Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 Magnetization reversal asymmetry in Fe/MgO(0 0 1) thin "lms J.L. Costa-KraKmer *, J.L. MeneHndez , A. Cebollada , F. Briones , D. GarcmHa , A. Hernando Centro Nacional de Microelectro&nica-IMM (CSIC), Isaac Newton 8 (PTM), Tres Cantos, Madrid 28760, Spain Instituto de Magnetismo Aplicado, P.O. Box 155, 28230 Las Rozas, Madrid, Spain Received 20 May 1999; received in revised form 13 September 1999 Abstract We study the in-plane magnetization process in 200 As Fe(0 0 1) thin "lms grown by sputtering at normal incidence. In spite of this growth geometry, a small uniaxial in plane magnetic anisotropy, whose origin is not totally understood, is found superimposed to the expected cubic biaxial one. This has a dramatic e!ect both on the reversal process and the domain structure. A combined longitudinal and transversal Kerr study shows the di!erent switching processes (1803 walls along the main easy axis versus 903 along the secondary easy axis) depending on the relative orientation of the magnetic "eld with respect to the Fe crystallographic axes. Remarkably, this two- and sometimes three-step switching process appears only when the "eld is applied along certain crystallographic directions. These "ndings are corroborated by domain observations. 2000 Elsevier Science B.V. All rights reserved. PACS: 81.15.Cd; 75.70.Ak; 75.70.Kw Keywords: Thin "lms; Fe; Domains; Anisotropy; Sputtering 1. Introduction tering are the standard growth techniques, while for the magnetic characterization, magneto-optic Kerr The study of growth and magnetic properties of e!ect has proven to be a very accessible and versa- epitaxial Fe thin "lms on insulator and semicon- tile tool. All these "lms usually exhibit the expected ductor substrates has received much attention. in-plane four-fold magnetic anisotropy [5,6], due Since the pioneering work of Prinz and Krebs [1], to the cubic structure of Fe, but often an additional single-crystal Fe thin "lms have been grown on uniaxial anisotropy is found superimposed [7]. many di!erent semiconductor substrates (for This additional component of the anisotropy has example, GaAs, ZnSe or Ge) [2] and insulators been attributed to di!erent origins, mainly caused (such as MgO [3] or diamond [4]). Molecular by the geometry of the deposition system, through beam epitaxy (MBE) and magnetron or RF sput- the lattice distortion induced by the angle of inci- dence of the deposited Fe atoms (therefore of mag- neto-elastic origin) [8}11], but also to the intrinsic * Corresponding author. Tel.: 34-91-8060700; fax: 34-91- anisotropy of the dangling bonds present in the 8060701. growth on GaAs(0 0 1) [12]. On the other hand, E-mail address: kramer@imm.cnm.csic.es (J.L. Costa-KraKmer) the magnetization reversal process and magnetic 0304-8853/00/$ - see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 8 8 5 3 ( 9 9 ) 0 0 6 9 8 - 8 342 J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 domain structure has also been studied in similar to the one described by Postava et al. [21]. Fe/Ag(0 0 1) [13] or Fe/GaAs(0 0 1) [14,15] as- We use either white light or a diode laser linearly grown "lms, as well as on patterned Fe/GaAs(0 0 1) polarized and focused on the sample, and a single [16,17] and Fe/MgO(0 0 1) structures [18]. photodiode in front of which a linear polarizer is Here we present a study on the in-plane anisot- either placed or removed in case of performing ropy and magnetization reversal process for longitudinal or transversal Kerr measurements, re- Fe/MgO(0 0 1) thin "lms grown by triode sputter- spectively. The magnetic "eld was swept at a fre- ing at normal incidence. In spite of the non oblique quency of 2 Hz using Helmholtz coils. The growth con"guration, a small uniaxial anisotropy photodiode data shown corresponds to a single- is found superimposed to the expected four-fold "eld cycle without any electronic "ltering or aver- cubic anisotropy. A combination of longitudinal aging. The sample was mounted in a goniometer and transversal Kerr investigation and magnetic with 13 precision. The "eld direction was "xed. domain observations allows us to discern the do- Magnetic domain imaging was performed using main structure and magnetization reversal pro- a home made low-"eld longitudinal Kerr setup cesses in these Fe single crystal "lms. including also a CCD camera, a PC video card and home-made software. 2. Experiment 3. Growth and structure The deposition of the "lms was performed in a triode-sputtering system with a base pressure in The growth of crystalline Fe was checked in situ the low 10\ mbar range. The substrate was "rst with RHEED and ex situ with XRD. Fig. 1(a) chemically cleaned in successive ultrasonic baths of shows RHEED patterns for Fe "lm along [1 0 0] tricloroethane, acetone, ethanol and distilled water. and [1 1 0] directions. Similar information is ex- After loading into the growth chamber, it was out- tracted by recording phi scans of both Fe(1 1 0) and gassed overnight at 1003C and then annealed at MgO(2 2 0) asymmetric peaks (Fig. 1(b)). Besides, 7003C for 1 h. Nominally, 200 As of Fe were depo- the expected 453 in-plane rotation of the Fe lattice sited at 503C with an Ar pressure of 4;10\ mbar, with respect to the MgO one is con"rmed. Fe yielding growth rates of about 0.2 As/s. A TiN cap- lattice parameters are obtained from data taken by ping layer nominally 15 As thick was deposited on the combination of symmetric and asymmetric top by reactive sputtering to avoid oxidation of the XRD scans yielding to values of (2.86$0.01) and Fe "lm. The structure was checked in situ by re#ec- (3.03$0.01) As for the in-plane and perpendicular tion high-energy electron di!raction (RHEED) and spacings, respectively. This means that Fe grows ex situ by X-ray di!raction (XRD). XRD experi- relaxed in the "lm plane but expanded in the ments were performed using two di!erent con"g- growth direction by 5.8% when compared with the urations: the Bragg}Brentano con"guration with bulk lattice parameters. This result is in contrast 1/43 slits plus a Cu secondary mono- with what has been previously observed by other chromator (Cu K radiation) for high angle scans; groups, which either do not comment on any dis- and, the high-resolution con"guration with tortion on the Fe lattice or observe growth under a Ge(2 2 0) four crystal monochromator (Cu a slight compression, yielding a perpendicular dis- K radiation) with 15 arcs beam divergence plus tortion of the Fe lattice parameter of about  a 1/83 receiving slit for the low angle scans. The !0.5% to !0.6% either for sputtering [6] or resulting re#ectivity measurements were "tted us- MBE growth [9]. Nevertheless, Lairson et al. [22] ing a code based on the recursion relation formal- have previously observed this vertical expansion in ism of the Fresnel equations developed by Parratt sputtered Fe/MgO(0 0 1), and correlated it with an [19,20]. Hysteresis loops with the magnetic "eld islanded like growth for thicknesses less than 15 applied in the "lm plane were taken by the use of monolayers. Even though we are studying "lms 10 a combined transversal/longitudinal Kerr setup times thicker, one can assume a similar initial J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 343 tice during the TiN capping process [23]. An esti- mate of the crystalline coherence length along the growth direction yields an average value of 70 As, applying Scherrer's equation to the width of the Fe(2 0 0) di!raction peak, the mosaic spread of the same peak being 2.153. Accurate values of "lm thickness and interface roughness are obtained through "ttings to the X-ray re#ectometry data (Fig. 1(c)), yielding to values of 198 As Fe, 19 As TiN and typical interface roughness of 10 As (rms) be- tween substrate and Fe, and 8 As (rms) between Fe and TiN and between TiN and air. 4. Magnetic characterization As mentioned in the introduction, our 200 As single-crystal BCC Fe thin "lm displays a magnetic behaviour consistent with the cubic crystalline structure. This is observed with in-plane magneto- optic hysteresis loops and with domain observa- tions at the surface, both types of measurements correlating well. The in-plane magnetic remanence as a function of the applied "eld angle with respect to the in- plane Fe crystallographic directions is shown in Fig. 2, together with the corresponding hysteresis loops. Note that the reduced remanence (M/M) is 1 when the "eld is applied along Fe [1 0 0] direc- tions and (M/M)+0.7+(cos 453) when the "eld is applied along the Fe [1 1 0] direction. This is expected for preferred domain orientations point- ing along the easy [1 0 0] direction. When the "eld is applied along these easy directions the reversal Fig. 1. Structural characterization: (a) RHEED pattern for the takes place by the nucleation and subsequent Fe/MgO(1 0 0) for two di!erent azimuths; (b) XRD phi scan for MgO(2 2 0) and Fe(1 1 0) asymmetric di!raction peaks; (c) X- propagation of domain walls. When the "eld is ray re#ectometry curve (open circles) and best "t (solid line). applied along the hard [1 1 0] directions, this irre- versible switch in the easy direction is followed by a continuous reversible rotation of the magnetiz- situation to the one described by Lairson at the "rst ation from the [1 0 0] to the [1 1 0] direction. The stages of growth, maintaining a big amount of obtained value for the anisotropy constant K residual strain as the "lm growth proceeds up to , assuming the bulk saturation magnetization value, the total thickness. In our case, we have observed is of the expected order of magnitude for Fe. Thus, a systematic variation of this lattice expansion on the domain distribution at remanence at 453 with changing growth conditions, work that will be pre- respect to the applied "eld direction produces the sented in a subsequent paper. Other possible ex- observed M planation for this lattice expansion involves the /M value. An important feature, not easily observable from the polar plot, but evident incorporation of interstitial Nitrogen in the Fe lat- from the observation of the di!erent hysteresis 344 J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 Fig. 2. Longitudinal hysteresis loops with the magnetic "eld applied along di!erent in plane directions; Center: polar plot showing M/M versus the applied "eld angle. loops, is that applying the magnetic "eld along ence of a small uniaxial anisotropy of unclear ori- equivalent crystallographic directions does not gin. This extra anisotropy is not evident in the produce equivalent hysteresis loops. This is obvi- polar plot, which allows us to give an upper bound ous when comparing loops obtained for symmetric for the anisotropy value of the order of the error orientations, such as 603 and 303, 753 and 153, 903 bar which is about 3% of the bulk Fe magnetocrys- and 03, and so on. talline anisotropy value. Its direction would be In order to investigate this asymmetry further perpendicular to the quadrants where the two- and a set of low-"eld Kerr magneto-optic loops were sometimes three-stage switching takes place, i.e., recorded. In addition to the Kerr re#ectivity chan- parallel to the [1 0 0] direction. Some possible ges, proportional to the magnetization compo- origins of this uniaxial anisotropy have been men- nent parallel to the applied "eld, the Kerr rotation tioned in the introduction and, in addition, residual was also measured to obtain the magnetization magnetic "elds present in the chamber could be component perpendicular to the applied "eld. This also responsible. However, test experiments de- procedure obtains both projections of the average signed to check this possibility produced a negative magnetization as a function of the applied "eld and result. The substrate texture is, in our opinion, the allows us to discern the operative magnetization most possible source for the extra anisotropy in this processes. As mentioned above, the low-"eld be- system, such as steps, as previously demonstrated for haviour is investigated in detail due to the presence Fe on stepped W(0 0 1) [24]. The low-"eld loops are of reversal anomalies that we attribute to the pres- shown in Fig. 3, where the relative orientation of the J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 345 Fig. 3. Hysteresis loops showing the longitudinal and transverse magnetization components for di!erent azimuths. magnetic "eld with the crystallographic axes is and transversal components, the last one of oppo- varied continuously. The "rst loop corresponds to site sign to the one detected along the hard axis a situation where the magnetic "eld is applied [1 1 0] as expected. On applying the magnetic "eld along the hard [1 1 0] direction. Both longitudinal  along the [1 1 0] hard direction the magnetization and  perpendicular components of the magnetiz- no longer switches in a single jump, an anomalous ation switch simultaneously, pointing to 1803-do- behaviour which is more apparent in the transverse main wall movement as the operative magneti- component. Note the di!erence with the [1 1 0] zation process. This switching occurs at 453 with loop that is, in principle, an equivalent  direction. respect to the "eld direction. When the magnetic This behaviour becomes more pronounced and "eld is applied along the [1 0 0] direction, the lon- reaches a maximum when the "eld is applied along gitudinal component switches again in one clear the [0 1 0] easy direction. Now, the longitudinal Barkhausen jump, but this time no transverse sig- magnetization switches in two clear Barkhausen nal is detected during the "eld cycle. This evidences jumps. These are separated by a plateau which as well 1803 wall propagation, but with the magnet- constitutes an intermediate state that is accom- ization now aligned parallel to the "eld. On increas- panied by a high transversal signal observed at ing the angle further, at 303 with respect to the easy the same "eld values. This implies that at the axis, we recover the detection of both longitudinal plateau the magnetization lies mostly along the 346 J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 perpendicular [1 0 0] easy direction, pointing out to 903 domain wall propagation as the operative magnetization process between saturation and pla- teau. This is in contrast with the loop obtained along the [1 0 0] direction, where the magnetiz- ation switching in a single 1803 Barkhausen jump within our experimental resolution. This two-stage process is more clear-cut in the transverse magnet- ization. For intermediate directions, such as the labelled 753, 1053 and 1203, the magnetization switching does not occur in a single step, but in two, and sometimes three distinct processes that can be explained by intermediate states where the magnet- ization lies perpendicular to the [0 1 0] direction as explained above. This behaviour is described in detail also in Ref. [14]. Note that this multiple step switching is only observed in the quadrant between 453 and 1353. This behaviour clearly separates from the one expected for a simple biaxial in-plane an- isotropy, and can be explained by the existence of a superimposed uniaxial anisotropy, which makes behave unequivalently, from the magnetic point of Fig. 4. (a) Kerr hysteresis loop taken with the magnetic "eld view, directions that are equivalent from the crys- applied in-plane along the main easy axis; (b) and (c) Kerr tallographic point of view. domain images taken in the above con"guration for magnetic The magnetic behaviour reported so far would "elds of !6.5 and !7.5 Oe, respectively. be consistent with a single-domain picture with coherent magnetization rotations and reversals. In the sample; in both cases it is due to macroscopic order to investigate the domain structure asso- morphological defects on the sample surface. These ciated to the di!erent magnetization reversal pro- results corroborate the 1803 domain wall picture of cesses, we performed MOKE domain observations the magnetization reversal along this direction. at selected loop positions. Figs. 4 and 5 show On the other hand, the hysteresis loop corre- hysteresis loops together with the corresponding sponding to the secondary easy axis (Fig. 5(a)) magnetic domains, for the situations where the shows the already mentioned plateaus at low "elds. magnetic "eld is applied along the two un- Detailed domain observations, indicate that the equivalent easy axis. The hysteresis loop corre- very "rst stages of magnetization reversal are char- sponding to the main easy axis (Fig. 4(a)) has acterized by large, up to 1 mm domains magnetized a perfect square shape with single and abrupt mag- along the main easy axis (white domains in the netization switching at about 7 Oe. Zones I and II "gure) that nucleate, propagate via 903, walls and correspond to the situations just before and after pin at surface structural defects. Further increase of switching of the magnetization direction. The asso- magnetic "eld produces the appearance reversed ciated domain distribution is shown in Fig. 4(b) and black domains that grow from the pinning sites. (c), for an applied magnetic "eld of !6.5 (zone I) These experimental facts have allowed us to assign and !7.5 Oe (zone II), respectively. The remanent the magnetization orientations in Fig. 5(b)}(e) state (I) exhibits a single-domain distribution. Do- which show the domain distribution corresponding main distribution for the state (II) shows almost to the selected zones in the hysteresis loop (III}IV). complete reversal of magnetization, except for III is indicative of a remanent state, IV and V cor- a narrow line that crosses almost completely the respond to intermediate states and VI to the rever- sample surface and some areas close to the edges of sed magnetization state. For the remanent state J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 347 magnetization distribution, corresponding to satu- ration, again except of some macroscopic mor- phological defects. 5. Conclusions The magnetization reversal process has been studied for Fe thin "lms grown by triode sputtering at normal incidence. The remanent state is well described by a cubic biaxial anisotropy, with the magnetization along [1 0 0] directions indepen- dently of the applied "eld direction. However, the low-"eld magnetization dynamics re#ects the exist- ence of a small uniaxial anisotropy superimposed over the biaxial crystalline anisotropy, leading to the inequivalency of crystalline easy axes. A single- jump transition driven by 1803 domain wall move- ment is observed when the magnetic "eld is applied along the main easy axis. A two-stage process driven by 903 domain wall movement is observed when the "eld is applied along the secondary easy axis. In addition, there is an angular range where two- and sometimes three-stage processes driven by 903 domain wall movement are observed. In all these cases, the reversal is via an intermediate state in which at least three-domain orientations coexist. This demonstrates again the inequivalency of the easy axes and the dramatic e!ect it has on both the magnetization reversal and the domain structure. The most possible cause, in our opinion, for this Fig. 5. (a) Kerr hysteresis loop taken with the magnetic "eld additional uniaxial anisotropy is the surface texture applied in-plane along the secondary easy axis. (b), (c), (d) and present in the MgO substrate. (d) Kerr domain images taken in the above con"guration for magnetic "elds of 0, !4, !6 and !16 Oe, respectively. See text for the domain direction assignment. Acknowledgements (III), the domain distribution is again uniform char- This work was carried out under "nancial sup- acteristic of a single domain. Zones IV and V basi- port of the Spanish Commission of Science and cally exhibit three di!erent regions: grey areas, Technology. One of us, J. L. MeneHndez, would like corresponding to domains whose magnetizations to acknowledge Comunidad de Madrid's Con- have not reversed yet, black areas corresponding to sejermHa de EducacioHn y Cultura for "nancial sup- reversed magnetization zones, and light areas cor- port. responding to domains with magnetizations per- pendicular to the secondary axis, i.e., along the main one and perpendicular to the applied mag- References netic "eld. Further application of magnetic "eld leads to zone VI, with a uniform single-domain [1] G.A. Prinz, J.J. Krebs, Appl. Phys. Lett. 39 (1981) 397. 348 J.L. Costa-Kra(mer et al. / Journal of Magnetism and Magnetic Materials 210 (2000) 341}348 [2] G.A. Prinz, Magnetic metal "lms on semiconductor [13] R.P. Coburn, J. FerreH, S.J. Gray, J.A.C. Bland, Phys. Rev. substrates, in: B. Heinrich, J.A.C. Bland (Eds.), Ultrathin B 58 (1998) 11507. Magnetic Structures, Springer, Berlin, 1994, pp. 1}42 and [14] E. Gu, J.A.C. Bland, C. Daboo, M. Gester, L.M. Brown, references therein. R. Poessl, J.N. Chapman, Phys. Rev. B 51 (1995) 3596. [3] P. Etienne, J. Massies, S. Lequien, R. Cabanel, F. Petro!, [15] C. Daboo, R.J. Hicken, E. Gu, M. Gester, S.J. Gray, J. Crystal Growth 111 (1991) 1003. D.E.P. Eley, E. Ahmad, J.A.C. Bland, R. Ploessl, [4] D.P. Pappas, J.W. Glesener, V.G. Harris, Y. Idzerda, J.N. Chapman, Phys. Rev. B 51 (1995) 15964. J.J. Krebs, G.A. Prinz, Appl. Phys. Lett. 64 (1994) 28. [16] U. Ebels, A.O. Adeyeye, M. Gester, C. Daboo, R.P. [5] Yu.V. Goryunov, N.N. Garif'yanov, G.G. Khaliullin, I.A. Cowburn, J.A.C. Bland, J. Appl. Phys. 81 (1997) 4724. Garifullin, Phys. Rev. B 52 (1995) 13450. [17] U. Ebels, A.O. Adeyeye, M. Gester, R.P. Cowburn, [6] B. Li, J.R. Fermin, A. Azevedo, F.M. de Aguiar, S.M. C. Daboo, J.A.C. Bland, Phys. Rev. B 56 (1997) 5443. Rezende, Appl. Phys. Lett. 72 (1998) 2760. [18] H. Ja!re s, L. Ressier, J.P. Peyrade, A.R. Fert, P. Gogol, A. [7] R.P. Cowburn, S.J. Gray, J. Ferre, J.A.C. Bland, J. Miltat, Thiaville, A. Schul, F. Nguyen Van Dau, J. Appl. Phys. 84 J. Appl. Phys. 78 (1995) 7210. (1998) 4375. [8] J.R. Childress, R. Kergoat, O. Durand, J.M. George, [19] T.A. Rabedeau, M.F.Toney, unpublished. P. Galtier, J. Miltat, A. Schuhl, J. Magn. Magn. Mater. 130 [20] L.G. Parrat, Phys. Rev. 95 (1954) 359. (1994) 13. [21] K. Postava, J.F. Bobo, M.D. Ortega, B. Raquet, [9] O. Durand, J.R. Childress, P. Galtier, R. Bisaro, A. Schuhl, H. Ha!res, E. Snoek, M. Goiran, A.R. Fert, J.P. Redoules, J. Magn. Magn. Mater. 145 (1995) 111. J. Pistora, J.C. Ousset, J. Magn. Magn. Mater. 163 (1996) [10] R.P. Coburn, S.J. Gray, J. FerreH, J.A.C. Bland, J. Miltat, 8. J. Appl. Phys. 78 (1995) 7210. [22] B.M. Lairson, A.P. Payne, S. Brennan, N.M. Rensing, [11] K. Postava, H. Ja!res, A. Schuhl, F. Nguyen Van Dau, B.J. Daniels, J. Appl. Phys. 78 (1995) 4449. M. Goiran, A.R. Fert, J. Magn. Magn. Mater. 172 (1997) 199. [23] K. Uchiyama, R.C. O'Handley, IEEE Trans. Magn. 35 [12] M. ZoKl#, M. Brockmann, M. KoKhler, S. Kreuzer, (1999) 2024. T. SchweinboKck, S. Miethaner, F. Bensch, G. Bayreuther, [24] H.J. Choi, Z.Q. Qiu, J. Pearson, J.S. Jiang, D. Li, J. Magn. Magn. Mater. 175 (1997) 16. S.D. Bader, Phys. Rev. B 57 (1998) R12713.