APPLIED PHYSICS LETTERS VOLUME 77, NUMBER 14 2 OCTOBER 2000 Switching of the exchange bias in FeÕCr 211... double-superlattice structures S. G. E. te Velthuis,a) J. S. Jiang, and G. P. Felcher Argonne National Laboratory, Argonne, Illinois 60439 Received 28 June 2000; accepted for publication 2 August 2000 The reversal of the direction of the exchange bias in a ``double-superlattice'' system which consists of an Fe/Cr antiferromagnetic AF superlattice which is ferromagnetically coupled with an Fe/Cr ferromagnetic F superlattice through a Cr spacer layer, is observed. Magnetometry and polarized neutron reflectometry show that a switch in the bias direction occurs at a field 447 Oe well below the field 14 kOe necessary to saturate the AF superlattice and well below the field 2 kOe where the AF superlattice initiates a spin­flop transition. The switching of the exchange bias cannot be explained in terms of a model of uniform rotation, but rather by breakdown into domains and reversal of the AF layers. The transparency of magnetic behavior of the double superlattice may be useful in understanding the behavior of traditional exchange bias systems. © 2000 American Institute of Physics. S0003-6951 00 00240-0 Exchange bias, first discovered in 1956 by Meiklejohn layers are again collinear, and are aligned along the easy axis and Bean1 in assemblies of Co­CoO particles, refers to the and H. However, for modest negative fields, the FM layers occurrence of a unidirectional magnetic anisotropy that turn toward the field while the magnetic structure of the AF manifests itself in shifted hysteresis loops for coupled ferro- layers remains unaltered. The exchange bias for descending magnet F ­antiferromagnet AF field cooled through the fields of the order of 40 Oe has the value expected4 for the Ne´el temperature TN of the AF. Once the system is cooled, ferromagnetic coupling between the two superlattices. The the exchange bias field is frozen. Applying laboratory fields goal of the present experiment was to determine at which in the opposite direction does not change the orientation of field the bias switches. the bias. The only way to reverse the bias is to warm up the A series of magneto-optic Kerr effect MOKE minor sample above the Ne´el temperature or, more exactly, the loops were taken between the turning points Hmax and Hmin , blocking temperature. No systematic study has been made of where Hmax was greater than the saturation field, while Hmin applying increasingly high field just below the blocking tem- was increased in magnitude from 45 Oe to 14 kOe in perature. Some experiments, however, are proceeding along steps of 5 Oe. Figure 2 gives the measured MOKE signal as this direction.2 a function of field for two critical values of Hmin . The figure We studied the reversal of the exchange bias in a shows that for Hmin 406 Oe, the loop is biased around ``double superlattice.'' This system consists of one F super- HE 38.5 Oe, and the magnetization in the F superlat- lattice of Fe and Cr 211 layers, and one AF superlattice tice is reversed back to its original orientation at H HE obtained similarly but with a different Cr thickness tCr since Hc 33.6 Oe. This loop is representative for all loops the interlayer exchange coupling oscillates with tCr . The measured with values of Hmin between 34 and 406 Oe. coupling between the AF and F superlattices is governed by However, for Hmin 447 Oe to 14 kOe, the F superlattice the value of tCr between the two superlattices. The sample magnetization does not reverse back until H HE Hc has a layer sequence Fe(50 Å)/Cr(20 Å) F5 / Fe(14 Å)/ Cr(11 Å) AF20 with tCr 20 Å between the F and AF super- lattices, to provide a ferromagnetic intersuperlattice cou- pling. A uniaxial anisotropy is introduced by epitaxially growing the sample via dc magnetron sputtering onto single- crystal MgO 110 substrates.3 This artificial exchange bias system was constructed in order to attain a predictable and controllable exchange bias. We recall here the main mag- netic properties,4­6 and show how the switching of the ex- change bias was obtained. Figure 1 gives the magnetization of the sample at room temperature with the field H applied along the easy axis. Above 14 kOe the magnetic moments of all layers in both superlattices are aligned with H in Fig. 1 the magnetization is normalized to this value . For descending H the magneti- zation decreases as the Fe layers in the AF superlattice first enter a spin-flop state,7,8 and then become antiferromagneti- FIG. 1. The normalized magnetization curve measured with a superconduct- cally aligned. Below 2 kOe the magnetic moments of all ing quantum interference device magnetometer and with decreasing field. The sets of four arrows indicate the magnetic orientation in the F superlat- tice top two and in the AF superlattice bottom two , at different stages of a Electronic mail: tevelthuis@anl.gov the magnetization curve. 0003-6951/2000/77(14)/2222/3/$17.00 2222 © 2000 American Institute of Physics Downloaded 28 Feb 2001 to 148.6.169.65. Redistribution subject to AIP copyright, see http://ojps.aip.org/aplo/aplcpyrts.html Appl. Phys. Lett., Vol. 77, No. 14, 2 October 2000 te Velthuis, Jiang, and Felcher 2223 FIG. 2. Minor magnetization curves measured by means of MOKE. Both curves are measured for descending field from H 14 kOe down to Hmin , with a H FIG. 3. Measured and calculated polarized neutron reflectivity for a H min 406 and b 447 Oe, and then ascending back up to H 14 kOe. The sets of four arrows indicate the magnetic orientation in the F 285 and b 940 Oe. The measured reflectivity is given for neutron superlattice top two and in the AF superlattice bottom two , at different with an initial and reflected polarization parallel to the applied field (R ). stages of the magnetization curve. The calculations are exactly the same as presented in Ref. 5. Since it is assumed all moments are collinear, R R . 40.7 Oe, indicating a change in the direction of the bias, which implies that the AF superlattice has reversed its orien- successively, after saturation at 15 kOe. All measurements tation. At intermediate values of H were performed at room temperature. min , the loop was either the same as for H The spin-flip reflectivities (R ,R ) are equal to 0, min 406 Oe, or it was an average of the two loops. within the resolution of the experiment, indicating that all The reversal of the AF superlattice was directly deter- magnetizations are collinear with the applied field. Figure 3 mined by polarized neutron reflectivity PNR . The measure- gives the measured reflectivities R as a function of mo- ments were performed at Argonne's intense pulsed neutron mentum transfer (qz) for both fields. The Bragg reflection at source. The spin-dependent neutron reflectivity gives infor- qz 0.09 Å 1 arises from interference between the Fe layers mation about the magnetic and structural profile perpendicu- in the F superlattice. The reflection at 0.12 Å 1 arises from the interference between the Fe layers within the AF super- lar to the surface. R and R denote reflectivities for neu- lattice, and corresponds to a periodicity twice that of the trons polarized parallel and antiparallel to H, respectively. If structural ordering. In the q region between total reflection polarization analysis of the reflected beam is performed, four and the Bragg reflections there are clear differences in R intensities are measured, two nonspin flip: R , R , and for the two applied fields. In Ref. 5, the measured reflectivi- two spin flip: R , R , reflectivities, where R R ties could be described by calculations based on the two R and R R R . If the magnetization of all configurations of collinear moments the magnetization in layers is collinear to H, then R R 0. In this case R the F superlattice is either parallel or antiparallel to the top is an optical transform of n(z) m(z), where n is a depth- layers in the AF superlattice . In Fig. 3 these same calcula- dependent nuclear scattering amplitude, m is the depth- tions are presented along with the new data. For H dependent magnetization, and R is an optical transform of 285 Oe, the magnetization of the top layer of the AF n(z) m(z). By alternatively measuring with neutrons in ei- superlattice is antiparallel to that of the F superlattice and to ther spin state, the magnitude and direction of the layer-by- that of the applied field, just as was the case for H layer magnetization can be determined. If R R 0, 72 Oe. However, the measured reflectivity at H there are components of the magnetization perpendicular to 940 Oe agrees with the configuration where the magne- H. tization of the top layer of the AF superlattice is parallel to It has been shown4,5 that the polarized neutron reflectiv- that of the F superlattice and that of the applied field. In both ity for the two magnetic configurations measured at H cases the antiferromagnetic arrangement in the AF superlat- 166 and 72 Oe after saturation in H 30 kOe) illustrate tice is maintained. the reversal of the magnetization in the F superlattice. Here A first model to understand the mechanism of the bias PNR measurements, performed with polarization analysis, reversal is that of uniform rotation of the magnetization in are presented that were made with H 285 and 940 Oe, the AF superlattice. After the magnetization of the F has Downloaded 28 Feb 2001 to 148.6.169.65. Redistribution subject to AIP copyright, see http://ojps.aip.org/aplo/aplcpyrts.html 2224 Appl. Phys. Lett., Vol. 77, No. 14, 2 October 2000 te Velthuis, Jiang, and Felcher been switched and is again aligned with the applied field, the the top layer of the AF superlattice. In the second case, the magnetization of the first layer of the AF superlattice and surface spin flop will start on the free side of the AF super- that of the adjacent F superlattice are opposite. Although the lattice, which means the F superlattice has little influence. state in which they are parallel has lower energy, the transi- The calculated surface spin-flop field is approximately tion has to take place through the spin-flopped state, where that inferred from the magnetization measurements Fig. 1 . the moments are roughly at 90° with the field, which has The reversal of the AF superlattice magnetization and the considerably higher energy. However, if the applied field is switch in the bias direction is observed at a much lower field increased, always in the negative direction, the energy of the than calculated. Therefore there must be another mechanism two extremal states, with the magnetization aligned along the that is driving this transition at this low field of 447 Oe. field, stay the same, while that of the spin-flopped state de- The fact that a uniform rotation model does not explain creases until, at a certain field, the switching of the bias is the switch of the AF is not totally unexpected. The width of permitted. Only at higher fields, the AF structure settles in a the FM hysteresis loop is only 5 Oe, which is much spin-flopped state. Although qualitatively this model ex- smaller than the anisotropy field of the ferromagnetic super- plains the sequence of observed magnetic transitions, its va- lattice. This indicates that in the double superlattices the lidity has to be tested on the basis of a quantitative compari- magnetization reversal of the F superlattice is not by coher- son. ent rotation, but rather by nucleation and growth of reverse The energy of the AF superlattice without the F super- magnetic domains.4,10 The present experiment indicates that lattice can be written as9 a similar magnetization reversal takes place for the AF su- perlattice layers. A scenario of nucleation and growth of re- 1 N 1 N N E verse domains in exchange bias systems is discussed briefly 2 cos i i 1 cos2 i h cos i . 1 i 1 2 i 1 i 1 in Ref. 11. Furthermore, Stiles and McMichael12 suggest that Here the energy is normalized on g AF AF in the case of AF domains of limited size, it is possible for BHE S, where HE is the exchange field between the layers of the AF superlattice. the moments to rotate out of the plane, decreasing the field The first term accounts for the exchange interaction between for the reversal. adjacent layers, the second is the anisotropy energy with In summary, we have shown the reversal of the direction H AF of the exchange bias in a double-superlattice system. The A/HE as the normalized uniaxial anisotropy field, and the third is the Zeeman energy with h H/HAF switch in the bias direction is the result of the reversal of the E as the nor- malized applied field. magnetization in the AF superlattice and takes place via do- i is the angle between the magnetiza- tion of layer i in the superlattice and the easy axis. It is main nucleation and growth at about H 447 Oe, well be- assumed that H is directed along the easy axis. low the surface spin-flop transition of the AF superlattice. An AF system becomes unstable,9 coinciding with the One of the descriptions of the exchange bias in traditional onset of the surface spin­flop transition, when the determi- exchange coupled systems involves the existence of laterally nant of the matrix m composed of the second derivative of limited antiferromagnetic domains.13,14 Our work seems to the energy with respect to indicate that, in adequate magnetic fields, it should be pos- i of layer i(mi j d2E/d id j), with both sublattices aligned in the field direction, becomes sible to switch those domains. The shape of the hysteresis zero. loop will reflect the distribution of effective domain bias For the double superlattice a term is added to the energy fields with their population. of the AF superlattice: E* E /2 cos( F 1), where F This work was supported by US DOE, BES-MS Con- is the angle between the magnetization in the F superlattice tract No. W-31-109-ENG-38. and the easy axis and is the exchange field across the 1 interface between the AF and F superlattice normalized on W. H. Meiklejohn and C. P. Bean, Phys. Rev. 105, 904 1957 . 2 J. Nogue´s, L. Morellon, C. Leighton, M. R. Ibarra, and I. K. Schuller, the exchange field in the AF superlattice. Again, the first Phys. Rev. B 61, R6455 2000 . zero point of the determinant of the matrix of second deriva- 3 E. E. Fullerton, M. J. Conover, J. E. Mattson, C. H. Sowers, and S. D. tives m * is used as a criterion for the instability of the col- Bader, Phys. Rev. B 48, 15755 1993 ; J. Appl. Phys. 75, 6461 1994 . 4 linear AF structure. The determinant is now calculated nu- J. S. Jiang, G. P. Felcher, A. Inomata, R. Goyette, C. Nelson, and S. D. Bader, Phys. Rev. B 61, 9653 2000 . merically for increasing magnetic fields, using values for , 5 S. G. E. te Velthuis, G. P. Felcher, J. S. Jiang, A. Inomata, C. S. Nelson, , and h obtained from Ref. 3. A. Berger, and S. D. Bader, Appl. Phys. Lett. 75, 4174 1999 . Starting with a top layer magnetization of the AF super- 6 L. Lazar, J. S. Jiang, G. P. Felcher, A. Inomata, and S. D. Bader unpub- lattice opposite to that of the F superlattice and the field, the lished . 7 R. W. Wang, D. I. Mills, E. E. Fullerton, J. E. Mattson, and S. D. Bader, determinant of m * becomes zero at H* 2169 Oe. While if Phys. Rev. Lett. 72, 920 1994 . the top layer is along the F superlattice magnetization then 8 S. Rakhmanova, D. L. Mills, and E. E. Fullerton, Phys. Rev. B 57, 476 H* 2470 Oe. The latter value is quite close to that obtained 1998 . 9 for the surface spin flop of the AF superlattice alone- A. L. Dantas and A. S. Carric¸o, Phys. Rev. B 59, 1223 1999 . 10 J. S. Jiang, G. P. Felcher, A. Inomata, R. Goyette, C. Nelson, and S. D. uncoupled to the ferromagnet. The difference between these Bader, J. Vac. Sci. Technol. A 18, 1264 2000 . two conditions may seem obvious, since the surface spin flop 11 T. C. Schulthess and W. H. Bulter, Phys. Rev. Lett. 81, 4516 1998 . will start on the side that is antiparallel to the field. In the 12 M. D. Stiles and R. D. McMichael, Phys. Rev. B 59, 3722 1999 . 13 first case this side is adjacent to the F superlattice, which due A. P. Malozemoff, Phys. Rev. B 35, 3679 1987 . 14 K. Takano, R. H. Kodama, A. E. Berkowitz, W. Cao, and G. Thomas, to the exchange interaction is increasing the effective field on Phys. Rev. Lett. 79, 1130 1997 . Downloaded 28 Feb 2001 to 148.6.169.65. Redistribution subject to AIP copyright, see http://ojps.aip.org/aplo/aplcpyrts.html