PHYSICAL REVIEW B VOLUME 62, NUMBER 5 1 AUGUST 2000-I Enhancement of perpendicular and parallel giant magnetoresistance with the number of bilayers in FeÕCr superlattices M. C. Cyrille,1 S. Kim,1 M. E. Gomez,1,* J. Santamaria,1, Kannan M. Krishnan,2 and Ivan K. Schuller1 1Department of Physics, University of California-San Diego, La Jolla, California 92093-0319 2Materials Sciences Division, National Center for Electron Microscopy, Lawrence Berkeley Laboratory, University of California, Berkeley, California 94720 Received 28 December 1999 We have correlated a detailed quantitative structural analysis by x-ray diffraction, transmission electron microscopy, and high spatial resolution electron energy-loss spectroscopy imaging, with the magnetization and anisotropic magnetotransport properties in sputtered Fe/Cr superlattices. To accomplish this, we developed a technique for magnetotransport measurements in metallic superlattices with the current perpendicular to the plane of the layers CPP . Using microfabrication techniques, we have fabricated microstructured Fe/Cr pillars embedded in SiO2 and interconnected with Nb electrodes. Because of the uniform current distribution in the Nb electrodes and the minimization of the superlattice-electrode contact resistance, the method allows a simple and independent measurement of the superlattice resistance and giant magnetoresistance GMR . Structural and magnetic characterization of Fe 3 nm /Cr 1.2 nm N superlattices where N is the number of repetitions indicate that the roughness is correlated and increases cumulatively through the superlattice stack with no significant change in the antiferromagnetic coupling. Both the current in-plane and CPP GMR increase with N as the roughness increases. I. INTRODUCTION section, the resistance of the sample is in the m range which can be measured with conventional techniques. Sec- Since the discovery of giant magnetoresistance GMR in ond, because of the superconducting Nb electrodes, the cur- magnetic superlattices,1 much interesting experimental and rent distribution is uniform in the pillars and computer simu- theoretical work has been done to understand this phenom- lation is not necessary to access the superlattice resistance.9 enon. The GMR amplitude depends on several physical Third, the contact resistance between the superlattice and the properties: the magnetic structure via the interlayer ex- electrodes has been minimized. Because the smallest spuri- change coupling , the spin dependent electronic band struc- ous contact resistance induced by the fabrication process will ture and the spin dependent electron scattering.2­5 GMR contribute to the measured resistance, this third point is the measurements are usually carried out with a current in-plane key to have independent access to the intrinsic superlattice CIP geometry, which is a straightforward technique. MR resistance and magnetoresistance. Therefore, because the measurements in the current perpendicular to the plane contact resistance is negligible, no adjustments or corrections CPP geometry are not easily achievable because of the have to be done to the measured resistance. Finally, this small resistivity of the superlattices, although they allow the method can be applied to any superlattice system with no deconvolution of the electron scattering occurring in the fer- limitation on the layer thicknesses, provided that the satura- romagnetic bulk from those occurring at the interfaces.6,7 tion field is smaller than 1.2 T. We have used this method to Several groups have developed methods to measure the CPP determine the CIP and CPP magnetoresistance of Fe 3 nm / GMR in magnetic superlattices. Pratt et al.8 developed a su- Cr 1.2 nm ]N superlattices, where N is the number of rep- perconducting contacting technique together with a sensitive etitions, grown by dc magnetron sputtering. superconducting quantum interference device SQUID Because most of theoretical and experimental work un- based system to measure the small resistance of their derscore the importance of interfacial roughness, detailed samples. Gijs, Lenczowski, and Giesbers9 have measured the structural characterization is critical to further understand the CPP GMR in Fe/Cr superlattices up to room temperature GMR. We have developed an interesting imaging technique using microfabrication techniques with normal electrodes, using high-resolution electron energy-loss spectroscopy Gijs et al.10 and Ono and Shinjo11 have used V-groove sub- EELS in a transmission electron microscope TEM to strates to measure the GMR with the current at an angle to quantify the interfacial roughness in metallic superlattices. plane CAP geometry and Piraux et al.,12 Blondel et al.,13 Hence the superlattices interfacial disorder was characterized and Liu et al.14 have fabricated multilayered nanowires by quantitatively using two complementary techniques: low- electrodeposition. angle x-ray diffraction LAXRD , and energy-filtered imag- We report on a different method to measure the magne- ing using cross-section samples in an analytical TEM and totransport properties of metallic superlattices in the CPP was correlated with the superlattices magnetization and mag- geometry. Using microfabrication techniques, we have fabri- netotransport properties. LAXRD and EELS analysis of cated microstructured pillars interconnected with Nb elec- Fe 3 nm /Cr 1.2 nm N provide evidence that the rough- trodes. This method's advantages are as follows: first, be- ness is correlated and increases cumulatively through the su- cause of the high number of columns and their small cross perlattice stack with no significant effect on the antiferro- 0163-1829/2000/62 5 /3361 7 /$15.00 PRB 62 3361 ©2000 The American Physical Society 3362 M. C. CYRILLE et al. PRB 62 the Nb top layer is etched away. A final Nb layer is then immediately deposited in situ by molecular beam epitaxy MBE and patterned to form the connection between the pillars. When the Nb is superconducting, the current distri- bution in the Nb electrodes is uniform and the current flows perpendicular to the plane of the substrate inside the pillars. Figure 1 b shows an optical micrograph of a typical sample. The sample consists of 100 pillars in series and has a ``me- anderlike'' structure. Several contact pads can be used to perform a four-lead measurement. To measure the resistance and magnetoresistance in the CIP configuration, the superlattices are deposited directly on top of a Si substrate. Then a 40- m-wide bridge is defined using optical photolithography and the superlattice is etched away, allowing a four-lead measurement to be carried out in a well-defined geometry. dc and ac magnetotransport measurements were per- formed in a helium cryostat equipped with a 9-T supercon- ducting solenoid. The measurement temperature is 2.0 0.1 K and the applied field is always parallel to the sub- strate plane. The superlattices structure was thoroughly characterized by low angle x-ray diffraction using a Rigaku rotating anode diffractometer with Cu K radiation. The specular spectra were fitted with the SUPREX refinement program17 in order to estimate the layers interfacial roughness. A quantitative structural analysis of the superlattices has been achieved with TEM and high spatial resolution EELS in the cross-sectional FIG. 1. a Schematic cross section of a CPP sample; arrows geometry. Fe and Cr have similar lattice parameters and also labeled 1 and 2 correspond to the patterns defined by the first exhibit very close scattering factors for elastic scattering of and second etches, respectively; b optical micrograph of a typical electrons. Hence conventional diffraction contrast and/or sample. Arrows show the current path. phase contrast imaging in a TEM will neither resolve the layers nor the details of their interface structure. However, magnetic coupling. The current in plane CIP GMR and the energy-filtered imaging using characteristic inner shell current perpendicular to the plane CPP GMR were found to excitations18 make it possible to image the Fe and Cr layers increase with N as the roughness increases cumulatively. separately and at sufficient resolution to quantify the local structural roughness of the layers. Analytical electron mi- II. EXPERIMENTAL DESCRIPTION croscopy investigations were carried out using a Philips CM20-FEG TEM equipped with a Gatan imaging filter, ca- A schematic cross section of the structure developed to pable of obtaining both electron energy-loss spectra and measure the CPP resistance and magnetoresistance is given energy-filtered images in real time at high spatial resolution. in Fig. 1 a . A Nb-superlattice-Nb sandwich is first deposited Samples suitable for imaging by TEM were prepared in the in situ by dc magnetron sputtering onto a Si substrate at cross-section geometry following the customary treatment of room temperature. Both Nb layers are 100 nm thick and the polishing, dimpling, and low angle less than 10° ion mill- superlattice consists of Fe 3 nm /Cr 1.2 nm N . The Cr ing in order to get large electron transparent regions thin thickness has been chosen to correspond to the first antifer- enough for investigations by EELS without any multiple romagnetic AF coupling peak reported for polycrystalline scattering. 110 Fe/Cr superlattices.15,16 The sandwich is then com- pletely etched down to the substrate in the form of 50 150- m2 pillars. The Nb is etched using reactive ion etch- ing in a mixed CCl III. RESULTS AND DISCUSSION 2F2 and O2 atmosphere and the superlat- tices are etched with a mixture of HCI, H3PO4 and water. Figures 2 a and b present the low angle x-ray diffrac- Then, only the top Nb layer and the superlattice are etched to tion LAXRD specular and nonspecular rocking curves form 30 30- m2 pillars. The pillars are embedded in a spectra taken on a series of superlattices Fe 3 nm /Cr SiO2 film deposited by rf magnetron sputtering and lifted off 1.2 nm ]N with N 20, 40, 60, grown on top of a 100-nm- using the same photoresist mask. A second SiO2 film is de- thick Nb buffer layer. The rocking curves where measured at posited by lift-off to complete the isolation and avoid any a 2 value of 1°, this ensures high scattered intensity over a short circuit and a 10 10 m2 via is defined on top of the wide angular range, and gives information about the layers columns. To prevent any parasitic contact resistance due to close to the surface. the oxidation of the top Nb layer during the lithography pro- Despite the roughness induced by the Nb layer, the specu- cess, an ion milling step is performed and the first 10 nm of lar spectra show clear superlattice peaks up to the second PRB 62 ENHANCEMENT OF PERPENDICULAR AND PARALLEL . . . 3363 FIG. 3. Cr mapping taken on Fe 3 nm /Cr 1.2 nm 60 grown on Nb , a on the first 25 bilayers and b the last 25 bilayers. 40, and 60 bilayers at a 2 angle of 1° as a function of the offset. The central peak is due to specular scattering and the background is due to nonspecular reflection. It is worth not- ing that although synchrotron light is usually necessary to ensure high intensity in the nonspecular scattering experi- ments, the high intensity obtained in our samples with a laboratory x-ray source shows the high degree of correlation of the interfaces from layer to layer as it will be proven later FIG. 2. a LAXRD specular spectra taken on a series of super- from EELS experiments . It is known that20 resonance ef- lattices Fe 3 nm /Cr 1.2 nm fects can appear in the nonspecular reflection when the mo- N with N 20, 40, 60, grown on top of a 100-nm-thick Nb buffer layer. Lines are fits using the SUPREX mentum change of incident rays normal to the film, equals a refinement program; b LAXRD rocking curves from the same reciprocal-lattice vector and the structure is correlated from samples. Spectra are offset for clarity. layer to layer. Since resonance effects govern the x-ray pat- terns of the multilayer samples at low angles, obtaining order and finite-size oscillations due to the buffer layer thick- quantitative information usually requires the assumption of ness for N 20. As N increases, the superlattice Bragg peaks self-affine interfaces21 or a certain growth model.22 However, broaden and their intensity decreases which indicate that the it can be qualitatively observed that the ratio of specular to roughness increases cumulatively with N.17 The lines in Fig. diffuse scattered intensity decreases with N in agreement 2 a are the fit to the data obtained with the SUPREX refine- with an increase of the roughness with the number of layers. ment program.17 The structural model assumes a roughness Moreover, ``angel wings'' or Yoneda scattering23 are ob- that increases cumulatively with the bilayer index M ac- served in all three samples. These are due to the increase in cording to a power law aM where a is the roughness the diffuse scattering when is equal to the critical angle. of the first bilayer, and an exponent describing the evolu- Interestingly, angel wings occur at smaller q when the num- tion of the roughness. The final values of the fitting param- ber of bilayers is increased, again pointing to a greater eters were checked to produce minima in the confidence fac- roughness in the vertical direction or to a shorter lateral tor of the fit 2. The superlattice modulation length roughness correlation length. t(Fe) t(Cr) where t is the layer thickness , extracted Complementary, quantitative structural analysis of the su- from the fit is in good agreement with nominal values de- perlattices has also been obtained with TEM and high spatial rived from deposition rates. The roughness parameters ob- resolution energy-filtered imaging in the cross-sectional ge- tained a 0.37 0.02 nm, 0.4 , a 0.37 0.02 nm, ometry on the same series of samples. Due to the low con- 0.4 , and a 0.36 0.02 nm, 0.4 for N 20, 40, trast in scattering power and similar lattice parameters be- and 60, respectively, are in good quantitative agreement im- tween Fe and Cr, brightfield TEM pictures taken on these plying that the roughness of the first bilayer is the same in all samples show only limited diffraction contrast running along samples as it should be and increases cumulatively with the the growth direction indicating a columnar growth of the number of layers. Roughness error bars were those produc- superlattices. ing a 10% increase in 2. On the other hand, EELS pictures showed an enhanced Similar results were obtained for a similar series of super- contrast. Cr maps using the L3,2 edge 2p-3d transitions, lattices grown directly on Si substrates. All the spectra show following dipole selection rules of the first 25 bilayers and superlattice peaks up to the third order and the roughness the last 25 bilayers of a Fe 3 nm /Cr 1.2 nm 60 superlattice parameters from SUPREX refinement are: a 0.19 grown on Nb are shown in Figs. 3 a and b , respectively. 0.02 nm, 0.4 , a 0.17 0.02 nm, 0.4 , and a The first set of bilayers appear much smoother than the last 0.17 0.02 nm, 0.4 for N 20, 40, and 60, respec- set of layers, indicating a dramatic increase of roughness tively. Note that the growth on the buffer layer introduces across the superlattice stack. Note that the roughness is some additional roughness in the superlattice,19 i.e., the highly correlated. Element both Cr and Fe intensity profiles roughness parameters a , are significantly bigger for the were taken on a series Fe 3 nm /Cr 1.2 nm N with N superlattices deposited on top of Nb. 20, 40, 60 along a section perpendicular to the substrate Figure 2 b shows the rocking curves for samples with 20, plane with an integration width of 27 nm. These profiles 3364 M. C. CYRILLE et al. PRB 62 FIG. 5. Modulation length fluctuations for a N 20, b N 40, c N 60. FIG. 4. a Roughness of each individual bilayer as a function of roughness. The histograms can be fitted to Gaussian curves the bilayer index M for N 20 , N 40 , N 60 . Line is lines in Fig. 5 meaning that the thickness fluctuations are a fit to the data with ( aM ); b superlattice modulation perfectly random. The bilayer thickness fluctuations given by length as a function of the bilayer index M for N 20 , N 40 half the full width at half maximum of the Gaussian curves , N 60 . Errors bars are the standard deviation of the mean are about 0.45, 0.44, and 0.48 nm for N 20, 40, and 60, value for each bilayer. Note that the relative error is much smaller respectively, which is just about two Fe or Cr unit cells. that the errors bars of the absolute value of each individual sample. That is additional proof of the roughness correlation. Be- cause the Gaussian fluctuation of layer thickness is also an were taken every 5.4 nm along a 100-nm lateral length assumption used in the along the multilayer surface . For each Cr Fe profile, the SUPREX refinement for the intensity calculation of the LAXRD spectra,17 this result supports the position of the maximum EELS intensity was determined. validity of the refinement method of the LAXRD data. For each bilayer, the roughness is defined as the standard Therefore LAXRD and EELS analysis provide roughness pa- deviation of the Cr intensity maxima over a 100-nm lateral rameters in good quantitative agreement and evidence that length. Figure 4 a shows the quantitative roughness of each the roughness is cumulative. Similar results were obtained individual Cr layer inside the superlattice stack for all three for the series of superlattices grown directly on Si substrates. superlattices. Note the superposition of the three curves Magnetization measurements were performed on within the error bar margins given by the spatial resolution of the EELS mapping and that the roughness increases with Fe 3 nm /Cr 1.2 nm N with a SQUID magnetometer at 10 the bilayer index. The line is a fit to the cumulative rough- K. Figure 6 presents the dependence of the remnant magne- ness model ( tization (MR) normalized by the saturation magnetization aM ) used to extract the roughness pa- rameters from the LAXRD specular spectra. Although the fit (MS) with N. This quantity gives an estimate of the sample is not perfect, the parameters obtained fraction which is not antiferromagnetically aligned at 0 field. a 0.36 nm, 0.42 , are very close to the one obtained from the refine- ment of the LAXRD spectra for Fe 3 nm /Cr 1.2 nm N grown on a Nb buffer layer. We have extracted the superlattice modulation length t(Fe) t(Cr) as a function of bilayer index for the same series of samples by measuring the distance between Cr maxima. As shown in Fig. 4 b , the modulation length is in good agreement with nominal values derived from deposi- tion rates, is completely independent of the bilayer index and therefore insensitive to the dramatic increase of roughness. Note that the relative error on is much smaller than the absolute error bar indicated in the figure for one of the samples. The statistical histograms in Fig. 5 show the fluctua- FIG. 6. Remnant magnetization (MR) normalized by the satura- tions for N 20, 40, 60. Note that the thickness fluctuations tion magnetization (MS) as a function of the number of bilayers are almost the same for all samples in spite of the increasing N . Line is a guide for the eyes. PRB 62 ENHANCEMENT OF PERPENDICULAR AND PARALLEL . . . 3365 FIG. 7. Perpendicular resistance at 2 K of a series of Fe 3 nm /Cr 1.2 nm N measured at zero field squares and at saturation triangles . The solid line is a linear fit. Absolute error bars for each sample are about 15% of the measured resistance. A slight 10% decrease of MR /MS is observed with increas- ing N, over a 10­60 bilayers range. MR /MS was estimated to be 0.38 0.02. The relative insensitivity of this ratio to the FIG. 8. a CPP resistivity number of bilayers provides further evidence for the degree AP of Fe 3 nm /Cr 1.2 nm N , CPP of roughness correlation established from structural probes in P , CIP AP , and CIP P as a function of N. b GMR ratio measured in the CPP and CIP configuration as previous paragraphs. a function of the number of bilayers. Because variations of the GMR amplitude have been re- ported with interfacial roughness for both CIP and CPP geometry,24­27 we measured the resistance and GMR of a the two magnetic configurations P and AP . Figure 8 b series of Fe 3 nm /Cr 1.2 nm N superlattices with 10 N shows the GMR ratio defined as (RAP RP)/RP of the two 40 in both configurations. In the perpendicular geometry, series. A first observation is that the CPP resistivity is always the Nb electrodes exhibit a superconducting critical tempera- larger than the CIP resistivity. Figure 8 a shows in the CPP ture (TC) of 7.5 K, which is depressed due to the patterning configuration a rapid increase of AP with N, while P stays and the proximity with Fe layers.28 Below TC , the 100 pil- constant. In the CIP configuration, both P and AP increase, lars in series provide a total resistance in the m range, AP increasing faster. which can be measured with conventional techniques. Figure 8 b shows that the GMR measured in the CPP Figure 7 presents the perpendicular CPP resistance mea- configuration is always greater up to a factor 4 than in the sured at zero field (RAP) and at saturation (RP) of the series CIP configuration, as predicted2,5,6 and measured by of superlattices. The ``saturation'' field where ferromagnetic others.8,9 The GMR ratio measured in both configurations alignment is achieved is always smaller than the upper criti- increases with N. Between 20 and 40 bilayers, the GMR ratio cal field of our Nb thin films at 2 K 1.2 T for a 100-nm-thick increases by a factor of 1.5 and 1.25 for CPP and CIP con- Nb film . RP is linear with N, with the Y-axis intercept close figurations, respectively, and shows no sign of saturation. to the origin, implying that the contact resistance with the Nb The increase of the CIP resistivity and GMR with N can electrodes is negligible. A linear fit through the data gives a be explained by the increase of correlated roughness in the Y-axis intercept of 0.05 0.1 m . Note that the error bar is superlattice stack. With increasing N, the correlated rough- estimated using the error in the absolute value of each mea- ness becomes high enough to have a substantial contribution surement. The least-square fit through the data which gives from currents flowing across the interfaces. Therefore the an intercept of 0.05 0.05 m indicates that the relative measurement configuration become closer to a CAP than a error is much smaller. In any case, this contact resistance is true CIP geometry and the resistivity and GMR are expected small compared with the intrinsic resistance of our Fe/Cr to increase with roughness. Note that in the diffusive regime, samples. The intercept corresponds to a contact resistance- not only the electrons propagating along the voltage direc- cross-sectional area product 2RA for two interfaces of 1.1 tions contribute to the resistivity. Even for flat interfaces, 2 f m2 which is within error bar of other reported Nb- electrons are scattered and therefore have somewhat random ferromagnetic metal interface resistances.8,29 A much smaller walks, which on the average give a current along the voltage error is obtained if only the scatter in the data is taken into directions. Therefore the current distribution is not uniform account (1.1 1 f m2). The uniform current distribution in and there will always be some electrons crossing the inter- the superconducting Nb electrodes and the negligible faces. superlattice-electrode contact resistance assure that the mea- There are two simple extrinsic explanations which can be sured resistance is intrinsic. This is very important in order to ruled out as causing the changes in CPP magnetotransport obtain independent resistivity and magnetoresistance mea- observed here: changes in i magnetic coupling and ii cur- surements. The saturation resistivity extracted from the slope rent flow direction with respect to the interfaces. Magnetic of the linear fit is P 34 cm. coupling can be ruled out as possible explanation of the mag- Figure 8 a presents the resistivities AP and P of netotransport changes because: i loss of antiferromagnetic Fe 3 nm /Cr 1.2 nm N in the CPP and CIP geometries for coupling often expected with increased roughness 24,25 3366 M. C. CYRILLE et al. PRB 62 would decrease the GMR, contrary to what is observed Fig. IV. SUMMARY 8 ; ii Fig. 6 shows no significant change in magnetic cou- pling with increasing N within the range N 10­ 40 , and We have developed a method to measure the perpendicu- iii EELS analysis show that the roughness is correlated lar magnetotransport in metallic superlattices using micro- without any evidence for an increase in the density of pin- fabrication techniques. Because of the minimization of the holes. The change in current flow direction with respect to contact resistance with the electrodes, high number of pillars the interfaces induced by the high correlated roughness can in series, and the current uniformity in the structures, the also be ruled out because it changes the measurement geom- method provides simple, direct, and independent access to etry from true CPP to CAP. As a consequence, the GMR the superlattice perpendicular resistance and magnetoresis- would decrease contrary to what is observed Fig. 8 . There- tance. CPP and CIP GMR of Fe 3 nm /Cr 1.2 nm N super- fore a more sophisticated explanation is needed. lattices were investigated. The CPP GMR is up to four times A possible explanation for the increase in CPP GMR is higher than the CIP GMR and both GMR ratios were found the increase of spin dependent scattering with roughness. At to increase with the number of bilayers N as the interfacial saturation, the transport is dominated by electrons of one roughness increases through the stack. The CIP GMR in- spin orientation. crease is likely due to an enhanced contribution from cur- P being constant with N indicate that those electrons are weakly scattered by the increasing roughness. rents flowing across the interfaces of the superlattices while On the other hand the CPP GMR increase is due to an increased spin dependent AP and the GMR increase with N while electrons of both spin direction contribute to the transport. scattering. Consequently, the majority electrons are strongly scattered ACKNOWLEDGMENTS by the interfaces disorder and the correlated roughness acts as a highly spin-selective scattering potential. A similar con- The authors thank Y. Jaccard, A. Fert, and P. Levy for clusion was obtained from CIP measurements in Fe/Cr 001 their fruitful comments. This work was supported by the superlattices with negligible bulk scattering26 and annealing U.S. Department of Energy. J. Santamaria thanks the Funda- induced interface defects. Note that comparative CIP and cion Jaime del Amo and Fundacion Flores Valles for its CPP measurements place this conclusion on a much firmer support. Maria Elena Go´mez thanks Universidad del Valle footing. and COLCIENCIAS. *On leave from Universidad del Valle, A.A. 25360 Cali, Colombia. 17 I. K. Schuller, Phys. Rev. Lett. 44, 1597 1980 ; W. Sevenhans, On leave from U. Complutense, 28040 Madrid, Spain. M. Gijs, Y. Bruynseraede, H. Homma, and I. K. Schuller, Phys. 1 M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, F. Rev. B 34, 5955 1986 ; E. E. Fullerton, I. K. Schuller, H. Petroff, P. Etienne, G. Creuzet, A. Friederich, and J. Chazelas, Vanderstraeten, and Y. Bruynseraede, ibid. 45, 9292 1992 ; D. Phys. Rev. Lett. 61, 2472 1988 . M. Kelly, E. E. Fullerton, J. Santamaria, and I. K. Schuller, Scr. 2 Y. Asano, A. Oguria, and S. Maekawa, Phys. Rev. B 48, 6192 Metall. Mater. 33, 1603 1995 . 1993 . 18 R. Rennekamp, J. Thomas, B. Arnold, and K. Suenaga, Fresenius 3 R. Q. Hood, L. M. Falicov, and D. R. Penn, Phys. Rev. B 49, 368 J. Anal. Chem. 361, 621 1998 . 1994 . 19 A series of superlattices Fe 3 nm /Cr 1.2 nm 20 was grown on 4 J. Barnas and Y. Bruynseraede, Phys. Rev. B 53, 5449 1996 . top of Nb buffer layers with thicknesses varying from 0 to 100 5 P. Zahn, I. Mertig, M. Ritcher, and H. Eschrig, Phys. Rev. Lett. nm. SUPREX simulations of the low angle x-ray spectra show a 75, 2996 1995 . monotonic increase of a with Nb thickness. 6 S. Zhang and P. M. Levy, J. Appl. Phys. 69, 4786 1991 . 20 D. G. Stearns, D. P. Gaines, D. W. Sweeney, and E. M. Gullik- 7 T. Valet and A. Fert, Phys. Rev. B 48, 7099 1993 . son, J. Appl. Phys. 84, 1003 1998 . 8 W. P. Pratt, Jr., S.-F. Lee, J. M. Slaughter, R. Loloee, P. A. 21 S. K. Sinha, E. B. Sirota, S. Garoff, and H. B. Stanley, Phys. Rev. Schroeder, and J. Bass, Phys. Rev. Lett. 66, 3060 1991 . B 38, 2297 1988 ; A. L. Barabasi and H. E. Stanley, in Fractal 9 M. A. M. Gijs, S. K. J. Lenczowski, and J. B. Giesbers, Phys. Concepts in Surface Growth Cambridge University Press, Cam- Rev. Lett. 70, 3343 1993 . bridge, 1995 , p. 19. 10 M. A. M. Gijs, M. T. Johnson, A. Reinders, P. E. Huisman, R. J. 22 S. F. Edwards and D. R. Wilkinson, Proc. R. Soc. London, Ser. A M. van de Veerdonk, S. K. J. Lenczowski, and R. M. J. van 381, 17 1982 . Gansewinkel, Appl. Phys. Lett. 66, 1839 1995 . 23 Y. Yoneda, Phys. Rev. 131, 2010 1963 . 11 T. Ono and T. Shinjo, J. Phys. Soc. Jpn. 64, 363 1995 . 24 J. M. Colino, I. K. Schuller, V. Korenivski, and K. V. Rao, Phys. 12 L. Piraux, J. M. George, J. F. Despres, C. Leroy, E. Ferain, R. Rev. B 54, 13 030 1996 ; M. Velez and I. K. Schuller, J. Magn. Legras, K. Ounadjela, and A. Fert, Appl. Phys. Lett. 65, 2484 Magn. Mater. 184, 275 1998 . 1994 . 25 R. Schad, P. Belien, G. Verbanck, K. Temst, V. V. Moshchalkov, 13 A. Blondel, J. P. Meier, B. Doudin, and J.-Ph. Ansermet, Appl. Y. Bruynseraede, D. Bahr, J. Falta, J. Dekoster, and G. Phys. Lett. 65, 3019 1994 . Langouche, Europhys. Lett. 44, 379 1998 ; R. Schad, P. Belien, 14 K. Liu, K. Nagodawithana, P. C. Searson, and C. L. Chien, Phys. G. Verbanck, C. D. Potter, H. Fischer, S. Lefebvre, M. Bessiere, Rev. B 51, 7381 1995 . V. V. Moshchalkov, and Y. Brunynseraede, Phys. Rev. B 57, 15 E. E. Fullerton, D. M. Kelly, J. Guimpel, I. K. Schuller, and Y. 13 692 1998 . Bruynseraede, Phys. Rev. Lett. 68, 859 1992 . 26 R. Schad, P. Belien, G. Verbanck, V. V. Moshchalkov, Y. 16 S. S. P. Parkin, N. More, and K. P. Roche, Phys. Rev. Lett. 64, Bruynseraede, H. E. Fischer, S. Lefebvre, and M. Bessiere, 2304 1990 . Phys. Rev. B 59, 1242 1999 . PRB 62 ENHANCEMENT OF PERPENDICULAR AND PARALLEL . . . 3367 27 W. C. Chiang, W. P. Pratt, Jr., M. Herrold, and D. V. Baxter, Goryunov, I. A. Garifullin, and G. G. Khaliullin, ibid. 55, 8945 Phys. Rev. B 58, 5602 1998 . 1997 . 28 G. Verbanck, C. D. Potter, V. Metlushko, R. Schad, V. V. Mosh- 29 C. Fierz, S. F. Lee, W. P. Pratt, Jr., and P. A. Schroeder, J. Phys.: chalkov, and Y. Bruynseraede, Phys. Rev. B 57, 6029 1998 ; Condens. Matter 2, 9701 1990 . Th. Muhge, K. Westerholt, H. Zabel, N. N. Garifyanov, Yu. V.