Vacuum 63 (2001) 337}344 Interface modeling in Cr/Fe/Cr sandwiches studied by CEMS M. Kubik , T. SDle9zak , M. Przybylski , W. KarasH , J. Korecki * Department of Solid State Physics, Faculty of Physics and Nuclear Techniques, University of Mining and Metallurgy, Al. Mickiewicza 30, 30-059 Krako&w, Poland Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences, ul. Niezapominajek 8, 30-239 Krako&w, Poland Abstract Conversion electron MoKssbauer spectroscopy (CEMS) was applied to model the buried interfaces in epitaxial 1}14 monolayer (ML) Fe(0 0 1) "lms sandwiched between Cr(0 0 1) layers. The local arrangement of Fe and Cr was derived by analysis of the experimental hyper"ne "eld distributions determined from 80 K CEMS spectra. A discontinuity in magnetic properties for the 6 ML Fe "lm suggests that a modi"cation of the magnetic structure occurs at this thickness. The Fe probe layer analysis showed that Cr/Fe and Fe/Cr interfaces are non-equivalent. The lower interface showed up to be smoother than the upper one. The Fe concentration pro"le obtained form the CEMS analysis displayed a considerable interface Fe}Cr alloying for "lms deposited at the room temperature. 2001 Elsevier Science Ltd. All rights reserved. PACS: 75.70.Ak; 76.80.#y Keywords: CEMS; Fe; Cr; Ultra thin epitaxial "lms; Interface structure; Alloying 1. Introduction tween ferromagnetic layers separated with a non- magnetic spacer depends strongly on the atomic The last decade brought a tremendous interest in structure of the interfaces. It has been suggested magnetic systems of nanometer thickness [1], that roughness and atomic interface intermixing mainly due to their application in recording tech- between Fe and Cr is responsible for this e!ect [3]. nique. Particularly, the indirect coupling phenom- Interface roughness suppresses the short-range enon that leads the ferro- or antiferromagnetic (with the period of 2 Cr monolayers) coupling oscil- arrangements of the sub-layer spins is of special lations, leaving the long-range ones (with the peri- attention and importance because the last is od of 12 Cr monolayers). Additionally, for the accompanied by the giant magnetoresistance Fe/Cr systems, the intrinsic and intricate anti- (GMR) observed for the "rst time in the Fe/Cr/Fe feromagnetism of chromium [4], strongly modi"ed system [2]. The indirect exchange coupling be- in layered structures [5], is involved in the coup- ling. The model studies using single crystalline whisker substrates [6] are of a great importance for * Corresponding author. Faculty of Physics and Nuclear Tech- niques, University of Mining and Metallurgy, Al. Mickiewicza 30, understanding the basic phenomena but often they 30-059 Krakow, Poland. Fax: #48-12-643-1247. do not relate directly to other systems, in which E-mail address: korecki@uci.agh.edu.pl (J. Korecki). the resulting structure depends crucially on many 0042-207X/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 2 - 2 0 7 X ( 0 1 ) 0 0 2 1 0 - X 338 M. Kubik et al. / Vacuum 63 (2001) 337}344 technological parameters. That is why we have tors with the accuracy of about 0.2 ML. In situ undertaken detailed studies of Fe/Cr multilayers MoKssbauer measurements were performed using an grown on the MgO(0 0 1) substrates, aimed to "nd e$cient 80}500 K CEMS spectrometer based on the correlation between magnetic and structural channeltron detection [8], with a 150 mCi properties. Here we report on the conversion elec- Co(Rh) source. The spectrometer geometry tron MoKssbauer spectroscopy (CEMS) experiments settled a "xed angle of 453 between the direction of for single Fe "lms sandwiched between Cr layers. the ray propagation and the sample normal. The CEMS o!ers a unique possibility to analyze both, room temperature spectra could be measured also the structural and the magnetic properties of buried ex situ with a #owproportional He/CH interfaces at atomic scale. Moreover, the isotopic  detector and with the ray propagation direction parallel to sensitivity of the method allows depth pro"ling of the sample normal. For all samples, a 20 nm Cr iron "lms using a Fe probe layer embedded dur- bu!er layer was used, on which Fe "lms of the ing the growth in a "lm consisting otherwise of thickness varying between 1 and 14 (0 0 1) atomic Fe. The probe layer concept is here especially layers were deposited. Finally, the Fe "lms were useful for verifying an asymmetry suggested for the capped with 5 nm Cr. Fe/Cr and Cr/Fe interfaces [7]. The atomic model of Fig. 1 shows typical LEED patterns for the the single "lm interfaces obtained from the present Mg(0 0 1) substrate (Fig. 1a) and for the Cr bu!er analysis sets a basis for further studies of the coupl- layer (Fig 1b). A lowintensity background and ing phenomena in the Fe/Cr multilayer systems. sharp spots in the whole range of the electron energy for the MgO substrate contrast with the feature observed for the Cr surface. To our know- 2. Sample preparation and characterization ledge, the epitaxy of Cr on MgO(0 0 1) was not reported before in the literature. The lattice spac- The sample preparation and the characterization ings 0.407 nm for Cr along [1 1 0] and 0.420 nm for was performed in situ, using a multi-chamber UHV MgO along [1 0 0] are well matched when the both system [8] with the base pressure 1;10\ mbar. lattices are rotated by 453. Accordingly, our LEED The system is equipped with a load-lock facility, patterns clearly demonstrate the epitaxial growth a universal sample mounting and transfer system, with the geometrical relations identical as in the standard surface characterization methods (a 4- case of Fe on MgO(0 0 1), i.e. Cr(0 0 1) [001] grid electron optics for LEED and AES), a MBE # MgO(0 0 1) [0 1 1]. With increasing electron en- system for deposition of several metals including ergy, all spots vary between sharp and di!used Fe and Fe isotopes and Cr and a CEMS shape periodically. This behavior is typical for ir- spectrometer. All samples were deposited on regular steps with randomly distributed step MgO(0 0 1) substrates that were cleaved ex situ in heights, widths or edge orientation [9]. By analogy pure N atmosphere prior to the introduction into with STM studies of iron bu!er layers deposited in the UHV system, where they were annealed at the same conditions [10] and basing on crystallo- 4003C for 1 h. Such treatment resulted in a clean graphic and electronic Fe and Cr similarity, we surface showing only traces of carbon contamina- suppose that monoatomic steps with randomly tion and a perfect background-free 1;1 LEED varying terrace width are responsible for the ob- pattern. The whole structure of the Cr/Fe/Cr sand- served LEED patterns. For the electron energy wiches was deposited on the as cooled substrates 95 eV ful"lling the out-of-phase conditions for (01) kept at about 330 K. Fe and Cr were deposited spots (Fig. 1b), the spot width is a measure of the from BeO crucibles heated from wrap-around average terrace width. A crude estimation, based on tungsten coils. The crucible assemblies were embed- the comparison with simulations for steps with ded in a water-cooled shroud and the pressure a Gaussian distribution of the terrace width [9], during the deposition was maintained in the gives us the edge atom fraction of the order of 20% 10\ mbar range. The "lm thickness was control- and respectively, assuming square terraces, the led during the deposition by quartz thickness moni- linear terrace dimension of a fewnanometers. M. Kubik et al. / Vacuum 63 (2001) 337}344 339 Fig. 1. LEED patterns and pro"les for (a) the MgO(0 0 1) substrate at the electron energy 85 eV and (b) 20 nm Cr(0 0 1) bu!er layer at 95 eV. The LEED characteristic did not change sub- 2a and b, respectively. All presented room temper- stantially for the further deposited Fe layer and ature spectra were measured with the proportional also for the Cr cap layer. From the qualitative counter, whereas the 80 K spectra were taken using LEED analysis it seems however that upon the Fe the in situ UHV CEMS spectrometer. For the thin- deposition the surface becomes smoother as evid- nest Fe "lms (N"1,2), pronounced temperature enced by narrowing of the spots in the out-of-phase e!ects are seen. Obviously, the Curie temperature conditions and by the lower level of the di!used for these "lms is below(for N"1) or near the room background intensity. temperature (for N"2). The temperature-induced In the following text, the symbol N denoting e!ects for all other samples are only quantitative, a sample Cr/Fe,/Cr is the number of the mono- observed as a several percent increase of the hyper- atomic layers in the "nal sandwich structure: "ne magnetic "elds. However, already by the visual MgO(0 0 1)/20 nm Cr(0 0 1)/N ML Fe(001)/5 nm inspection strong di!erentiation of the spectra is Cr(0 0 1). The thickness is given assuming that 1 Fe seen with increasing the Fe "lm thickness. Up to (Cr) (0 0 1) ML corresponds to 0.143 nm (0.144 nm) N"5 the spectra are characterized by a broad as follows from the bulk lattice constants. distribution of the hyper"ne magnetic "eld. Then, abruptly, at N"6, distinct spectral components with relatively narrow lines appear. Increasing the 3. Results and discussion of MoKssbauer analysis "lm thickness further makes the spectra dominated by a sharp bulk like component accompanied by The CEMS spectra measured for Cr/Fe,/Cr a broader one (with a smaller hyper"ne splitting), sandwiches at 300 and 80 K are summarized in Fig with the intensity decreasing going to the thicker 340 M. Kubik et al. / Vacuum 63 (2001) 337}344 Fig. 2. The CEMS spectra for Cr/Fe,/Cr sandwiches measured at (a) 300K and (b) 80K. The solid lines are the numerical "ts. "lms. Obviously, in the simplest picture, the "lm where n interior contributes to the bulk-like component,  and n are the numbers of the nearest and next nearest Cr neighbors of the given Fe atom and whereas the broad distribution comes from inter- c facial Fe atoms. The broad distribution of the mag-  and c are proportionality constants. The pre- viously reported experimental c netic hyper"ne "elds may originate either from  and c values di!er only slightly for bulk [13] and for interfaces a structural and chemical disorder or from a speci- [12,14]. It has to be pointed-out that they depend "c magnetic structure. The magnetic hyper"ne "eld on temperature and, especially for thinnest "lms, B at the Fe nucleus is determined by the chem- they should be derived from lowtemperature data. ical coordination. It has been commonly assumed At the interfaces, in the ultra thin "lms, the number that for the Fe}Cr interface, the alloy analogy ap- of allowed con"gurations is strongly limited by the plies quite well [11,12] and the B distribution planar geometry, however the interface roughness re#ects the distribution of the local atomic arrange- and alloying may smear out the chemical order ments of the Fe and Cr atoms. In this model, the characteristic for the sharp interface. The sudden B value at a given Fe nucleus is lowered from change of the spectrum character observed for the bulk Fe value by each Cr atom in the "rst and N"6 could suggest that interface sharpening the second coordination shell: occurs for this "lm thickness. It is however not accompanied by any detectable changes in the B (n,n)"B (0,0)}cn}cn, (1) corresponding LEED patterns. Therefore, we M. Kubik et al. / Vacuum 63 (2001) 337}344 341 Fig. 3. The results of numerical "t for Cr/Fe,/Cr sandwiches; (a) distribution of B for selected samples. For N"6, the distribution maxima are labeled with corresponding atomic con"gurations, according to the notation described in the text; (b) average 1B 2 values versus the "lm thickness N; (the inset in (b)) dependence of B/B"[1B (80K)2!1B (300K)2]/1B (80K)2 value on the "lm thickness. incline toward the interpretation of a magnetic and 5, with sharper maxima around 20 T, becomes origin of the transition observed at N"6. It can be peaked out at six well de"ned B due to a change of the magnetic structure (type of values for N"6. A transition in magnetic properties is also apparent ordering, magnitude of the magnetic moments) in from the dependence of the average 1B the Cr layers. Such phenomenon is highly probable 2 values on the "lm thickness N obtained from the "tted in viewof the sensitivity of the Cr magnetic struc- distributions (Fig. 3b). The 1B ture to the size e!ects [5], structural modi"cations 2 dependence ex- hibits notable cusps. The "rst one, at N"2, corres- [15] or proximity of a magnetic layer [16]. ponds to the abrupt change of the con"guration The spectra were analyzed numerically by "tting between 1 and 2 ML bcc "lms. In the (0 0 1) mono- a hyper"ne "eld distribution (HFD) using the layer "lm, neglecting the deviation from the layer Voigt-based method of Rancourt and Ping [17]. growth, all the Fe nearest neighbors (n.n.) are Cr The method describes the HFD by a sum of Gaus- atoms. For the 2 ML (0 0 1) "lm the number of the sian components for the B , isomer shift (IS) and Fe n.n. increases by 4 and for the 3 ML (0 0 1) "lm quadrupole splitting (QS) distributions. Only a lin- all Fe n.n. atoms are already Fe atoms. The two ear correlation between B and IS or QS can be "rst -Fe-like coordination shells are restored only used in the numerical procedure, which might be an in the 5 ML (0 0 1) "lm. Filling the coordination oversimpli"cation, leading to a systematical error shells by the Fe atoms explains the cusps in the when a broad range of B is analyzed. The number 1B of Gaussian components was increased gradually 2 versus N dependence but it cannot be responsible for the change in the character of the from 1 such that a minimum number of "tting B parameters were introduced. Once a statistically distribution. Even more intriguing is the in#u- ence of temperature on the average hyper"ne mag- ideal "t was obtained, increasing the number of the netic "eld visualized by the inset in Fig. 3b as the component did not change the distribution or any dependence of the 1B of the essential "t parameters. The numerical "ts 2 reduction between 80 and 300 K normalized to the 80 K value. The only are shown as the solid lines in Fig. 2 and their plausible explanation of the discontinuities at results are summarized in Fig. 3. Fig. 3a shows N"3 and 6 involves a change of the coupling selected characteristic distribution of B . It is between the Fe "lm and the Cr layers. This issue clearly seen that the broad distributions for N"3 (it will be discussed further elsewhere [18]) 342 M. Kubik et al. / Vacuum 63 (2001) 337}344 emphasizes the importance of the interplay be- interfaces studies [11,12,14]. The most striking dif- tween the structural and magnetic properties, ference is that our analysis gives c which must be considered in the interpretation of 'c, whereas in the previous studies the reverse relation was the MoKssbauer spectra in terms of a structural assumed basing on the data for dilute FeCr alloys model. Therefore, for the interface modeling, the [13]. Our methodology applies strictly to the ultra 6 ML "lm has been chosen, for which the magnetic thin "lm interfaces, giving consistent results for structure seems to be well established but still inter- a series of samples, it considers also the temper- facial atoms contribute predominantly to the MoKs- ature e!ects, which may alter surface and sub- sbauer spectrum. surface B The character of the B components in a di!erent way. Thus, we distribution for the question using the bulk alloy parameters for the 6 ML sample can be explained using Eq. (1). Six description of the interfacial hyper"ne magnetic maxima in the B distribution for N"6 (Fig. 3a) "eld with Eq. (1). Relation between hyper"ne "elds are distinctly resolved. The high "eld maximum and atomic con"gurations in thin "lms may be corresponds to the central part of the "lm, where considerably di!erent than in the bulk. the Fe atoms do not have Cr atoms as the n.n.'s and Using the derived relation between the hyper"ne the next n.n.'s. Such con"guration is denoted magnetic "eld and the atomic con"guration, the (n,n)"(0,0), accordingly to the notation used in interface modeling was made by adjusting the con- Eq. (1). The sharp interface would yield for the centration pro"le at the interfaces to reproduce the 6 ML "lm only two additional con"gurations: (4,1) intensity of the peaks in the experimental B for the surface interface layer and (0,1) for the distribution. A simple "lm model, assuming only subsurface one. The three additional maxima in the monoatomic steps, was not able to reproduce the B distribution for N"6 in Fig. 3a, corresponding experimental B to the con"gurations (3,1), (2,1) and (1,1), are then distribution. Therefore, the inter- face zone was allowed to spread over few (2 or 3) due to a deviation from the perfect interface caused atomic layers, where random two-dimensional by steps or Fe}Cr interfacial alloying. From the Fe}Cr alloy was formed. Then the con"guration maxima positions, the constants in Eq. (1) can probabilities were calculated accordingly to the bi- be precisely determined as: B (0,0)"34.5T, nomial distribution and converted to the B c distri- "2.99 T and c"2.66 T. The c and c values bution using Eq. (1). Neither this procedure gave di!er remarkably from those used in the previous a satisfactory description. The best solution could Fig. 4. Fe probe layer analysis of the interfaces. The CEMS spectra of 6 ML Fe samples, in which at "rst 3 ML Fe (Fe) monolayers were deposited on the Cr bu!er followed by 3 ML Fe (Fe) monolayers. The B histograms resulting from numerical "ts are shown with the spectra. M. Kubik et al. / Vacuum 63 (2001) 337}344 343 came reasonable and the contributions of di!erent spectral components could be reproduced within $3%. The resulting concentration pro"le of Fe in the "lm of the nominal thickness 6 ML is shown in Fig. 5. The lower interface is relatively sharp and the Cr atoms are found in two layers only, whereas the upper interface is formed by at least three layers, in which the Cr concentration is consider- able. Till now, such enhanced Fe}Cr intermixing was reported only at elevated preparation [21] or annealing [20] temperatures. The previous data concerned however a single crystalline substra- Fig. 5. The concentration pro"le of Fe in the "lm of the nominal te*the Fe whiskers [21] or the Cr single crystals thickness 6 ML. [20]. The presently observed interfacial alloying at the room temperature is favored by the step-like be found assuming that both the interfaces, the structure of the Cr bu!er layer. It is in the close lowera one, formed when the Fe layer is deposited agreement with the STM observation by Choi et al. on the Cr bu!er, and the uppera one are non- [20] that steps or island edges work as the reaction equivalent. This assumption has been veri"ed ex- sites for incorporation of the Fe adatoms into the perimentally using the Fe probe layer concept. Cr substrate. Two special 6 ML Fe "lms were prepared, in which at "rst 3 ML Fe (Fe) layers were deposited on the Cr bu!er, and then followed by 3 ML Fe 4. Conclusion (Fe). In the "rst samples the lowera, in the sec- ond one the uppera interface were probed in the The conversion electron MoKssbauer spectro- following CEMS measurements. The results of the scopy at lowtemperatures is a unique tool for probe layer CEMS measurements are showin Fig. studying the local structure of buried interfaces in 4. Indeed, the both interfaces are di!erent as seen the ultra thin Fe (0 0 1) "lms sandwiched between by the MoKssbauer spectroscopy. The lower inter- Cr(0 0 1) layers. The data analysis for the 1}14 ML face showed up to be smoother than the upper one. "lms, based on the correlation between the distri- The CEMS spectrum of the upper interface is dom- bution of the hyper"ne parameters and the "lm inated by the broad interfacial components, where- structure, revealed a discontinuity of the magnetic as for the lower one, the contribution of the bulk- properties for the "lm consisting of the 6 Fe mono- like component is not far from the value  layers. This e!ect was interpreted in terms of  expected for the perfectly sharp and smooth boundary be- a modi"cation of the magnetic structure in the Cr tween Cr and Fe. The relative contribution of the layers. The Fe probe layer method proved that spectral components cannot be taken directly as Cr/Fe and Fe/Cr interfaces are non-equivalent. The the measure of the composition because isotope lower interface showed up to be smoother than the intermixing can take place during the deposition. upper one. The Fe concentration pro"le obtained Nevertheless, the probe layer analysis clearly pro- from the CEMS analysis displayed a considerable ved that the growths of Fe on Cr and Cr on Fe are interface Fe}Cr alloying for the "lms deposited at di!erent, as postulated previously from STM stud- the room temperature. ies of the Cr growth on Fe whiskers [19] and of the Fe growth on the Cr single crystal [20], and from Acknowledgements the CEMS studies of the Fe/Cr multilayers [7]. When non-equivalent interfaces were introduced This work was supported by the Polish State to our interface model, the agreement between the Committee for Scienti"c Research, Grants No. experimental and simulated distributions of B be- 7 T08C 002 16 and 2 P03B 142 17. 344 M. Kubik et al. / Vacuum 63 (2001) 337}344 References [11] Landes J, Sauer Ch, Brand RA, Zinn W, Mantl S, Kajcsos Zs. J Magn Magn Mater 1990;86:71. [1] Bland JAC, Heinrich B. Ultrathin magnetic structures. [12] Klinkhammer F, Sauer Ch, Tsymbal E, Handschuh S, Berlin: Springer, 1994. Leng Q, Zinn W. J Magn Magn Mater 1996;161:49. [2] Baibich MN, Broto JM, Fert A, Ngyuen Van Dau F, [13] Dubiel SM, Zukrowski J. J Magn Magn Mater Petro! F, Etienne P, Creuzet G, Friederich A, Chazelas J. 1981;23:214. Phys Rev Lett 1988;61:2472. [14] Zukrowski J, Liu G, Fritzsche H, Gradmann U. J Magn [3] Schad R, Barnas P, Belien P, Verbanck G, Potter CD, Magn Mater 1995;145:57. Fischer H, Lefebvre S, Bessiere M, Moshchalkov VV, [15] Demuynck S, Meersschaut J, Dekoster J, Swinnen B, Bruynseraede Y. J Magn Magn Mater 1996;156:339. Moons R, Vantomme A, Cottenier S, Rots M. Phys Rev [4] Fawcett E. Rev Mod Phys 1998;60:209. Lett 1998;81:2562. [5] Sonntag P, Bodeker P, Schreyer A, Zabel H, Hamacher K, [16] Fuchs P, Petrov VN, Totland K, Landolt M. Phys Kaiser H. J Magn Magn Mater 1998;183:5. Rev B 1996;54:9304. [6] Unguris J, Celotta RJ, Pierce DT. Phys Rev B [17] Rancourt DG, Ping JY. Nucl Instr and Meth B 1991; 1992;69:1125. 58:85. [7] Shinjo T, Keune W. J Magn Magn Mater 1999;200:598. [18] Kubik M, KarasH W, Korecki J. to be published. [8] Korecki J, Kubik M, Spiridis N, SDle9zak T. Acta Phys Pol A [19] Davies A, Stroscio Joseph A, Pierce DT, Celottaa RJ. 2000;97:129. Phys Rev Lett 1996;76:4175. [9] Ibach H, Henzler M. Electron spectroscopy for surface [20] Choi YJ, Jeong IC, Park J-Y, Kahng S-J, Lee J, Kuk Y. analysis. Berlin: Springer, 1997. Phys Rev B 1999;59:10918. [10] Spiridis N, Korecki J. unpublished. [21] Venus D, Heinrich B. Phys Rev B 1995;53:R1733.