VOLUME 87, NUMBER 15 P H Y S I C A L R E V I E W L E T T E R S 8 OCTOBER 2001 Metallic-Type Oscillatory Interlayer Exchange Coupling across an Epitaxial FeSi Spacer R. R. Gareev, D. E. Bürgler,* M. Buchmeier, D. Olligs, R. Schreiber, and P. Grünberg Institut für Festkörperforschung, Forschungszentrum Jülich GmbH, D-52425 Jülich, Germany (Received 2 November 2000; published 21 September 2001) We study interlayer exchange coupling in epitaxial Fe Fe0.56Si0.44 Fe trilayers. Iron-silicide spacers with high structural and compositional homogeneity for thicknesses up to 34 Å are grown by coevapo- ration from two electron-beam sources. The coupling strength oscillates with spacer thickness for tem- peratures from 20 to 300 K with two antiferromagnetic maxima at 12 and 26 Å, and it clearly increases with decreasing temperature down to 80 K. We conclude that the coupling across ordered Fe12xSix (x 0.5) is described by the conventional theory of interlayer coupling across metallic spacers. DOI: 10.1103/PhysRevLett.87.157202 PACS numbers: 75.70.­i, 73.61.At, 75.30.Et Since the discovery of the antiferromagnetic (AF) ex- well defined composition x. According to Ref. [12] the in- change coupling between ferromagnetic layers separated terdiffusion at Fe FeSi interfaces is strongly suppressed in by a nonmagnetic metal spacer [1], this phenomenon has comparison to Fe Si interfaces. Hence we prepare epitax- been observed for a wide range of metallic spacer materi- ial Fe Fe12xSix Fe trilayers with a nominal composition als [2]. A further increased interest in this field has been of the spacer layer close to x 0.5 of the stoichiometric stimulated by the finding of AF coupling through non- B2 phase with CsCl structure or the B20 (e-FeSi) phase. metallic amorphous Si [3] and strong AF coupling across Epitaxial Fe FeSi-wedge Fe sandwiches are grown iron-silicide spacers [4]. Precise structural measurements in an ultrahigh vacuum multichamber molecular beam proved that metallic iron silicides with an epitaxially sta- epitaxy system by e-gun evaporation onto GaAs 100 bilized cubic CsCl (B2) structure are preferably formed in Fe 1 nm Ag 150 nm substrate-buffer systems described a spacer as a result of a strong interdiffusion at the Si Fe elsewhere [13,14]. In order to minimize segregation of interfaces [5­7]. It is now well established that the inter- Ag [15] the first 4 ML of the bottom 50-Å-thick Fe layer layer coupling through metallic spacers is connected with are grown at room temperature (RT) and the remaining at indirect RKKY-type exchange and that it oscillates from 473 K. All thicknesses and deposition rates are controlled ferromagnetic (FM) to AF as a function of spacer thick- by calibrated quartz-crystal monitors. The wedge-shaped ness [8]. Hence, an oscillatory behavior of the coupling FeSi spacers are prepared at 473 K, too. Two separate through metallic iron-silicide spacers is expected. How- electron guns are used to codeposit Fe and Si at equal ever, in spite of intense research dealing with coupling atomic flux to yield Fe phenomena across crystalline iron silicides no evidence 0.5Si0.5. The thickness of the alloy layer is then given by of oscillatory exchange coupling was observed so far for these materials. Recently, de Vries et al. [9] reported an dFeSi 1.06dSi 0.67 dFe 1 dSi , (1) exponential decrease of the coupling strength with spacer where dFe and dSi are the quartz-crystal readings for Fe thickness mediated by metallic FeSi spacers in epitaxial and Si, respectively [9]. The spacer thickness dFeSi varies Fe FeSi Fe trilayers and concluded to have found a new along the wedge linearly from zero to 34 Å (dFe 1 dSi type of interlayer coupling. Moreover, a strong enhance- 0 50 Å). Finally, an upper 50-Å-thick Fe layer and a ment of the coupling strength through highly resistive FeSi 500-Å-thick ZnS coating are deposited at RT. with more than 80% content of Si in Fe FeSi superlattices The composition and the structure of the Fe FeSi Fe tri- [10] demonstrate that the mechanism of exchange coupling layers are verified in situ by Auger electron spectroscopy through iron silicides is still far from being understood and (AES) and low-energy electron diffraction (LEED), re- therefore needs a further investigation. spectively. A well-defined LEED (00) spot (at 75 eV elec- The essential demand for a proper investigation of oscil- tron energy) could be observed for both Fe layers and for latory interlayer coupling through binary compounds is a the whole range of dFeSi indicating epitaxial growth. The high degree of structural and compositional homogeneity spacer composition is calculated from Fe and Si deposi- of the spacer layer for all thicknesses of interest. However, tion rates as well as from Auger spectra. In Fig. 1 a typi- epitaxial growth of Fe FeSi Fe trilayers has previously cal dependence of the nominal FeSi composition on spacer been observed only for spacer thicknesses dFeSi # 20 Å, thickness is shown (open symbols). The spacer composi- and additionally iron silicides of different composition and tion is also calculated from the Auger spectral intensities structure were usually formed by interdiffusion at the in- of Fe (703 eV) and Si (92 eV) lines taking into account terfaces [6,9]. Furthermore, Fe Si and Si Fe interfaces ap- the contribution of the bottom Fe layer to the Fe Auger peared to be inequivalent with respect to the iron-silicide signal due to the finite information depth of AES (filled formation [11]. In this work we employ codeposition of Fe symbols in Fig. 1). Both methods to determine the Si con- and Si to obtain more homogeneous Fe12xSix spacers with tent agree well and confirm the homogeneous composition 157202-1 0031-9007 01 87(15) 157202(4)$15.00 © 2001 The American Physical Society 157202-1 VOLUME 87, NUMBER 15 P H Y S I C A L R E V I E W L E T T E R S 8 OCTOBER 2001 dFe + dSi (Å) 1 (a) (b) 0 10 20 30 40 50 55 dFeSi= 12 Å dFeSi= 12 Å from deposition rates S T = 300 K T = 20 K 0 %) 50 from AES M/M 45 -1 1 (c) (d) 40 Silicon content (at. dFeSi= 26 Å dFeSi= 26 Å S T = 300 K T = 20 K 35 0 0 10 20 30 M/M Spacer thickness dFeSi (Å) FIG. 1. Si content x of a wedge-shaped Fe0.56Si0.44 spacer layer as a function of spacer thickness. Open symbols: determined -1 from the deposition rates measured by means of two quartz- crystal monitors. Filled symbols: calculated from AES intensity -500 0 500 -500 0 500 H (Oe) H (Oe) ratios measured at different positions along the wedge taking into account contributions from the bottom Fe layer (inelastic FIG. 2. Longitudinal MOKE hysteresis loops of a wedge-type mean free paths of the Auger electrons: lSi 2 ML and lFe Fe 5 nm Fe0.56Si0.44 dFeSi Fe 5 nm trilayer for (a) dFeSi 5 ML [16]). Abscissa: dFe 1 dSi is the sum of the nominally 12 Å and 300 K, (b) d deposited thicknesses of Fe and Si, and d FeSi 12 Å and 20 K, (c) dFeSi 26 Å FeSi is the thickness of and 300 K, and (d) d the resulting alloy layer according to Eq. (1). FeSi 26 Å and 20 K. Dashed lines indi- cate the switching fields HS. of the spacer for dFeSi up to 34 Å. The deposition rates samples, then we cannot derive an analytical relation yield an average nominal composition x 0.44 6 0.01 between the saturation field and the coupling strength. (dashed line in Fig. 1). Hence, we can exclude the forma- However, we can define for all our MOKE loops a tion of Fe3Si. Spacers intentionally prepared with lower Si switching field HS (see, e.g., dashed lines in Fig. 2) content (x 0.36) did not show AF or 90± coupling. We where the magnetizations of the Fe films jump either explain this with the onset of FM order [17]. to saturation or to a symmetric alignment with both The magnetic properties are checked by longitudinal magnetizations slightly deviating from the field axis. Con- magneto-optic Kerr effect (MOKE) measurements with the sidering the anisotropy K, bilinear (J1) and biquadratic external field applied in the sample plane. The MOKE (J2) coupling, and the external field H hysteresis loops setup based on an optical cryostat allows temperature de- can be obtained by minimizing for each value of H pendent measurements in the range from 20 to 300 K and the total areal energy density of the system, has previously been described in Ref. [18]. Easy-axis MOKE hysteresis loops for the spacer thick- E 2HMSt cos q1 1 cos q2 nesses dFeSi 12 and 26 Å taken at 20 and 300 K are Kt shown in Fig. 2. The asymmetries and the peaks around 1 sin2 2q 4 1 1 sin2 2q2 zero field are caused by second-order MOKE effects and temperature-dependent relative contributions of the two Fe 2 J1 cos q1 2 q2 2 J2 cos2 q1 2 q2 , (2) layers to the MOKE signal. Our analysis given below is with respect to the orientation of the magnetizations of based on switching fields and on the presence and absence the two films given by the angles q1 and q2 which are of remanent magnetization and hence is not influenced by measured relative to the field axis. The magnetization these effects [18,19]. M H is then given by For very large in-plane, fourfold anisotropy K or M negligible K the saturation field can be used as a measure M H S cos qmin 1 H 1 cos q min 2 H . (3) of the coupling. We measured the anisotropy field H 2 K of our sample in the FM coupled regions (i.e., at thicknesses The Fe films are taken to be of equal thickness t. Vary- dFeSi 3 10, 16 20, 32 Å) from hard-axis MOKE loops ing the coupling constants J1 and J2 we find from the and found K HKMS 2 4.5 6 0.4 3 104 J m3 in simulated hysteresis loops that the switching field HS is good agreement with the Fe bulk anisotropy constant. in good approximation proportional to the total coupling MS 1.714 3 106 A m is the saturation magnetization J J1 1 J2. of Fe. If Kt is of the same order of magnitude as the The dependence of HS on the spacer thickness is mea- coupling J as it will turn out to be the case for our sured for different temperatures ranging from 20 to 300 K. 157202-2 157202-2 VOLUME 87, NUMBER 15 P H Y S I C A L R E V I E W L E T T E R S 8 OCTOBER 2001 Figure 3 shows the resulting coupling curves taken at about 80 K and then levels off or even slightly decreases 20 and 300 K. We observe an oscillatory behavior of again. In order to understand the different temperature de- the coupling versus spacer thickness in the whole tem- pendence of J 1 and J 2 we note that all MOKE loops perature range. Two distinct regions with clear coupling taken in the first maximum show remanent magnetization maxima are found near dFeSi 12 and 26 Å. The [e.g., Figs. 2(a) and 2(b)], whereas the remanence is zero positions of the maxima are not dependent on tempera- in the second peak [e.g., Figs. 2(c) and 2(d)]. In Fig. 5 ture. The oscillations are more pronounced at higher we show the results of simulations for the remanent state, temperatures where the hysteresis loops suggest a FM i.e., H 0 in Eqs. (2) and (3), for different J values while coupled thickness range (dFeSi 16 20 Å) between the varying the relative strength of the bilinear and biquadratic two coupling maxima. We note that these data represent coupling, J1 J2. Remanence appears when J2 dominates the first observation of oscillatory interlayer coupling over J1. A qualitatively similar statement has been given in across an FeSi spacer layer. The positions of the first peak Ref. [21] for the case K 0, where an analytical expres- matches the results of de Vries et al. [9], but the second sion for the dependence of MR on J1 and J2 is obtained. peak contradicts their exponential decay of the coupling. Thus, the temperature dependence of J 2 is dominated by We relate this discrepancy to (i) the more homogeneous the bilinear coupling while for J 1 biquadratic coupling spacer of our samples and (ii) the larger epitaxial spacer is prevailing. Therefore, the different shape of the curves thickness range accessible in our experiments. Both in Fig. 4 reflect different temperature dependencies of J1 advantages arise from the preparation of the FeSi spacer and J2. by coevaporation instead of interdiffusion. A mechanism The temperature behavior of J2 depends on its cause. for how structural disorder in a metallic spacer can lead We exclude intrinsic, higher order contributions [8] as to an exponential thickness dependence of RKKY-type the cause of J2 because J2 is comparable or even big- interlayer coupling is described in Ref. [20]. Obviously, ger than J1. On the other hand, the observed increase only the fact that dFeSi is not limited to values smaller of the total coupling J upon cooling is on the order of 3 than 20 Å (dFe 1 dSi 30 Å)- as it is the case in [Fig. 4]. The loose spins model favored in Ref. [22] pre- Ref. [9]-allows one to observe the oscillatory behavior. dicts a stronger exponential temperature dependence of J2 The temperature dependence of the coupling at the first [23]. Hence, it is likely that the big biquadratic contri- J 1 and second J 2 coupling maximum is shown in Fig. 4. bution is caused by spatial or compositional fluctuations We have determined the total coupling strength J J1 1 at interfaces [13,21,24]. A more detailed analysis of the J2 by simulating hysteresis loops that reproduce the mea- origin of the biquadratic coupling is beyond the scope of sured switching fields. Note that the values of J are of the this Letter. However, for all known biquadratic coupling same order of magnitude as Kt. An unequivocal separation mechanisms -intrinsic, higher order term [8], loose spins of J1 and J2 is possible for MOKE loops that exhibit three model [23,25], and fluctuation model [21,24]-J2 mono- plateaus. An example is Fig. 2(c) for which we obtain tonically increases upon cooling and eventually saturates J1 20.14 mJ m2 and J2 20.07 mJ m2. These cou- at low temperatures. Therefore, our data indicate that J1 pling constants are in the typical range obtained for many levels off below 80 K. However, this shows up only in the other metallic spacer layers. J 1 exhibits a monotonic in- total coupling when J1 is dominant (curve J 2 in Fig. 4) crease upon cooling, whereas J 2 first increases down to but is hidden when J2 prevails (curve J 1 in Fig. 4). dFe + dSi (Å) 8000 10 20 30 40 50 -0.6 J(1) first maximum T = 300 K -0.5 600 2 ) (Oe) T = 20 K second maximum S -0.4 (mJ/m 400 2-0.3 + J 1 J(2) -0.2 200 J = J Switching field H -0.1 0 0 0 10 20 30 0 50 100 150 200 250 300 Spacer thickness dFeSi (Å) Temperature (K) FIG. 3. Switching field HS versus spacer thickness dFeSi of a FIG. 4. Total coupling strengths J 1 (open symbols) and J 2 wedge-type Fe 5 nm Fe0.56Si0.44 dFeSi Fe 5 nm trilayer mea- (filled symbols) at the first and second coupling maxima, respec- sured at 300 and 20 K, respectively. tively, as a function of temperature. 157202-3 157202-3 VOLUME 87, NUMBER 15 P H Y S I C A L R E V I E W L E T T E R S 8 OCTOBER 2001 ers without the need to claim a new type of coupling for 0.5 J = -0.1 mJ/m2 J = -0.5 mJ/m2 this specific material. S This work is supported by the HGF-Strategiefonds- / M project "Magnetoelectronics." R M Kt = 0.23 mJ/m2 00 0.5 1.0 1.5 2.0 J *Author to whom correspondence should be addressed. 1 / J2 Electronic address: D.Buergler@fz-juelich.de FIG. 5. Dependence of the remanent magnetization MR on the [1] P. Grünberg et al., Phys. Rev. Lett. 57, 2442 (1986). relative strength of the bilinear and biquadratic coupling J1 J2 [2] S. S. P. Parkin, Phys. Rev. Lett. 67, 3598 (1991). obtained from simulations based on Eqs. (2) and (3) for H 0. [3] S. Toscano, B. Briner, H. Hopster, and M. Landolt, J. Magn. The anisotropy is kept constant, Kt 0.23 mJ m2, and the total Magn. Mater. 114, L6 (1992). coupling J is varied in the range covered in Fig. 4. [4] E. E. Fullerton et al., J. Magn. Magn. Mater. 117, L301 (1992). [5] A. Chaiken, R. P. Michel, and M. A. Wall, Phys. Rev. B 53, The temperature dependence of the bilinear coupling 5518 (1996). strength for metallic and insulating spacers can be de- [6] G. J. Strijkers, J. T. Kohlhepp, H. J. M. Swagten, and scribed by the quantum interference model formulated by W. J. M. de Jonge, Phys. Rev. B 60, 9583 (1999). Bruno [8]. Oscillations and an increase of the coupling [7] M. Fanciulli et al., Phys. Rev. B 59, 3675 (1999). strength upon cooling with saturation at low temperatures [8] P. Bruno, Phys. Rev. B 52, 411 (1995). are found for metallic spacers. In contrast, for insulating [9] J. J. de Vries et al., Phys. Rev. Lett. 78, 3023 (1997). spacers the coupling strength is expected to exponentially [10] Y. Endo, O. Kitakami, and Y. Shimada, Phys. Rev. B 59, decay with spacer thickness and to decrease with decreas- 4279 (1999). [11] R. Kläsges et al., Phys. Rev. B 56, 10 801 (1997). ing temperature. We emphasize that below 80 K the cou- [12] Y. Endo, O. Kitakami, and Y. Shimada, J. Magn. Soc. Jpn. pling is also metallic-type, because we observe oscillatory 21, 541 (1997). coupling behavior in Fig. 3 down to 20 K. [13] M. Schäfer et al., J. Appl. Phys. 77, 6432 (1995). The temperature dependence of J1 might be connected [14] D. E. Bürgler et al., Surf. Sci. 366, 295 (1996). with the resistance behavior of the FeSi spacer. The [15] D. E. Bürgler et al., Phys. Rev. B 56, 4149 (1997). temperature-dependent resistance of thin FeSi films is [16] M. P. Seah, in Practical Surface Analysis, edited by complex because different crystalline phases with differ- D. Briggs and M. P. Seah (John Wiley & Sons, Sussex, ent resistivity behaviors compete: Fe England, 1990), Vol. 1, Chap. "Quantification of AES and 12xSix with x 0.5 grows either in the epitaxially stabilized B2 phase [26] XPS," pp. 201­255. or -depending on thickness and deposition temperature [17] D. Berling et al., J. Magn. Magn. Mater. 191, 331 (1999). [18] S.-S. Yan, R. Schreiber, P. Grünberg, and R. Schäfer, [7,27]-in the bulk stable B20 phase [28]. For instance, J. Magn. Magn. Mater. 210, 309 (2000). a significant reduction of the number of carriers in the [19] M. Goiran et al., J. Magn. Magn. Mater. 172, 199 (1997). spacer near the Fermi level as a function of temperature [20] D. E. Bürgler et al., Phys. Rev. Lett. 80, 4983 (1998). in the B20 phase could lead to the observed behavior of [21] E. E. Fullerton and S. D. Bader, Phys. Rev. B 53, 5112 the bilinear coupling [8,29]. (1996). In conclusion, we have demonstrated for the first time [22] G. J. Strijkers, J. T. Kohlhepp, H. J. M. Swagten, and oscillatory interlayer coupling across FeSi spacer layers. W. J. M. de Jonge, Phys. Rev. Lett. 84, 1812 (2000). The epitaxial growth obtained by coevaporation allowed [23] J. C. Slonczewski, J. Appl. Phys. 73, 5957 (1993). us to produce samples with superior structural and com- [24] J. C. Slonczewski, Phys. Rev. Lett. 67, 3172 (1991). positional homogeneity up to 34 Å spacer thickness. The [25] M. Schäfer et al., J. Appl. Phys. 77, 6432 (1995). coupling strength increases with decreasing temperature- [26] E. G. Moroni, W. Wolf, J. Hafner, and R. Podloucky, Phys. at least down to 80 K. The temperature dependence, the Rev. B 59, 12 860 (1999). [27] K. A. Mäder, H. von Känel, A. Baldereschi, Phys. Rev. B oscillatory behavior, and the order of magnitude of the 48, 4364 (1993). coupling strength all imply that the coupling across or- [28] Z. Schlesinger et al., Phys. Rev. Lett. 71, 1748 (1993). dered FeSi can be understood in terms of the conventional [29] Z.-P. Shi, R. P. Singh, and B. M. Klein, Europhys. Lett. 29, models for interlayer coupling across metallic spacer lay- 585 (1995). 157202-4 157202-4