Journal of Magnetism and Magnetic Materials 401 (2016) 242–247

Contents lists available at ScienceDirect

Journal of Magnetism and Magnetic Materials

journal homepage: www.elsevier.com/locate/jmmm

Effect of interface roughness on superconducting transition

temperatures of Nb/Co multilayers
L.Y. Liu a,n, U.D. Chacón Hernandez b, Y.T. Xing c, N.M. Suguihiro a, D. Haeussler d,
E. Baggio-Saitovitch b, W. Jäger d, I.G. Solórzano a
a
DEQM, Pontifícia Universidade Católica do Rio de Janeiro, Rio de Janeiro, Brasil
b
Centro Brasileiro de Pesquisas Físicas, Rio de Janeiro 22290-180, Brasil
c
Instituto de Física, Universidade Federal Fluminense, Niterói 24210-346, Brasil
d
Institute for Materials Science, Christian-Albrechts Universittzu Kiel, Germany

art ic l e i nf o a b s t r a c t

Article history: Superconductor (SC)/ferromagnet (FM) Nb/Co multilayers have been produced by magnetron-sputtering
Received 10 August 2015 with a 100 nm thickness of Nb and 5, 10, and 20 nm of Co. The superconducting properties have been
Received in revised form investigated by electric transport measurements. It was found that the thicker Co layers decrease the
14 September 2015
superconducting transition temperature (Tc) less than the thinner ones. In order to understand this
Accepted 14 October 2015
Available online 20 October 2015
unexpected behavior, the microstructure of the layers has been investigated by means of Atomic Force
Microscopy (AFM), Transmission Electron Microscopy (TEM) and by energy-dispersive X-ray spectro-
Keywords: scopy line scan analyses in scanning TEM (STEM) mode. It was found that the decisive parameter which
Superconductivity determines the effect of magnetic layers on Tc of the superconducting layers is not only the roughness
Ferromagnetism
(R), but the ratio of the roughness to thickness (dCo) of the magnetic Co layer, δ = R/dCo . For δ > 1 the
Multilayers
magnetic stray field of the magnetic layers is the main reason for the Tc reduction.
Cross section characterization
Interface roughness & 2015 Elsevier B.V. All rights reserved.

1. Introduction due to the interaction between SC and FM, an external magnetic

It is well known that an external magnetic field usually de-
creases the critical parameters of a superconductor, for example,
the critical current density (Jc) and critical temperature (Tc). There
mainly are three mechanisms for the Tc decrease of SC/FM mate-
rials: proximity effect [1], formation of spontaneous vortices [2–4]
and spin flip scattering [5]. The proximity effect occurs at the in-
terface of a SC and normal metal due to Andreev reflection. If the
FM in a SC/FM nanocomposite consists of magnetic nanoparticles,
the stray field can induce spontaneous vortices, which will de-
crease Tc of the superconductor stronger than the proximity effect
[4]. The main reason for this is that the diameter of the vortices is
approximately 2ξ (ξ: coherence length), which is much larger than
the diameter of the magnetic nanoparticles. The strongest effect of
decreasing Tc, however, is spin flip scattering in the case that there
are single atoms of a ferromagnetic material in the supercondu-
ctor [6].
Recent studies of nanocomposites made from superconducting
and ferromagnetic materials have shown some new phenomena:
in some SC/FM nanocomposites, for example, it was found that
n
Corresponding author.
E-mail address: liu@aluno.puc-rio.br (L.Y. Liu).
field (Bext) can increase the critical parameters of the super-
conductor as long as it cannot reverse the magnetic moment of the
FM component. The resulting increased Tc is usually called mag-
netic field induced superconducting transition [7–9]. It was also
reported for thicknesses of the magnetic layers in the range of
several nanometers that Tc increases and decreases periodically
with the increase of the thickness of the magnetic layers [10]. This
is due to the Andreev reflection at the interface of the SC and
FM [1].
In this paper we will report on another unusual Tc change in
Nb/Co multilayers, i.e., sample with thicker Co layers has higher Tc,
which is difficult to be explained by one of the mechanisms de-
scribed above. In order to find out what is responsible for the
unusual Tc change we studied the surface roughness and the mi-
crostructures of the SC/FM, here Nb/Co, multilayers.
2. Experimental
[Nb (100 nm)/Co (dCo)]
5/Nb multilayers (dCo ¼5, 10 and
20 nm), which are denoted as Co5, Co10 and Co20, respectively,
have been prepared by magnetron sputtering at room tempera-
ture on Si(100) substrates. The base pressure of the system
was ∼5 × 10−8 mbar and the deposition was performed in an Ar

http://dx.doi.org/10.1016/j.jmmm.2015.10.048
0304-8853/& 2015 Elsevier B.V. All rights reserved.
 L.Y. Liu et al. / Journal of Magnetism and Magnetic Materials 401 (2016) 242–247
atmosphere with a pressure of 2 × 10−3 mbar . After deposition, the
electrical transport properties, microstructure and surface rough-
ness have been investigated. The electrical transport properties
have been measured by a four-probe method with a 1 mA AC
current of 16 Hz. The surfaces of the samples have been analyzed
with atomic force microscopy (AFM) in order to study the surface
roughness on micrometer scale. For the microstructure analysis,
cross-section TEM samples were prepared from the Co5 sample
system by sectioning, followed by mechanical dimpling and con-
ventional ion beam thinning methods using Ar ions. The micro-
structure of the multilayer, its interfaces, and the elemental dis-
tributions of Co and Nb across the multilayer were investigated by
conventional transmission electron microscopy (TEM) methods,
transmission electron diffraction (TED), high-angle annular dark
field (HAADF) imaging in scanning TEM (STEM) mode, and energy-
dispersive X-ray spectroscopy line scan analyses in STEM mode.
The analyses were carried out at an operating voltage of 300 kV
with a FEI Tecnai F30 analytical transmission electron microscope.
3. Results and discussions
Fig. 1 shows the electrical transport measurements for the
three samples. One can see that the Tc of Co5, Co10 and Co20 are
6.1, 6.4 and 6.7 K, respectively. Comparing with the Tc of pure Nb
thin film (∼9 K) prepared by the same method [11], all the samples
suffer a Tc reduction due to the existence of Co layers. However,
the sample with thinner Co layers has even lower Tc, which is
unexpected. As we have discussed above, assuming that the in-
terfaces of Nb and Co are the same for the three samples the
proximity effect should be the same and we would expect that the
Co20 samples have the lowest Tc due to the biggest FM thickness,
which has a stronger magnetic moment.
In our samples, the thickness of the magnetic layers is much
larger than that range where the periodic changes have been ob-
served in Ref. [10]. In addition, the variation of Tc is much higher
than the amplitude of the observed periodic changes [10]. Our
result, therefore, cannot be explained by the effect of Andreev
scattering. In order to understand our observations, the roughness
of the Nb surface of our samples has been investigated by AFM and
the results are shown in Fig. 2. AFM can give only the roughness of
the surface of the samples but not the roughness of the interface of
Nb/Co directly. Here we suppose that the roughness of the surface
of the sample is in the same order of magnitude as the roughness
Fig. 1. Electrical resistance as a function of temperature for Co5, Co10 and Co20.
The sample with thickest Co layer (Co20) has highest Tc and with thinnest layer of
Co (Co5), lowest Tc.
243
at the interface of Nb/Co since our sample is a multilayer system
and the roughness of the interface depends on the condition of the
previously deposited layer. From Fig. 2 one can see that the surface
roughness of sample Co5 [Fig. 2(a)] is much larger than that of the
other two samples [Fig. 2(b) and (c)]. We performed a quantitative
analysis of the surface roughness of the three samples and the
results are given in Table 1. In this table, Rq, Ra and Rmax are the
root mean-square-average, the arithmetic mean value and the
maximum value, respectively, of the sample surfaces. The values of
Rq and Ra are given by the following equation:
n n
1 1
Rq = ∑ Ri2 R a = ∑ |R i |
n i=1
n i=1
here Ri is the roughness at each pixel of the images, i is the order
number of the pixels and n is the total number of the pixels of the
image. From this table one can see that the average roughness is
quite small. Even for Co5, the sample with maximum roughness,
Rq, is only 1.4 nm, still smaller than the measured thickness of the
Co layers. Comparing the three samples, Co5 has the highest
roughness and the value is almost double the value of Co10 and
Co20. Keeping in mind that the difference of the three samples is
only the thickness of the Co layers, the differences in roughness
obtained among the three samples in this work have to be due to
the different thicknesses of the Co layers. It means that with the
addition of Co, the roughness decreases and it saturates when the
Co layer reaches 10 nm. The value of Rmax, on the other hand, is
quite big, as can be seen from Table 1. For sample Co5, Rmax
(16 nm) ⪢dCo(5 nm); for Co10, Rmax (11 nm) EdCo(10 nm) and for
Co20, Rmax (14 nm) <dCo (20 nm). In order to have a direct view of
the roughness of the surface we made three-dimensional images
from the two-dimensional data for the three samples as shown in
Fig. 3. From this figure we can directly see that the roughness of
Co5 is much larger than those of the other two samples and there
are several high peaks on the surface. The grains are also larger
than for the other two samples.
In order to have a direct view of the roughness of the interface,
the microstructure of one sample (Co5), its interfaces, and the
elemental distributions of Co and Nb across the multilayer system
have been studied by means of different analytical transmission
electron microscopy (TEM) methods applied to cross-section
samples.
The general structure of the Nb/Co multilayer system is illu-
strated by the bright-field (BF) TEM micrograph taken from the
multilayer cross-section sample Co5 [Fig. 4(a)]. The Nb/Co multi-
layer system is characterized by five thin Co layers and six thicker
Nb layers. The multilayer and the individual Nb and Co layers are
rather uniform in thickness, with an average thickness of the Nb
layers of ∼100 nm and a thickness of the Co layers of ∼5 nm. These
results indicate that the nominal conditions during sputter de-
position of the layer system have been controlled quite well.
The microstructure of the Nb layers is polycrystalline and
contains grain boundaries and crystalline defects within grains
most of which are likely to be dislocations [Fig. 4(b)]. The poly-
crystalline nature of the layers is also reflected by the selected-
area transmission electron diffraction pattern (Fig. 4, inset). Fur-
thermore, this pattern indicates a texture-like intensity distribu-
tion indicating a narrow angular range of lattice orientations for
some of the grains of the individual Nb layers. The distribution of
lattice orientations of the individual grains in directions parallel to
the substrate surface has not been investigated. Fig. 4(b) shows a
dark-field (DF) TEM image taken with the intensity of the re-
ciprocal space region marked by the aperture. The micrograph
shows some of the Nb layer regions with texture-like lattice or-
ientation relationships as image areas of bright contrast. Based
on the projected shapes of individual grains [Fig. 4(b)], it can be
244 L.Y. Liu et al. / Journal of Magnetism and Magnetic Materials 401 (2016) 242–247

Fig. 2. AFM surface profile and Fourier Transformation spectra of a line profile for (a) Co5; (b) Co10; and (c) Co20. The white lines on the images show the position of the line
profile for each spectrum.

assumed that these crystalline regions within the Nb layers are
originating from the columnar grain growth within a narrow an-
gular range along the surface normal of the Si substrate. The
widths of such columnar grains are of the order of ∼30–50 nm. In
this DF micrograph, contrast due to the grain boundaries and the
crystalline defects of the Nb layers is seen as well.
The individual Co layers appear uniform in thickness and are
interconnected on a larger scale [Fig. 4(a)]. The interfaces between
the individual Co and Nb layers appear abrupt and smooth on a
small scale (Figs. 4 and 5). In order to estimate the roughness of
the Nb/Co interface on a nanometer scale, cross-section high re-
solution TEM (HRTEM) imaging of the Co layers and the interfaces
were taken in Si [110] zone axis orientation for the topmost layers
of the Nb/Co multilayer system [Fig. 5(a) and (b)]. The micrograph
 L.Y. Liu et al. / Journal of Magnetism and Magnetic Materials 401 (2016) 242–247 245

Table 1 confirms that the Co layer is interconnected. The average Co layer

The surface roughness of the three samples obtained from the AFM images. Rq, Ra thickness is of the order of 5 nm [see marker distance in Fig. 5(b)].
and Rmax are the root mean-square-average, the arithmetic mean value and the
The grainy contrast of the Co layer is likely to result from a su-
maximum value of the roughness, respectively.
perposition of small nanocrystals [Fig. 5(b)]. Lattice images of in-
Samples Rq ( ± 0.05 nm) Ra ( ± 0.05 nm) Rmax ( ± 0.05 nm) dividual crystallites are absent which is compatible with this as-
sumption. The HRTEM images of such layers reflect the mis-
Co5 1.4 1.1 16 orientation of nanocrystals and the superposition of crystallites
Co10 0.74 0.59 11
along the imaging beam direction and therefore do not provide
Co20 0.85 0.74 14
clearly discernible lattice contrast for nanometer-sized crystallites.
The chosen imaging conditions can still be utilized for a quanti-
tative determination of the local (projected) interface roughness
on the accessible length scales. The projected average roughness of
the Nb/Co interface is estimated from this cross-section micro-
graph [Fig. 5(b)]. It amounts to a roughness of the order of the size
of the nanocrystals ( ≤1 – 2 nm ). The image regions of the adjacent
Nb layers show Moiré fringe contrast due to overlap of crystal
regions and contrast due to the presence of defects.
An example of a high angle annular field (HAADF) STEM image
of the Co5 multilayer system is shown in Fig. 6. The image contrast
is determined by differences of the atomic number, the volume
density of the materials, and the sample thickness. The continuous
change of the overall contrast which is related to the Nb layers can
be attributed to changes of the local TEM specimen thickness
across the multilayer region. This is in accordance with the results
of the EDX profile measurements which show a corresponding
change in the relative signal intensity of the Nb K line across the
multilayer (Fig. 6). Clearly visible is the HAADF STEM image in-
tensity of the thinner Co layers, reflected by the lower contrast of
the element with lower atomic number. Fig. 6 shows the drift-
corrected EDX spectrum profiles for Nb-K (upper) and Co–K
(lower) line intensities. The line scan profile plots are aligned with
the STEM-HAADF image. The elemental profiles clearly show the
relative increase of the Co signal and the relative decrease of the
Nb intensity at the positions of the Co layers.
Based on the above results and discussion, we propose that the
effect of the roughness of the interface between Co and Nb layers
plays an important role for the unusual behavior of Tc in our
samples. In order to understand the effect of roughness R, we need
to take also into account the thickness d of the magnetic layer. We
define a new parameter δ = R/d to demonstrate two limiting si-
tuations, when δ⪢1, i.e., R⪢d and when δ⪡1, i.e., R⪡d . As shown in
Fig. 7(a), when δ⪢1, the amount of second deposited material (Co)
is too small to form a continuous layer. In this case all the material
mainly forms isolated islands. Within creasing deposition of ma-
terial, some part of the islands can connect between each other
and form bigger isolated islands as shown in Fig. 7(b). If depositing
even more materials δ⪡1, i.e., the roughness now is much less than
the layer thickness, and the effect of the roughness can be ignored
and one can consider the interface is flat as shown in Fig. 7(c).
In SC/FM multilayers, the effect of magnetic layers is very dif-
ferent for the three cases since the distribution of the stray field
changes from one sample to another. When δ⪢1, the Co layer forms
magnetic islands. The magnetic field between two islands is much
stronger than that of the connected layers and is in the same di-
rection with the magnetic moment of them when the moments of
them are aligned in-plane, which is most likely the case of Co5
since the distance between two Co layers is much bigger than the
distance of the islands. For this situation it has been reported that
the stray field of magnetic particles can induce spontaneous vor-
tices in this kind of sample [2] and it can lead to a much stronger
decrease of the superconducting Tc than the proximity effect [4].
When δ⪡1, the sample can be considered as multilayers with
completely flat interfaces. For such a structure, the magnetic field
in the Co layers is very strong since all the magnetization is as-
Fig. 3. 3D topological AFM images of the sample surface of (a) Co5; (b) Co10; and
(c) Co20.
sumed to be in-plane. The magnetic field between the Co layers,
however, is low. This is the reason why in SC/FM multilayers
246 L.Y. Liu et al. / Journal of Magnetism and Magnetic Materials 401 (2016) 242–247

Fig. 4. (a) Bright-field (BF) TEM micrograph of multilayer cross-section sample Co5, showing a Nb/Co multilayer system with uniform thickness of the Nb and the Co layers.
The individual Nb layers are polycrystalline. The grains contain crystalline defects. The Co layers appear uniform in thickness and are interconnected on a larger scale.
(b) Dark-field (DF) TEM image, showing the size and shape of single crystallites and the texture-like lattice orientation of some of the Nb grains (see areas of bright contrast).
Strain field contrast due to dislocations is visible in some of the grains. The transmission electron diffraction pattern (inset) shows the polycrystalline nature of the Nb/Co
multilayer. White circle: position and size of aperture chosen for TEM DF imaging [Fig. 5(b)]. Dark arrow: beam stop blocking the central electron beam.

Fig. 5. Cross-section HRTEM images taken in Si [110] zone axis showing the topmost layers of the Nb/Co multilayer system. (a) The Nb layers show Moiré fringe contrast and
contrast due to the presence of defects. The Co layer appears to be interconnected. (b) Same specimen region as in (a), however, imaged under slightly different imaging
condition. The Co layer image shows grainy contrast indicating that it is likely to originate from the superposition of small nanocrystals. Local roughness of the Nb/Co
interface of the order of the size of the nanocrystals is visible more clearly in this imaging condition. The average Co layer thickness is of the order of 5 nm or less (see marker
distance). The projected average layer thickness and interface roughness are estimated from these cross-section micrographs.

usually there is no evidence for the existence of spontaneous
vortices [12].
With the above discussion we can understand the decrease of Tc
in our samples with increasing d. The variation of the surface
roughness amplitudes of Co5 sample is between  5 nm and þ5 nm
and in this level of the amplitude there are many peaks on the sur-
face. These peaks are much higher than the thickness of Co layers
(5 nm) and they can break the Co layers in this sample. As a result,
the Co layers may form islands or a network, and decrease Tc due to
the presence of a magnetic stray field or even spontaneous vortices
similar to those for Co magnetic nanoparticles in Pb [2].
The average roughness (Ra, Rq) and maximum roughness (Rmax)
of Co10 are smaller than that of Co5. From the spectrum of the
cross-section one can see that the variation of the surface is mainly
between  2 nm and þ2 nm, but the Rmax is very close to the
thickness of Co layers. As we discussed above, there should be very
few spontaneous vortices induced by the magnetic layers com-
pared with Co5 and the stray field of the Co layers should be
weaker than for Co5. So the decrease of Tc of Co10, due to magnetic
stray field of the Co layers should be smaller than that for Co5. The
roughness of Co20 is very similar to Co10 despite dCo of Co20 is
double of Co10. As a result, the Tc decrease of Co20 was expected
to be even smaller, as observed experimentally.
4. Conclusions
In summary, we have observed that Tc of Nb/Co samples with
thinner magnetic Co layers is lower than the one with thicker Co
layers while the thickness of the superconducting Nb layers is the
same. Based on studies of the microstructure and interface rough-
ness, we propose that this unexpected behavior depends on the ratio
(δ) of the interface roughness and the thickness of the Co layer.
Depending upon the values of δ, the Co layers can be in the form of
islands (Co5), rough layers (Co10) or flat layers (Co20). The different
forms of Co layers determine the Tc change. Co islands decrease Tc
more than connected layers, which is the reason why we observe the
lowest Tc in sample Co5.
 L.Y. Liu et al. / Journal of Magnetism and Magnetic Materials 401 (2016) 242–247 247

Fig. 6. (Above) HAADF-STEM image of sample Co5 and (below) drift-corrected EDX spectrum profiles for Nb-K (upper) and Co-K (lower) line intensities. The line scan is
aligned with the STEM-HAADF image. The scans show clearly the increase of the Co signal and the decrease of the Nb intensity at the positions of the Co layers. In
accordance with this result, the HAADF STEM image intensity is low reflecting the presence of the lower atomic number element. The window marks the region used for
drift correction, the line depicts the positioning of the electron beam for the EDXS line scan. The center of the image is brighter than the edges of the image is due to the
different thickness of the sample after the preparation.

We are thankful to J. Litterst and Hans Micklitz for critical reading of

this paper. Thanks are also due to Dr. Li Feng for his help of AFM
R

experiments.
d

Co5: Tc~6.2 K
References

[1] A.I. Buzdin, Rev. Mod. Phys. 77 (2005) 935, http://dx.doi.org/10.1103/

Co10: Tc~6.5 K
Co20: Tc~6.7 K
Fig. 7. Sketch of the interface between two layers with the thickness of upper layer
is (a) much smaller than the roughness ( δ = R/d⪢1); (b) smaller than the roughness
( δ = R/d ≈ 1) and (c) much bigger than the roughness ( δ = R/d⪡1).
Acknowledgments
This research is supported by CNPq, FAPERJ and CAPES (Brazil).
RevModPhys.77.935.
[2] Y.T. Xing, H. Micklitz, T.G. Rappoport, M.V. Milošević, I.G. Solórzano-Naranjo,
E. Baggio-Saitovitch, Phys. Rev. B 78 (2008) 224524, http://dx.doi.org/10.1103/
PhysRevB.78.224524.
[3] Y.T. Xing, H. Micklitz, E. Baggio-Saitovitch, T.G. Rappoport, Phys. Rev. B 80
(2009) 224505, http://dx.doi.org/10.1103/PhysRevB.80.224505.
[4] Y.T. Xing, H. Micklitz, W. Herrera, T.G. Rappoport, E. Baggio-Saitovitch, Eur.
Phys. J. B 76 (2010) 353, http://dx.doi.org/10.1140/epjb/e2010-00183-5.
[5] M. Vélez, M.C. Cyrille, S. Kim, J.L. Vicent, I.K. Schuller, Phys. Rev. B 59 (1999)
14659–14662, http://dx.doi.org/10.1103/PhysRevB.59.14659.
[6] A.A. Abrikosov, L.P. Gor'kov, Sov. Phys. JETP 12 (1961) 1243.
[7] M. Lange, M.J.V. Bael, A.V. Silhanek, V.V. Moshchalkov, Phys. Rev. B 72 (2005)
052507, http://dx.doi.org/10.1103/PhysRevB.72.052507.
[8] M.J. Van Bael, K. Temst, V.V. Moshchalkov, Y. Bruynseraede, Phys. Rev. B 59
(1999) 14674–14679, http://dx.doi.org/10.1103/PhysRevB.59.14674.
[9] M.V. Milošević, G.R. Berdiyorov, F.M. Peeters, Phys. Rev. Lett. 95 (2005)
147004, http://dx.doi.org/10.1103/PhysRevLett.95.147004.
[10] I. Garifullin, J. Magn. Magn. Mater. 240 (2002) 571–576, http://dx.doi.org/
10.1016/s0304-8853(01)00849-6.
[11] A.I. Gubin, K.S. Il'in, S.A. Vitusevich, M. Siegel, N. Klein, Phys. Rev. B 72 (2005)
064503, http://dx.doi.org/10.1103/PhysRevB.72.064503.
[12] C. Monton, F. de la Cruz, J. Guimpel, Phys. Rev. B (Condens. Matter Mater.
Phys.) 75 (2007) 064508, http://dx.doi.org/10.1103/PhysRevB.75.064508.