JOURNAL OF APPLIED PHYSICS VOLUME 89, NUMBER 1 1 JANUARY 2001 Interface morphology of a Cr 001...ÕFe 001... superlattice determined by scanning tunneling microscopy and x-ray diffraction: A comparison C. M. Schmidt, D. E. Bu¨rgler,a) D. M. Schaller, F. Meisinger, and H.-J. Gu¨ntherodt Institut fu¨r Physik, Universita¨t Basel, Klingelbergstrasse 82, CH-4056 Basel, Switzerland K. Temst Laboratorium voor Vaste-Stoffysika en Magnetisme, Katholieke Universiteit Leuven, Celestijnenlaan 200 D, B-3001 Leuven, Belgium Received 19 April 2000; accepted for publication 11 October 2000 A Cr 001 /Fe 001 superlattice with ten bilayers grown by molecular beam epitaxy on a Ag 001 substrate is studied by in situ scanning tunneling microscopy STM and ex situ x-ray diffraction XRD . Layer-resolved roughness parameters determined from STM images taken in various stages of the superlattice fabrication are compared with average values reported in the literature or obtained from the fits of our XRD data. Good agreement is found for the rms roughnesses describing vertical roughness and for the lateral correlation lengths characterizing correlated as well as uncorrelated interface roughness if peculiarities of STM and XRD are taken into account. We discuss in detail i the possible differences between the STM topography of a free surface and the morphology of a subsequently formed interface, ii contributions due to chemical intermixing at the interfaces, iii the comparison of XRD parameters averaged over all interfaces versus layer-resolved STM parameters, and iv the question of the coherent field of view for the determination of rms values. © 2001 American Institute of Physics. DOI: 10.1063/1.1330770 I. INTRODUCTION atomic force microscopy ATM -provides very detailed direct-space information about a free surface that only after Magnetic multilayers reveal fascinating physical proper- measuring-upon deposition of a subsequent film of a dif- ties as giant magnetoresistance GMR 1,2 and oscillatory ferent material-will transform into a buried interface. magnetic interlayer exchange coupling.3­5 Both phenomena Hence, the information gained by STM describes one single are known to be highly structure sensitive. In particular, interface and strongly relies on the assumption that the mor- thickness fluctuations of the nonferromagnetic interlayers re- phology of a buried interface is sufficiently well described by sulting from uncorrelated interface roughness crucially affect the exchange coupling properties6,7 or give rise to biqua- the corresponding initial free surface. The high degree of dratic coupling,8 and GMR depends on the quality of the structural detail information obtained by STM comes at the interfaces due to its origin from spin-dependent interface cost of low statistics compared to XRD due to its nature as a scattering.9 Interface roughness in general must be character- near-field technique. Obviously, the combined use of both ized by a whole set of parameters such as rms roughness complementary techniques STM and XRD promises strong always associated to some sampling length measured within advantages to get more reliable interface characterizations, the plane of the interface , in-plane correlation lengths, ter- e.g., by using a roughness model derived from STM images race sizes and shapes, profiles of atomic intermixing, atomic as the starting point of the fitting procedure for the XRD data displacements, and many more depending on the specific in- analysis. terface. In this article we present a combined in situ STM and ex Two main courses for the characterization of the inter- situ XRD study of a Cr 001 /Fe 001 10 multilayer. We dis- faces in metallic layered structures have been followed: i cuss interface roughness parameters deduced from direct- x-ray diffraction XRD is a widely spread technique that space images of the various surfaces occurring during sample allows the characterization of buried interfaces, but usually fabrication and complement our findings with subsurface requires a minimum number of the order of ten interfaces to sensitive diffraction measurements performed with the very yield sufficient signal intensities. The resulting interface pa- same sample after completion of its superlattice structure. rameters correlation lengths, interface widths, and chemical intermixing profiles represent an averaged interface and also The Cr 001 /Fe 001 superlattice is a suitable and physi- depend on the model assumptions plugged into the fitting cally relevant model system for such a comparative study: procedure. ii Imaging by scanning tunneling microscopy Oscillatory magnetic interlayer coupling and GMR have both STM -or any other scanning probe technique such as been discovered in Cr/Fe layered structures, and since then Cr/Fe has served as a model systems in the field of thin film magnetism. Much effort has already been put into the char- a Author to whom correspondence should be addressed; Present address: acterization of the interfaces by different techniques, among Institut fu¨r Festko¨rperforschung, Forschungszentrum Ju¨lich GmbH, Ger- many; electronic mail: D.Buergler@fz-juelich.de them STM6,7 and XRD,11 and interesting properties such as 0021-8979/2001/89(1)/181/7/$18.00 181 © 2001 American Institute of Physics Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp 182 J. Appl. Phys., Vol. 89, No. 1, 1 January 2001 Schmidt et al. interface alloying affecting the phase and strength of inter- layer coupling,12,13 subtle correlations between interface morphology and interlayer coupling,6,7 and an increase of the GMR effect with increasing interface roughness14,15 have been reported. II. EXPERIMENT Sample preparation and all measurements, with the ex- ception of XRD, are performed in an UHV system with a base pressure of 5 10 11 mbar that is equipped with a mo- lecular beam epitaxy deposition system, UHV­STM, low- energy electron diffraction LEED , Auger and x-ray photo- emmission electron spectroscopy AES, XPS , and an in situ FIG. 1. Longitudinal MOKE magnetization curve of the Cr 001 /Fe 001 10 sample in the intermediate trilayer stage of preparation see Ref. 10 . The magnetooptical Kerr effect MOKE setup that we operate in external magnetic field is applied parallel to a 100 magnetic easy axis of the longitudinal configuration. the Fe 001 layers. Arrows indicate the relative orientation of the magneti- A 150-nm-thick Ag 001 buffer layer grown on Fe- zations of the Fe layers. precovered GaAs 001 wafers at TS 380 K and postan- nealed at TA 570 K serves as substrate system for the mag- III. RESULTS netic multilayer. We have previously presented a detailed investigation of the morphological properties of the Ag 001 Figure 1 shows a MOKE loop in units of the saturation buffer layer:16 STM images reveal terraces with a mean magnetization MS measured in situ after stopping the prepa- width of approximately 35 nm that are separated by mon- ration process at the trilayer level. The contour of the mag- atomic steps. Most of these steps originate from screw dis- netization curve is typical for Fe 001 /Cr 001 /Fe 001 locations which are found to be the representative kind of trilayers grown at room temperature on the Ag 001 /Fe/ defect in this substrate system. Meanwhile we have been able GaAs 001 substrate system with a Cr thickness in the range to extend the average Ag terrace width by about a factor of between 2 and 3 nm.7 It reveals a characteristic plateau at three by using GaAs 001 wafers which are passivated by an M/MS 0 reflecting antiferromagnetic interlayer exchange amorphous As cap instead of oxidized GaAs 001 coupling at small external fields and two other plateaus at substrates.17 M/MS 0.5 resulting from 90° alignment of the magneti- The multilayer itself consists of ten repetitions of zation vectors at intermediate fields as indicated by the pairs Cr 001 /Fe 001 grown at room temperature at a deposition of arrows. rate of 0.01 nm/s. We intermit the preparation process at The comparison of the LEED patterns of the Ag 001 various stages to take STM, MOKE, and/or LEED data. The substrate Fig. 2 a and of the completed Cr 001 / nominal layer thickness in contrast to the one measured by Fe 001 10 superlattice Fig. 2 b confirms the single crys- ex situ XRD is monitored by a quartz microbalance; for the talline quality of the entire structure. The epitaxial relation- Fe films the nominal thickness reads 5 nm, whereas for the ship reads as follows: the bcc­Cr 001 100 axes and the Cr layers it amounts to 2.5 nm. Fe 001 /Cr 001 /Fe 001 bcc­Fe 001 100 axes are parallel to each other and also trilayers with wedge-shaped Cr spacers grown at elevated parallel to the fcc­Ag 001 110 axes. temperatures are expected to exhibit interlayer exchange STM overview and detail images are recorded from the coupling oscillations with a periodicity in Cr thickness very sample in various stages of preparation: An STM overview close to 2 ML 0.29 nm.5 By verifying these 2 ML oscilla- image i.e., an image with a scan area of 400 400 nm2 of tions using MOKE measurements on trilayer samples pre- the bottom Fe film is shown in Fig. 3 a . The shape and pared accordingly7 we estimate the absolute error of the nominal thickness measurement to be on the order of 10%, whereas the relative reproducibility proves to be better than 5%. The cleanness of the layers is confirmed by XPS and AES. All morphological, chemical, and magnetic character- izations are performed at room temperature. For the ex situ XRD analysis the sample is coated with a 5-nm-thick Ag protection layer. The XRD experiments are carried out by ­ 2 scans on a Rigaku diffractometer with a 12 kW rotating anode and using Cu K radiation ( 0.154 nm . The diffractometer is equipped with a post- sample crystal monochromator and a Ni filter. The multilayer samples are mounted on a thin-film attachment. The step size FIG. 2. (1 1) LEED patterns taken at 50 eV: a Ag 001 substrate and b of the measured data was 0.01°, and a scintillation counter top Cr 001 film of the complete Cr 001 /Fe 001 was used as a detector. 10 multilayer see Ref. 10 . The patterns are displayed with an arbitrary relative orientation. Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp J. Appl. Phys., Vol. 89, No. 1, 1 January 2001 Schmidt et al. 183 FIG. 3. STM overview images image size: 400 400 nm2 of various Fe FIG. 4. STM overview images image size: 400 400 nm2 of various Fe and Cr surfaces occurring in the preparation process of the and Cr surfaces occurring in the preparation process of the Cr 001 /Fe 001 10 multilayer grown on the Ag 001 /Fe/GaAs 001 sub- Cr 001 /Fe 001 10 multilayer grown on the Ag 001 /Fe/GaAs 001 sub- strate system. Insets: detail images 50 50 nm2 . a Fe (z range: 1.0 nm , strate system. Insets: detail images 50 50 nm2 . a Fe/ Cr/Fe 5 (z range: b Cr/Fe (z range: 1.0 nm . The derivative along the fast scan direction has 1.5 nm , b Cr/Fe 10 (z range: 2.0 nm . The derivative along the fast scan been added to the plane-subtracted raw data for contrast enhancement see direction has been added to the plane-subtracted raw data for contrast en- Ref. 10 . hancement see Ref. 10 . arrangement of the large-scale contrast are very similar to the where step structure of the bare Ag 001 substrate, and hence, it is induced by the substrate. However, the terraces in between 1 two substrate-induced steps are neither structureless nor flat. H r, H r A z z r d2 2 A The surface is covered with hillocks as revealed by the detail image i.e., an image with a scan area of 50 50 nm2 in the is the two-dimensional height-height correlation function de- inset. We statistically quantify the vertical roughness of rived from the surface profiles z(r) of STM images. Thus, R overview and detail images by calculating the rms value corresponds to the mean separation between typical features, z2 , and the lateral roughness of detail images by cal- i.e., the average distance of two adjacent hillocks. The offset culating the lateral correlation length R. The latter quantity is of z(r) is such that z 0. Therefore, with the normalization determined by the position of the nearest-neighbor maximum chosen in Eqs. 1 and 2 PCF(0) 2 holds. in the pair correlation function As discussed in Ref. 18 Fe grows on Ag 001 at room temperature as a continuous, single-crystalline film with a 1 2 PCF r rough surface. For the data presented in the inset of Fig. 3 a 2 H r, d , 1 0 R detail Fe 6.2 nm and Fe 0.13 nm. Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp 184 J. Appl. Phys., Vol. 89, No. 1, 1 January 2001 Schmidt et al. FIG. 5. High-angle x-ray diffraction spectrum of the Cr2.5 nm /Fe5 nm 10 multilayer grown on the Ag 001 /Fe/GaAs 001 substrate system. The left FIG. 6. Schematic drawing of the structure model used for the simulation vertical axis belongs to the upper graph measurement , the right one to the and fitting of the XRD patterns. vertically shifted best fit. The fit nicely reproduces position, number, and relative intensity of the bilayer satellites see Ref. 10 . In Fig. 5, for clarity the fitted curve is vertically shifted below the measured curve by one order of magnitude verti- Figures 3 b and 4 a depict images of the first Cr layer cal axis on the right-hand side . An additional spectrum ob- and the sixth Fe layer, respectively, and Fig. 4 b shows the tained from a pure Ag 001 /Fe/GaAs 001 sample allows to tenth Cr layer, i.e., the surface after completion of the separately fit and subtract the substrate contribution. Taking multilayer. Generally, from the appearance of the STM data, into account only the Cu K 1 line and using the STM data in it is impossible to distinguish between Fe and Cr surfaces: all setting the range of reasonable roughness values for the x-ray morphologies are dominated by a rough, irregular structure data, we arrive at a stable solution and a fair agreement be- due to growth hillock as demonstrated by the respective de- tween data and fit in the sense that the position and the num- tail images in the insets. ber of bilayer satellites are nicely reproduced as well as their The upper curve in Fig. 5 shows the measured ­ 2 relative intensities. The asymmetry in sharpness of the satel- high-angle XRD spectrum of the sample on a logarithmic lite peaks on the left and right-hand side of the Cr/Fe 002 scale vertical axis on the left-hand side . The highest apex in peak is recognized, too. The fit produces identical layer sepa- the pattern is produced by the GaAs 001 substrate. Its split- rations in the growth direction for Fe and Cr, namely 0.143 ting into two peaks comes from the x-ray beam that is not nm. From the fit the average thickness of the Fe layers is perfectly monochromatic but includes contributions from determined to be 4.05 nm, whereas the average Cr thickness both the Cu K 1 and the Cu K 2 lines, which differ in is 2.12 nm, i.e., the fitted average bilayer thickness measures wavelength by 0.25%. Note that these two peaks are almost 6.17 nm, which is 82% of the nominally deposited thickness. two orders of magnitude stronger than the peaks produced by Calculating the model curve using the nominal thickness val- the multilayer. To the left of the GaAs signal the bcc­Cr/ ues derived from the quartz microbalance does not reproduce Fe 002 fundamental peak can be observed. The periodic the experimental pattern in a satisfactory manner. From the modulation of the superlattice is demonstrated by the equi- best fit the average rms roughness for the Fe surfaces can be distant first to third-order superlattice peaks that are visible calculated as ¯XRD Fe 0.431 nm. The corresponding quantity on both sides of the main peak. for the Cr surfaces is ¯XRD 0.345 nm. The lower graph displays the best fit to the data, which Cr The low-angle measurements taken from our multilayer has been analyzed using the Suprex modeling and fitting do not exhibit distinct multilayer peaks, and it has not been program described in Ref. 19. Imperfections in the possible to fit the data. The problem arises probably i from multilayer are included by introducing a number of param- the fact that the atomic scattering powers of Fe and Cr are eters, as schematically shown in Fig. 6. For crystalline lay- very close to each other, which diminishes the contrast be- ers, roughness is included by assuming the presence of ran- tween the two materials and makes it difficult for the super- dom variations in the number of monolayers in the lattice structure to show up clearly, and ii from the strong crystalline layer indicated by NA and NB for materials A and substrate contribution to the total intensity that cannot be B, respectively . These fluctuations are, therefore, named dis- unambiguously separated from the weak multilayer signal. crete, i.e., quantized in steps equal to the lattice spacing and are presumed to have a Gaussian distribution. Furthermore, it is assumed that there can be a fluctuation of the interface IV. DISCUSSION distance, i.e., the vertical distance between two dissimilar First, we would like to discuss the assumption that STM atoms at the interface between two layers. For high-angle images of a free surface represent the morphology of a sub- XRD data, Suprex allows a fitting of the patterns by relying sequently formed interface. In the case of intermixing inter- on a one-dimensional kinematical structure model, implying faces may change during growth. It was shown by Davies that lateral correlations are not included. et al.12 by means of STM and scanning tunneling spectros- Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp J. Appl. Phys., Vol. 89, No. 1, 1 January 2001 Schmidt et al. 185 copy that the initial Cr growth on Fe 001 whiskers leads to the formation of a Cr­Fe alloy within the first monolayers with temperature dependent composition. At a preparation temperature TS 570 K, the first predominantly Cr layer oc- curs at a Cr coverage of 2­3 ML. Using angular resolved AES, Heinrich et al.13 found that for TS 570 K the interface intermixing is mostly confined to the two topmost atomic Fe layers and that the degree of intermixing is nearly 50%. Therefore, intermixing is likely to occur in our sample al- though to a lesser degree than observed by these authors because of the lower preparation temperature, TS 300 K. However, even in the case of moderate intermixing a STM image of an Fe surface may still be regarded as an approxi- mation of the resulting Cr/Fe interface morphology upon progressing in the multilayer fabrication, if we assume the FIG. 7. Lateral correlation lengths R as a function of multilayer thickness chemically diffuse Cr/Fe interface to be centered around the calculated via the pair correlation function Eqs. 1 and 2 from STM STM representation of the topography of the Fe layer. Con- detail images. Gray rhombuses stand for Fe data points and white triangles cerning our multilayer structure with alternating Cr/Fe and mark Cr data points. The dashed line is fitted to the data points and goes with the multilayer thickness to the power of 0.2. Inset: PCF calculated from Fe/Cr interfaces, STM data of Cr surfaces regarded as ap- the top Cr surface of the multilayer see Ref. 10 . proximations of Fe/Cr interfaces are supposed to excel the ones of the Cr/Fe interfaces by far, since chemically sharp interfaces are reported for the growth of Fe on Cr 001 .13 tained from small-angle XRD scans of a similar room tem- Hence, a detailed and quantitative comparison of STM and perature Cr1.7 nm /Fe5.2 nm 9 sample. In agreement with our XRD-derived parameters characterizing the interface mor- STM findings, they encounter the presence of two lateral phology seems legitimate. length scales. The smaller one ( 5 nm is connected with A qualitative description of the STM images involves uncorrelated roughness and corresponds to our hillock struc- two different lateral length scales. The steps that cause the ture with R's in the range between 6 and 11 nm. The larger large-scale image contrast in Figs. 3 and 4 separate Ag buffer lateral length scale of Schreyer et al. ( 200 nm is linked to layer terraces which are on average 100 nm wide and propa- a high degree of correlation and is attributed to the Ag sub- gate through the Cr/Fe layer stack, with their distinctness strate template, too. As mentioned by Schreyer et al.11 all vanishing during the growth of the multilayer: The sharp step their absolute values may only be considered as rough order structures visible in Figs. 3 a and 3 b transform upon pro- of magnitude estimates. Hence, the STM­XRD comparison gressing through the superlattice into modulations with a of the lateral interface roughness parameters yields satisfac- comparable vertical dynamic range and a wavelength of the tory agreement. order of several times the mean terrace width in Figs. 4 a The STM and XRD parameters for the vertical rough- and 4 b . Locally, the morphologies are dominated by ness are displayed in Fig. 8. The averaged rms roughness growth hillocks; the mean hillock separation of the order of values derived from our XRD measurements, ¯XRD Fe and only a few nanometers increases steadily by roughly a factor ¯XRD Cr , are shown as horizontal lines smallest rhombuses of 2 from the bottom Fe surface to the topmost Cr surface. A Fe and triangles Cr together with the layer-resolved data quantitative analysis of the lateral correlation lengths R con- points Fe,Cr PCF(0) derived from the STM detail over- firms the latter trend: In Fig. 7, R is plotted versus the nomi- nal multilayer thickness.20 Gray rhombuses indicate Fe sur- faces and white triangles symbolize Cr surfaces. Independent of the respective surface material present, R gets larger with increasing layer thickness-approximately proportional to the total multilayer thickness to the power of 0.2 dashed curve in Fig. 7 . An exemplary PCF function calculated from the topmost Cr surface Fig. 4 b is provided in the inset of Fig. 7. From the inequality of the R's we can directly con- clude that the interface roughnesses cannot be correlated across the layers on the lateral length scale of the growth hillocks, i.e., a few nanometers. Therefore, layer thickness fluctuations within each layer must be present. This is indi- rectly confirmed in our MOKE data Fig. 1 by the clear observation of 90° coupling in the trilayer state: In the framework of Slonczewski's model8 spacer layer thickness fluctuations are a necessary precondition for 90° coupling. FIG. 8. rms roughnesses obtained from XRD measurements small sym- Schreyer et al.11 have arrived at fair approximations for bols , STM overview images medium-sized symbols , and STM detail im- ages large symbols , plotted against the nominal multilayer thickness see the correlated and uncorrelated lateral correlation lengths ob- Ref. 10 . Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp 186 J. Appl. Phys., Vol. 89, No. 1, 1 January 2001 Schmidt et al. view images largest medium sized rhombuses Fe and 2 ¯XRD 2 ¯XRD 2. 3 triangles Cr . As it was the case for the lateral correlation Fe Cr length, the STM rms values increase with increasing We obtain 0.258 nm. Assuming that chemical intermix- multilayer thickness regardless of whether the respective ing only occurs at the Cr/Fe interface and that it causes the layer surface is Fe or Cr. All values with the exception of full difference between ¯XRD XRD Fe and ¯Cr we can estimate an overview upper bound of the effect of chemical intermixing. Depend- Cr of the complete multilayer lie below the correspond- ing average encountered by XRD ( ¯XRD ing on the detailed assumptions about the intermixing the Fe,Cr). For a comparison with the XRD values one has to compute the average of all value of corresponds to a thickness of the FeCr alloy layer-resolved interface roughnesses including those of the layer at the Cr/Fe interface of approximately 3 ML. This is interfaces that have not been imaged by STM . Evidently certainly a reasonable value for an upper limit12,13 indicating one finds that the averaged STM-derived rms values are sys- that the surface does not undergo significant geometric tematically smaller than those derived from XRD: overview changes when a subsequent layer is deposited even when Fe,Cr ¯XRD modest intermixing occurs. The center of the alloy layer Fe,Cr . We explain this difference with three arguments. First, a vertical length scale is always associated to some shows similar geometric fluctuations as the initial free sur- face, and chemical fluctuations due to the interface alloy lateral sampling length that defines the longest wavelength of simply add to the geometric fluctuations. This scenario con- the roughness which is taken into account for the determina- firms our assumption of a chemically diffuse interface which tion of roughness parameters. In STM we can tune this field is centered at the topography of the initial free surface. of coherent view very easily just by varying the scan range. Additional XRD simulations with a single rms value and The medium-sized symbols in Fig. 8 display the rms rough- assuming an intermixed Fe nesses as calculated from the STM overview images of Figs. xCr1 x layer show that the main features in the XRD pattern are more sensitive to the rough- 3 and 4. We always find detail overview. The main differ- ness parameters than to interdiffusion parameters. They do ence between the two sets of data is that the latter reflects to not provide a unique determination of thickness and compo- a much larger extent the Ag substrate contribution. For large sition x of the alloy layer. This is due to the fact that the multilayer thicknesses, overview does not approach detail im- XRD pattern is not changing very much and that a compara- plying that the influence of the Ag substrate steps does not tively large number of new parameters enters the problem. vanish, but rather smears out upon growth, as described be- The analysis is complicated because the interdiffusion causes fore. In agreement with Ref. 11 this scenario involves a high slight changes in the intensity of the satellite peaks, which degree of correlated roughness on the length scale of the can be compensated in the fit by the very influential back- substrate terrace width. A problem with roughness param- ground from the GaAs substrate peak. eters determined by XRD is precisely that the field of coher- In conclusion, roughness parameters derived from STM ent view is not well-known. In textured Nb/Cu multilayers images taken in various stages of superlattice fabrication Temst et al.21 have encountered a sampling length of about compare well with parameters obtained from the fitting of the grain size 45 nm by an ex situ XRD and ex situ AFM XRD spectra of the completed structure and agree with pre- comparison. In our single-crystalline samples the coherent vious XRD results of Schreyer et al.11 The comparison field of view in the XRD experiment might still be larger shows that STM images of free surfaces indeed yield valu- than the one connected with the STM overview images 400 able information about the morphology of subsequently nm and may thus explain the larger rms values encountered formed interfaces. Chemical intermixing occurring during in- in XRD. terface formation leads to an alloy layer which follows the Second, as has also been noted before in Ref. 22, the topography of the initial free surface. The width of the alloy vertical roughness values obtained from XRD refinement layer leads to an additional contribution to the rms roughness procedures measures deviations from the ideal multilayer measured by XRD but not by STM. The steady increase of structure consisting of i interface roughness and ii layer the layer-resolved rms roughness ( ) and lateral correlation thickness variations from one layer to the next along the length R with the number of layers in the superlattice indi- multilayer. The second contribution tends to increase the rms cates that XRD-derived parameters can only be understood values derived from XRD as compared to the STM-derived as averages over interfaces of with widely spread 's and parameters. R's. The XRD rms roughness is larger than the average of Third, as stressed before, surface topographies might the STM-derived values even for the interface type showing change geometrically and/or chemically when turning into no chemical intermixing ( ¯XRD overview Cr Cr ). This fact points interface morphologies upon deposition of additional layers. out that the coherent field of view in XRD is larger than the In particular, the Cr/Fe interface-in contrast to the Fe/Cr STM image size of 400 400 nm2. interface-is well known to exhibit chemical intermixing. ¯XRD XRD Fe ¯ Cr could reflect the chemical broadening of the Cr/Fe interface regions. The STM images do not show this ACKNOWLEDGMENTS trend since the actual interfaces with the chemically inter- Financial support from the Swiss National Science mixed regions are formed after the STM measurement. Foundation and the Swiss Kommission fu¨r Technologi- Hence, chemical intermixing seems to occur. The additional etransfer und Innovation is gratefully acknowledged. K. T. is roughness contribution XRD XRD to ¯ Fe compared to ¯Cr can be a Postdoctoral Research Fellow of the Fund for Scientific calculated as Research-Flanders FWO . Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp J. Appl. Phys., Vol. 89, No. 1, 1 January 2001 Schmidt et al. 187 1 G. Binasch P. Gru¨nberg, F. Saurenbach, and W. Zinn, Phys. Rev. B 39, 13 B. Heinrich, J. F. Cochran, D. Venus, K. Totland, D. Atlan, S. Govorkov, 4828 1989 . and K. Myrtle, J. Appl. Phys. 79, 4518 1996 . 2 M. N. Baibich et al., Phys. Rev. Lett. 61, 2472 1988 . 14 R. Schad, P. Belie¨n, G. Verbanck, V. V. Moshchalkov, Y. Bruynseraede, 3 P. Gru¨nberg, R. Schreiber, Y. Pang, M. B. Brodsky, and H. Sowers, Phys. H. E. Fischer, S. Lefebvre, and M. Bessiere, Phys. Rev. B 59, 1242 Rev. Lett. 57, 2442 1986 . 4 S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, 2304 1999 . 15 1990 . D. Olligs, Ph.D. thesis, Universita¨t zu Ko¨ln, 1999. 5 16 J. Unguris, R. J. Celotta, and D. T. Pierce, Phys. Rev. Lett. 67, 140 1991 . D. E. Bu¨rgler, C. M. Schmidt, J. A. Wolf, T. M. Schaub, and H.-J. 6 D. T. Piere, J. A. Stroscio, J. Unguris, and R. J. Celotta, Phys. Rev. B 49, Gu¨ntherodt, Surf. Sci. 366, 295 1996 . 14564 1994 . 17 D. E. Bu¨rgler and F. Meisinger unpublished . 7 C. M. Schmidt, D. E. Bu¨rgler, D. M. Schaller, F. Meisinger, and H.-J. 18 D. E. Bu¨rgler, C. M. Schmidt, D. M. Schaller, F. Meisinger, R. Hofer, and Gu¨ntherodt, Phys. Rev. B 60, 4158 1999 . H.-J. Gu¨ntherodt, Phys. Rev. B 56, 4149 1997 . 8 J. C. Slonczewski, Phys. Rev. Lett. 67, 3172 1991 . 19 E. E. Fullerton, I. K. Schuller, H. Vanderstraeten, and Y. Bruynseraede, 9 S. S. P. Parkin, Phys. Rev. Lett. 71, 1641 1993 . 10 Phys. Rev. B 45, 9292 1992 . The pictograph included in the figure indicates the level of sample prepa- 20 ration at which the displayed measurement is performed. Each Fe Cr I.e., the thickness as measured by the quartz microbalance during growth. 21 layer is symbolized by a dark gray light gray colored rectangle. K. Temst, M. J. Van Bael, B. Wuyts, C. Van Haesendonck, Y. 11 A. Schreyer, J. F. Ankner, Th. Zeidler, H. Zabel, M. Scha¨fer, J. A. Wolf, Bruynseraede, D. G. de Groot, N. Koeman, and R. Griessen, Appl. Phys. P. Gru¨nberg, and C. F. Majkrzak, Phys. Rev. B 52, 16066 1995 . Lett. 67, 3429 1995 . 12 A. Davies, J. A. Stroscio, D. T. Pierce, and R. J. Celotta, Phys. Rev. Lett. 22 I. Heyvaert, K. Temst, C. Van Haesendonck, and Y. Bruynseraede, J. Vac. 76, 4175 1996 . Sci. Technol. B 14, 1121 1996 . Downloaded 02 Oct 2001 to 148.6.178.100. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp