JOURNAL OF APPLIED PHYSICS VOLUME 85, NUMBER 8 15 APRIL 1999 Biquadratic coupling in sputtered Fe/Cr/Fe still in need of a new mechanism S. M. Rezende,a) C. Chesman, M. A. Lucena, M. C. de Moura, A. Azevedo, and F. M. de Aguiar Departamento de Fi´sica, Universidade Federal de Pernambuco, Recife PE 50670-901, Brazil S. S. P. Parkin IBM Almaden Research Center, San Jose, California 95120-6099 The bilinear (J1) and biquadratic (J2) exchange coupling constants were measured in sputtered trilayers of 100 Fe 40 Å /Cr s /Fe 40 Å for several Cr spacer layer thicknesses in the range s 8­35 Å and as a function of temperature T, using magneto-optical Kerr effect magnetometry, Brillouin light scattering, and ferromagnetic resonance. In the samples in the range s 8­13 Å, corresponding to the first antiferromagnetic peak of J1 , J2 follows J1 with a room temperature ratio J2 /J1 0.1, while in the range 25­35 Å, corresponding to the second antiferromagnetic peak, J2 also follows J1 but with a much larger ratio J2 /J1 1. This result, as well as the temperature dependence of J2 in all samples but the one with s 15 Å, cannot be explained by any of the intrinsic or extrinsic mechanisms that have been proposed for the origin of the biquadratic exchange coupling in Fe/Cr/Fe. © 1999 American Institute of Physics. S0021-8979 99 68008-2 INTRODUCTION some attributed to intrinsic properties of the spacer layer and others to extrinsic factors, such as the presence of impurities Magnetic multilayers, consisting of stacks of ferromag- or interface roughness. Intrinsic mechanisms seem not to ac- netic layers separated by nonmagnetic metallic layers, have count for the observed coupling strengths and signs, nor for attracted considerable attention due to their unique physical its temperature dependence.18 On the other hand, if the properties and potential for technological applications. Many mechanisms based on extrinsic factors prevails in the BEC, multilayer systems exhibit a coupling between the magnetic one expects the value of the coupling constant J2 to be quite layers mediated by the nonmagnetic spacers, which oscillates sensitive to details of the sample preparation conditions. In- periodically between ferromagnetic FM and antiferromag- deed, there is a considerable spread in the values of J2 mea- netic AFM as the spacer-layer thickness varies in the range sured by different groups for nominally the same system, of 5­50 Å.1­5 Due to its central role in the properties of indicating the dominance of extrinsic mechanisms. However, magnetic multilayers, the coupling between the magnetic there are conflicting reports on the temperature and spacer layers through the nonmagnetic metallic spacer has been the layer thickness dependence of J2 and only in very few cases subject of extensive investigations for nearly ten years. Ex- there seems to be a reasonable connection between data and perimentally this coupling is more conveniently studied in a some specific extrinsic mechanism.22 The fact is that the trilayer structure, formed by two magnetic thin films sepa- whole question of the origin of the BEC is still quite rated by a nonmagnetic layer, Fe/Cr/Fe being the most stud- controversial23 and deserves further investigation. In this ar- ied system.4­12 ticle we present new data on the temperature dependence of The coupling between the magnetic layers is usually J1 and J2 in the prototype system 100 Fe/Cr/Fe for varying dominated by a mechanism which can be modeled by an Cr spacer layer thickness and show that in only one sample interaction energy of the form J1m1­m2 , where m1and m2 the exchange fluctuation mechanism accounts for the experi- are the unit magnetizations of the two magnetic layers and J1 mental data. is the bilinear exchange coupling constant. The origin of this coupling lies in the interaction between the s electrons in the metallic spacer and the d electrons in the magnetic layers,13­15 and is currently well understood.16 However, EXPERIMENTAL RESULTS more recently it was observed5,6,17 that under certain condi- The samples investigated are single-crystal trilayer struc- tions the magnetic moments of the two layers tend to align at tures of 100 Fe 40 Å /Cr s /Fe 40 Å grown by magnetron 90° with respect to each other. This alignment may be ac- sputter deposition, as described in Ref. 25 on MgO 100 counted for through an interaction energy described by a substrates. All samples have the same Fe layer thickness, phenomenological biquadratic exchange coupling BEC d 40 Å, and a thin Cr cap layer. Initial characterization of J2(m1­m2)2, where J2 is the biquadratic coupling con- the coupling was made by magneto-optical Kerr effect stant. Over the last few years several mechanisms have been MOKE with a Cr wedged sample, with 0 s 70 Å. Then proposed for the origin of the biquadratic coupling,18­21 a series of samples was prepared with uniform Cr thickness varying from 5 to 35 Å, a range that corresponds to the first a Electronic mail: smr@df.ufpe.br two antiferromagnetic peaks. 0021-8979/99/85(8)/5892/3/$15.00 5892 © 1999 American Institute of Physics J. Appl. Phys., Vol. 85, No. 8, 15 April 1999 Rezende et al. 5893 FIG. 1. Room-temperature exchange coupling constants measured in sput- FIG. 2. Temperature dependence of the exchange coupling constants in tered 100 Fe 40 Å /Cr s /Fe 40 Å by MOKE, BLS, and FMR. Fe 40 Å /Cr s /Fe 40 Å . The symbols represent the data for the samples with s 11 circles , 13 squares , and 15 Å triangles and the solid lines are fits with theoretical predictions. In order to obtain reliable values for J1 and J2 , we have used three independent techniques, namely, MOKE magne- Figure 2 shows the temperature dependence of J1 and J2 tometry, Brillouin light scattering BLS , and ferromagnetic measured in three samples with Cr layer thickness s 11, 13, resonance FMR . The data were fitted with a phenomeno- and 15 Å. Qualitatively the data are similar to results previ- logical energy model including bilinear and biquadratic ex- ously obtained in several systems, both exchange constants change couplings, as well as surface and crystalline cubic decrease with increasing temperature. However, a detailed anisotropy contributions. Details of the measuring techniques analysis of the temperature dependence contains important and the procedures used to extract the values of J1 and J2 are clues on the mechanisms responsible for the coupling be- presented elsewhere.24,25 tween the magnetic layers. Figure 1 shows the room-temperature values for J1 and J2measured in 14 samples with varying Cr spacer thickness. DISCUSSION The vertical bars represent the uncertainties due to the esti- mated errors in each fitting plus the spread in the values In order to discuss the origin of the biquadratic coupling obtained with the various techniques. Two AF peaks alter- in our Fe/Cr/Fe samples, we start looking at the behavior of nating with one FM peak are observed in J the bilinear coupling J 1 in the thickness 1 . There is general agreement today range 5 Å s 35 Å, a well known result which has been that the bilinear coupling originates in the interaction be- obtained by many authors. The maximum absolute value of tween the s electrons in the Cr layer and the d electrons in J the Fe layers, the so-called intrinsic mechanism. Calculations 1 in the first AF peak is 0.59 erg/cm2, for s 9.5 Å, a value somewhat smaller than J taking into account the full electronic structure of the metals 1 1 erg/cm2 reported for some mo- lecular beam epitaxy MBE grown samples,3,4,9 but similar show26,27 that for perfectly sharp interfaces the behavior of to those reported for other MBE7 and sputtered11 100 Fe/ J1 with the spacer layer thickness is entirely dominated by Cr/Fe trilayers. short period oscillations with amplitude decaying with in- The result for J creasing thickness. The maximum negative value of J 2 is not so well known. In fact, to our 1is ap- knowledge, this is the first measurement of J proximately 7 erg/cm2, which is an order of magnitude larger 2 vs spacer- layer thickness in the second AF peak. The data show that J than the measured values. This discrepancy is accounted for 2 is negative in the whole range, and that its ratio to J by the existence of roughness, interdiffusion, vacancies, and 1 varies considerably with s. In most of the first AF peak, J steps in the real sample, which smooth out the short period 2 follows a dependence with s similar to that of J oscillations and drastically reduce the peak value.27 While 1 , with J2 /J1 0.1. However, near the crossing from AF to FM s 15 Å , this comparison between theory and experimental data for the ratio increases to J strength of the coupling is not satisfactory, the same is not 2 /J1 0.3, which is similar to that mea- sured in a structure Fe 28 Å /Cr 15.8 Å grown by MBE on true for the temperature dependence of J1 . Consider the the- a 100 Fe whisker.8 Throughout the Cr thickness range oretical prediction for the intrinsic mechanism21 J1(T) s 16­24 Å, corresponding to the second FM peak, the ratio J1(0)f1(T), where f1(T) (T/T0)/sinh(T/T0) . J2 /J1 remains in the range of 0.2­0.3. Surprisingly, in the The solid lines in Fig. 2 a represents the fits of this second AF peak the ratio increases to J2 /J1 1, so that the function to the experimental data, obtained with T0 390, antiferromagnetic phase ceases to exist.24,25 214, and 122 K for the samples with Cr layer thickness 5894 J. Appl. Phys., Vol. 85, No. 8, 15 April 1999 Rezende et al. s 11, 13, and 15 Å, respectively. Note that T0 decreases This yields an exponent a 0.25 0.10, implying that the with increasing s, and although it does not follow the 1/s law temperature dependence of J2 for the s 15 Å sample is con- of the simple theory, the good fits indicate that the intrinsic sistent with the prediction of the exchange fluctuation mechanism accounts for the origin of the bilinear exchange mechanism. However none of the proposed mechanisms for coupling. the BEC12,18 can account quantitatively for the data in the Regarding the origin of the biquadratic coupling J2 , we other Fe/Cr/Fe samples. Therefore, the present results add first note that it cannot be attributed to intrinsic mechanisms evidence to previous12,23 conclusions that further theoretical for two reasons: the predicted oscillation period for J2 is and experimental work is necessary to fully explain the bi- smaller than for J1 , whereas the data of Fig. 1 shows J2 quadratic exchange coupling in magnetic multilayers. following J1 ; theory21 predicts a rapid decay of J2 with in- creasing s, which is certainly not the case of the data. For the ACKNOWLEDGMENTS samples under investigation here, among the various extrin- sic sources proposed for J This work has been supported by the Brazilian federal 2 , the most plausible one is the Slonczewski's exchange fluctuation mechanism18 caused by agencies CNPq, FINEP, PADCT, and CAPES and the Per- interface roughness. According to the model, J nambuco state agency FACEPE. The work at IBM was par- 2 arises from the combined effect of the rapid oscillation in the intrinsic J tially supported by the Office of Naval Research. 1 and the variation in spacer layer thickness in the form of terraces. If J 1 P. Gru¨nberg, R. Schreiber, Y. Pang, M. B. Brodsky, and H. Sowers, Phys. 1 varies in steps of 2 J1 , the first order con- tribution of this mechanism to the BEC is18 Rev. Lett. 57, 2442 1986 . 2 S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, 2304 4 J 1990 . 3 J 1 2L See, for example, B. Heinrich and J. F. Cochran, Adv. Phys. 42, 523 2 coth d , 1 3A L 1993 ; Ultrathin Magnetic Structures, edited by B. Heinrich and J. A. C. Bland Springer, Berlin, 1994 . where A is the exchange stiffness constant of the Fe layer, L 4 J. J. Krebs, P. Lubitz, A. Chaiken, and G. A. Prinz, Phys. Rev. Lett. 63, is the terrace width, and d is the Fe layer thickness. Equation 1645 1989 ; J. Appl. Phys. 67, 5920 1990 . 5 J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 67, 140 1991 . 1 predicts that J2 is always negative, favoring the 90° 6 M. Ruhrig, R. Scha¨fer, A. Hubert, R. Mosler, J. A. Wolf, S. Demokritov, alignment, as observed in the experiments, and that its and P. Gru¨nberg, Phys. Status Solidi A 125, 635 1991 . strength varies with the square of J 7 U. Ko¨bler, K. Wagner, R. Wichers, A. Fuss, and W. Zinn, J. Magn. Magn. 1 . Considering that J1 is a step change in the bilinear coupling arising from the Mater. 103, 236 1992 . 8 M. From, L. X. Liao, J. F. Cochran, and B. Heinrich, J. Appl. Phys. 75, short period oscillation, and that the measured coupling rep- 6181 1994 . resents an average of J 9 1 , Eq. 1 predicts for J2 a tempera- R. J. Hicken, C. Daboo, M. Gester, A. J. R. Ives, S. J. Gray, and J. A. C. ture dependence following J2 Bland, J. Appl. Phys. 78, 6670 1997 . 1(T)/A(T). In order to verify 10 this prediction, it is necessary to take into account the tem- A. J. R. Ives, J. A. C. Bland, R. J. Hicken, and C. Daboo, Phys. Rev. B 55, 12428 1997 . perature variation of the exchange stiffness.23 Thus we have 11 M. Grimsditch, S. Kumar, and E. E. Fullerton, Phys. Rev. B 54, 3385 determined A T by measuring the volume mode frequencies 1996 . in a 250 Å thick single film of 100 Fe/MgO as a function of 12 S. O. Demokritov, J. Phys. D 31, 925 1998 . 13 temperature using BLS. Y. Wang, P. M. Levy, and J. L. Fry, Phys. Rev. Lett. 65, 2732 1990 . 14 D. M. Edwards, J. Mathon, R. B. Muniz, and M. S. Phan, Phys. Rev. Lett. We now argue that the present model of the exchange 67, 493 1991 . fluctuation mechanism cannot by itself explain the measured 15 P. Bruno and C. Chappert, Phys. Rev. Lett. 67, 1602 1991 ; Phys. Rev. B BEC in the whole range of Cr spacer layer thickness. The 46, 261 1992 . 16 A. T. Costa, Jr., J. d'Albuquerque e Castro, and R. B. Muniz, Phys. Rev. first argument is that Eq. 1 predicts that the amplitude of J2 B 56, 13697 1997 . decays with increasing s following J2 17 1. This is in complete C. J. Gutierrez, J. J. Krebs, M. E. Filipkowski, and G. A. Prinz, J. Magn. disagreement with the data, which show that while the peak Magn. Mater. 116, L305 1992 . 18 amplitude of J J. Slonczewski, Phys. Rev. Lett. 67, 3172 1991 ; J. Magn. Magn. Mater. 1 does decrease with increasing s, the ratio 150, 13 1995 . J2 /J1 is 0.1 in the first AF peak and 1 in the second AF 19 J. Barna´s and P. Gru¨nberg, J. Magn. Magn. Mater. 121, 326 1993 . peak. The second argument is based on the temperature de- 20 R. P. Erickson, K. B. Hathaway, and J. R. Cullen, Phys. Rev. B 47, 2626 pendence of the coupling constants. The solid lines in Fig. 1993 . 21 2 b are the fits of the J D. M. Edwards, J. M. Ward, and J. Mathon, J. Magn. Magn. Mater. 126, 2(T) data with f 1(T) b, obtained 380 1993 . with the values b 11.8, 6.6, and 1.7 for the samples with Cr 22 M. Scha¨fer, S. Demokritov, S. Mu¨ller-Pfeiffer, R. Scha¨fer, M. Schneider, layer thickness s 11, 13, and 15 Å, respectively. This shows P. Gru¨nberg, and W. Zinn, J. Appl. Phys. 77, 6432 1995 . that only for the sample with s 15 Å the temperature varia- 23 Z. Celinski, B. Heinrich, and J. F. Cochran, J. Magn. Magn. Mater. 145, tion of J 2 L1 1995 . 2 is close to the J1(T) dependence predicted by the 24 A. Azevedo, C. Chesman, S. M. Rezende, F. M. de Aguiar, X. Bian, and exchange fluctuation mechanism. In order to verify if this S. S. P. Parkin, Phys. Rev. Lett. 76, 4837 1996 . dependence really applies to the s 15 Å sample, one needs 25 S. M. Rezende, C. Chesman, M. A. Lucena, A. Azevedo, F. M. de Aguiar, to take into account the temperature variation of the ex- and S. S. P. Parkin, J. Appl. Phys. 84, 958 1998 . 26 change stiffness. So we fitted the BLS data with A(T) M. D. Stiles, Phys. Rev. B 54, 14679 1996 ; and references therein. 27 A. T. Costa, Jr., J. d'Albuquerque e Castro, and R. M. Muniz unpub- f 1(T) a,, with T0 122 K appropriate for this sample. lished .