PHYSICAL REVIEW B VOLUME 53, NUMBER 17 1 MAY 1996-I Canted coupling of buried magnetic multilayers V. Chakarian* and Y. U. Idzerda Naval Research Laboratory, Code 6345, Washington, D.C. 20375 H.-J. Lin AT&T Bell Laboratories, 600 Mountain Avenue, Murray Hill, New Jersey 07974 C. Gutierrez Department of Physics, South West Texas State University, San Marcos, Texas 78666 G. A. Prinz Naval Research Laboratory, Code 6345, Washington, D.C. 20375 G. Meigs and C. T. Chen AT&T Bell Laboratories, 600 Mountain Avenue, Murray Hill, New Jersey 07974 Received 12 October 1995 Soft-x-ray magnetic circular dichroism is used as an element-specific magnetometer to determine the mag- netic behavior of a buried 4.3 monolayer Mn film in a Fe25Co75 /Mn/Fe25Co75 trilayer, which exhibits a 90° coupling between the two ferromagnetic films. By measuring element-specific magnetic hysteresis curves for Fe, Co, and Mn along directions parallel and perpendicular to an applied magnetic field, the magnetization behavior of each element is described, indicating an anomalous in-plane canting of the net Mn moment with respect to the Fe and Co moments by 23°. The properties of buried and embedded layers, whether overcome this coupling. Because this coupling angle is de- covered by a passivation layer or simply a component in a pendent on the thickness and perhaps roughness of the Mn heterostructure, are often device defining. The task of study- layer, this unique magnetic configuration of the two alloy ing these strategic layers becomes much more difficult when films must be reflected in an atypical magnetic structure of the measurement signal associated with the embedded layer the embedded Mn layer. is much smaller than the concurrent signal of the embedding Although these earlier works have described the magnetic media. Magnetic investigations of heteromagnetic multilay- behavior of the ferromagnetic FeCo alloy films, no experi- mental determination of the magnetic structure of the inter- ers are representative of this class of problem. Ferromagnet/ layer Mn, the central mechanism of the coupling, was made. metal/ferromagnet trilayers and superlattices exhibit exciting To establish the magnetic structure of the buried Mn layer by and technologically useful properties which are mainly de- conventional magnetic characterization techniques is not fined by the interlayer.1 The prototypical system is Fe/Cr/Fe possible because the magnetic signatures of the FeCo alloy which first demonstrated interlayer coupling2,3 and giant layers are 2­3 orders of magnitude larger than the signature magnetoresistance,4 and continues to yield new develop- of the Mn film. Perhaps the only method which is capable of ments. For a review of the Fe/Cr/Fe system, see Ref. 5. The separating the magnetic Mn signal from those of the Fe and overall magnetic behavior of these types of trilayers is pre- Co is soft-x-ray magnetic circular dichroism MCD , an dicted to be controlled to a large extent by the interlayer element-specific magnetic spectroscopic tool where the dif- properties,6 motivating a variety of studies to focus on the ference in the absorption of left- and right-circular polarized magnetic structure of simple overlayers of the interlying ma- photons is measured at the absorption edges of the constitu- terial as a starting point to understanding the trilayer ent elements. Soft-x-ray MCD is capable of determining the coupling.7­10 magnetic order of overlayers10,14,15 and buried layers16 with Recently a new multilayer system, essentially bcc Co/ high sensitivity. MCD measurements made in an applied Mn/Co the actual composition of the trilayer is magnetic field can be used to generate element-specific mag- Fe25Co75 /Mn/Fe25Co75), has been found to have quite a netic hysteresis curves which in turn can be used to dissect unique coupling behavior,11 consistent with a recent theoreti- complicated total-moment hysteresis curves into their el- cal model for systems with an ordered antiferromagnetic emental components,17 extract magnetoresistance values in- interlayer.12 Instead of simply exhibiting aligned or anti- dependent of multidomain and incomplete moment align- aligned configuration of the two ferromagnetic films, the re- ment effects,18 and can even be used as an element-specific sults of magnetometry, ferromagnetic resonance,11 and neu- magnetometer ESM ,19 where all three components of the tron scattering studies13 of these trilayers indicate the magnetic moment vector are determined for each element or magnetic moments of the two single-crystal FeCo alloy lay- layer. ers are strongly coupled at a fixed angle with respect to one In this Brief Report, by using the ESM technique, we another, requiring a very large magnetic field 15­20 kG to demonstrate that surprisingly, the Mn interlayer possesses a 0163-1829/96/53 17 /11313 4 /$10.00 53 11 313 © 1996 The American Physical Society 11 314 BRIEF REPORTS 53 net magnetic moment which exhibits a nearly field- independent canting of 23° relative to the average moments of Fe and Co. From the evolution of these moments in re- sponse to an arbitrary applied field, a detailed description of Mn magnetization behavior can be constructed to help un- derstand the unique coupling found in this system. The magnetic multilayer, representative of the 15 differ- ent trilayers generated, consists of two single crystal 100 Å Fe25 Co75 alloy films20 deposited at 175 °C, separated by a thin 4.3 monolayer ML 8.7 Å Mn interlayer grown below room temperature (0 °C to minimize Mn clustering. The trilayer was grown on a thick 660 Å ZnSe 001 buffer layer and capped with a 30 Å Al film to prevent oxidation.11 Although these trilayers display a distribution of coupling angles between the two high quality FeCo alloy films, this trilayer structure displays both a large Mn MCD signal and a coupling angle of 90°. The ESM experiments were conducted at the NRL/NSLS U4B beamline located at the National Synchrotron Light Source.21­23 The element-specific magnetic hysteresis curves were determined by monitoring the partial x-ray fluorescence yield, as a function of the applied magnetic field, at the L3 edges of the relevant elements.17 A liquid-nitrogen cooled FIG. 1. Mx and My hysteresis loops for Fe where the points are electromagnet capable of reaching 3 kG is mounted with the data and the solid curves are provided as a guide to the eye. The the sample placed between its poles such that the applied arrows indicate increasing and decreasing magnetic field for the magnetic field is parallel to the sample surface. The samples Mx solid arrows and My dashed arrows magnetization curves. were kept at room temperature and the incidence angle of the Insets: measurement configurations for the determination of Mx and incoming photon beam, , was fixed at 60° from the surface My and a schematic view of the trilayer. normal with a degree of circular polarization of 75%. The ESM studies of the Fe, Co, and Mn showed that each Mx Fe , shows a gradual reversible decrease while My Fe film of the trilayer exhibits only an in-plane magnetization, increases indicating a near coherent rotation of the magneti- i.e., M Mxx Myy at all applied fields directed along the zation vector away from the applied field direction. The x axis. Briefly,19 in order to determine the Mx and My com- Mx Fe curve displays a single, irreversible transition from ponents of M, the intensity of the L3 absorption peak was b to c) at the critical fields of 55 Oe. The intensity of measured as a function of magnetic field for circularly polar- My Fe , on the other hand, is nearly unchanged during the ized photon beam directions in the x z and y z planes of abrupt change in Mx Fe , and displays a decidedly less the sample as indicated by k1 and k2 in the inset of Fig. 1; abrupt transition beginning at 100 Oe. z direction defined along surface normal and at 45° to these These two interdependent hysteresis loops can be com- two planes. The former two orientations yield the Mx and bined to form a single two-dimensional 2D parametric rep- My hysteresis curves since Mz 0), while the latter yields resentation of M Fe , shown in Fig. 2. This 2D representa- (Mx My)/ 2 and serves as an internal normalization of the tion can be superimposed on the principal crystallographic Mx and My curves. The lack of an Mz component was veri- directions of the surface to create an accurate portrayal of fied directly by measuring a null MCD signal for photon M Fe , both in magnitude and direction. As the magnetic beam incident normal to the film plane. field is scanned from 2.2 kOe to 2.2 kOe, M Fe Although our intent is to study the magnetic behavior of traverses the lower half of the loop in the counter-clockwise the Mn interlayer, comparison to the behavior of the ferro- ccw direction, as indicated by the arrows. For the opposite magnetic layers will prove to be instructive. In Fig. 1, the field variation, M Fe traverses the upper half of the loop. relative Mx and My magnetization curves for Fe are shown. Several positions on the loop have been marked with letter The magnetic field, which is always applied along the symbols which correspond to the points on Fig. 1 with the 110 direction (x axis , is varied between 2.2 kOe. These same labeling letter. normalized curves were obtained in less than 3 h and have By taking the 90° coupling between the two FeCo films been corrected for the small saturation effects.24 The larger into account, the 2D parametric loop can be used to model concentration of Co requires larger saturation effect correc- the behavior of the magnetic moments of each FeCo layer as tions, therefore, the Co magnetization curves, although iden- a function of the applied field shown above Fig. 2 . At the tical to the Fe curves, were not used. Note that, due to large maximum applied fields, the average moment is aligned with probing depth of the fluorescence yield measurements, the Fe the field direction, but since the field is far smaller than the magnetization curves shown in Fig. 1 represent the response coupling strength, the two individual films are each oriented of the average M, M Fe , of the Fe in both FeCo films 45° to the applied field point a). As the field intensity is which are coupled at 90° to each other, and not the indi- reduced, but prior to its reversing sign, the two individual vidual moments of a single film. As the field intensity is moments move away from the 100 directions toward the reduced, the average Fe moment in the x direction, 110 directions point b) indicating that, for the individual 53 BRIEF REPORTS 11 315 FIG. 3. Mx and My hysteresis loops for Mn points . The solid curves are the corresponding Fe loops generated by the inclusion of FIG. 2. 2D parametric representation of average Fe moment a rotation of 23° to the data of Fig. 1. generated from the data of Fig. 1. Also shown above the figure is a vector model describing the moment reversal process. similar MCD measurements of Fe/Cr/Fe trilayers showed that the Cr interlayer had a completely antiferromagnetic films, the 110 axes are magnetically easy. The magnetic structure and therefore no net moment.10 The difference be- configuration of the two films abruptly changes by 90° so tween the Cr and Mn magnetic behavior is all the more re- that, for two films of nearly equal thickness, the y component markable in that MCD studies of overlayers of Cr on bcc of the average magnetization is nearly conserved while the Fe 001 Ref. 10 and overlayers of Mn on bcc Co 001 x component changes sign. Unlike the nearly coherent rota- found identical overlayer behavior. For submonolayer cover- tion of the magnetization vectors in going from point a to age, the overlayer moments were antialigned to the substrate point b, the transition from point b to point c is accom- moment, and as the thickness of the overlayer was increased, plished by the formation and coalescence of magnetic do- the MCD signal was continually reduced. This is consistent mains, clearly expressed by the overall loss in magnitude of with the bulk type-I antiferromagnetic ordering which, for the average magnetization vector. the bcc 001 orientation, consists of alternating ferromag- As the field is further increased, a second, decidedly less netic sheets antialigned with the adjoining layers as is experi- abrupt transition occurs, again, mediated by domain dynam- mentally observed for the Cr 001 surface.25,26 Due to the ics. In this case the domain coalescence is quite slow, occur- statistical roughness of the overlayer, for coverage 3 ML, ring over a large field range (c to d), and a simultaneous this antiferromagnetic structure results in no net overlayer rotation of the average moment toward the applied field di- moment as measured by MCD. Although, there is no net rection occurs point d). A further field increase results in a moment for the Mn overlayer, the deposition of a second continued magnetization rotation until alignment with the ap- ferromagnetic layer reestablishes a net Mn moment within plied field is accomplished point e). As the field is swept in the trilayer film, indicating that although the Mn interlayer the opposite direction, the magnetization process is not re- has an antiferromagnetic basis, its actual structure is not versed i.e., a clockwise rotation , but rather M1 and M2 simple. continue this ccw rotational scheme due to a growth induced At first glance, the behavior of the Mn hysteresis loops 6 uniaxial magnetic anisotropy which makes one crystallo- h of acquisition seem much more complex than the Fe or Co graphic axis preferred.19,20 To assure that this ccw rotation loops, displaying multiple jumps in both the Mx and My was not due to a experimental artifacts, e.g., sample and/or magnetization. Only after constructing the 2D Mx vs My magnetic field misalignments, the measurements were re- parametrized curve of the M Mn and superimposing that of peated with the applied field at an angle of 10° away from the M Fe shown in Fig. 4 do we recognize that the seem- the 110 direction in an attempt to force the moment rota- ingly complicated Mn hysteresis loops are the result of a tion to reverse direction. The results were identical, experi- nearly rigid 23° rotation of M Mn with respect to M Fe . mentally confirming the presence of the uniaxial magnetic This rotation is most clearly demonstrated in Fig. 3 when we anisotropy. overlay the Mn data shown as the dots with the Fe Mx and It is interesting to contrast this behavior with that of the My data after a rotation of 23° solid line . All the elements Mn magnetization, shown in Fig. 3. The fact that the Mn of the magnetization behavior of the Mn are reproduced by interlayer is hysteretic indicates that the Mn possesses a net the Fe spectra with the inclusion of the rotation . The rota- ferromagnetic moment. This alone is quite unique because tion angle, however, is not completely field independent and 11 316 BRIEF REPORTS 53 canting continues through the Mn layers until the last layer is only slightly rotated away from the second Fe Co moment direction, which is at 90° with respect to the first. Because of this rotation, the magnetic moment of each Mn layer is not completely compensated for by the moment of the subse- quent Mn layer as occurs in the bulk antiferromagnetic structure resulting in a residual net Mn moment with an overall rotation with respect to the net Fe moment of the two FeCo films. If the trilayer structure is symmetric equal roughness at both interfaces , this simple model would predict a symmet- ric distribution of Mn moment directions for the Mn layers and cannot predict a residual Mn moment at 23° from the net Fe moment. But a realistic description of the structure of the trilayer must account for the nonintegral number of Mn lay- ers and the different degrees of roughness present in the two Mn/FeCo interfaces the second Mn interfaces are much rougher than the first . As an approximation, the trilayer can be straightforwardly modeled with a smooth Mn interface FIG. 4. 2D parametric representations of M Fe and M Mn followed by a rough Mn interface, with an average of a total clearly showing the canting of M Mn with respect to M Fe . of 4.5 layers of interlying Mn. This asymmetric structure results in a net Mn moment with a direction dependent on the varies from 25° at low fields to only 21° at high fields. The actual distribution of Mn thicknesses. Choosing a realistic 23° rotation is inconsistent with an interdiffusion of the Mn distribution of thicknesses generated from a Poisson distri- into the FeCo alloy layers. Such an interdiffusion results in bution which is accurate if no diffusion of the Mn occurs a an antialigned (180°) Mn moment to that of Fe Co . canting angle of 23° of the net Mn moment direction from Both the presence of a net Mn moment and its rotation the net Fe moment direction can be obtained. Furthermore, from the net Fe moment can be understood if the micro- this helical model predicts that as the coupling angle of the scopic magnetic structure of the Mn interlayer is an antifer- two enclosing ferromagnetic films is decreased at higher romagnetic helix structure, consistent with the proposed he- applied fields , the net Mn moment would both be reduced in lical structure for a system with an ordered antiferromagnetic its magnitude and its rotation angle, as is experimentally ob- interlayer bounded by ferromagnetic sheets.12 This structure served. Other micromagnetic descriptions of the Mn inter- is similar to the bulk antiferromagnetic ordering for the layer can be proposed ferrimagnetic Mn clusters, layer de- bcc 001 orientation described earlier but with the added fea- pendent Mn moment, etc. , but they must be consistent with ture of a slight rotation of the magnetic moment of each the observed hysteretic behavior found here. adjacent antialigned Mn layer, forming a helix. The first Mn One of the authors V.C. was supported by the Office of layer is slightly rotated from being antialigned to the first Fe Naval Research. Work done at National Synchrotron Light Co moment direction. The second Mn layer is antialigned Source was supported by DOE, under Contract No. DE- and slightly rotated 180 ) from the first Mn layer. This AC02-76CH00016. *Mailing address: NSLS Bldg. 725A/U4B, Brookhaven National 14 J. H. Tjeng et al., J. Magn. Magn. Mater. 109, 288 1992 . Laboratory, Upton, NY 11973. 15 J. G. Tobin et al., Phys. Rev. Lett. 68, 3642 1992 . 1 L. M. Falicov et al., J. Mater. Res. 5, 1299 1990 . 16 Y. Wu et al., Phys. Rev. Lett. 69, 2307 1992 . 2 P. Gru¨nberg, J. Appl. Phys. 57, 3673 1985 . 17 C. T. Chen et al., Phys. Rev. B 48, 642 1993 . 3 P. Gru¨nberg et al., Phys. Rev. Lett. 57, 2442 1986 . 18 Y. U. Idzerda et al., Appl. Phys. Lett. 64, 3503 1994 . 4 M. N. Baibich et al., Phys. Rev. Lett. 61, 2472 1988 . 19 V. Chakarian et al., Appl. Phys. Lett. 66, 3368 1995 . 5 B. Heinrich and J. F. Cochran, Adv. Phys. 42, 523 1993 . 20 C. J. Gutierrez et al., J. Appl. Phys. 61, 2476 1992 . 6 J. C. Slonczewski, Phys. Rev. Lett. 67, 3172 1991 . 21 C. T. Chen and F. Sette, Rev. Sci. Instrum. 60, 1616 1989 . 7 R. Jungblut et al., J. Appl. Phys. 70, 5923 1991 . 22 C. T. Chen, Rev. Sci. Instrum. 63, 1229 1992 . 8 J. Unguris et al., Phys. Rev. Lett. 69, 1125 1992 . 23 9 C. T. Chen et al., Phys. Rev. B 42, 7262 1990 . T. G. Walker et al., Phys. Rev. Lett. 69, 1121 1992 . 24 10 Y. U. Idzerda et al., Nucl. Instrum. Methods Phys. Res. A 347, Y. U. Idzerda et al., Phys. Rev. B 48, 4144 1993 . 11 134 1994 . M. E. Filipkowski et al., Phys. Rev. Lett. 75, 1847 1995 . 25 12 J. C. Slonczewski, J. Magn. Magn. Mater. 150, 13 1995 . L. E. Klebanoff et al., Phys. Rev. B 32, 1997 1985 . 26 13 J. F. Ankner et al. unpublished . R. Wiesendanger et al., Phys. Rev. B 65, 247 1990 .