidation method that we pro- posed to realize CoO/Co interface proved to be better than oxidation in air as concerns the homo- geneity and roughness of the oxide layer. The differ- ence in the values of the unidirectional exchange Fig. 7. Magnetization curves performed at 5 K in polar config- coupling obtained for the in-plane and perpendicu- uration along the hard direction (normal to surface of the lar field directions in the hysteresis loops can be sample) before and after oxidation. The fit with a coherent accounted for by the fact that for in-plane field the rotation model is shown in solid lines. unidirectional exchange field is locally suppressed by alternation of the magnetization at monoatomic steps. This can be an indirect method to check the where defines the angle between the normal axis interfacial roughness or to test any method in- and the magnetization. tended to improve the interfacial roughness. The first term is the effective perpendicular an- isotropy of the Co layer, the second corresponds to the Zeeman energy and the last term accounts for Acknowledgements the exchange coupling energy at the CoO/Co inter- face. Before oxidation, the experimental magneti- We acknowledge the services of the Ecole Na- zation curve is easily fitted without considering tionale SupeŽrieure de Chimie and the Laboratoire the exchange coupling. We find K equal to de Chimie de Coordination de Toulouse for ESCA !5;10 erg/cm. After UV oxidation, the two and SQUID measurements. parameters deduced from the best fit, by minimi- zing Eq. (3) with respect to , are K "!2.64; 10 erg/cm and J "0.67 erg/cm. The effective References perpendicular anisotropy affects the slope of the Co magnetization in low magnetic field, whereas, the [1] J.S. Moodera, L.R. Kinder, J. Appl. Phys. 79 (1996) 4724. strength of the exchange coupling causes the diffi- [2] P. Leclair, J.S. Moodera, R. Meservey, J. Appl. Phys. 76 cult saturation of the cobalt. It is surprising to (1994) 6546. [3] M.J. Carey, A.E. Berkowitz, Appl. Phys. Lett. 60 (1992) remark that the exchange coupling deduced from 3060. the polar configuration is increased by a factor 6, [4] T. Lin, C. Tsang, R.E. Fontana, J.K. Howard, IEEE Trans. compared to the exchange energy measured in Magn. 31 (1995) 2585. a longitudinal configuration. In fact, the experi- [5] E.Y. Chen, S. Tehrani, T. Zhu, M. Durlam, H. Goronkin, J. ment in polar configuration allows to estimate the Appl. Phys. 81 (1971) 3992. [6] J. Bransky, I. Bransky, A.A. Hirsch, J. Appl. Phys. 41 real intensity of the coupling, independently of the (1970) 183. magnetic multidomain structure of the CoO in- [7] B.J. Garcia, C. Fontaine, A. Munoz-Yaguše, Appl. Phys. duced by the interface steps. A simple calculus Lett. 66 (1995) 610. shows that, despite the cooling under 50 kOe, 40% [8] R. Mamy, Surf. Sci. 322 (1995) 337. of the exchange surface between CoO/Co has got [9] E. Bedel, A. Munoz-Yaguše, C. Fontaine, C. Vieu, Mater. Sci. Eng. B 21 (1993) 157. adjacent CoO moments opposite to the applied [10] M. Goiran, N. Ne gre, S. Jaren, C. Meyer, R. Barbaste, A.R. positive magnetic field, due to the interfacial atomic Fert, J.C. Ousset, S. AskeŽnazy, Physica B, to be published. 48 B. Raquet et al. / Journal of Magnetism and Magnetic Materials 184 (1998) 41-48 [11] E.C. Stoner, E.P. Wohlfarth, Philos. Trans. A 240 (1948) [18] B. Raquet, R. Mamy, J.C. Ousset, unpublished. 599. [19] L. Smardz, U. Košbler, W. Zinn, J. Appl. Phys. 71 (1992) [12] B. Dieny, J.P. Gavignan, J.P. Rebouillat, J. Phys.: Con- 5199. dens. Matter 2 (1990) 159. [20] T.J. Chuang, C.R. Brundle, D.W. Rice, Surf. Sci. 59 (1976) [13] V. Grolier, Ph.D. Thesis, Orsay, 1994. 413. [14] J.A.C. Bland, M.J. Baird, H.T. Leung, A.J.R. Ives, K.D. [21] C.R. Brundle, T.J. Chuang, D.W. Rice, Surf. Sci. 60 (1976) Mackay, H.P. Hughes, J. Magn. Magn. Mater. 113 (1992) 286. 178. [22] T.A. Carlson, G.E. McGuire, J. Electron. Spectrosc. Relat. [15] M. Gester, C. Daboo, R.J. Hicken, S.J. Gray, A. Ercole, Phenom. 7 (1975) 175. J.A.C. Bland, J. Appl. Phys. 80 (1996) 347. [23] K. Wandelt, Surf. Sci. Rep. 2 (1982) 1. [16] M. Albrecht, T. Furubayashi, M. Przybylski, J. Korecki, [24] W.H. Meiklejohn, C.P. Bean, Phys. Rev. 102 (1956) 1413. U. Gradmann, J. Magn. Magn. Mater. 113 (1992) [25] T. Ambrose, C.L. Chien, Phys. Rev. Lett. 76 (1996) 1743. 207. [26] H. Danan, H. Gengnabel, J. Steinert, A. Linzen, J. de Phys. [17] J. Fassbender, C. Mathieu, B. Hillebrands, G. Gušntherodt, 32 (1971) C1. R. Jungblut, M.T. Johnson, J. Magn. Magn. Mater. 148 [27] M. Takahashi, A. Yanai, S. Tagushi, T. Suzuki, Jpn. J. (1995) 156. Appl. Phys. 19 (1980) 1093.