Influence of interstitial nitrogen on the structural and magnetic properties of
FeCoV/TiNx multilayers
M. Senthil Kumar and P. Böni
Citation: J. Appl. Phys. 91, 3750 (2002); doi: 10.1063/1.1450259
View online: http://dx.doi.org/10.1063/1.1450259
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v91/i6
Published by the American Institute of Physics.
Related Articles
Influence of dynamical dipolar coupling on spin-torque-induced excitations in a magnetic tunnel junction
nanopillar
J. Appl. Phys. 111, 07C906 (2012)
Quasi-omnidirectional electrical spectrometer for studying spin dynamics in magnetic tunnel junctions
Rev. Sci. Instrum. 83, 024710 (2012)
Thermal-magnetic-electric oscillator based on spin-valve effect
J. Appl. Phys. 111, 044315 (2012)
Dependence of spin-transfer switching characteristics in magnetic tunnel junctions with synthetic free layers on
coupling strength
J. Appl. Phys. 111, 07C905 (2012)
Electrical manipulation of spin polarization and generation of giant spin current using multi terminal spin injectors
J. Appl. Phys. 111, 07C505 (2012)
Additional information on J. Appl. Phys.
Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
JOURNAL OF APPLIED PHYSICS
VOLUME 91, NUMBER 6
15 MARCH 2002
Influence of interstitial nitrogen on the structural and magnetic properties
of FeCoVÕTiNx multilayers
M. Senthil Kumara),b)
Department of Physics, Indian Institute of Technology Bombay, Mumbai 400 076, India
P. Bönia)
Physik Department E21, Technische Universität München, D-85748 Garching, Germany
共Received 3 July 2001; accepted for publication 16 December 2001兲
FeCoV/TiNx and FeCoV/Ti multilayers having t FeCoV⫽30– 700 Å prepared by dc magnetron
sputtering are investigated by x-ray diffraction, stress, and magnetization measurements. The x-ray
diffraction data of the FeCoV/TiNx system show the presence of interstitial N atoms in the FeCoV
layers due to reactive sputtering of Ti with nitrogen. The interstitial N causes an expansion of the
FeCoV lattice in FeCoV/TiNx for small t FeCoV . However, for the samples with large t FeCoV , no
lattice expansion is observed. In addition to the lattice expansion caused by the intake of N atoms,
a change in the crystalline texture of FeCoV layers is also observed as indicated by the enhancement
of the FeCoV共200兲 peaks. The magnetic hysteresis measurements on the samples show that the easy
direction of magnetization lies in the plane of the layers. They further show that there are easy and
hard axes of magnetization within the plane of the FeCoV layers. The stress anisotropy present in
the plane of the samples induces a magnetic anisotropy through magnetostrictive effects leading to
the formation of the in- plane easy axis. The hysteresis and stress measurements carried out on these
samples clearly show the influence of N on the in-plane magnetic anisotropy. The magnetoelastic
energy in the case of the FeCoV/TiNx system, calculated from the stress data and from the
magnetization measurements as a function of t FeCoV is found to agree over a large range of
thickness, whereas the curves deviate significantly for small layer thickness. This deviation may be
due to the role of the FeCoVNx phase. Hysteresis measurements also show that the remanence is
about 95% for all the samples of the FeCoV/TiNx system. In contrast, the coercivity increases
linearly with increasing t FeCoV in this system. The coercivity of the FeCoV/Ti system is larger and
increases more rapidly with t FeCoV , as compared with the FeCoV/TiNx system. This behavior is
attributed to a smaller grain size in the FeCoV/TiNx system due to the reactive sputtering of the Ti
layers. However, there is no significant influence of N on the saturation magnetization of both
systems. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1450259兴
I. INTRODUCTION
in their remanent state after saturating the sample magnetization in low magnetic fields. These properties enhance the
performance of these multilayers in the devices. Because of
these reasons, the FeCoV/TiNx system is ideally suited for
the polarization of neutrons and hence it is used as coating in
polarizing neutron mirrors. More details can be found in our
previous publications.1,2 Though our multilayers are directly
used in neutron optical devices, the magnetic properties investigated in this article also throw light on the behavior of
the general class of magnetic multilayers in which the magnetic layers are separated by nonmagnetic spacer layers.
These types of magnetic multilayers are used in magnetic
random access memory, magnetoresistive read heads, and
magnetic sensors.4 – 6
In this article, we present the investigations carried out
on the structure, stress, and magnetic hysteresis of
FeCoV/TiNx and FeCoV/Ti multilayers. We have investigated FeCoV/Ti multilayers in order to compare them with
the properties of reactively sputtered Ti layers. This comparison helps us to understand and to improve the performance
of the FeCoV/TiNx layers.
FeCoV/TiNx multilayers are interesting systems 共here
FeCoV stands for Fe50Co48V2 兲 for investigating the influence
of reactive sputtering, layer thickness, and grain size on the
structural, magnetic, and magnetoelastic properties of the
multilayers. The main objective of the present investigation
is to study these multilayers for their use in neutron optical
devices for the polarization of neutrons. These multilayers
exhibit a high neutron reflectivity due to the high scattering
contrast between the magnetic FeCoV layer and the nonmagnetic TiNx , for spin up neutrons. The concentration of nitrogen in the Ti layers is adjusted such that the nuclear scattering length densities of TiNx and FeCoV are equal leading to
a vanishing contrast for spin down neutrons.1,2 Furthermore,
in these multilayers, due to the presence of in-plane stress
anisotropy and large magnetostriction of FeCoV a significant
magnetoelastic coupling is observed.3 The presence of high
remanence and coercivity makes it possible to use the layers
a兲
Formerly at the Laboratory for Neutron Scattering, ETH Zurich and Paul
Scherrer Institute, CH-5232 Villigen PSI, Switzerland.
b兲
Electronic mail: [email protected]
0021-8979/2002/91(6)/3750/9/$19.00
3750
© 2002 American Institute of Physics
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
M. S. Kumar and P. Boni
3751
II. EXPERIMENTAL DETAILS
The samples are prepared by dc magnetron sputtering
using a Leybold Z600 sputtering system equipped with three
targets.7 In this system, the deposition of the layers takes
place by moving substrates linearly under the targets. Prior to
the deposition of each layer, the gas pressure in the chamber
is stabilized for 1 min, plasma is ignited at the required target, and then the target is ‘‘presputtered.’’ We call the ‘‘presputtering time’’ as the time during which the targets are
sputtered such that the substrates are kept away from the
targets. Thus, no deposition of the layers on the substrates
takes place during this time. The deposition of the layers
takes place when the substrates come directly under the targets when they are moved linearly. In this article, we use the
symbols 储 and ⬜ for indicating the parallel and perpendicular
directions, respectively, relative to the substrate movement
because we have noticed that the stress and magnetic properties are directional dependent. The commercial alloy
Fe50Co48V2 共Vacoflux50®兲 supplied by Vacuumschmelze
GmbH, Hanau, Germany is used for our investigations. The
multilayers are deposited on both glass and Si substrates.
Three series of samples, namely FeCoV/Ti, FeCoV/TiNx
with a presputtering time of 1 min for the FeCoV layers and
FeCoV/TiNx with a presputtering time of five minutes for the
FeCoV layers are prepared for the investigations. The presputtering time for Ti is 1 min for all the samples of each
series. The sputtering of the FeCoV layers of all the three
series is carried out in Ar at a pressure p Ar⫽0.73
⫻10⫺3 mbar. The Ti layers of FeCoV/Ti are sputtered at
p Ar⫽0.44⫻10⫺3 mbar. The Ti layers of FeCoV/TiNx for
both the second and third series are sputtered at partial pressures of Ar and N2 that are p Ar⫽0.44⫻10⫺3 mbar and p N2
⫽0.02⫻10⫺3 mbar, respectively. The layers sputtered
in pure Ar 共first series兲 are 关 FeCoV(30 Å)/Ti(30 Å) 兴 75 ,
关 FeCoV(50 Å)/Ti(50 Å) 兴 75 , 关 FeCoV(100 Å)/Ti(80 Å) 兴 75 ,
and 关 FeCoV(700 Å)/Ti(150 Å) 兴 10 . A Ti top layer of 300 Å
is deposited on these samples for protection from oxidation. The reactively sputtered samples 共second series兲
are
关 FeCoV(30 Å)/TiNx (30 Å) 兴 75 ,
关 FeCoV(50 Å)/
TiNx (50 Å)] 75 ,
关 FeCoV(100 Å)/TiNx (80 Å) 兴 75 ,
TiNx (110 Å)] 25 ,
关 FeCoV(200 Å)/
关 FeCoV(300 Å)/
TiNx (120 Å)] 25 , and 关 FeCoV(700 Å)/TiNx (150 Å) 兴 10 . A
300 Å top layer of TiNx is deposited on these samples. In
the case of the third series, samples prepared are
关 FeCoV(30 Å)/TiNx (30 Å) 兴 75 ,
关 FeCoV(50 Å)/
TiNx (50 Å)] 75 ,
and
关 FeCoV(100 Å)/TiNx (80 Å) 兴 75 ,
关 FeCoV(200 Å)/TiNx (110 Å) 兴 25 . A 300 Å top layer of TiNx
is deposited on these samples as well. The thicknesses and
the number of bilayers in the multilayers are chosen in such
a way that they are representative of the sequences occurring
in polarizing supermirrors.1,2
X-ray diffraction studies are carried out using a Siemens
D500 powder diffractometer using Cu K ␣ radiation. The
stress measurements are performed using a Tencor P2 Long
scan profilometer. The stress in the layers along the 储 and ⬜
directions is determined from the radii of curvatures of the Si
wafers before and after coating.8 Magnetic measurements are
performed using an extraction magnetometer that is part of
FIG. 1. 共a兲 X-ray diffraction data of FeCoV/Ti multilayers of the first series
of the samples. These samples are prepared with a presputtering time of 1
min for both FeCoV and Ti layers. The small peaks around the main
FeCoV共110兲 and Ti共002兲 peaks are the satellite peaks due to the coherence
between the atomic layers of the individual FeCoV and Ti at the interfaces.
共b兲 X-ray data of FeCoV/TiNx multilayers of the second series. These
samples are also prepared with a presputtering time of 1 min for both FeCoV and Ti layers. The appearance of FeCoVNx (110) peaks is due to the
presence of nitrogen in the FeCoV layers. 共c兲 X-ray data of FeCoV/TiNx
multilayers of the third series. These samples are prepared with a presputtering time of 5 min for FeCoV and 1 min for Ti layers. The appearance of
the FeCoVNx (110) peaks is due to the presence of nitrogen in the FeCoV
layers.
the physical property measurements system of Quantum Design equipped with a superconducting magnet delivering
magnetic fields of 0–9T.
II. RESULTS AND DISCUSSION
The results obtained on the FeCoV/TiNx and FeCoV/Ti
multilayers are presented in this section. X-ray diffraction
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
3752
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
M. S. Kumar and P. Boni
共XRD兲 studies show changes in texture and d spacings of the
FeCoV layers. Intrinsic stress in the layers is investigated in
order to study its influence on the magnetic properties
through magnetoelastic coupling. Hysteresis studies show
high remanence and interesting coercive properties. We have
observed that the structural, mechanical, and magnetic properties are intimately related and they are significantly modified by the reactive sputtering of Ti. In the following, we
discuss these properties.
A. Structure
XRD data obtained for FeCoV/Ti and FeCoV/TiNx systems are shown in Fig. 1. We have normalized the data by
dividing the measured intensity of each sample by the total
thickness of the magnetic FeCoV layers in that sample because the thickness of the FeCoV layers and number of bilayers in the samples are different. In the case of the first
series of the samples, the XRD pattern of FeCoV共700Å兲/
Ti共150Å兲 shows an intense FeCoV共110兲 peak of bodycentered-cubic 共bcc兲 FeCoV indicating a strong 共110兲 texture
as shown in Fig. 1共a兲. Additionally, very weak peaks due to
FeCoV共200兲 and Ti共002兲 are also visible. When the layer
thickness t FeCoV is reduced, superlattice peaks around the
main Ti共002兲 and FeCoV共110兲 peaks appear. These superlattice peaks indicate some degree of coherence of the atomic
layers at the interfaces.9,10 Furthermore, a gradual broadening of the reflections due to the decrease in t FeCoV is also
seen.
Figure 1共b兲 shows the XRD patterns for the FeCoV/TiNx
system of the second series. The data for the
FeCoV共700 Å兲/TiNx (150 Å) sample is nearly the same as
for the FeCoV共700 Å兲/Ti共150 Å兲 of Fig. 1共a兲 except a small
change of the FeCoV共110兲 intensity due to the enhancement
of the FeCoV共200兲 peak. It can also be seen that the enhancement of the FeCoV共200兲 peak is rather high for small
t FeCoV . Additionally, a peak around 2 ␪ ⫽42.5° emerges. Its
intensity gradually increases with decreasing t FeCoV . A careful examination shows that this extra peak does not belong to
FeCoV, Ti, or TiNx . It is also seen that this peak appears at
lower 2␪ values indicating larger d spacings than that of the
pure FeCoV共110兲. Similar results are also obtained for the
third series of the samples as can be seen from Fig.1 共c兲.
Our interpretation of the appearance of the extra peak is
as follows: The extra peak appears due to the reactive sputtering of Ti in N2 . Before we proceed further, let us discuss
the influence of reactive sputtering in Ni/Ti multilayers for
comparison. We have observed variations in the intensities of
peaks in the case of Ni共50 Å兲/Ti共50 Å兲 multilayers in which
the Ni layers are reactively sputtered in a mixture of Ar and
synthetic air.11 The composition of the synthetic air is ⬃80%
N2 and ⬃20% O2 . For the layers sputtered in pure Ar, the
Ni共111兲 peak is the strongest. When reactive sputtering is
employed the Ni共200兲 peak is enhanced. In this case, we
have found that the Ni layers contain two Ni phases when the
Ni layers are sputtered in low partial pressures of the reactive
gas. One is pure Ni phase where the grains with 共111兲 orientation are virtually free from reactive gas atoms. Thus, the
Ni共111兲 peak shows no lattice expansion. Another is an Ni
phase in which the reactive gas atoms occupy the octahedral
FIG. 2. 共a兲 d spacings determined from FeCoV共110兲, FeCoVNx (110), and
FeCoV共200兲 peaks for FeCoV/TiNx system 共second series兲 as a function of
t FeCoV . d spacings calculated from TiNx (002) are also plotted. The d spacing of Ti共002兲 determined from the FeCoV共700 Å兲/Ti共150 Å兲 sample is
plotted for comparison. 共b兲 The change in d spacings for FeCoV/TiNx 共second series兲 system determined from FeCoVNx (110) peaks for various
samples. For FeCoV layers the increase in d spacings are calculated with
respect to d FeCoV(110) ⫽2.015 Å for FeCoV共700 Å兲/Ti共150 Å兲. For TiNx layers the increase in d spacings of d TiNx(002) are calculated with respect to
d Ti(002) of FeCoV共700 Å兲/Ti共150 Å兲.
interstitial sites of face-centered-cubic Ni lattice thereby
showing an expansion of the lattice. This expansion is clearly
evident from the Ni共200兲 peaks arising from the grains with
共200兲 orientation. The existence of such two phases is in
agreement with the Mössbauer data on bulk Ni hydrides.12
As the interstitial occupancies of N, H, O, and C in metals
both in bulk and thin film forms are similar to a certain
extent, we have compared the metal hydrides with the interstitial occupancy of N and O. In the case of the Ti layers,
though the diffusion of N and O atoms into the Ti layers is
also expected, we could not determine the d-spacings of the
Ti peaks because the Ti layers are amorphous.
In the present FeCoV/TiNx system, from Figs. 2共a兲 and
2共b兲, we see that the TiNx (002) peak shows a lattice expansion of the Ti layers. The data shown in Fig. 2 are obtained
from the second series of samples. The increase in d spacing
of TiNx (002) planes due to the occupancy of N atoms at the
interstitial sites is about 2.5% to 3% when compared with the
d spacing of Ti共002兲 determined from the Ti共002兲 peak in
FeCoV共700 Å兲/Ti共150 Å兲 sample. In the case of the FeCoV
layers, the reactive sputtering of Ti leads to the presence of N
atoms at the interstitial sites of the neighboring FeCoV layers
as evident from the appearance of the extra peak and the
changes in the intensities of the peaks. These interstitial N
atoms expand the FeCoV lattice resulting in an increased d
spacings corresponding to the FeCoVNx peaks. The d spacings of the extra peak, designated as FeCoVNx (110), lie between 2.106 and 2.136 Å. These d spacings are considerably
larger when compared with 2.015 Å which is the d spacing
determined from the pure FeCoV共110兲 peak observed for
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
FeCoV共700 Å兲/Ti共150 Å兲. The d spacings calculated from
the observed FeCoVNx (110) reflections are plotted in Fig.
2共a兲. It is seen that the expansion of the FeCoVNx (110)
planes is about 4.5%– 6%. It can further be seen from Fig.
1共b兲 and 1共c兲 that the FeCoV共110兲 and FeCoVNx (110)
phases are simultaneously present in some samples with intermediate t FeCoV .
The diffraction data of Fig. 1共b兲 and 1共c兲 also provide
further information on the presence of N in FeCoV. Let us
compare the intensities of the FeCoVNx (110) peaks of the
samples from the second and the third series, from figs. 1共b兲
and 1共c兲. In the case of the samples with small t FeCoV , the
FeCoVNx (110) peak is present and the FeCoV共110兲 peak is
absent. Then, with increasing t FeCoV the nitride peak gradually decreases and finally it almost vanishes for large t FeCoV .
There are three possibilities for the presence of the extra
FeCoVNx (110) peaks. In the first possibility, nitrogen can
react with the FeCoV sputtering target when Ti is reactively
sputtered. In our sputtering system, immediately after the
deposition of a particular layer, say an FeCoV layer in pure
Ar, we evacuate the sputtering chamber for 2 min by stopping all the gas supply to the chamber. After the evacuation,
the required gases, say Ar and N2 for TiNx layer, are supplied
into the chamber and then about a minute is allowed for the
stabilization of the pressure in the chamber. During the 2 min
time before the gas supply the FeCoV target is expected to
reach near ambient temperatures because of the excellent
cooling provided for the targets. Thus, we presume that the
FeCoV target surface absorbs nitrogen during the gas stabilization and presputtering times of the TiNx layers, at ambient temperatures. This may be the main cause for the formation of FeCoVNx phase at the FeCoV target in our case.
Thus, when we try to deposit the next FeCoV layer, the
FeCoVNx phase is transferred from the FeCoV target to the
film and hence the extra peak. Because a large number of N
atoms can occupy the interstitial sites, a large lattice expansion of the FeCoVNx phase can be observed. In the second
possibility, the adsorbed N atoms on an already grown TiNx
layer diffuse into the growing FeCoV when we deposit a
fresh FeCoV layer. As these adsorbed atoms are at the surface of Ti, they can be readily absorbed by the FeCoV. However, this absorption phenomenon will not result in a large
lattice expansion of the entire FeCoV layer because the concentration of the adsorbed N is so small that only the first
few atomic layers of FeCoV can absorb N. In the third possibility, an already deposited FeCoV layer absorbs N atoms
during the gas stabilization and presputtering times of the
next TiNx layer. In this case, a considerable lattice expansion
of FeCoV is expected due to the occupancy of the interstitial
sites by N atoms if the content of N is large. This absorption
phenomenon is similar to that described herein for the presence of N atoms at the FeCoV target surface. Hence, the first
and the third possibilities are mainly responsible for the large
increase in d spacings of FeCoVNx (110) determined from
the XRD data.
In order to clearly distinguish the significant roles of the
first and third possibilities, we have prepared the third series
of samples. They are prepared under identical conditions as
those of the second series except that the presputtering time
M. S. Kumar and P. Boni
3753
FIG. 3. 共a兲 Schematic of the FeCoV/TiNx multilayers having thick FeCoV
layers. The dots shown in the FeCoV layers are the nitride nuclei that form
by absorbing the N atoms that are adsorbed on the already grown TiNx layer.
The density of the nuclei decreases as we go away from the bottom of the
FeCoV layer and the nuclei are completely absent at the top part. 共b兲 Schematic of the FeCoV/TiNx multilayers having thin FeCoV layers. In this case,
a considerable number of nuclei is present at the top part of the FeCoV layer
thus favouring the absorption of N when the next TiNx layer is sputtered.
of the individual FeCoV layers in this case is 5 min instead
of 1 min for the second series. The XRD data of these
samples are shown in Fig. 1共c兲. As can be seen from Fig.
1共c兲, the FeCoVNx (110) peak vanishes for t FeCoV⫽200 Å.
However, for the second series as shown in Fig. 1共b兲, we see
that this peak is present for t FeCoV⫽200 Å, indicating the
existence of the nitride phase in the FeCoV layer due to the
transfer of the nitride phase from the FeCoV target. Hence,
in the case of the second series, most of the N atoms come
from the target and less is coming from the N2 partial pressure during the stabilization and presputtering times.
In the case of the third series, as seen from Fig. 1共c兲, it is
also clear that even after the longer presputtering time of 5
min strong FeCoVNx (110) peaks are present for t FeCoV⫽30
and 50 Å. We presume that the FeCoVNx phase forming at
the FeCoV target surface is sputtered away during the presputtering time. In such a case, the pure FeCoV layer alone
grows on the substrates. Thus, the third possibility, i.e., the
diffusion of N atoms into the already grown FeCoV layer
during the gas stabilization and presputtering times for the
next TiNx layer, also seems to exist.
A possible mechanism for the absorption of the N atoms
by the FeCoV layer at ambient temperatures, is explained as
follows. When an FeCoV layer starts to grow on a just finished TiNx layer, the adsorbed N atoms on the TiNx layer
diffuse into the FeCoV lattice thereby creating nuclei of
FeCoVNx (110) phase. In general, it is known that such nuclei of the nitride/hydride/carbide phases initially form. The
final homogeneous interstitial phase forms when these nuclei
enlarge and finally coalesce. In our case, the FeCoVNx nuclei
are formed by the intake of the adsorbed N on the TiNx layer.
Because the concentration of N in this case is small the
FeCoVNx (110) nuclei do not spread in the entire FeCoV
layer. A schematic diagram of the distribution of the nuclei in
the FeCoV layers is shown in Figs. 3共a兲 and 3共b兲 for thick
and thin FeCoV layers, respectively. In Figs. 3共a兲 and 3共b兲,
the solid circles indicate the nuclei and it can be seen that the
concentration of the nuclei gradually decreases as we go
from the bottom to the top of the FeCoV layer. When the
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
3754
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
process of depositing the next TiNx layer starts, we take
about a minute to stabilize the partial pressure of the nitrogen
gas in the sputtering chamber. Then presputtering of the target is performed before the actual deposition of the TiNx
layer begins. When the presputtering is in progress, the substrates are kept away such that the deposition of the layers on
the substrates does not takes place though the target is continuously sputtered. During the stabilization and presputtering times the FeCoVNx (110) nuclei enlarge due to the diffusion of N atoms into the FeCoV lattice. This is the
diffusion mechanism in the FeCoV layers for t FeCoV⭐50 Å,
for the third series of the samples.
For t FeCoV⫽200 Å, as can be seen from Fig. 1共c兲, the
FeCoVNx (110) peak vanishes. This indicates that the concentration of N in the FeCoV layer of this sample is negligible. From Fig. 3共b兲, we can see that the top portion of a
thick FeCoV layer is free from FeCoVNx (110) nuclei. Due
to the lack of these nuclei the absorption of the N atoms does
not occur during the deposition of the next TiNx layer and
hence the growth of the FeCoVNx phase is not favorable. As
the volume fraction of the FeCoVNx (110) nuclei is very
small the XRD data do not show the FeCoVNx (110) peak
for t FeCoV⫽200 Å in the case of the third series. However,
instead of the peak there is a slight increase in background
due to these nuclei. Furthermore, it shall be pointed out that
there is a strong FeCoVNx (110) peak for t FeCoV⫽100 Å of
the third series. If the presence of this peak is due to the first
possibility, then for t FeCoV⫽200 Å of the same third series
we shall be able to see a weak FeCoVNx (110) peak since the
first 100 Å part of the 200 Å thick layer should contain the
nitride transferred from the FeCoV target. But the complete
absence of the FeCoVNx (110) peak for this sample is in
support of the third possibility, i.e., the absorption of N atoms by the FeCoV layers at ambient temperatures.
In the following, we discuss the enhancement of the FeCoV共200兲 peak in FeCoV/TiNx system. It is known that a
growing surface of a film generally tries to keep the highest
in-plane atomic density.13 In the case of the pure bcc FeCoV,
the 共100兲, 共110兲, and 共111兲 planes have the atomic densities 1
atom per a 2 , 1.414 atoms per a 2 , and 0.577 atom per a 2 ,
respectively. Here a is the lattice parameter. In the case of
FeCoV/Ti system, as can be seen from Fig.1共a兲, the 共110兲
reflections are the strongest since the 共110兲 planes have the
highest atomic density. When the growth of the layers takes
place in the presence of reactive atoms like N, O, H, and C,
the atomic density has to be computed by considering the
reactive gas atoms occupying the interstitial sites.14,11 In bcc
Fe, there are tetrahedral and octahedral interstitial sites
available.15 It is generally observed that the N atoms occupy
octahedral sites in many Fe nitrides.16,17 Since the composition of FeCoV is closer to Fe50Co50 which has a bcc structure
similar to pure Fe 共Ref. 18兲 we are comparing FeCoVNx and
Fe–N. In bcc Fe, there are four tetrahedral sites in each 共100兲
plane and therefore there are 12 such sites in a unit cell.
Similarly there are five octahedral sites in each 共100兲 plane
and therefore there are nine octahedral sites per unit cell.
Even though there are many interstitial sites available, at low
N concentrations, it can be expected that N atoms occupy
only one or two interstitial sites per unit cell, probably in a
M. S. Kumar and P. Boni
FIG. 4. 共a兲 Grain size versus t FeCoV determined from FeCoV共110兲 and
FeCoVNx (110) peaks in both FeCoV/Ti and FeCoV/TiNx 共second series兲
systems. The grain sizes for t FeCoV⫽30, 50, and 100 Å in FeCoV/Ti system
are only indicative. The exact values could not be deduced due to the presence of satellite peaks. 共b兲 Grain size versus t FeCoV determined from FeCoV共200兲 peaks in both FeCoV/Ti and FeCoV/TiNx 共second series兲 systems.
random fashion. In our case, it is possible to have the highest
in plane atomic density for the 共100兲 planes if one octahedral
site at the face center of the bcc unit cell is occupied. On the
other hand, even when only one tetrahedral site of the bcc
FeCoV unit cell is occupied by N, the in-plane atomic density for 共100兲 planes is the highest since these tetrahedral
sites are not located in the 共110兲 or 共111兲 planes where Fe
atoms are located. We, therefore, conclude that the enhancement of the FeCoV共200兲 peak in the XRD patterns in the
case of FeCoV/TiNx system is due to the interstitial occupancy of N atoms. However, we can not determine whether
the N atoms occupy tetrahedral or octahedral sites, from our
experimental data.
Another possible explanation for the changes in the intensities of the peaks is the loss of texture of the FeCoV
layers. In the case of bulk Fe50Co50 , the Fe共110兲 peak is the
strongest and the ratio of the intensities of Fe共110兲 and
Fe共200兲 is 100:8.18 In our FeCoV/TiNx system, for t FeCoV
⫽100, 200, and 300 Å, the FeCoV共200兲 peak is the strongest
indicating that a simple loss of texture does not occur. These
observations support the fact that the enhancement of the
intensities of the FeCoV共200兲 peak occurs to keep the highest in-plane atomic density.
In Fig. 4共a兲, the grain sizes obtained from the FeCoV共110兲 and FeCoVNx (110) peaks in FeCoV/TiNx system
共second series兲 are plotted for all the samples. The grain
sizes are determined by using Scherrer formula.19 We infer
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
M. S. Kumar and P. Boni
3755
FIG. 6. 共a兲 Thickness dependence of stress ␴ in FeCoV/Ti 共first series兲 and
FeCoV/TiNx 共second series兲 multilayers. ␴ 储 and ␴⬜ are the stress measured
along and perpendicular to the substrate movement. 共b兲 The difference stress
⌬ ␴ ⫽ ␴⬜ ⫺ ␴ 储 determinedfrom Fig. 6共a兲 as a function of t FeCoV .
grains are virtually free from oxygen and nitrogen atoms that
were used as reactive gases.11 In the following sections, the
influence of N on the physical properties is discussed.
B. Stress and induced magnetic anisotropy
FIG. 5. 共a兲 In-plane hysteresis loops for FeCoV/Ti system 共first series兲. 共b兲
In-plane hysteresis loops for FeCoV/TiNx 共second series兲 system. The solid
and open symbols represent the measurements along the 储 and ⬜ directions,
respectively. The thickness t of the FeCoV layers are shown.
from Fig. 4共a兲 that the expansion of the FeCoVNx (110) is
associated by a sharp reduction in grain size when compared
with the grain size determined from the FeCoV共110兲 peaks.
This reduction in grain size is due to the large lattice distortion caused by the N atoms at the interstitial sites.
It can further be seen from Fig. 2共a兲 that the d spacing of
the FeCoV共200兲 planes is nearly the same for all the samples
indicating the absence of any lattice expansion. Additionally,
from Fig. 4共b兲, we can see that the grain size as a function of
t FeCoV determined from the FeCoV共200兲 peaks of the first
and second series behaves in a similar fashion. This indicates
that the size of the grains with 共200兲 orientation is not effected by the reactive sputtering. The absence of lattice expansion indicates that the concentration of N in the FeCoV共200兲 grains is rather small. Since the content of N in
these grains is small, they grow like pure FeCoV phase so
that their characteristic grain size is similar to that observed
in the FeCoV/Ti system, as seen in Fig. 4共b兲. We have a
similar observation in Ni/Ti multilayers where the Ni共111兲
We have performed stress and magnetic hysteresis measurements to study the magnetic anisotropy observed in the
FeCoV layers. The hysteresis measurements at room temperature are performed using an extraction magnetometer.
We have performed the magnetic measurements on the first
and second series of samples only. The third series is prepared only for structural studies. From the magnetic measurements, we have found that the magnetization lies in the
plane of the layers. Figure 5 shows the in-plane hysteresis
loops for some samples. From Fig. 5, we can see that there
are easy and hard axes of magnetization within the plane of
the layers indicating the presence of an in-plane easy axis of
magnetization. The in-plane easy axis of magnetization in
our samples arises due to the stress in the layers that acts
through magnetostrictive effects.3,20 All the FeCoV/Ti and
FeCoV/TiNx samples prepared by us exhibit a large intrinsic
stress anisotropy. The stress in the layers determined for various samples is plotted in Fig. 6. The stress measured parallel
and perpendicular to the substrate movement are designated
as ␴ 储 and ␴⬜ , respectively. It can be seen from Fig. 6 that
the stress as a function of t FeCoV shows a significant variation. Almost all the reactively sputtered samples display a
smaller stress than the nonreactively sputtered samples. In
spite of these variations the difference stress ⌬ ␴ ⫽ ␴⬜ ⫺ ␴ 储
shown in Fig. 6共b兲 is almost constant for all samples indicating that the reactive sputtering does not alter ⌬␴ signifi-
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
3756
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
M. S. Kumar and P. Boni
FIG. 7. Magnetoelastic energy as a function of t FeCoV for FeCoV/Ti 共first
series兲 and FeCoV/TiNx 共second series兲 system. The solid symbols represent
the energy obtained from magnetization measurements. The open symbols
represent the energy calculated from the stress data.
cantly. Since ⌬␴ is tensile and FeCoV has a positive magnetostriction coefficient, the easy direction of magnetization is
perpendicular to the substrate movement.
We have calculated the magnetoelastic energy from the
stress data for the FeCoV/TiNx system. The magnitude of
this energy is equal to 1.5␭ s (⌬ ␴ ) for random polycrystalline
samples.21 The value for the saturation magnetostriction coefficient ␭ s for the bulk Fe50Co50 alloy is about 83.4⫻10⫺6
共Ref. 21兲. Since this alloy composition is very close to the
composition of the FeCoV alloy, we have used this value for
calculating the magnetoelastic energy and the calculated data
is shown in Fig. 7. The magnetoelastic energy is also determined from the hysteresis data and it is plotted in Fig. 7. In
this case, this energy is equal to the difference in the area
determined from the area under the curves 共in the first or
third quadrant兲 for the hysteresis measurements along the ⬜
and 储 directions.22 For the FeCoV/Ti system, the energies
determined from the magnetization and stress data show
similar trends though the magnitudes differ considerably for
smaller t FeCoV . The large discrepancy at smaller t FeCoV may
be due to the role of interfaces where ␭ s may be considerably
different from its bulk value. The second factor that is responsible for the deviation may be the textured nature of the
FeCoV layers. In the case of FeCoV/TiNx , there is a considerable disagreement of the data for smaller tFeCoV and the
curves show opposite trends. We presume that these opposite
trends may be due to presence of the FeCoVNx phase that
should have a ␭ s considerably different from that of the FeCoV. As the volume fraction of the FeCoVNx phase decreases as t FeCoV increases, a good agreement between the
data is observed for larger t FeCoV .
In the following, we discuss the magnetic anisotropy observed in the plane of the layers in some magnetic systems.
Tomaz et al.23 have reported that in Fe/Cr multilayers, the
in-plane anisotropy of the samples correspond to their epi-
FIG. 8. 共a兲 Remanence M r /M s as a function of t FeCoV in FeCoV/Ti and
FeCoV/TiNx 共second series兲. M r /M s is around 95% for FeCoV/TiNx system. 共b兲 H c of FeCoV/Ti and FeCoV/TiNx 共second series兲 systems as a
function of t FeCoV . 共c兲 H c of FeCoV/Ti and FeCoV/TiNx 共second series兲
systems as a function of grain size determined from the FeCoV共200兲 peaks.
All the data shown in 共a兲, 共b兲, and 共c兲 are obtained from the first and the
second series of samples.
taxial orientation of the magnetic layers. They grew three
multilayers with orientations 共100兲, 共211兲, and 共110兲 which
were simultaneously sputtered onto MgO共100兲, MgO共211兲,
and Al2 O3 (112̄0) substrates, respectively. The 共100兲 samples
showed a four-fold 共two in-plane easy axes separated by
90°兲, while the 共110兲 and 共211兲 samples showed a two-fold
symmetry. However, they could not determine which inplane crystallographic direction is associated with the easy
axis in each sample. Haeiwa et al.24 have observed in-plane
magnetic anisotropy in Fe/Al system. For t Fe /t Al⭓3, the
easy direction coincided with the direction of substrate motion. The Fe layers of their Fe/Al system have 共110兲 texture
which is similar to our FeCoV/Ti system. But we observe the
easy direction along the ⬜ direction. Thus, it appears that the
crystallographic orientation of Fe or FeCoV is not the impor-
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
M. S. Kumar and P. Boni
3757
tant factor in determining the easy direction, at least for the
textured samples prepared by sputtering. The good agreement between magnetoelastic energy determined by us from
the stress and magnetization measurements indicates that the
magnetostrictive effect plays a major role in determining the
easy axis.
C. Hysteresis behavior
From the hysteresis data of the samples, the remanence
M r and coercivity H c are determined and they are plotted
against t FeCoV in Figs. 8共a兲 and 8共b兲, respectively. It can be
seen from the Fig. 8 that almost all the reactively sputtered
samples show higher remanence than the samples sputtered
in pure Ar. This indicates that the reactive sputtering improves the remanence of the FeCoV layers. The values of M r
for FeCoV/TiNx system is about 95% and hence these multilayers are used as the coatings for polarizing supermirrors
that operate in the remanent state.1,2
Figure 8共b兲 shows the variation of coercivity with t FeCoV
for both systems. From Fig. 8共b兲, we can see that the reactive
sputtering reduces the coercivity considerably. The coercivities for the FeCoV/TiNx system are quite low and follow a
linear behavior. In the case of the FeCoV/Ti system, H c is
large and linear behavior is observed for t FeCoV
⫽30– 100 Å. For t FeCoV⫽700 Å, H c is nearly the same as
that for t FeCoV⫽100 Å. We attribute the large difference in
H c and the linear behavior observed for both systems to
grain size effect.
In the following, we discuss the behavior of H c . Let us
first consider the linear behavior observed in both systems.
Various reports on the variation of H c in multilayers are
available in the literature. In as-deposited Fe–N/Al multilayers, Barnard et al.25 have observed an increase in H c with
t Fe–N from 2 Oe for 32 Å to about 115 Oe for 130 Å. Gao
et al.26 also reported an increase of H c in Co/Cu and
Co95B5 /Cu systems with t Co and t Co95B5 , respectively. Similar behavior is also noticed by Dirne et al.27 for Fe/CoNbZr
multilayers. Most of the aforementioned results were obtained on samples with a constant thickness of the nonmagnetic layers. In our samples, however, t TiNx is varied. In spite
of this, the behavior of our samples is comparable with the
multilayers previously mentioned.
There are reports that contradict this behavior by displaying a decrease in H c with increasing thickness of the
magnetic layers. Matsumoto et al.28 have reported that in the
case of Fe/Ti, Fe/Ta, Fe/Al, Fe/Ag, and Fe/Cu systems, H c
decreases with increasing t Fe . Haftek et al.29 have reported a
decrease in H c from 80 to 7 Oe with increasing t Ni in the
Ni/Ti system. Lafford et al.30 have also reported that H c decreases with increasing t FeCo in FeCo/Ag multilayers.
Löffler et al.31 have reported coercivity in nanostructured Fe. In this case, H c increases from 5 Oe for the grain
size of 100 Å to 100 Oe for about 350 Å. The H c decreases
above 350 Å and attains a value of 10 Oe for 1000 Å. Thus,
a maximum of H c at about 350 Å is observed. They have
explained the behavior of H c for the grain sizes between 100
and 350 Å by a random anisotropy model.31 For random
orientation of the anisotropy axes of the grains, the effective
FIG. 9. Plot of the magnetization of FeCoV layers in an M s t FeCoV versus
t FeCoV representation. A good agreement between the data for the samples
prepared by reactive sputtering and in pure Ar is seen. The data shown are
obtained from the first and second series of samples.
anisotropy is remarkably reduced. This model shows a decrease of the anisotropy energy and consequently a decrease
of H c with decreasing grain size.
In Fig. 8共c兲, H c as a function of the grain size determined from FeCoV共200兲 peaks in both FeCoV/TiNx and
FeCoV/Ti is shown. It can be seen that H c does not vary
linearly with grain size. Furthermore, in the case of the
FeCoV/Ti system, H c for t FeCoV⫽700 Å is almost the same
as that for t FeCoV⫽100 Å. This shows that there is a maximum of H c between t FeCoV⫽100 and 700 Å, i.e., for the
grain size between 100 and 200 Å. It can be noted that the
H c values observed for both systems are much larger than
the value H c ⬇3 Oe for bulk FeCoV 共data supplied by the
manufacturer Vacuumschmelze GmbH, Hanau, Germany兲.
Hence, for t FeCoVⰇ700 Å, H c should tend towards a smaller
value of the order of the bulk value of 3 Oe. In the case of
the FeCoV/TiNx , the maximum may be observed for t FeCoV
⬎700 Å. Löffler et al.31 have explained the increase in H c
with decreasing grain size until the maximum, in nanostructured Fe, by domain wall pinning at grain boundaries that
becomes progressively more efficient as the volume fraction
of the grain boundaries increases. In our case, the pinning
effect can come from both FeCoV grain boundaries and interfaces of the multilayer structure.
We can further see from Figs. 8共b兲 and 8共c兲 that there is
a large difference in H c between the two systems. From Fig.
4共b兲, it can be seen that the grain size determined from
FeCoV共200兲 peaks for both systems is nearly the same indicating the absence of any influence due to reactive sputtering. But the grain size obtained from FeCoV共110兲 and
FeCoVNx (110) peaks show a considerable reduction in the
grain size of the FeCoV layer in the FeCoV/TiNx system due
to reactive sputtering. Thus the smaller grain size of
FeCoVNx (110) phase leads to a much reduced H c in the
FeCoV/TiNx system.
The thickness dependence of magnetization and the influence of reactive sputtering on the magnetization are dealt
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
3758
J. Appl. Phys., Vol. 91, No. 6, 15 March 2002
with in the following. The saturation magnetization determined for all the samples sputtered with and without N are
plotted as M s t FeCoV versus t FeCoV in Fig. 9. This plot
is based on the phenomenological expression M s ⫽M 0 关 1
⫺(2 ␦ /t FeCoV) 兴 where ␦ is the thickness of the dead magnetic
layers at the interfaces and M 0 is the saturation magnetization of the bulk FeCoV. M s is the saturation magnetization of
the FeCoV layers in our multilayers. Figure 9 clearly shows
that the data for both systems follow a linear behavior. The
slope of the straight line fit yields M 0 and the x intercept
gives the dead layer thickness. The deduced value for ␦ from
the fit is 2.1 Å. A good agreement of the data for all the
samples in both systems indicates that the content of N in the
FeCoV layers and interfaces has no detectable influence on
the magnetization of the layers.
IV. CONCLUSIONS
XRD data of FeCoV/TiNx and FeCoV/Ti systems show
that the FeCoV layers have preferred 共110兲 orientation when
Ti is sputtered in pure Ar. But reactive sputtering of the Ti
results in the following: 共a兲 Lattice expansion of the Ti layers
as shown by the TiNx (002) peak, 共b兲 lattice expansion in the
FeCoV layers as indicated by the appearance of an extra
peak corresponding to FeCoVNx grains with 共110兲 orientation, 共c兲 enhancement of the FeCoV共200兲 peak in order to
keep highest in-plane atomic density in the growing atomic
layers. The changes observed in the FeCoV layers are due to
presence of interstitial N atoms. Magnetic hysteresis measurements show that the magnetization lies in the plane of
the layers. Within the plane of the layers an easy axis of
magnetization is observed in the perpendicular direction of
the substrate movement. We have shown that the in-plane
easy axis arises due to magnetostrictive effects through mechanical stress in the layers. The hysteresis measurements
further show that the reactive sputtering improves the remanence. Remanence of about 95% for the samples of
FeCoV/TiNx system is observed. The coercivities of the
FeCoV/Ti and FeCoV/TiNx systems are remarkably different. We have attributed this difference to grain size effect.
The increase in coercivity with t FeCoV for both systems is
consistent with the random anisotropy model. The coercivity
data also display the role of FeCoV grain boundaries and
interfaces in pinning the domain walls. The hysteresis curves
also show that there is no observable influence of the reactive
sputtering on the saturation magnetization of the FeCoV layers.
ACKNOWLEDGMENTS
The authors thank Professor P. Selvam, IIT Bombay for
valuable discussions. The technical assistance provided by
M. S. Kumar and P. Boni
M. Horisberger, Laboratory for Neutron Scattering ETHZ &
PSI, Switzerland, is gratefully acknowledged.
P. Böni, Physica B 234, 1038 共1997兲.
P. Böni, D. Clemens, M. Senthil Kumar, and S. Tixier, Physica B 241,
1060 共1998兲.
3
A. Vananti, Diploma thesis, ETH Zurich, 1995.
4
D. Wang, M. Tondra, J. M. Daughton, C. Nordman, and A. V. Pohm, J.
Appl. Phys. 85, 5255 共1999兲.
5
K. M. H. Lenssen, A. E. T. Kuiper, and F. Roozeboom, J. Appl. Phys. 85,
5531 共1999兲.
6
R. Zuberek, A. Wawro, H. Szymczak, A. Wisniewski, W. Paszkowicz, and
M. R. J. Gibbs, J. Magn. Magn. Mater. 214, 155 共2000兲.
7
O. Elsenhans, P. Böni, H. P. Friedli, H. Grimmer, P. Buffat, K. Leifer, J.
Söchtig, and I. S. Anderson, Thin Solid Films 246, 110 共1994兲.
8
D. Clemens, A. Vananti, C. Terrier, P. Böni, B. Schnyder, S. Tixier, and M.
Horisberger, Physica B 234, 500 共1997兲.
9
B. M. Clemens, J. Appl. Phys. 61, 4525 共1987兲.
10
S. Tixier, P. Böni, and H. Van Swygenhoven, Thin Solid Films 342, 188
共1999兲.
11
M. Senthil Kumar, P. Böni, and D. Clemens, J. Appl. Phys. 84, 6940
共1998兲.
12
F. E. Wagner and G. Wortmann, in Hydrogen in Metals I, edited by G.
Alefeld and J. Völkl 共Springer, Berlin, 1978兲 p. 131.
13
M. B. Stearns, Mater. Sci. Eng., B B3, 409 共1989兲.
14
H. Grimmer, P. Böni, O. Elsenhans, H. P. Friedli, I. S. Anderson, K. Leifer,
and P. Buffat, Mater. Sci. Forum 126, 615 共1993兲.
15
B. Stalinski, Ber. Bunsenges. Phys. Chem. 76, 724 共1972兲.
16
A Handbook of Lattice Spacings and Structures of Metals and Alloys,
edited by W. B. Pearson 共Pergamon, New York, 1958兲 p. 984.
17
W. Kiriake, K. Kuwahara, H. Iwanaga, and H. Fujiyama, Surf. Coat. Technol. 98, 1293 共1998兲.
18
JCPDS, card No. 44-1433 共ICDD, 1998兲.
19
X-ray Diffraction, edited by B. E. Warren 共Addison–Wesley, Reading,
MA兲, p. 253.
20
M. Senthil Kumar, P. Böni, and M. Horisberger, IEEE Trans. Magn. 35,
3067 共1999兲.
21
S. Chikazumi, Physics of Magnetism 共Wiley, New York, 1964兲, p. 161.
22
M. T. Johnson, P. J. H. Bloemen, F. J. A. den Broeder, and J. J. de Vries,
Rep. Prog. Phys. 59, 1409 共1996兲.
23
M. A. Tomaz, W. J. Antel, Jr., W. L. O’Brien, and G. R. Harp, Phys. Rev.
B 55, 3716 共1997兲.
24
T. Haeiwa, H. Negoro, and M. Matsumoto, J. Appl. Phys. 69, 5346 共1991兲.
25
J. A. Barnard, M. Tan, A. Waknis, and E. Haftek, J. Appl. Phys. 69, 5298
共1991兲.
26
C. Gao and M. J. O’Shea, J. Appl. Phys. 69, 5304 共1991兲.
27
F. W. A. Dirne, F. J. A. den Broeder, J. A. M. Tolboom, H. J. de Wit, and
C. H. M. Witmer, Appl. Phys. Lett. 53, 2386 共1988兲.
28
M. Matsumoto, T. Haeiwa, and H. Negoro, IEEE Transl. J. Magn. Jpn. 7,
149 共1992兲.
29
E. Haftek, M. Tan, A. Waknis, and J. A. Barnard, J. Appl. Phys. 70, 6442
共1991兲.
30
T. A. Lafford, M. R. J. Gibbs, and C. Shearwood, J. Magn. Magn. Mater.
132, 89 共1994兲.
31
J. F. Löffler, J. P. Meier, B. Doudin, J. -P. Ansermet, and W. Wagner, Phys.
Rev. B 57, 2915 共1998兲.
1
2
Downloaded 26 Feb 2012 to 14.139.97.73. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions