PHYSICAL REVIEW B VOLUME 61, NUMBER 14 1 APRIL 2000-II Exchange-bias effect in FeÕCr 211... double superlattice structures J. S. Jiang, G. P. Felcher, A. Inomata, R. Goyette, C. Nelson, and S. D. Bader Argonne National Laboratory, Argonne, Illinois 60439 Received 29 November 1999 Shifted hysteresis loops characteristic of the exchange-bias effect between a ferromagnet F and an anti- ferromagnet AF are demonstrated in ``double-superlattice'' structures. Utilizing the well-established oscilla- tory interlayer exchange coupling in Fe/Cr, we have constructed Fe/Cr AF/Cr/ Fe/Cr F double superlattices where Fe/Cr superlattices with appropriate Cr-spacer thickness represent the F and the AF. The double super- lattices are 211 -oriented epitaxial films sputter grown on single-crystal MgO 110 substrates. The AF/F interface is coherent compared to conventional exchange-bias interfaces consisting of dissimilar AF and F phases. Magnetization results show that AF/F exchange coupling affects the nucleation of reverse magnetic domains, and that the magnitude of the exchange-bias field is given directly by the classic formula for collinear spin structures. The collinear spin distribution is confirmed by polarized neutron reflectivity. The exchange-bias effect is a well-known but still unre- compensated spins for CoO-Permalloy bilayers at the AF/F solved phenomenon.1 Discovered in 1956 by Meiklejohn and interfaces. However, experimental observations do not al- Bean in Co-CoO particle systems, it refers to the occurrence ways agree on the effect of interfacial disorder, as both of a unidirectional magnetic anisotropy that manifests itself increased6 and decreased7 exchange bias due to interfacial in strikingly shifted hysteresis loops for coupled ferromagnet disorder have been reported. F /antiferromagnet AF systems cooled through the Nee´l To our knowledge, to date there has not been an experi- temperature in the presence of a magnetic field.2 However, mental study that can ascertain the interfacial atomic and despite extensive research effort on various AF/F systems,3­8 spin structures in an exchange-bias system. Suitable atomi- and important technological applications such as magnetore- cally flat surfaces do not tend to exist for such studies; there sistive read heads that utilize exchange bias,9 a clear under- is always at least atomic-scale roughness at the AF/F inter- standing of the microscopic origin of the phenomenon has face. Since the interface is buried and therefore inaccessible yet to emerge. to most surface probes, the AF spin structure at the interface Since its discovery exchange biasing has been interpreted is often assumed to be the same as that of the bulk, while in as the result of the exchange interaction at AF/F interfaces: reality there could be a spin rearrangement at the the magnitude of the exchange-bias field is given by balanc- interface.5­7 The reduced lateral coherence due to interfacial ing the gain in Zeeman energy with the energy cost of inter- roughness or random AF domains renders scattering experi- facial exchange when the ferromagnet reverses its magneti- ments ineffective.17 In view of these unresolved issues, it is zation. In the earliest model10 it was assumed that the F and beneficial to construct a system where the exchange-bias ef- AF spin structures are rigid, and that the AF/F interface is fect can be realized and examined in detail with minimal perfectly flat and uncompensated. However, such an intuitive materials-related complexities. picture gives estimated exchange-bias fields that are nearly In this paper, we demonstrate the exchange-bias effect in two orders of magnitude larger than what is typically ob- Fe/Cr double superlattice structures. The exchange coupling served experimentally. Two models have been proposed to of ferromagnetic transition-metal layers across a nonmag- address this difficulty: the domain wall model of Mauri netic spacer allows for the creation of magnetic structures et al.11 in which an AF domain wall parallel to the interface with desired magnetic configurations.18 The interlayer- formed during the magnetization reversal of the ferromagnet exchange coupling between Fe layers across a Cr spacer is reduces the interfacial energy, and the random-field model of oscillatory, with a ``long'' period of 18 Å in Cr thickness.19 Malozemoff12 in which interfacial disorder such as rough- Thus, a double superlattice structure with the configuration ness is treated as a random field giving rise to in-plane AF Fe/Cr AF/Cr/ Fe/Cr F, where the superscripts denote anti- domains and a reduced but statistically nonvanishing interfa- ferromagnetic and ferromagnetic coupling within the base cial energy for a finite system. Extending the domain wall Fe/Cr superlattices, constitutes an exchange-bias system with model of Mauri et al., Koon13 was able to account for the the center Cr layer delineating the AF/F interface. The req- exchange-bias effect observed in fully compensated AF/F in- uisite magnetic anisotropy in the AF for exchange bias is terfaces with perpendicular i.e., spin-flop coupling.14 How- represented by a growth-induced uniaxial anisotropy. It has ever, by solving the full equation of motion during magnetic been shown that 211 -oriented Fe/Cr superlattices epitaxi- reversal, Schulthess and Butler15 showed that spin-flop cou- ally grown on the MgO 110 substrates have a uniaxial, in- pling alone leads to a uniaxial rather than a unidirectional plane, surface magnetic anisotropy, with the easy axis along anisotropy. They further argued that domain wall pinning by the Fe/Cr 01¯1 direction.20 The AF/F interfacial coupling, interfacial defects is necessary to establish exchange bias. i.e., the intersuperlattice coupling, in the double superlattice The Malozemoff theory was corroborated by Takano et al.16 system is governed by the thickness of the center Cr layer. who showed the relation between exchange bias and net un- Since the 18-Å period of the interlayer coupling is relatively 0163-1829/2000/61 14 /9653 4 /$15.00 PRB 61 9653 ©2000 The American Physical Society 9654 J. S. JIANG et al. PRB 61 FIG. 1. Room-temperature magnetization curve of an FIG. 2. Minor hysteresis loops of the Fe/Cr double superlattice Fe(14 Å )/Cr(11 Å ) of Fig. 1 after alignment at 20 kOe. The solid line is measured by 20 /Cr(9 Å )/ Fe(50 Å )/Cr(20 Å) 5 double superlattice. The arrows mark spin-flop transitions. Inset: Schematic SQUID magnetometry and the dashed line by means of the illustration of a double superlattice structure. The dark layers rep- magneto-optic Kerr effect. The magnetization is normalized to the resent magnetic layers. full saturation value. long compared to the range of the interatomic exchange oc- tice, while the AF superlattice contributes zero net magneti- curring at conventional AF/F interfaces, the exchange cou- zation. The kinks in magnetization marked by arrows pling between the AF and F superlattices in our double su- identify the spin-flop transitions in the AF superlattice.22 perlattice structures is less sensitive to roughness and can be With increasing field, the Fe moments in the AF rotate from considered uniform across the interface. The double super- a spin-flopped state toward parallel alignment and the mag- lattice structure is different from the spin valves where a netization gradually increases. The field values for the spin- synthetic antiferromagnet replaces the pinned layer,21 be- flop transition 2 kOe and for saturation 14 kOe are iden- cause in those spin valves the sensing layer is not coupled to tical to those of the AF superlattices in Ref. 20 with the same the synthetic antiferromagnet. layer thicknesses. Of present interest are double superlattice structures with In a conventional AF/F exchange-bias system, cooling in the AF superlattice having a configuration Fe(14 Å)/ a field through the Nee´l temperature of the AF is required to Cr(11 Å) establish a unidirectional magnetic anisotropy. However, this 20 , while the F superlattice is Fe(50 Å)/ Cr(20 Å) is not necessary for our AF/F double superlattice structures. n with n F F 2, 3, 5, and 10. The numbers inside Figure 2 shows a minor hysteresis loop of the same double the parentheses denote the layer thicknesses, and the sub- superlattice measured in fields between 200 Oe, after a scripts denote the number of repetitions of the Fe/Cr bilayer large field of 20 kOe had been applied to align all Fe layers unit. The Cr layer between the AF and F superlattices is in both F and AF superlattices. The minor loop is displaced 20-Å thick and gives rise to ferromagnetic intersuperlattice from zero in the negative field direction by 34.4 Oe.23 The coupling. The Fe/Cr double superlattices were grown via dc shifted hysteresis loop is indicative of the unidirectional an- magnetron sputtering onto single-crystal MgO 110 sub- isotropy. The aligning field breaks the symmetry and leaves strates. A 200-Å Cr buffer layer was first deposited at 400 °C the interfacial Fe layer of the AF superlattice necessarily to establish epitaxy with the substrate. The double superlat- parallel to the alignment direction. The exchange interaction tice structure was then grown at 100 °C, followed by a 50-Å between the F superlattice and the interfacial Fe layer then Cr cover layer. Samples with only a single AF or F Fe/Cr causes the hysteresis loop of the F superlattice to shift to- superlattice were also prepared similarly for benchmarking. ward the negative direction. Note that the width of the hys- The structures were characterized by x-ray diffraction using teresis loops is only 10 Oe, which is much smaller than Cu K radiation. The crystal structure is bcc. For the single the anisotropy field. This indicates that the magnetization Fe/Cr superlattices, high-angle superlattice diffraction peaks reversal of the F superlattice is not by coherent rotation, but up to third order were observed. Asymmetric azimuthal rather by nucleation and growth of reverse magnetic do- scans confirmed the expected in-plane epitaxial relations: mains. Also shown in Fig. 2 is the minor loop measured Fe/Cr 01¯1 MgO 001 and Fe/Cr 1¯11 MgO 11¯0 . The using the magneto-optic Kerr effect. Since the Kerr effect is anisotropy constant determined from the hard-axis magneti- sensitive to the magnetization on the scale of the optical zation curves agrees with the published value KS penetration depth ( 200 Å ), which is roughly the thickness 0.06 erg/cm2.20 The anisotropy fields are 1.6 kOe for 14-Å of the F superlattice, the single-stepped switching in Kerr Fe layers, and 450 Oe for 50-Å Fe layers. intensity indicates that all of the Fe layers, the F superlattice, Shown in Fig. 1 is the room-temperature magnetization reverse their magnetization simultaneously. The sharpness of curve of a double superlattice with nF 5 measured by su- the switching indicates pinning-free domain wall motion. perconducting quantum interference device SQUID magne- Therefore, the exchange coupling manifests itself as a bias tometry along the easy direction. The magnetization is nor- field at the onset of domain reversal. It is worth noting that malized with respect to the full saturation value. Since the Fe the models of Refs. 11, 12, and 13 imply a coherent rotation moment in the F superlattice comprises 47% of the total of the F magnetization, and that the scenario of nucleation moment, the transition between 0.47 and 0.47 in the nor- and growth of reverse domains in exchange-bias systems is malized magnetization in low field represents the F superlat- discussed only indirectly in Ref. 15. PRB 61 EXCHANGE-BIAS EFFECT IN Fe/Cr 211 DOUBLE . . . 9655 FIG. 3. The exchange-bias field HE as a function of the number of Fe layers in the F superlattice, nF . The solid curve is the calcu- lated exchange-bias field as described in the text. In Fig. 3, the values of the exchange-bias field HE for several double superlattices are shown as a function of nF , the number of Fe layers in the F superlattices. With increas- ing nF , HE decreases monotonically. The classic formula for FIG. 5. Top: Spin asymmetry P for the double superlattice of the magnitude of the exchange-bias field as applied to sys- Fig. 1 in a descending field of H 21 Oe. Bottom: P for the tems of collinear spin structures is same sample in an ascending field H 35 Oe. The curves are calculations assuming a collinear spin profile. The diagrams illus- HE Jint /tFMF , 1 trate the spin configurations near the AF/F interface. The parallel where J arrows indicate the magnetization directions of the Fe layers in the int is the interfacial exchange-coupling energy, and t F superlattice and the antiparallel arrows indicate those in the AF F and M F are the thickness and magnetization of the ferro- magnet, respectively. In the present Fe/Cr AF/Cr/ Fe/Cr F superlattice. double superlattices, the equivalent interfacial exchange in- teraction is the coupling across the center Cr layer, and the highly ideal AF/F interfaces in double superlattices per- t F F mit unambiguous determination of Jint . Note that whereas FM F nFdFeM Fe , where dFe is the Fe layer thickness in the F superlattice, and M the classic formula overestimates the exchange-bias field by Fe is the saturation magnetization of Fe. Using the previously determined interlayer coupling energy two orders of magnitude in conventional AF/F systems, the across a 20-Å Cr spacer layer J data points in Fig. 3 are well described by Eq. 1 with the F 0.07 erg/cm2 Ref. 20 , (J exact value for Jint . However, since exchange biasing occurs int 2JF , since JF was defined as the coupling strength per Fe layer in a bilayer structure24 , dF at the onset of domain reversal, the quantitative agreement Fe 50 Å , and M Fe between measured and calculated exchange-bias fields in the 1700 emu/cm3, we have calculated the expected double superlattices advocates that the region of significance exchange-bias field from Eq. 1 and plotted it as the solid for exchange bias includes only the volume of the nucleated curve in Fig. 3. Such a comparison is possible only because reverse domain in the F and the part of the AF that is ex- change coupled to it, rather than the entire volume of the AF/F system. Polarized neutron reflectivity PNR measurements were taken in order to determine the layer-by-layer magnetization of the double superlattice, both in size and orientation.25 The momentum transfer of the neutron perpendicular to the sur- face qz 4 sin / , where is the angle of the neutron beam with respect to the surface plane, and is the neutron wavelength. As a rule of thumb the spatial resolution is the inverse of the maximum value of qz that has been measured. The PNR measurements were taken at the ``POSY I'' instru- ment at Argonne's Pulsed Neutron source. The sample had nF 5 ferromagnetic layers and a surface area of 6 6 mm2. Two scans were taken at room temperature in the two branches of a minor loop after aligning the sample in a field of 20 kOe. They were, respectively, in a field of 21 Oe, with the ferromagnet magnetized in the direction FIG. 4. Measured and calculated polarized neutron reflectivity of the aligning field; and 35 Oe after cycling to for the double superlattice of Fig. 1 in a field H 35 Oe for 120 Oe, where the ferromagnet is magnetized in the oppo- neutrons with spin parallel to H full points/full line and antiparal- site direction. Figure 4 shows the reflectivities for neutrons lel to H open points/dashed line . polarized parallel (R ) and antiparallel (R ) to the applied 9656 J. S. JIANG et al. PRB 61 magnetic field H 35 Oe. It is interesting to describe the evant magnetic and nuclear amplitudes are in quadrature , physical significance of the main features of the spectra. The this result provides the most direct confirmation hitherto ob- strong spin dependence of the reflectivity indicates the pres- tained of a collinear spin configuration in an exchange-bias ence of large magnetic induction fields in the sample, paral- system. lel to the applied field. At the left side of Fig. 4, the critical In conclusion, we have demonstrated exchange-bias be- angle is characteristic of the MgO substrate, while at the havior in double-superlattice structures that utilizes oscilla- right side, the broad ferromagnetic peak appears the first AF tory interlayer exchange coupling. The exchange-bias field peak is out of the qz range presented here . The most pro- agrees quantitatively with the classical formula and polarized nounced interference fringes of the polarized neutrons cor- neutron reflectivity measurements confirm the collinear spin respond to the total thickness of the F superlattice. Also in- distribution. While there is no straightforward way to char- dicated in Fig. 4 is the reflectivity calculated assuming a acterize and manipulate the interfacial coupling in conven- collinear distribution of the spins of the F and the AF tional exchange-bias systems, our double superlattice struc- components-with the magnetization of the first AF layer tures have highly ideal AF/F interfaces. The flexibility in opposite to that of the F superlattice. The spin asymmetry configuration, and tunable coupling strength and magnetic P (R R )/(R R ) is shown in Fig. 5 for the two magnetization branches. The measurements show a pro- anisotropy offered by the double superlattice structures nounced difference at q should create new opportunities to elucidate the underlying z 0.05. This is the region where the calculated asymmetries are most sensitive to the reversal of physics of the exchange-bias phenomenon. the magnetization in the F superlattice. Since noncollinear This work was supported by U.S. DOE, BES-MS, Con- configurations do not contribute to the asymmetry the rel- tract No. 31-109-ENG-38. 1 J. Nogue´s and I. K. Schuller, J. Magn. Magn. Mater. 192, 203 15 T. C. Schulthess and W. H. Butler, Phys. Rev. Lett. 81, 4516 1999 , and references therein. 1998 . 2 W. H. Meiklejohn and C. P. Bean, Phys. Rev. 105, 904 1957 . 16 K. Takano, R. H. Kodama, A. E. Berkowitz, W. Cao, and G. 3 J. S. Kouvel, J. Phys. Chem. Solids 16, 132 1960 . Thomas, Phys. Rev. Lett. 79, 1130 1997 . 4 A. E. Berkowitz and J. H. Greiner, J. Appl. Phys. 36, 3330 17 S. S. P. Parkin, V. R. Deline, R. O. Hilleke, and G. P. Felcher, 1965 . Phys. Rev. B 42, 10 583 1990 ; G. P. Felcher, Y. Y. Huang, M. 5 R. Jungblut, R. Coehoorn, M. T. Johnson, Ch. Sauer, P. J. van der Carey, and A. E. Berkowitz, J. Magn. Magn. Mater. 121, 105 Zaag, A. R. Ball, Th. G. S. M. Rijks, J. aan de Stegge, and A. 1993 . 18 Reinders, J. Magn. Magn. Mater. 148, 300 1995 . S. S. P. Parkin and D. Mauri, Phys. Rev. B 44, 7131 1991 . 19 6 T. J. Moran, J. M. Gallego, and I. K. Schuller, J. Appl. Phys. 78, S. S. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, 2304 1887 1995 . 1990 ; J. Unguris, R. J. Celotta, and D. T. Pierce, ibid. 67, 140 7 J. Nogue´s, D. Lederman, T. J. Moran, I. K. Schuller, and K. V. 1991 . 20 E. E. Fullerton, M. J. Conover, J. E. Mattson, C. H. Sowers, and Rao, Appl. Phys. Lett. 68, 3186 1996 . 8 S. D. Bader, Phys. Rev. B 48, 15 755 1993 ; J. Appl. Phys. 75, N. J. Go¨kemeijer, T. Ambrose, and C. L. Chien, Phys. Rev. Lett. 6461 1994 . 79, 4270 1997 . 21 V. Speriosu, B. A. Gurney, D. R. Wilhoit, and L. B. Brown un- 9 C. Tang, J. Appl. Phys. 55, 2226 1984 ; B. Dieny, V. S. Spe- published . riosu, S. S. P. Parkin, D. R. Wilhoit, and D. Mauri, Phys. Rev. B 22 R. W. Wang, D. L. Mills, E. E. Fullerton, J. E. Mattson, and S. D. 43, 1297 1991 . Bader, Phys. Rev. Lett. 72, 920 1994 . 10 W. H. Meiklejohn, J. Appl. Phys. 33, 1328 1962 . 23 To correct for the remanent field due to trapped flux in the super- 11 D. Mauri, H. C. Siegmann, P. S. Bagus, and E. Kay, J. Appl. conducting magnet of the SQUID magnetometer, a second mi- Phys. 62, 3047 1987 . nor loop was measured with the sample rotated by 180°. The 12 A. P. Malozemoff, Phys. Rev. B 35, 3679 1987 . midpoint between the centers of the two minor loops defines the 13 N. C. Koon, Phys. Rev. Lett. 78, 4865 1997 . zero-field point. 14 T. J. Moran, J. Nogue´s, D. Lederman, and I. K. Schuller, Appl. 24 W. Folkerts, J. Magn. Magn. Mater. 94, 302 1991 . Phys. Lett. 72, 617 1998 . 25 H. Zabel, Physica B 198, 156 1994 .