VOLUME 79, NUMBER 24 P H Y S I C A L R E V I E W L E T T E R S 15 DECEMBER 1997 Magnetic Structure of Cr in Exchange Coupled Fe Cr(001) Superlattices A. Schreyer,1 C. F. Majkrzak,2 Th. Zeidler,1 T. Schmitte,1 P. Bödeker,1 K. Theis-Bröhl,1 A. Abromeit,1 J. A. Dura,2 and T. Watanabe2 1Experimentalphysik (Festkörperphysik), Ruhr-Universität Bochum, 44780 Bochum, Germany 2National Institute of Standards and Technology, Gaithersburg, Maryland 20899 (Received 20 February 1997) We demonstrate how the noncollinear exchange coupling between the Fe layers in Fe Cr(001) superlattices is caused by a frustrated spiral modulation of the Cr moments not observed in bulk. The noncollinear coupling vanishes above the Néel temperature of this commensurate antiferromagnetic Cr order. This clarifies the essential contribution of Cr ordering to the coupling in the regime of smallest thicknesses where no incommensurate Cr spin density wave can form. For larger Cr thicknesses we observe a predicted incommensurate to commensurate transition with temperature. [S0031-9007(97)04817-5] PACS numbers: 75.70.Cn, 75.25.+z, 75.30.Fv The oscillatory exchange coupling between ferromag- found in Fe Cr(001) [18] and FeCo Mn [21], no direct netic (FM) layers over a non-FM interlayer [1], the related information on the magnetic structure of the interlayers in giant magnetoresistance effect [2], and the biquadratic or such NC coupled superlattices exists. noncollinear (NC) exchange coupling between FM layers Here we present high angle neutron scattering data, giv- [3,4] were important recent discoveries in the field of thin ing access to the magnetic structure of the Cr interlay- film magnetism [5]. Fe Cr(001) layered structures played ers in Fe Cr(001) superlattices. The results are correlated the key role in these findings. Since Cr is antiferromag- with polarized neutron reflectometry (PNR) and magneto- netic (AF) in bulk [6], it is of fundamental importance optic Kerr effect (MOKE) data on the coupling of the Fe to understand the role of the Cr magnetism in this model layers. For the smallest Cr thickness investigated, strong system. NC exchange coupling between the Fe layers occurs as After the discovery of the long range period of the ex- soon as a frustrated, commensurate structure of AF Cr change coupling in Fe Cr(001) as a function of Cr thick- spirals forms below its Néel temperature. This highlights ness [1], an additional short two monolayer oscillation the essential contribution of Cr order to the coupling. period was found [7,8] exhibiting phase slips [7]. These Thus, the correlation between Cr order and coupling is phase slips are consistent with the same nesting vector exactly opposite to the one reported previously [15,22,23] which gives rise to the incommensurate spin density wave for larger Cr thicknesses, where an ISDW forms. For this (ISDW) antiferromagnetism in bulk Cr. Current models regime we confirm the predicted [24], previously over- for oscillatory exchange coupling [9­12] readily explain looked transition from an incommensurate to a commen- the short period oscillation and the phase slips as a re- surate SDW with increasing temperature. sult of this nesting vector, independent of the existence The samples were grown in a molecular beam epitaxy of any magnetic long range order (LRO) of the Cr [13], (MBE) system at 300±C at a pressure #10210 mbar using in the same way as for nonmagnetic interlayers. Initial growth rates of 0.16 Å s for Cr and 0.2 Å s for Fe on a experiments [14,15] yielded contradictory results on the Cr Nb buffer system on Al2O3 (1¯102) substrates [25]. To existence of any LRO in Cr interlayers. maximize the amount of Cr, large 5 3 5 cm2 substrates The importance of the structure of the Fe Cr inter- were used and up to 200 Fe Cr bilayers were grown. faces has been pointed out by many authors [5,7,16]. For Reflection high energy electron diffraction performed example, it was demonstrated that for a given substrate during growth indicated a smooth growth front with steps the lateral length scale of the interface fluctuations in- and a smoothing during growth. High angle x-ray data creases strongly with the growth temperature, drastically with sharp superlattice peaks up to third order are proof of modifying the coupling behavior [5,17,18]. In the re- a coherent superlattice structure. Energy dispersive x-ray cent proximity magnetism model [19] for NC coupling, (EDX) analysis yielded the relative Fe and Cr contents such interface fluctuations and the intrinsic magnetism which, combined with the measured superlattice period, of AF interlayers such as Cr or Mn were taken into ac- provided the layer thicknesses. X-ray scattering and EDX count, yielding a parabolic expression for the exchange spectra taken from the center and near the edges of the energy causing an asymptotic approach to the saturated samples confirmed a perfect lateral homogeneity obtained state as a function of applied field, in contradiction to the by continuous rotation of the sample during growth. usual bilinear-biquadratic [3] exchange coupling formal- High angle neutron measurements were performed at ism. Recent band structure calculations confirm this pic- NIST on the SPINS spectrometer. Bulk Cr is AF be- ture [20]. Although the asymptotic approach has been low its Néel temperature TN,ISDW 311 K, exhibiting an 4914 0031-9007 97 79(24) 4914(4)$10.00 © 1997 The American Physical Society VOLUME 79, NUMBER 24 P H Y S I C A L R E V I E W L E T T E R S 15 DECEMBER 1997 ISDW. Depending on polarization (longitudinal or trans- designate the layer thicknesses in Å, whereas the one verse) and propagation, the ISDW causes a double peak outside gives the number of bilayers. To characterize structure around certain 001 positions [6]. A commen- the magnetic order of the Fe layers we used hysteresis surate AF order causes single peaks at the 001 positions. loops measured by MOKE and PNR with polarization Our findings are summarized in Fig. 1 as a function of Cr analysis as detailed in Ref. [18]. A PNR scan of the interlayer thickness DCr and temperature for DFe 20 Å. sample taken at room temperature [27] near remanence The insets indicate schematically the observed intensity on the reflectometer BT-7 at NIST is shown in the left around (001) with the arrows pointing along the [001] inset of Fig. 2. The splitting in the 11 and 2 2 data, (out of plane) direction. ISDW double peaks were ob- together with the existence of a first order peak at about served for DCr 80 Å at low temperatures, indicating 0.1 Å21 2p LSL in the 12 and 21 scattering, a transverse ISDW propagating along [001], i.e., with indicate a net magnetization in the sample. In addition, the moments in the film plane [6,26]. Between about half order peaks at about 0.055 Å21 are observed in all 175 and 310 K we observe a gradual transition (marked cross sections, indicating that the magnetic period LFe,NC T in Fig. 1) from incommensurate I to commensurate is doubled with respect to the chemical superlattice period AF order C . Furthermore, we find a Néel temperature LSL. Since collinear AF order could not yield such a net TN,COM 500 K of this C phase, above which the Cr be- magnetization, the individual Fe magnetizations must be comes paramagnetic P . Naively, the I to C transition NC oriented [18], consistent with our MOKE results. could be correlated with the increase of the ISDW period In the main part of Fig. 2 a scan along the out- with temperature, as it occurs in bulk: For a given DCr, a of-plane 00l direction through 010 taken from the single ISDW period cannot form above a certain tempera- same sample at 30 K is shown. Clearly, on both sides ture and the LRO becomes C. For thinner DCr 42 Å of the fundamental 010 peak, additional satellites are we find only strong, modulated C peaks which again visible. These are proof for a modulation of the C AF vanish at TN,COM 500 K. Thus, below a DCr,min the structure with a period LCr,AF. The separation DQ ISDW cannot form even for the lowest temperature mea- 2p LCr,AF between the satellites is the same as the one sured. TN,COM is significantly enhanced with respect to observed between the first and half-order peaks in the the bulk TN,ISDW. Such high TN,COM have been observed PNR spectrum in the inset. Thus, we find LCr,AF in strained C Cr in bulk [6]. Our observations are quali- LFe,NC 2LSL. This would result from the coupling tatively consistent with a recent theoretical study [24] of between Fe and Cr at each interface being equal: In Fe Cr with ideal interfaces predicting the C to I transi- the right inset of Fig. 2 the Fe magnetizations are shown tions as a function of DCr and T observed here. Finally, we point out that we always find strong intensity at or around a (001) reflection below TN,COM. Since neutron scattering is sensitive only to magnetic moment compo- nents perpendicular to the scattering vector, this indicates an in-plane orientation of the magnetic moments in the DCr range discussed here. As we show elsewhere [26], a reorientation occurs for larger DCr. Now we focus on the sample Cr42 Fe19 200, with the C order below DCr,min. The subscripts in the brackets FIG. 2. Left inset: PNR of the sample Cr42 Fe19 200 at RT in B 14 G, indicating NC coupling of the Fe layers with 2LSL LFe,NC. The full squares indicate 11 , the empty ones 2 2 and the up and down triangles 12 and 21 scattering. 1 2 designate the neutron spin state up (down) before and after the sample, respectively. The 12 and 21 data are shifted by 1022. Main part: High angle l scan through the (010) position at T 30 K for the same FIG. 1. Phase diagram summarizing the magnetic structure of sample. Also shown is a fit to the data using five Gaussians the Cr interlayers. Diamonds denote commensurate SDW C , separated by DQ 2p LCr,AF. Right inset: schematic of the squares denote incommensurate SDW I , and circles denote magnetic structure deduced from the data in the figure fulfilling paramagnetic P regions. T is a transition region. The insets LCr,AF LFe,NC 2LSL. The empty and filled small arrows schematically indicate the measured intensity (dots) around indicate an opposing sense of rotation of the Cr moments (001) with the arrows denoting the [001] scan direction. between the NC coupled Fe layers (large arrows). 4915 VOLUME 79, NUMBER 24 P H Y S I C A L R E V I E W L E T T E R S 15 DECEMBER 1997 as large arrows in a NC structure with LFe,NC 2LSL Slonczewski [19] in his proximity magnetism model for as found by PNR. The AF Cr moments in between NC coupling. Comparison of the schematics in Figs. 2 are represented by the smaller arrows (filled and empty and 3 shows that this arbitrary variation of the sense of for two senses of rotation). The spiral-type modulation rotation strongly reduces the coherence of the modulation of the AF Cr interlayers is the topologically simplest of the AF Cr in the out-of-plane (l) direction. Structure assumption with the boundary condition of AF coupling factor calculations confirm the model structure depicted [28] at each Fe Cr interface. A consequence of this in Fig. 3. Specifically, these calculations also reproduce equal coupling is LCr,AF LFe,NC as observed [29]. The the line shape in the main part of Fig. 2, confirming higher order satellites in the data of Fig. 2 indicate rather the insensitivity of such a scan to the vertical disorder. sharp interfaces of the AF Cr structure. Furthermore, the small intensity of the central peak in The coherence length of the magnetic order of the Fig. 3 is explained as a destructive interference due to Cr along the growth direction can be determined from fluctuating Fe thicknesses. The lateral disorder depicted scans through 001 along 00l . As shown in Fig. 3, in the model structure of Fig. 3 is fully confirmed, e.g., we again observe a multicomponent intensity around the by recent diffuse x-ray scattering data from similarly 001 position. The peak can be decomposed into a prepared Fe Cr superlattices [18]. Also it is obvious sharp fundamental reflection and one satellite peak on that the in-plane magnetic structure must be limited in each side [30]. As in Fig. 2, these satellites are offset coherence as well due to the lateral fluctuation. From the with respect to the fundamental reflection by DQ width of a scan along the in-plane 0k0 direction through 2p LCr,AF. However, they are much broader. This 010 , we obtain a coherence length of about LjjC indicates an out-of-plane coherence length of only about 150 Å. The Cr spiral structure, which we postulated from 60 Å for the AF Cr modulation. Thus, we observe topology arguments, can be tested with polarized neutrons coherence of the NC order of the Fe layers over many and polarization analysis. Repeating the scan of Fig. 3 bilayers (from the PNR data) and of the C AF structure we find the same intensity within statistics in all four of the Cr (from the sharp central peak in Fig. 3). On cross sections consistent with a spiral distribution of the the other hand, we find a short coherence of the AF Cr Cr moments. More details will be discussed elsewhere. modulation. Consequently, the schematic in the inset of Finally, we again point out that the observation of strong Fig. 2 is idealized. (001) intensity requires the in-plane orientation of the Cr In the following, we consider disorder. In the inset of moments assumed so far. Fig. 3 the effect of Cr interlayer thicknesses fluctuating Thus, we have demonstrated that all aspects of the by one monolayer is shown schematically. Comparing magnetic structure postulated in the proximity magnetism the left and right parts of the figure, we find that the model [19] are actually observed. In the model, the sense of rotation of the partial Cr spirals changes upon occurrence of NC coupling is explained by the opposing varying the number of Cr monolayers by one within torque which laterally coexisting partial spirals apply to the same Cr layer. This is a consequence of the equal the Fe layers. Since the Fe layers can neither couple AF coupling at each Fe Cr interface and was demonstrated by (even number of Cr layers) or FM (uneven number of Cr layers) the system compromises on a NC orientation. Slonczewski [19] also provides an upper limit on the lateral length scale Ljj of DCr fluctuations, above which the magnetic exchange stiffness of the Fe becomes too weak to prevent the formation of AF- and FM-coupled domains. This value is consistent with the measured LjjC which is directly connected to Ljj. So far we have correlated NC coupling between the Fe layers with a specific LRO in the Cr interlayers. Finally, we study the effect on the coupling between the Fe layers when the Cr order vanishes at TN,COM. In Fig. 4 the temperature dependence of the (001) peak intensity is shown for the same sample. Clearly, the LRO in the Cr interlayers vanishes at about 500 K. In the insets we show hysteresis loops measured by MOKE along an FIG. 3. High angle l scan through the 001 position at easy axis of the in-plane anisotropy at and below TN,COM. T 30 K for the same sample as in Fig. 2. Also shown is a Below TN,COM, the sample is not saturated at 2 kOe, fit to the data using three Gaussians which again are separated indicating strong exchange coupling. In passing, we note by DQ 2p LCr,AF. Inset: Modification of the model in that equivalent hysteresis loops measured up to 10 kOe Fig. 2 by inserting a vertically uncorrelated lateral Cr interlayer thickness fluctuation which reduces the coherence explaining indicate the asymptotic approach to saturation expected the broad satellites. The empty and filled small arrows again for the proximity magnetism model [18,21]. At TN,COM indicate opposing senses of rotation. we observe square hysteresis loops, indicating parallel 4916 VOLUME 79, NUMBER 24 P H Y S I C A L R E V I E W L E T T E R S 15 DECEMBER 1997 with sample preparation. Financial support was provided by NATO (CRG 901064), as well as the German BMBF (03-ZA4BC2 -3) and DFG via SFB 166. [1] P. Grünberg et al., Phys. Rev. Lett. 57, 2442 (1986); S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, 2304 (1990); S. Demokritov, J. A. Wolf, and P. Grünberg, Europhys. Lett. 15, 881 (1991). [2] M. N. Baibich et al., Phys. Rev. Lett. 61, 2472 (1988); G. Binasch, et al., Phys. Rev. B 39, 4828 (1989). FIG. 4. Neutron (001) intensity as a function of temperature [3] M. Rührig et al., Phys. Status Solidi (a) 125, 635 (1991). indicating T [4] B. Heinrich et al., Phys. Rev. B 44, 9348 (1991). N,COM 500 K. In the insets typical MOKE hysteresis loops measured below and at T [5] B. Heinrich and J. F. Cochran, Adv. Phys. 42, 523 (1993). N,COM are shown. At TN,COM the NC coupling between the Fe layers vanishes, [6] E. Fawcett, Rev. Mod. Phys. 60, 209 (1988). consistent with PNR results. [7] J. Unguris, R. J. Cellotta, and D. T. Pierce, Phys. Rev. Lett. 67, 140 (1991). [8] S. T. Purcell et al., Phys. Rev. Lett. 67, 903 (1991). [9] Y. Wang, P. M. Levy, and J. L. Fry, Phys. Rev. Lett. 65, alignment of the Fe layers. Consistently, the PNR half- 2732 (1990). order peak observed below TN,COM vanishes above TN,COM [10] P. Bruno and C. Chappert, Phys. Rev. Lett. 67, 1602 in a 60 G field. Thus, the strong NC exchange coupling (1991). observed below TN,COM vanishes with the Cr LRO. [11] D. M. Edwards et al., Phys. Rev. Lett. 67, 493 (1991). Finally, we discuss our results in the framework of [12] M. D. Stiles, Phys. Rev. B 48, 7238 (1993). other available data. Meersschaut et al. [14] found a [13] M. van Schilfgaarde and F. Herman, Phys. Rev. Lett. 71, longitudinal ISDW with the moments oriented out of the 1923 (1993); D. Stoeffler and F. Gautier, Phys. Rev. B 44, film plane for D 10 389 (1991). Cr above 75 Å and measured Cr to be nonmagnetic for thicknesses below 60 Å by perturbed [14] J. Meersschaut et al., Phys. Rev. Lett. 75, 1638 (1995). angular correlation spectroscopy. Here we clearly observe [15] E. E. Fullerton, S. D. Bader, and J. L. Robertson, Phys. C LRO down to smaller D Rev. Lett. 77, 1382 (1996). Cr , where no ISDW forms. [16] J. A. Wolf et al., J. Magn. Magn. Mater. 121, 253 (1993). Furthermore, in the whole DCr range discussed here, the [17] D. T. Pierce et al., Phys Rev. B 49, 14 564 (1994). Cr moments are predominantly in the film plane. Our [18] A. Schreyer et al., Europhysics Lett. 32, 595 (1995); Phys. results are confirmed by Fullerton et al. [15]. The solid Rev. B 52, 16066 (1995). lines separating the C and I regions in Fig. 1 were chosen [19] J. C. Slonczewski, J. Magn. Magn. Mater. 150, 13 (1995). such that the Cr thickness dependent TN,ISDW reported in [20] M. Freyss, D. Stoeffler, and H. Dreyssé, Phys. Rev. B 54, Ref. [15] essentially lies in our transition region T from R12 677 (1996). I to C order. However, in Ref. [15] a paramagnetic phase [21] M. E. Filipkowski et al., Phys. Rev. Lett. 75, 1847 (1995). above T [22] E. E. Fullerton et al., Phys. Rev. Lett. 75, 330 (1995). N,ISDW is reported, whereas here we find proof for C AF order in this regime, qualitatively consistent with [23] J. F. Ankner et al., J. Appl. Phys. 81, 3765 (1997). theoretical predictions [24]. [24] Z. P. Shi and R. S. Fishman, Phys. Rev. Lett. 78, 1351 In conclusion, we find strong NC coupling only below (1997). [25] For details, see S. Di Nunzio, K. Theiss-Bröhl and the ordering temperature of a frustrated C AF spiral struc- H. Zabel, Thin Solid Films 279, 180 (1996); I. Zoller et ture in the Cr layers. So far, spiral magnetic structures al., Phys. Rev. B (to be published). have only been observed in the rare earths [31]. The cor- [26] P. Sonntag et al., J. Magn. Magn. Mater. (to be published). relation between coupling and Cr order reported here is [27] Temperature dependent PNR measurements confirm that exactly opposite to the results for larger thicknesses in the the spectra remain essantially unchanged between 30 K I phase. In this case the absence of coupling has been and RT in case of field or zero-field cooling. correlated with the occurrence of Cr LRO [22,23]. Thus, [28] See, e.g., F. U. Hillebrecht et al., Europhys. Lett. 19, 711 for this important model system we have clarified the es- (1992). sential contribution of Cr order in the thickness regime, [29] This would also result either for FM coupling at all Fe Cr where all important aspects of exchange coupling were interfaces or for AF coupling at all top (bottom) and FM discovered. Furthermore, we find an I to C phase transi- coupling at all bottom (top) interfaces of each Cr layer, respectively. tion for large Cr thicknesses, as predicted by theory. [30] As opposed to Fig. 2, the fit yields negligible higher order We gratefully acknowledge discussions with H. Zabel satellite intensities. and J. Borchers. We thank V. Nunez and S. H. Lee for [31] See, e.g., M. B. Salomon et al., Phys. Rev. Lett. 56, 259 their support with SPINS, and I. Zoller for spending nights (1986), and references therein. 4917