E L S E V I E R Thin Solid Films 292 (1997) 61-68
lhi o
Microstructure of Co-Ni/Au multilayers studied by XRD
D. Rafaja a.,, M. Chl~idek a, V. Valvoda a, M. Seddat b, H. Lassri b, R. Krishnan b
Faculty of Mathematics and Physics, Charles University, Prague, Ke Karlovu 5, CZ-121 16 Prague 2, Czech Republic b Laboratoire de Magndtisme et Mat~riaux Magndtiques, CNRS, F-92195 Meudon, France
Received 11 January 1996; accepted 30 May 1996
Abstract
Several series of Co-Ni/Au multilayers were investigated using different X-ray diffraction methods. The compositions and thicknesses of the Co-Ni layers were varied, and the influence of temperature treatment of gold buffer on the quality of the multilayers was studied. According to the results, the quality of the multilayers is influenced primarily by the roughness of the substrate and the buffer. The Co-Ni/Au multilayer systems grow predominantly with a replicative, highly correlated roughness induced by the substrate and buffer roughness. A cumulative roughness was found only rarely. In contrast, a decrease in the original roughness towards the sample surface was found for certain deposition parameters.
Keywords: Cobalt; MuItilayers; Nickel; X-ray diffraction
1. Introduction
Many physical properties of materials are related to both their structure and microstructure. This is also true of multi- layers, which are now being studied intensively due to their interesting novel magnetic and magneto-optical properties.
Giant magnetoresistance, surface magnetic anisotropy and Kerr rotation are the most frequently reported effects [ 1-3 ].
Concerning Co~',tix _ J A u multilayers, we have recently shown [3] that with the addition of cobalt the surface ani- sotropy becomes strongly positive and the perpendicular magnetization is stabilized. We also found that the magnetic properties are quite sensitive to the deposition parameters, e.g. the attributes of buffer layer, whether annealed or not, its thickness, etc.
Multilayers that show interesting magnetic and magneto- optical properties are created by very thin layers usually with thicknesses below 2 nm. The crystallographic structures of such materials vary widely and, of course, differ from those of the bulk components. The real structure of multilayers is one factor that influences their physical properties; another is the quality of the interfaces, which is usually described in terms of the interface roughness. We have therefore carried out detailed X-ray diffraction studies of miscellaneous sam- ples of CoxNi~ _x/Au multilayers. The cobalt concentration (x), the thicknesses of the respective layers and the number
* Corresponding author.
of bilayers were varied, and different substrate materials were used. The Au buffer layers were annealed at various temper- atures before depositing the multilayers.
2. Sample preparation
The CoxNii_x/Au multilayers were prepared by dual e- beam evaporation under UHV and controlled conditions. The pressure during deposition was in the range 1-5 × 10 .9 Torr;
the deposition rate was typically 1 nm min - 1. The thickness of each layer was independently checked by a previously calibrated quartz crystal. The thickness of the gold layer, t(Au), was kept constant at 1.5 nm, but that of the magnetic layer was varied in the range 0.5-3.0 nm.
A gold buffer layer 10 nm thick was first deposited onto glass or Si substrates kept at room temperature. The multi- layer was then grown on this buffer layer. In some cases the gold buffer was annealed at 200°C for 2 h and then the Co~Nii _x/Au multilayer was grown. Three cobalt concentra- tions were used, x = 0, 0.35 and 0.70.
The magnetic alloy layer was evaporated from an alloy ingot. It was later verified by electron probe microanalysis that the layer composition corresponded to that of the ingot within our limits of accuracy. The magnetization of a single layer of the alloy about 50 nm thick also confirmed the above point.
0040-6090/97/$17.00 Copyright © 1997 Elsevier Science S.A. All rights reserved
P I I S O O 4 0 - 6 0 9 0 ( 9 6 ) O 8 9 7 8 - X
62 D. Rafaja et al./ Thin Solid Fibns 292 (1997) 61--68
3. Experimental
The microstructure of CoxNil_x/Au multilayers was investigated by means of X-ray diffraction (XRD). In terms of XRD, the multilayers are basically characterized by several kinds of periodicity that occurs in such systems. The first one corresponds to the average interplanar spacing of the multi- layer, d, which is a weighted d-spacing of the two materials present in the multilayer. In the diffraction pattern, the main Bragg peak located in the high-angle region (HAR, 20> 15 °) is directly related to the average interplanar spacing d. This d-spacing is modulated across the multilayer by the bilayer thickness, and thus satellite peaks appear near the main Bragg maximum. Their positions contain information on the thick- ness of one bilayer, A, which is also hidden in the positions of the diffraction maxima in the low-angle region (LAR, 20 < 15 °) The next quantity that can be investigated by means of XRD is the total thickness of the multilayer, t. Its projection into reciprocal space are the Kiessig interferences, which can be found in the LAR.
Besides the periodicity, which is calculated from the posi- tions of diffraction lines only, other physical parameters of the multilayer microstructure can be estimated from the inten- sity and shape of the diffraction profiles. Two representative parameters are the width of continuous interface roughness, c, and the fluctuations in the interplanar spacing of the respec- tive material [ 6], which are usually obtained from the HAR.
To obtain additional information on the quality of the inter- faces, reflectivity measurements must be performed in the LAR. These measurements yield information on the state of the surface and the interface roughness, as well as on the correlation between the roughnesses of the substrate and the surface.
The information on the interface quality contained in the continuous interface roughness is completed by the discrete interface roughness. The continuous interface roughness, which is obtained from the broadening of superlattice peaks in the HAR, is mainly related to the loss of long-range order.
This is usually caused by variations in the thickness of par- ticular layers, which originate in the unequal number of atomic planes deposited in each layer and by interface dis- order due to the incommensurate lattice mismatch. The dis- crete interface roughness, which is calculated from the decrease in the reflected intensity in the LAR, refers to lateral changes in the interface positions. The variances in interface positions include both the steps between neighbouring grains and the curvature measured over the whole irradiated area of the sample. Therefore, large differences of up to an order of magnitude can be observed in numerical values of the con- tinuous and discrete roughnesses [ 6].
Diffraction patterns in the HAR were measured using a conventional Bragg-Brentano diffractometer XRD-7 (Sei- fert & FPM) operating with nickel filtered radiation of a copper anode. Considering the sample size, the divergence of the primar~j beam was fixed to 0.2 °. The receiving slit was 0.15 mm wide. A Soller collimator was inserted into the
diffracted beam. This instrumental setup offers sufficiently good resolution in the HAR at relatively high diffraction line intensities. As a measure of the resolution, the full width at half maximum (FWHM) of the instrumental part of the dif- fraction profile is usually reported. The value of the FWHM was 0.065°20, as measured using LaB 6 powder produced by NIST as a standard for instrumental broadening. The angular interval 20-- 25°-55 ° was scanned repeatedly in step sizes of 2 0 = 0.05 °, with a total counting time of 40 s per step.
The measurements in the LAR were performed using a double-crystal diffractometer (BEDE). As a source of radi- ation, an 18 kW rotating anode generator was used. The copper radiation was monochromatized by a flat ( 111)-ori- ented germanium single crystal. The high intensity of the rotating anode allowed the use of very narrow slits that are necessary to obtain a good resolution. The first slit reduced the width of the primary beam to about 80 tzm, and the second receiving slit was 150 ~m wide to restrict the intensity of the scattered radiation. The reflectivity curves were taken in sym- metrical 20-0 scans. To complete the information on the multitayer structure, ~-scans and offset-scans were also done in the LAR. The diffraction patterns in the LAR were meas- ured in steps of 30-50 arcsec (20=0.017-0.028 °) with a counting time of 6 s per step.
4. Methods of data reduction
To evaluate the XRD diffraction measurements, four meth- ods of data reduction were applied. Starting with a plain technique, we performed the line fitting using an analytical function, the superlattice refinement from the HAR diffrac- tion pattern, a standard optical theory using a recursive Fres- nel formalism, and finally, diffraction theory based on the distorted wave Born approximation for the LAR refinement.
The first method approximates the diffraction profiles using an analytical function. This is especially useful for ab initio calculations of the mean d-spacings and the bilayer thicknesses from the HAR measurements. Besides the posi- tions of the diffraction maxima, the refinement also gives the line width (FWHM), the integral intensity and the shape of the separated diffraction lines. However, the obtained data must first be transformed into parameters that describe the microstructure of multilayers directly (the averaged-spacing, the bilayer thicknessl the strain and the coherent domain size). The profile analysis was performed using the computer program DIFPATAN [ 4 ]. In all cases, the Pearson VII func- tion was chosen.
For the direct calculation of microstructural parameters,
we used the program SUPREX (superlattice refinement by
X-ray diffraction) [5], which performs least-squares refine-
ments of the microstructure parameters using different
multilayer models. The diffraction pattern is calculated
according to the kinematic theory of diffraction. As a suitable
model, we chose the standard A/B superlattice, where A and
B are assumed to be single elements. The model includes
D. Rafaja et al. / Thin Solid Films 292 (1997) 61-68 63 strain profiles, discrete and continuous disorder [6].
Although the microstructure model implemented in SUPREX also allows us to determine the stage of interdiffusion, this utility was not used, because the interdiffusion parameters are strongly correlated with other microstructure parameters, such as the interference roughness. Moreover, the interdif- fusion of Co or Ni and Au atoms can be neglected.
From the reflectivity measurements, the microstructure characteristics of the multilayers were refined using the stan- dard optical theory that applies a recursive Fresnel formalism described by Underwood and Barbee [7] to a suitable model of the multilayer structure [ 1 ]. The SUPREX calculations in the LAR yielded the substrate, buffer, interface and surface roughnesses, and the thicknesses of all parts of the multilay- ers. The results of the SUPREX refinement were then com- pared with those obtained from the theory based on the distorted wave Born approximation (DWBA) [8] extended by Hol~ [9] for layered systems.
Both the LAR and the HAR refinements are generally applied for structural characterization ofwell-developedmul- tilayers. Such systems often have large thicknesses, as they consist of a large number of bilayers, and low interference roughness. A consequence of these phenomena is that the diffraction patterns are well pronounced and can be easily interpreted. We also tried to employ the HAR and LAR SUPREX and D W B A refinements to study the real structure of very thin and severely damaged multilayers. Although extremely thin magnetic multilayers have been investigated intensively to determine their magnetic and magneto-optical characteristics, in most cases structural studies have been inadequate.
5. Results
5. l . M u l t i l a y e r s o n d i f f e r e n t s u b s t r a t e s
Two identical multilayer systems Coo.35Nio.6JAu (1.8 nm/1.5 nm) lo were deposited onto glass and single-crystal oriented (111) silicon substrates. The nominal thicknesses of the respective materials were obtained by a quartz crystal.
In both cases, the substrate was covered with a 10 nm gold buffer layer.
Fig. 1 compares the reflectivity curves and diffraction pat- terns measured in the HAR. The values of the parameters inspected in the HAR, i.e. bilayer thickness, thicknesses of individual materials, average interplanar spacing, continuous interface roughness and strain, are similar for both substrates.
The real thickness of one bilayer calculated from the HAR pattern was ( 3 . 5 9 + 0 . 0 4 ) nm. This was confirmed by the LAR refinement, where the bilayer thickness (3.60 __ 0.05) nm was found. The SUPREX analysis yielded the thicknesses t ( C o - N i ) = (2.17_+0.03) nm and t(Au) = (1.42_+0.04) nm.
The strain, which is defined as the relative change in the d-spacing, is larger in the Co--Ni composite than in gold in
_=
108 107 106 105 10 4 10 3 10 2 101 100
(a) Angle (°20)
90 I
75 60
4s !
30 15 0
25 30 35
%
410 415 ' 510 ' 55
(b) Angle (°20)
Fig. 1. (a) Reflectivity curves and (b) high-angle diffraction patterns meas- ured on Coo.35Nio.65/Au (1.8 nm/l.5 nm)~o multilayers deposited onto silicon (circles) and glass (triangles) substrates. The respective SUPREX fits are drawn as solid lines.
both cases. The strain values in Co-Ni and in gold have opposite signs, as the deformation near the interface is caused by the lattice mismatch. In addition, the strain profiles of both Co-Ni and Au are strongly asymmetric. Although the differ- ences in the HAR diffraction patterns are small, the results indicate that multilayer deposited on the silicon substrate is less regular than that deposited on glass. The influence of the substrate material on the quality of the multilayer becomes clear from the reflectivity measurements. Although the sub- strate roughnesses yielded by SUPREX are the same, (0.75 +_ 0.05) nm, the buffer roughness is greater if the gold buffer is deposited on silicon, o-buff(Si) = (1.10 + 0.05) nm, than on glass, O-buff(glass)----(0.70+0.05) nm. The buffer roughness is then more or less replicated towards the sample surface. A slight cumulative roughness was observed only for the glass substrate. The replicative and highly correlated interface roughnesses were also confirmed by the large amount of diffuse scattering, which contributes to the meas- ured intensity and causes a discrepancy between the observed and calculated reflectivity curves. This effect is unmistakable at the position of the first Bragg maximum (Fig. la) because the diffuse scattering is not subtracted in the SUPREX refine- ment before fitting the specular reflectivity.
Concurrently with the relatively high roughness obtained
from the LAR measurements, we observed a relatively high
continuous disorder obtained from the HAR diffraction pat-
terns: c(glass) -- 0.033 nm and c(Si) -- 0.034 nm.
64 D. Rafaja et al. / Thin Solid Films 292 (1997) 61--68
107',
lO 6
105`
• ,~ 104 103
, g
,- 102 101
i
10 o 10-1
1 2 3 4 5 6 7
(a) Angle (°2®)
2250 2000 1750
"~. 1500 1250
•
~ 1000
750
-
500
250
0 34 36 38 40 42 44 46
(b) Angle (°28)
Fig. 2. Comparison of (a) reflectivity curves and (b) high-angle diffraction patterns for multilayers deposited on different buffers. Data: C%.TNio.3/Au (0.63 nm/1.5 nm)~6 multilayer deposited on a 10 nm non-annealed gold buffer (circles); Coo.7Nio.3/Au (0.54 nm/1.5 nm) ~6 multilayer deposited on a 10 nm annealed gold buffer (triangles); and on a 30 nm annealed gold buffer (squares).
5.2. Thermal treatment of the buffer
Two Coo.7Nio.3/Au (0.54 nm/1.5 nm) 16 multilayer spec- imens were deposited onto glass substrates covered with gold buffers of different thicknesses, 10 and 30 nm, both annealed at 200°C for 2 h. These samples were compared with a similar multilayer, COo.TNio.3/Au (0.63 nm/1.5 nm)1~, grown on a 10 nm thick non-annealed gold substrate. The HAR diffrac- tion patterns and the reflectivity curves of these samples are shown in Fig. 2. For the sample with the 10 nm thick annealed gold buffer, these measurements were accompanied by an offset scan carried out at A = 300 arcsec and an I)-scan taken at 7980 arcsec (Fig. 3). In the [2-scan, the detector was located at the position of the first Bragg maximum corre- sponding to the bilayer thickness.
The SUPREX refinement confirmed that the thickness of the gold layers in the system is ( 1.5 4- 0.1 ) nm, as obtained from measurements using a quartz oscillator. The real thick- ness of the Co-Ni layers differs insignificantly from the nom- inal value. SUPREX yielded the following data: t(Ccr- Ni) = ( 0 . 7 6 + 0 . 0 9 ) nm rather than 0.63 nm for the non- annealed buffer, and t ( C o - N i ) = ( 0 . 6 2 + 0 . 0 8 ) nm rather than 0.54 nm for the 30 nm thick gold buffer. The thickness of Co--Ni layers in the sample deposited onto a 10 nm gold
1 0 6 ~oo
4oo
10 5 3oo
2 ~ 8
~ 104 1~o
&
lO 3 102
100 1 2 3 4 5 6 7
Angle (°2@)
Fig. 3. Symmetrical 20-0 scan (solid line) and offset scan taken at A = 300 arcsec (crosses) for the Coo.7Nio,3/Au (0.54 n m / t . 5 nm)t6 multilayer deposited on a 10 nm annealed buffer. Inset: O-scan at the first Bragg
maximum corresponding to the bilayer thickness.
buffer, t ( C o - N i ) = (0.56-t-0.03) nm, agrees well with the nominal value of 0.54 rim.
The annealing process strongly affects the microstructure of the gold buffer. First, an increase in the coherent domain size in the gold buffer was observed. In different samples deposited on non-annealed buffers, the integral broadening of the diffraction line (111) of gold was 0.28-0.39 nm -~
despite the buffer thickness. The (111) reflection of gold is well separated from the satellite peaks corresponding to the multilayer structure at this bilayer thickness and at the given average d-spacing. Therefore, both the intensity and the broadening of the (111) gold line can be determined with high accuracy. Assuming a negligible microstrain in the gold buffer, the line broadening given above corresponds to a coherent domain size of 2.5-3.5 nm. After annealing, the size of coherent domains in buffer was comparable with the buffer thickness. The profile analysis yielded gold crystallites about 11 nm in size in the 10 nm gold buffer, and 31.5 nm in size in the 30 nm gold buffer. This indicates that the first layer grows coherently with the buffer. We speculate that the first Co-Ni layer is more damaged and thinner than it should be according to its nominal thickness. We assume that the cov- erage of the first Co-Ni layer is not continuous over the whole buffer, but grows in separated grains (islands).
After annealing, the gold buffer tends to be strongly { 111 } textured. An additional effect that occurs after annealing is a relaxation of the lattice parameter of the gold in the buffer towards its bulk value. The originally low value of the lattice parameter increases with both buffer thickness and annealing.
The annealing causes the relaxation of the strain and stress in the buffer; the increasing buffer layer thickness probably reduces the influence of the next Co-Ni layer on the buffer.
In comparison with the first series of samples, large vari-
ations in the microstructure are noticeable if buffers of dif-
ferent thickness are used, or if the buffer is annealed. Both
the annealing of the buffer and the use of a larger buffer
thickness results in lower strain in both materials. This was
indicated by a decrease in the broadening of the diffraction
lines corresponding to the multilayer structure and therefore
D. Rafaja et al. / T h i n Solid Films 292 (1997) 61-68 65
Table 1
Comparison of the continuous, interface and cumulative roughness for muttilayers deposited onto different buffers. A steep decrease in the buffer roughness against the basic substrate roughness with annealing and the buffer thickness is illustrated in the last column
Sample c / n m or(inter)/nm o-(cum)/nm o-(buffer) / o'(substrate)
Coo.7Nio.3/Au (6.3/15) 1~ 10 nm non-annealed 0.012 0.54 0.113 1.67
Coo.7Nio.3/Au (5.4/15) 16 10 nm annealed at 200 °C for 2 h 0.016 0.54 0 0.95
COo.TNio.3/Au (5.4/15) 16 30 nm annealed at 200 °C for 2 h 0.020 0.13 0 0.52
by a narrower distribution in interplanar spacing. Here also, the Co-Ni layers are more strained than the Au layers, but almost no asymmetry in the strain profile was found in this series of samples. Only a slight strain asymmetry was detected for the Co-Ni layers. The continuous interface roughness increases with both the annealing of the buffer and with increasing buffer thickness. Nevertheless, the continu- ous roughness is substantially lower than in samples with the inverse Co-Ni ratio, Coo.35Nio.6JAu (1.8 nm/1.5 nm)lo investigated in the previous case. The SUPREX refinement in the HAR yielded the continuous roughness values listed in Table 1. However, it follows from the reflectivity measure- ments that both the annealing of buffer and a larger buffer thickness reduce the overall interface roughness as well as the cumulative roughness. The cumulative interface rough- ness is defined as an increase in the interface roughness towards the sample surface according to
o-~ = cra(AB) + j . cr2(cum)
wherej is the number of the bilayer relative to the substrate, crj is the interface roughness at thejth bilayer, cr(AB) is the basic interface roughness at the substrate or buffer and o-(cum) is the cumulative interface roughness.
Annealing the buffer and applying a large buffer thickness, the continuous disorder in multilayers indicated by c values becomes more dominant, but the discrete disorder measured by the interface roughness in the LAR decreases rapidly. For the stage of the discrete interface disorder, the roughness at the top of buffer is crucial. The important role of the buffer quality is illustrated in Table 1 by means of the ratio o-(buffer)/o-(substrate). Starting with approximately the same substrate roughnesses in all three cases, the buffer roughness is higher than the substrate roughness if the buffer was not annealed. After annealing the buffer at 200 °C for 2 h, the original roughness was slightly reduced. Finally, by annealing the 30 nm thick substrate at 200 °C for 2 h, the original substrate roughness was reduced to half of that at the top of the buffer.
The different behaviours of the continuous and the discrete interface roughnesses can be explained by means of the estab- lished definitions of these two parameters. The continuous interface roughness is defined as the variation in the bilayer thickness measured along the direction of the diffraction vec- tor, whereas the discrete roughness is related to lateral fluc- tuations in the interface positions in neighbouring blocks. A broken multilayer structure, which is characterized by a high discrete roughness, can exhibit a low continuous roughness
if the neighbouring blocks are mutually independent. With the improvement in interface quality, indicated by a decrease in the discrete interface roughness, the lateral coherent size becomes larger and the interaction between neighbouring domains increases. A growth of the continuous roughness and an expansion of the strain profile can be observed as a consequence.
From the measurements performed in the LAR we can confirm the qualitative conclusions about the correlated roughness in the multilayer stack. The high reflected intensity observed in the offset scan carried out on the C o o . 7 N i o . 3 / A u
(0.54 nm/ 1.5 nm)16 sample (Fig. 3) shows a strongly cor- related roughness [10]. Such a distribution of diffracted intensity is caused by diffusely scattered radiation and is influenced by the roughness replication. It also corresponds to the occurrence of Kiessig fringes in the low-angle part of the offset scan. Consequently, we conclude that both the substrate and the buffer roughnesses are partially reproduced towards the multilayer surface. The f~-scan (inset in Fig. 3) shows a cut throughout this maximum and confirms the highly diffuse contribution to the total measured reflectivity curve. The diffuse scattering gives more than 60% of the total scattered intensity at the angle of 7980 arcsec at which the f~-scan was taken. Regarding the highly correlated rough- ness, we assume that the measured interface roughness in these multilayers originates at the buffer.
5.3. Thermal treatment of the multilayers
After annealing the sample Coo.35Nio.6s/Au (0.54 nm/1.5 nm) go deposited on glass substrate covered with 10 nm gold buffer at 200°C for 2 h, the multilayer structure was almost destroyed. Only very weak multilayer peaks were observed in the reflectivity curve. For the refinement of the micros- tructure parameters from the reflectivity measurement we used the simplified assumption that the system consists of one single layer with an average refractive index. The cal- culated thickness of the whole multilayer was (48.4 + 0.5) nm and the relative refractive index related to the averaged bulk values was (1.04 _+ 0.05). The substrate roughness was (0.65 _+ 0.08 ) nm and the surface roughness was ( 1.1 _+ 0.1 ) nm. The outcome of the DWBA refinement is shown in Fig. 4.
The result that the multilayer structure is almost completely
destroyed was also confirmed by the measurements carried
out in the HAR. The structure created by the annealing can
probably be described as incoherent islands of Co--Ni com-
posite dissolved in a gold matrix. However, the size of C o -
66 D. Rafaja et al. /Thin Solid FifthS 292 (1997) 61-68
O0
10 f 103 _~ 102
Bits of
1 o o , , ,
Angle (°2®)
Fig. 4. Reflectivity curve, indicating the destruction of the multilayer struc- ture after annealing. Only fragments of the original multilayer structure can be found. A model describing the system as a single layer with an average refractive index was used for the DWBA refinement. Inset: HAR pattern diffracted by the broken muttilayer structure. The weak peaks indicated by arrows are from the bits of the multilayer structure. The single high peak belongs to the recrystallized gold buffer.
Ni islands cannot be estimated because the potential peak from the Co--Ni composite is totally overlapped by the dif- fraction pattern from bits of the multilayer structure. Thelarge peak in the HAR diffraction pattern (inset in Fig. 4) belongs mainly to the gold buffer. During annealing, the buffer relaxes in a similar way to the case described in Section 5.2. Assum- ing a negligible strain contribution to the physical broaden- ing, the size of coherent domains in the gold buffer was estimated to be about 9 nm.
5.4. Effect of Co-Ni alloy composition
The different stoichiometry of the Co-Ni alloy influences predominantly its interplanar spacing and thus the average interplanar spacing of the multilayer. Fig. 5 compares the diffraction patterns measured on the Coo.7Nio.JAu ( 1,8 nm/
1.5 nm)9 and Coo.35Nio.6JAu (1.8 nm/1.5 nm)lo samples.
The interplanar spacings read from the positions of the dif- fraction lines in the HAR are (0.23124 _+0.00028) nm and (0.23268_+0.00029) nm for the first and second samples, respectively. In addition to the detected line shift, a variation in intensity can be seen in Fig. 5 (b). The insignificant abso- lute difference in intensities is caused mainly by the unequal numbers of bilayers in these systems. The relative change in intensity of neighbouring satellite peaks is given by different continuous interface roughnesses.
In another experiment, we observed completely different behaviours of samples with smaller thicknesses of the Co-Ni composite compared with the thickness of the Au layers. The multilayer structure was well developed in the sample with 70% cobalt, whereas in the samples containing more nickel the multilayer structure was poor. Advanced growth of mul- tilayers with 70% cobalt was also observed in samples with 1.8 nm Co-Ni and 1.5 nm Au, but the difference was not so striking here.
Comparing the results obtained from the reflectivity curve measurements, we can see the highest interface roughness in
Angle (°28) 10 7
10 e 10 5 10 4 10 3 10 2 I01 (a)
&
,e,
, e ,
m
