Spin orientation in an exchange coupled Fe/Cr/Fe trilayer determined by polarized neutron reflection J. A. C. Bland, H. T. Leung, and S. J. Blundell Cavendish Laboratory, Madingley Road, Cambridge CB3 OHE, United Kingdom V. S. Speriosu, S. Metin, and B. A. Gurney IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, California 95120 J. Penfold Rutherford-Appleton Laboratory, Chilton 0X11 0QX, United Kingdom We have used polarized neutron reflection to determine the layer-dependent spin orientations in an antiferromagnetically coupled 100 Å Cr/50 Å Fe/15 Å Cr/50 Å Fe/Si sandwich structure prepared by sputtering. At low field, the net Fe layer magnetic moments align in an asymmetric canted orientation with a near zero total magnetic moment for the sample. At high fields, a canted state, nearly symmetric with respect to the applied field direction is observed and the magnetization in each layer does not reach the bulk saturation value until the layers are ferromagnetically aligned. The behavior is discussed in the context of current theories of exchange coupling. © 1996 American Institute of Physics. S0021-8979 96 24908-6 Exchange coupling in ultrathin transition metal sandwich R was determined as a function of perpendicular wave vec- structures such as Fe/Cr/Fe has been intensively studied, in tor q for incident neutron spin parallel and antiparallel part because of the giant magnetoresistance GMR behavior to the applied field direction z axis . The measurements which can result.1­4 In addition to the Heisenberg-like bilin- were made at fixed orientation using the CRISP time of flight ear coupling, an additional biquadratic coupling mechanism, reflectometer17 at the ISIS facility in the UK Rutherford favoring a 90° spin alignment has also been found in several Laboratory. As the field is reduced the magnetic moments of epitaxial systems.5,6 Theoretical models of biquadratic cou- the layers move away from the z axis. Both and reflec- pling distinguish between an intrinsic mechanism due to the tivities are then dependent upon both of the in-plane compo- electronic properties of the perfect structure7­9 and extrinsic mechanisms which predict a strong dependence on the de- tails of the interface morphology and film structure.10­12 Neutron scattering studies of biquadratic coupling have been previously performed on both polycrystalline FeNi/Ag13 and Fe/Cr14 multilayers. Polarized neutron reflection provides an appropriate tool for probing the spin orientation in polycrys- talline single trilayers. We present in this article the results of a detailed polar- ized neutron reflection PNR study of a polycrystalline Fe/ Cr/Fe single-sandwich structure with AF coupling.15 In an earlier study,16 we were able to show that the spin orientation for such structures occurring at very low field departed sig- nificantly from the purely antiparallel structure expected for pure bilinear coupling. In this article the departure from the antiparallel state is quantified. The 100 Å Cr/50 Å Fe/15 Å Cr/50 Å Fe/Si sandwich structures were prepared by sputter- ing and the field-dependent magnetoresistance behavior in- vestigated as reported previously.15 X-ray diffraction studies confirm the polycrystalline structure with a preferred 110 texture. Magnetization measurements obtained using a vi- brating sample magnetometer VSM indicate the presence of antiferromagnetic coupling with saturation and coercive fields of 1.7 and 0.043 kOe, respectively solid lines, inset of FIG. 1. Magnetization measurements for the positive part of the M H loop Fig. 1 . The saturation value of the moment agrees with that solid line for the sputtered 100 Å Cr/50 Å Fe/15 Å Cr/50 Å Fe/Si sample calculated from the bulk magnetization. Measurements made obtained by VSM. The inset shows the full magnetization loop. The net magnetic moment parallel to the applied field deduced from the PNR fits are as a function of the in-plane orientation of the applied field shown for measurements made upon reducing the applied field from the confirm the absence of any significant magnetic anisotropy, positive saturation value open squares and upon increasing the applied as expected for such polycrystalline samples. field from the negative saturation value solid squares . The spin configura- PNR measurements17 were made at 300 K with the tions magnitudes and directions are drawn to scale as deduced from the coherent multidomain model. The thicker arrow represents the magnetiza- sample magnetized in-plane and the total specular reflectivity tion of the upper Fe layer. The applied field is vertical z axis . J. Appl. Phys. 79 (8), 15 April 1996 0021-8979/96/79(8)/6295/3/$10.00 © 1996 American Institute of Physics 6295 Downloaded¬09¬Sep¬2002¬to¬148.6.178.13.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp FIG. 2. The sample geometry for a multilayer sample with nonaligned in- plane magnetizations. The first layer magnetization vector M1 thick arrow and the second Fe layer magnetization vector M2 are constrained to the film plane. The angle 1 refers to the orientation of M1 with respect to the applied field direction vertical z axis and the angle 2 defines the angular separation of M1 and M2 narrower arrow , with positive angle correspond- ing to the anticlockwise sense. nents of the magnetization vector as described by a reflectiv- ity matrix.18 A similar description has been given by Felcher et al.19 The magnetic configuration of the sample is shown schematically in Fig. 2. An appropriate spatial averaging pro- cedure is also required since beneath the saturation field magnetic domains can develop.5 If the domains were larger than the coherence length, then each domain would contrib- FIG. 3. The spin asymmetry observed for the 100 Å Cr/50 Å Fe/15 Å Cr/50 ute incoherently to the reflected intensity. The coherence Å Fe/Si sputtered sandwich structure and the fits solid lines predicted by length projected in plane is estimated to be about 100 m at the coherent multidomain model as the applied field is reduced from the positive saturation value to the field strengths indicated. The fit parameters grazing incidence17 while Kerr microscopy studies have re- are given in Table I and the resulting spin orientations are shown in Fig. 1. vealed the existence of an irregular patch domain structure on the scale of a few microns at zero applied field for a trilayer structure with a Cr thickness in the vicinity of the In fitting the data in the coherent case we relax the re- first antiferromagnetic peak i.e., close to the Cr thickness of quirement that the average magnetization vectors of each our samples suggesting that coherent averaging applies in layer has the full saturation value Ms corresponding to a our experiment. In a first approximation the net magnetiza- single-domain state and allow both the orientation and size tion in each layer then corresponds to the spatial average of the magnetization vector in each layer to be determined by over the magnetizations of each domain within the coherence the result of a least-squares fit to the asymmetry data. The area. S(q) curve is computed with the size and orientation of the The spin asymmetry S (R R )/(R R ) for magnetization vector in each layer treated as independently Ha 2 kOe is consistent with the values calculated for par- variable parameters until a close fit solid line in Fig. 3 is allel alignment of the two Fe layer moments along the ap- found using a least-squares minimization procedure. This plied field direction.20 This measurement serves as a check method is found to fit the data very well for the appropriate on the experimentally determined layer thicknesses and also values of the magnetization vector magnitude and orienta- on the size of the Fe layer magnetizations which are found to tion in each layer, in contrast to the incoherent averaging correspond to the bulk value within experimental error. In method for which no fits could be obtained. The results are Fig. 3 we show the spin asymmetry observed evolving as the summarized in Table I and the resulting spin configurations applied field is reduced from the positive saturation value. are shown in Fig. 1. For each configuration shown an alter- The ferromagnetic alignment induces a pronounced peak in native symmetry related configuration is also possible corre- the asymmetry at a value of wave vector close to 2.9qc sponding to that obtained by reflecting about the applied field where qc is the critical wave vector for the Si substrate . As direction. The bold arrow refers to the magnetic moment of the applied field is reduced the peak asymmetry at low wave the top layer and the thin arrow refers to that of the bottom vector diminishes, indicating the expected reduction in the layer. The resulting component of the magnetization along degree of ferromagnetic alignment between the layers, and a the applied field M in Table I fitted using the above pro- second peak close to 6qc increases, as expected from simu- cedure is consistent with the magnetometry data on these lations of the spin asymmetry for antiferromagnetic ordering. samples, as shown in Fig. 1. We also find that the angular This peak is related to the antiferromagnetic Bragg diffrac- separation of the magnetic moment vectors is almost exactly tion peak that occurs in superlattices,21,22 although refraction preserved under physical rotation of the sample in low-field and spin orientation effects must be included in order to ac- applied fields see Table 1 and Fig. 1 . This serves as a useful curately fit the position and magnitude of the peak.18 check on our analysis. We see that upon reducing the applied 6296 J. Appl. Phys., Vol. 79, No. 8, 15 April 1996 Bland et al. Downloaded¬09¬Sep¬2002¬to¬148.6.178.13.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp TABLE I. The result of a least-squares fit to the set of field-dependent spin asymmetry data. The 1st column gives the applied field strength for positive values upon reducing the field from the positive saturation value. The measurement denoted (R) indicates that the sample has been rotated by 90° with respect to the applied field. The angles 1, 2 denote the angular position of the top layer magnetization with respect to the applied field and the angular separation of the layer magnetizations. The 4th and 5th columns show the magnitudes of the top and bottom Fe layer moments. The net components of the magnetization parallel and perpendicular to the applied field M and Mper are shown in the final two columns. H Oe 1 2 M1/Ms M2/Ms M /Ms Mper/Ms 12 33 160 0.27 0.27 0.032 0.004 0.034 0.039 12R 110 170 0.26 0.26 0.021 0.002 0.010 0.011 94 11 131 0.27 0.29 0.060 0.006 0.100 0.293 200 19 129 0.30 0.20 0.108 0.005 0.045 0.054 320 24 79 0.42 0.30 0.278 0.016 0.037 0.062 540 8 17 0.46 0.46 0.455 0.009 0.004 0.087 2000 0 0 1.00 1.00 1.000 0.000 field from the saturation value, the formation of magnetic biquadratic coupling. This view is supported by the sample domains is inferred i.e., M1 , M2 decreases with, after ini- rotation results. tial canting away from the applied field direction, the lower In conclusion, the results demonstrate the importance of Fe layer undergoing a rotation process to attain an overall determining the layer dependent spin orientation in exchange canted, near AF configuration at low applied fields. Each coupled structures, as is possible using PNR. layer has 25%­30% of its full saturation magnetization at low fields. We note that from Table I the total magnetic mo- 1 G. Binasch, P. Grunberg, F. Saurenbach, and W. Zinn. Phys. Rev. B 39, ment of the sample perpendicular to the applied field Mper 4828 1989 . is always zero, within experimental error. This is consistent 2 J. J. Krebs, P. Lubitz, A. Chaiken, and G. A. Prinz, Phys. Rev. Lett. 63, with the physical requirement that for stability there be no 1645 1989 . 3 A. Fert and P. Bruno, in Ultrathin Magnetic Structures, edited by B. net torque on the layers arising from the applied field since Heinrich and J. A. C. Bland Springer, Berlin, 1994 , Chap. 2.2. the magnetic anisotropy in the sample is negligible. The 4 S. S. Parkin, in Ultrathin Magnetic Structures, edited by B. Heinrich and magnetization reversal process is clearly inequivalent in the J. A. C. Bland Springer, Berlin, 1994 , Chap. 2.4. 5 two Fe layers, resulting at low field in an asymmetric mag- M. Ruhrig, R. Schafer, A. Hubert, R. Mosler, J. A. Wolf, S. Demokritov, and P. Grunberg, Phys. Status Solidi A 125, 635 1991 . netic orientation with respect to the applied field direction. 6 B. Heinrich and J. Cochran, Adv. Phys. 42, 523 1993 . This inequivalence may result from the inequivalent inter- 7 R. P. Erickson, K. B. Hathaway, and J. R. Cullen, Phys. Rev. B 47, 2626 faces of the two Fe layers the bottom layer is directly de- 1993 . 8 posited on Si . D. M. Edwards, J. M. Ward, and J. Mathon, J. Magn. Magn. Mater. 126, 380 1993 . The canting observed at low field could suggest that in 9 P. Bruno, J. Magn. Magn. Mater. 121, 248 1993 . addition to bilinear coupling favoring antiparallel alignment, 10 J. C. Slonczewski, Phys. Rev. Lett. 67, 3172 1991 . 11 biquadratic coupling comparable in strength is also present, J. C. Slonczewsi, J. Appl. Phys. 73, 5957 1993 . 12 S. Demokrtitov, E. Tsymbal, P. Grunberg, W. Zinn, and I. K. Schuller, since in the absence of anisotropy this configuration would Phys. Rev. B 49, 720 1994 . be otherwise unstable. In the fluctuation model10 biquadratic 13 R. Rodmacq, K. Dumesnil, Ph. Mangin, and M. Hennion, Phys. Rev. B coupling is expected to occur only for sufficiently large ter- 48, 3556 1993 . 14 race widths and appropriate roughness values. For sputtered M. Scha¨fer, J. A. Wolf, P. Gru¨nberg, J. F. Ankner, A. Schreyer, H. Zabel, and C. F. Majkrzak, J. Appl. Phys. 75, 6193 1994 ; A. Schreyer et al., structures very sharp interfaces can be achieved on the mac- Europhys. Lett. 1995 in press . roscopic scale but the correlation length is expected to be 15 B. A. Gurney, D. R. Wilhoit, V. S. Speriosu, and I. L. Sanders, IEEE reduced on a nm scale in comparison with epitaxial struc- Trans. Magn. MAG-26, 2747 1990 . 16 tures. On the basis of this model, weaker biquadratic cou- J. A. C. Bland, R. D. Bateson, N. F. Johnson, S. J. Blundell, V. S. Spe- riosu, S. Metin, and B. Gurney, J. Magn. Magn. Mater. 123, 320 1993 . pling would be expected in sputtered samples in contrast 17 J. A. C. Bland, in Ultrathin Magnetic Structures, edited by J. A. C. Bland with our findings. Other extrinsic models invoking dipolar and B. Heinrich Springer, Berlin, 1994 . coupling12 and the presence of ``loose'' spins11 in the spacer 18 S. J. Blundell and J. A. C. Bland, Phys. Rev. B 46, 3391 1992 . 19 layer are unlikely to result in the large biquadratic coupling G. P. Felcher, R. O. Hilleke, R. K. Crawford, J. Haumann, R. Kleb, and G. Ostrowski, Rev. Sci. Instrum. 58, 609 1987 . strength we observe. It is also possible that the antiferromag- 20 J. A. C. Bland, A. D. Johnson, R. D. Bateson, S. J. Blundell, H. J. Lauter, netism of the Cr layer is important in this context. Alterna- C. Shackleton, and J. Penfold, J. Magn. Magn. Mater. 93, 513 1991 . tively, differences between the Fe layers in net moment, pin- 21 A. Schreyer, K. Bro¨hl, J. F. Ankner, C. F. Majkrzak, Th. Zeidler, P. Bo¨de- ning sites or local anisotropy variations together with ker, N. Metoki, and H. Zabel, Phys. Rev. B 47, 15334 1993 . 22 Y. Y. Huang, G. P. Felcher, and S. S. P. Parkin, J. Magn. Magn. Mater. 99, bilinear coupling could explain the canting without invoking L31 1991 . J. Appl. Phys., Vol. 79, No. 8, 15 April 1996 Bland et al. 6297 Downloaded¬09¬Sep¬2002¬to¬148.6.178.13.¬Redistribution¬subject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp