PHYSICAL REVIEW B VOLUME 57, NUMBER 17 1 MAY 1998-I Interface selective vector magnetometry of FeNi/Cu/Co trilayer spin-valve structures J. A. C. Bland, C. Daboo, M. Patel, and T. Fujimoto Cavendish Laboratory, University of Cambridge, Madingley Road, Cambridge CB3 0HE, United Kingdom J. Penfold Rutherford Laboratory, Chilton, Oxon., OX11 OQX, United Kingdom Received 25 November 1997 Using polarized neutron reflectometry PNR together with superconducting quantum interference device magnetometry, the interface and interior magnetic moments have been determined for each of the ultrathin FeNi and Co layers within an epitaxial FeNi/Cu/Co trilayer structure, so demonstrating interface selectivity in layers of the same nominal chemical composition. The reduced moment found for the Co/Cu and FeNi/Cu interface regions are consistent with a model of enhanced electron spin-flip scattering at rough interfaces proposed to explain the temperature dependence of the giant magnetoresistance amplitude in FeNi/Cu/Co spin-valve structures. We further show that the layer-dependent vector moments can be determined by PNR with high precision. S0163-1829 98 00117-9 The spin configuration and magnetization at an interface of the vector magnetometry capability of PNR. The total between two layers within a magnetic multilayer structure sample moment deduced from PNR is tested against super- determine such key properties as the magnetoresistance be- conducting quantum interference device SQUID magne- havior, interlayer exchange coupling, and interface tometry measurements. A further motivation in studying this anisotropy.1 Testing the interface spin structure of overlayers system is provided by the results of recent studies of the and multilayers predicted by theoretical models, for example, temperature dependence of the giant magnetoresistance in Ref. 2, requires experimental techniques capable of selec- FeNi/Cu/Co spin valve structures which suggest that the tively probing buried single interfaces. Interest is now rap- spin-flip electron scattering which varies quadratically with idly growing in the application of polarized radiation tech- temperature T at low temperature is controlled by spin-wave niques such as x-ray magnetic circular dichroism XMCD ,3 excitations at the ferromagnetic/nonferromagnetic interfaces or second harmonic generation4 to the study of interface with a strength dependent on the interface structure.16 magnetic structure. For example, the element specific selec- Structures of the form Si 100 /Cu 500 Å /FeNi/Cu 60 Å / tivity of XMCD has been successfully used to determine the Co/Cu 60 Å were grown on HF-passivated p-Si 100 sub- spin structures in Mn layers within Fe/Mn/Fe trilayers.5 On strates in UHV at ambient temperature, with the pressure the other hand, polarized neutron reflection PNR and dif- during growth maintained below 5 10 8 mbar. The Co and fraction techniques6­10 can be used to probe the spin orien- FeNi layer thicknesses estimated from quartz microbalance tation within a magnetic multilayer structure with layer spe- measurements were typically in the range 20­40 Å. Prior to cific selectivity. Neutron scattering methods are the method deposition of the single trilayer, the substrate was heated to of choice when a determination of the absolute value of the 200 °C for 30 min after which a thick Cu buffer layer magnetization vector is required. In the case of multilayer ( 500 Å) was deposited at room temperature. The Cu structures, the interface structure can be obtained on a nm buffer layer is deposited to improve epitaxy17 and is of a scale or better and PNR studies of magnetic ordering and thickness chosen to give rise to oscillations in the neutron interfacial structure in Gd/Fe multilayers,11 Fe/Cr,12 and reflectivity at low wave vectors. In situ reflection high- Fe/Si Ref. 13 multilayers have been recently reported. For energy electron diffraction RHEED images obtained during simple overlayer structures, e.g., Gd/Fe Ref. 14 , the reso- growth confirm the epitaxial growth of the FeNi, Cu, and Co lution with which the interface structure can be determined is overlayers in the 100 orientation.17 Equivalent single reduced due to the limited wave vector range over which trilayer polycrystalline structures grown directly onto Si sub- reflectivity data can be obtained. However, in the case of strates exhibit uniaxial anisotropies and giant magnetoresis- magnetic multilayers, in order to analyze the PNR data, it is tance GMR .16 During growth the sample was rotated to usually necessary to assume a superlattice structure i.e., suppress uniaxial anisotropies and the resulting epitaxial equivalent repeat units within the structure . The presence of FeNi and Co layers both exhibit a fourfold anisotropy.18 unknown domain structures further complicates the analysis SQUID measurements reveal a significant reduction in the of field dependent data in practice. magnetization with increasing temperature in the range 1.5­ In the present work we have chosen to study FeNi/Cu/Co 300 K and thus the temperature range of the PNR measure- single trilayer spin valve structures using PNR in order to ments was extended to 1.5 K. The PNR measurements were probe the interface spin configuration in a system without a made using the CRISP time of flight reflectometer at the ISIS periodic structure. We have selected a simple trilayer since facility at the Rutherford Appleton Laboratory. The spin- the soft FeNi layer undergoes coherent spin rotation in an dependent specular reflectivity R was determined as a func- external field of appropriate strength15 and so permits a test tion of perpendicular wave vector q for spin parallel and 0163-1829/98/57 17 /10272 4 /$15.00 57 10 272 © 1998 The American Physical Society 57 BRIEF REPORTS 10 273 plateaux which occur following the reversal of the FeNi layer correspond to a near antiferromagnetic AF alignment of the FeNi and Co moments and thus can be used to esti- mate the magnetic moment of the FeNi and Co layers. The magnetic anisotropy of the Co layers is strong enough to constrain the Co moments at low fields. Figure 1 a shows spin-dependent reflectivity spectra in the low wave vector range obtained for an epitaxial trilayer held at 100 K with the layer magnetization aligned parallel by an applied field of 3 kG. Seven well pronounced reflec- tivity oscillations as a function of wave vector are seen within the wave vector range accessed. In fitting the data, two models of the structure were used: one in which the nominal sample structure is assumed and a second model in which additional intermixed FeNi-Cu and Co-Cu layers of variable composition and thickness are introduced at the in- terfaces. In both models, a Gaussian interface roughness of adjustable amplitude is introduced7 and the layer thicknesses are determined from the reflectivity data available through- out the temperature range studied whereas the magnetic mo- ments are fitted separately for each temperature. In the nomi- nal structure model, the bulk scattering lengths and densities are assumed for each nonmagnetic layer, while all layer thicknesses, the scattering densities for the magnetic layers, and the magnetic moments were freely varied. The result, while reproducing well the features of the reflectivity data in the low wave vector range19 is not consistent with the total moment measured by SQUID magnetometry and therefore FIG. 1. a The spin-dependent reflectivity spectra plotted as a can be excluded see Table I . In the interface model, the function of wave vector q for the epitaxial trilayer structure held at interior regions in the ferromagnetic layers are assumed to 100 K with the layer magnetizations aligned parallel by an applied have the full bulk moment of FeNi and Co and bulk scatter- field of 3 kG. The lines correspond to fits using the interface model ing densities. In the interface regions the densities, moments, see text . The inset shows the magnetic hysteresis loop for the and thicknesses are adjusted Table I . The resulting best fit Si 001 /Cu/FeNi/Cu/Co/Cu epitaxial spin-valve structure obtained by SQUID magnetometry. b The spin-dependent reflectivity spec- using the interface model is shown as solid spin-up and tra plotted as a function of wave vector q obtained at low wave dashed spin-down lines in Fig. 1 a and the corresponding vectors for an epitaxial trilayer held at 1.5 K with the layer magne- parameters are given in Table II. Layer thicknesses are de- tizations aligned parallel by an applied field of 3 kG. The lines termined in this way to a precision of 2­3 Å. It can be seen correspond to fits using the interface model. that the resulting fit closely reproduces all features of the data very well. antiparallel to the applied field parallel to the 100 di- Figure 1 b shows spin-dependent reflectivity spectra in rection easy axis for the FeNi layer as a function of tem- the low wave vector range obtained for an epitaxial trilayer perature with the sample magnetically saturated in plane and held at 1.5 K with an applied field of 3 kG. Three well at room temperature as a function of field orientation. Imme- pronounced reflectivity oscillations as a function of wave diately following the PNR measurements, magnetic hyster- vector are seen within the wave vector range accessed. The esis loops as a function of temperature were obtained from best fit solid lines again assumes the presence of interfacial SQUID magnetometry with the field applied along the easy layers as given in Table II and the resulting value for the axis direction. The room temperature loops show saturation total moment of the sample is found to be in good agreement fields for the Co layers of 200 Oe and reveal abrupt rever- within errors with the result of the SQUID measurements sal of the FeNi layers at low fields inset Fig. 1 a . The Table I . High quality fits are also obtained for the 300 K TABLE I. Moments fitted by PNR interfacial model and nominal layer models compared to moments fitted by SQUID magnetometry. The sample area used for the SQUID measurements is 0.165 cm2, leading to the following conversions: for Co, 1.376 10 6 emu/ B Å; for FeNi, 1.373 10 6 emu/ B Å Layer Moment ( 10 5 emu) at RT Moment ( 10 5 emu) at 100 K Moment ( 10 5 emu) at 1.5 K PNR SQUID PNR PNR SQUID PNR SQUID Interfacial Interfacial Nominal Interfacial Co 5.09 0.84 5.03 5.39 1.02 4.46 0.49 5.40 5.56 0.88 5.51 FeNi 2.87 0.48 2.86 3.16 0.79 3.17 0.42 3.13 3.23 0.83 3.21 10 274 BRIEF REPORTS 57 TABLE II. Thicknesses and moments in Bohr magnetons used to fit the PNR data using an interfacial layer model. The errors shown for the thicknesses are those determined for the room tem- perature fit and are typical of the values found for the other fits. Layer Thickness Å PNR Moment ( B) RT 100 K 1.5 K Cu 62.65 2.24 Cu-Co 10.25 1.13 1.44 0.21 1.52 0.13 1.57 0.26 Co 4.44 0.94 1.60 0.41 1.67 0.83 1.73 Co-Cu 10.60 1.14 1.42 0.22 1.53 0.20 1.58 0.31 Cu 64.45 2.05 Cu-FeNi 11.14 2.14 0.82 0.15 0.91 0.93 FeNi 5.06 1.93 0.86 0.06 0.94 0.96 FeNi-Cu 8.88 1.72 0.83 0.13 0.91 0.22 0.93 Cu 451.13 2.86 data using the interface model see Table II which again agree with the SQUID magnetometry measurements. As above, it is again possible to exclude a model in which the interface regions are not included. The average moments of the FeNi and Co based layers i.e., effective moments aver- aged over both the interior and interface regions are found to decrease by 11.1 and 7.4% Ref. 19 in the temperature range 1.5­300 K, in excellent agreement with the results from SQUID magnetometry of 10.5 and 8.9%, respectively. FIG. 2. The spin asymmetry spectra as a function of the relative The interface regions are found to extend over a thickness layer magnetization orientation at room temperature for a parallel of 9­11 Å and to have significantly lower effective magne- alignment and after the sample is rotated with respect to the applied tizations than the corresponding values expected for the pure field by approximately b 90° and c 180° causing the FeNi layer to rotate as shown in the schematic right insets . The lines corre- FeNi and Co layers. Such a reduction is consistent with spond to fits described in the text. chemical intermixing but while no significant moment is ex- pected for the Cu atoms, the values we obtain for the effec- tive moments of the mixed composition regions are slightly 3 kOe and then reducing the field to 50 Oe, which is enough larger than might be expected from the effect of simple di- to switch the FeNi layer but not the Co layer magnetization. lution by the Cu atoms. However, the existence of reduced Strong variations in the spin asymmetry as a function of coordination effects associated with atomic scale roughness wave vector are seen. A best fit is shown as a solid line can increase the local moment on the magnetic atoms as has obtained with all the magnetic layers aligned parallel to the been observed in epitaxial Fe/Ag 001 layers, see, for ex- applied field. It can be seen that all the features of the data ample, Ref. 20. The existence of such a diffuse interface are well reproduced by the model. The sample was then supports the interpretation of measurements of the tempera- physically rotated in the applied field by 90° Fig. 2 b ture dependence of the giant magnetoresistance amplitude and 180° Fig. 2 c . The spin asymmetry data display for FeNi/Cu/Co spin-valve structures which suggests the dramatic variations according to the direction of the sample presence of intermixed interface regions.16 For relatively with respect to the applied field, with pronounced oscillatory rough interfaces, as in the case of FeNi/Cu/Co spin-valve structure across the wave vector range. In fitting the data we structures, the interface exchange coupling is expected to be first assume that the FeNi layer has coherently rotated to weakened in comparison with the ideal interface and the angles of 90° and 180°, respectively, and fit the moments of spin-fluctuation intensity and electron spin-flip scattering ac- the layers; then we vary the angle of the FeNi layer keep- cordingly increased. ing the moments fixed. For the 180° data the fitted angle is A very important aspect of PNR is its sensitivity to the very close within 1° to 180° confirming that coherent rota- vector orientation of the magnetic moments of the layers tion occurs. The inequivalence of the spin orientations in within the sample.9 In the case of noncollinear structures the both orientations is clearly seen in the fit parameters. In each and reflectivities are both dependent on the in-plane orientation, very good agreement between all the features of components of the magnetization vector as described by a the data and the fitted curves is obtained indicating that to a reflectivity matrix.6 However, until now this capability has good approximation the FeNi layer magnetization indeed ro- not been tested to our knowledge in trilayer samples in tates as a single domain to become aligned parallel to the which the magnetic orientation can be accurately controlled. applied field and that to a good approximation the orientation Figure 2 a shows the spin asymmetry defined by S (R of the Co layer is unchanged, thus exhibiting ideal spin-valve R )/(R R ) at room temperature measured by first satu- behavior. For the 90° state the best fit parameters yield an rating the sample along an easy axis with an applied field of error of 8°­14° indicating that in this configuration the spin 57 BRIEF REPORTS 10 275 asymmetry is less sensitive to the spin-separation angle. We ing inferred from the strong temperature variation of the emphasize that in addition to the vector orientation of the GMR amplitude in FeNi/Cu/Co spin-valve structures. We layers, PNR also yields the absolute value of the total mo- show that the layer-dependent vector moments can be deter- ment of each layer. This is important in determining the ex- mined quantitatively with high precision from measurements tent to which the layers are uniformly magnetized. of the spin asymmetry as a function of wave vector. The In conclusion we have used polarized neutron reflectom- capability of PNR in quantitatively probing the interface spin etry measurements to demonstrate interface selective magne- structure of layers with a common magnetic element is likely tometry in an epitaxial FeNi/Cu/Co trilayer spin-valve struc- to provide an important complementary probe to XMCD ture. Interface and interior magnetic moments in each of the techniques and to play an important part in unraveling the FeNi and Co layers are obtained which agree within errors spin structure of magnetic interfaces in the future. with the results of SQUID magnetometry measurements of the total sample moment. This is the first time to our knowl- We are grateful to the Toshiba Corporation and to the edge that such a detailed comparison between SQUID mag- EPSRC for supporting this work, and we thank the Ruther- netometry and PNR has been successfully performed. The ford Appleton Laboratory for making reflectometry facilities results are consistent with the presence of chemical intermix- available. 1 For a review of these topics see, for example, Ultrathin Magnetic P. Bo¨deker, N. Metoki, and H. Zabel, Phys. Rev. B 47, 15 334 Structures, edited by J. A. C. Bland and B. Heinrich Springer- 1993 . Verlag, Berlin, 1994 , Vols. I and II. 11 W. Hahn, M. Loewenhaupt, Y. Y. Huang, G. P. Felcher, and S. S. 2 A. Vega, C. Demangeat, H. Dreysse´, and A. Chouairi, Phys. Rev. P. Parkin, Phys. Rev. B 52, 16 041 1995 . B 51, 11 546 1995 . 12 S. Adenwalla, G. P. Felcher, E. E. Fullerton, and S. D. Bader, 3 V. Chakarian, Y. U. Idzerda, G. Meigs, E. E. Chaban, J. H. Park, Phys. Rev. B 53, 2474 1996 . and C. T. Chen, Appl. Phys. Lett. 66, 3368 1995 ; C. T. Chen, 13 J. F. Ankner, C. F. Majkrzak, and H. Homma, J. Appl. Phys. 73, Y. U. Idzerda, H. J. Lin, G. Meigs, A. Chaiken, G. A. Prinz, and 6436 1993 . G. H. Ho, Phys. Rev. B 48, 642 1993 . 14 O. F. K. McGrath, N. Ryzhanova, C. Lacroix, D. Givord, C. 4 G. Spierings, V. Koutsos, H. A. Wierenga, M. W. J. Prins, D. Fermon, C. Miramond, G. Saux, S. Young, and A. Vedyayev, Abraham, and Th. Rasing, J. Magn. Magn. Mater. 121, 109 Phys. Rev. B 54, 6088 1996 . 1993 . 15 5 M. Patel, T. Fujimoto, E. Gu, C. Daboo, and J. A. C. Bland, J. V. Chakarian, Y. U. Idzerda, H. J. Lin, C. Gutierrez, G. A. Prinz, G. Meigs, and C. T. Chen, Phys. Rev. B 53, 11 313 1996 . Appl. Phys. 75, 6528 1994 . 16 6 G. P. Felcher, R. O. Hilleke, R. K. Crawford, J. Haumann, R. T. Fujimoto, M. Patel, E. Gu, C. Daboo, and J. A. C. Bland, Phys. Kleb, and G. Ostrowski, Rev. Sci. Instrum. 58, 609 1987 ; S. J. Rev. B 51, 6719 1995 . 17 Blundell and J. A. C. Bland, Phys. Rev. B 46, 3391 1992 . R. Naik, M. Ahmad, G. L. Dunifer, C. Kota, A. Poli, K. Fang, U. 7 J. A. C. Bland, in Ultrathin Magnetic Structures Ref. 1 , pp. Rao, and J. S. Payson, J. Magn. Magn. Mater. 121, 60 1993 . 18 305­343. A. Ercole, T. Fujimoto, M. Patel, C. Daboo, R. J. Hicken, and J. 8 R. Rodmacq, K. Dumesnil, Ph. Mangin, and M. Hennion, Phys. A. C. Bland, J. Magn. Magn. Mater. 156, 121 1996 . Rev. B 48, 3556 1993 . 19 M. Patel, T. Fujimoto, A. Ercole, C. Daboo, and J. A. C. Bland, J. 9 A. Schreyer, J. F. Ankner, T. Zeidler, H. Zabel, C. F. Majkrzak, Magn. Magn. Mater. 156, 53 1996 . M. Schafer, and P. Grunberg, Europhys. Lett. 32, 595 1995 . 20 J. A. C. Bland, C. Daboo, B. Heinrich, Z. Celinski, and R. D. 10 A. Schreyer, K. Bro¨hl, J. F. Ankner, C. F. Majkrzak, Th. Zeidler, Bateson, Phys. Rev. B 51, 258 1995 .