Very large unidirectional anisotropy constant of 2.2 erg/cm2 and the appearance of double-shifted loops, ascribed to the coexistence of positive and negative exchange bias, have been obs
Trang 1Journal of Magnetism and Magnetic Materials 304 (2006) 41–45
N.P Thuya,b, , N.A Tuanb, N.N Phuocb,c, N.T Namb, T.D Hienb, N.H Haid
a College of Technology, Vietnam National University, Hanoi, Vietnam
b International Training Institute for Materials Science, Hanoi University of Technology, Hanoi, Vietnam
c Information Storage Materials Laboratory, Toyota Technological Institute, Nagoya, Japan
d Laboratoire Louis Ne´el, C.N.R.S., Grenoble, France
Available online 3 March 2006
Abstract
A systematic study of exchange bias in MnPd/Co and MnPd/Co1xFex bilayers has been carried out Very large unidirectional anisotropy constant of 2.2 erg/cm2 and the appearance of double-shifted loops, ascribed to the coexistence of positive and negative exchange bias, have been observed The dependence of exchange bias, unidirectional anisotropy constant and coercivity on thickness, temperature, annealing regime and Fe content has been investigated and discussed
r2006 Elsevier B.V All rights reserved
PACS: 75.70.Cn; 75.70.–i; 75.25.+z; 75.30.Gw
Keywords: Exchange bias; Double-shifted loops; MnPd/Co 1x Fe x bilayers; Magnetic thin film
1 Introduction
Discovered by Meiklejohn and Bean (MB) on CoO/Co
system, the exchange bias (EB) effect has been explained by
these authors using a model based on the suggestion of the
existence of the unidirectional anisotropy originated from
the exchange coupling between uncompensated spins at the
interface between antiferromagnetic (AFM) and
ferromag-netic (FM) layers[1,2] This model has been considered so
far as a simple and ideal model due to the fact that the EB
field HE and the unidirectional anisotropy constant JK
predicted theoretically are two to three orders of
magni-tude larger than those observed on most of the AFM/FM
systems The JK value measured on the CoO/Co system
was 3 erg/cm2 at 10 K being rather unique and still far
smaller than the theoretical value of 10 erg/cm2[3]
Studies on EB in AFM/FM bilayers have therefore been
developed in two trends The first is to develop different
theories that can adequately explain the actual
experi-mental EB values[4,5] The second one, on the other hand,
is to bring the experimental values of HEand JKcloser to those predicted by MB theory The later trend has been realized by two ways: (i) by improving the preparation techniques of the well-known AFM/FM systems so that the crystallographic morphologies approach the perfect states [6,7] and (ii) by looking for novel materials which have larger unidirectional anisotropy [8,9]
Recently a large value of JK of 1.3 erg/cm2 at room temperature has been discovered by Tsunoda et al.[10,11]
on IrMn/Co1xFex system We have also reported quite large values of EB field at cryogenic temperatures in MnPd/Co [12] and MnCr/Co [13] bilayers This paper focuses on a more comprehensive study on MnPd/Co as well as recent experimental results on MnPd/Co1xFex
systems hoping to shed some light on the understanding of
EB effect
2 Experimental results MnPd/Co1xFexbilayer samples have been prepared by RF-sputtering onto Si (1 0 0) wafers in the sequence of Si/ AFM/FM As-sputtered samples have been subsequently annealed at temperatures varied from Tann¼240 to 320 1C
www.elsevier.com/locate/jmmm
0304-8853/$ - see front matter r 2006 Elsevier B.V All rights reserved.
doi:10.1016/j.jmmm.2006.02.063
Corresponding author International Training Institute for Materials
Science (ITIMS), Dai hoc Bach khoa, 1 Dai Co Viet, Hanoi, Vietnam.
Tel.: +84 04 8680787; fax: +84 04 8692963.
E-mail addresses: thuy@itims.edu.vn, thuynp@vnu.edu.vn
(N.P Thuy).
Trang 2in vacuum of 105m bar for 1 h followed by field cooling to
room temperature under a magnetic field of 5 kOe
Scanning electron microscope (SEM) patterns of two
MnPd films with thickness of 6 and 36 nm prepared
separately on the Si (1 0 0) substrates do not show any
structure that ensures the homogeneity of the AFM layers
in our present bilayers The energy dispersion X-ray
spectroscopy (EDS) carried out on these films showed that
their composition is around Mn30Pd70 The bulk material
of the same composition has a CuAu-I type structure with
the Ne´el temperature TNaround 200 K[14]
Magnetic properties have been characterized by
vibrat-ing sample magnetometers (VSM) in temperatures rangvibrat-ing
from 10 to 300 K
Shown in Fig 1a are the hysteresis loops measured at
123 K in the MnPd (tMnPd)/Co (10 nm) samples with
tMnPd¼2, 3, 6, 12, 18 and 36 nm which were annealed at
300 1C For the samples with tMnPdlarger than 18 nm only
the negative exchange bias, corresponding to a
single-shifted loop (SSL), is observed For samples with smaller
MnPd thickness, however, the double-shifted loop (DSL) behavior has been observed As discussed later this DSL can be considered to be the superposition of the subloops with negative exchange bias (NEB) and positive exchange bias (PEB) We can then denote HE1and HE2 as positive and negative EB fields respectively and use their average value The thickness dependence of this average HE is plotted inFig 1btogether with that of the unidirectional anisotropy constant JK derived from it by using the expression JK ¼|HEMStFM|, where MSand tFMare the ferromagnetic layer magnetization and thickness, respec-tively The coercivity HCof this sample series as a function
of tMnPd is presented in Fig 1c, in which the average HC value of those two sub-loops in the case of DSL has been used
The role of annealing regime has been studied on the MnPd (12 nm)/Co (10 nm) sample which were annealed at different temperature Tann¼240, 260, 280, 300, 320 1C Hysteresis curves together with the variation of the exchange biased field HE and the coercivity HC as a function of annealing temperature are presented inFigs 2a and b, respectively
Temperature dependence of the EB phenomenon has been investigated in two samples from the above series, one annealed at 240 1C and has the DSL and the other annealed
-1000 0 1000
H (Oe)
-2000 -1000 0 1000 2000
-1.0
-0.5
0.0
0.5
1.0
H E (Oe)
-1.0
-0.5
0.0
0.5
1.0
-2000 -1000 0 1000 2000
H (Oe)
0 5 10 15 20 25 30 35
0
200
400
600
800
0.4 0.6 0.8 1.0 1.2 1.4
0
200
400
600
800
tMnPd (nm)
H c
H E
J K
2 )
(a)
(b)
(c)
Fig 1 Magnetic measurement results at 123 K on MnPd (t MnPd )/Co
(10 nm) samples with t MnPd ¼ 2 to 36 nm: (a) hysteresis loops, (b) H E and
J K versus t MnPd and (c) H C versus t MnPd (a) quoted from our previous
paper [12] but with corrected Co thickness values.
1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5
500 600 700 800
900
H (Oe)
-H E
H C
T ann = 240° C
T ann = 260 ° C
T ann = 280° C
T ann = 300° C
T ann = 320 ° C
T a (°C)
(a)
(b) Fig 2 Magnetic measurement results on the MnPd (6 nm)/Co (10 nm) samples annealed for 1 h at T ann ¼ 240, 260, 280, 300 and 320 1C: (a) hysteresis loops, (b) H and H as a function of T
Trang 3at 320 1C which has only SSL behavior Magnetic
measurements in the temperature range from 10 K to
300 K (see Figs 3a and b) allowed us to evaluate
temperature dependence of exchange biased fields
HE1,(T), HE2(T) (for sample with DSL) and HE (T) (for
sample with SSL) as shown inFigs 3c and d, respectively
Hysteresis loops measured at 123 K in the MnPd
(30 nm)/Co1xFex(20 nm) sample series in which the alloy
composition of the Co1xFexlayers varied from x ¼ 0 to 1
show only SSL behavior The derived HE, JK and HC
values are plotted inFigs 4a and b as a function of iron
content x in the Co1xFexlayers
3 Discussion
The overall behavior of the systems studied in this paper
are very large EB fields found in samples with rather thick
FM layers, which results in a huge unidirectional
aniso-tropy constant Indeed, a JK value of 2.2 erg/cm2 is
obtained at 123 K for the MnPd (30 nm)/Co (20 nm)
samples (see Fig 4b) Noting that the maximal JK values
reported so far are of 3 erg/cm2 at 10 K on the CoO/Co
bilayer samples[3]and of 1.3 erg/cm2at room temperature
on the MnIr/ Co1xFexsystem[11], the considered MnPd/
Co1xFexbilayer systems thus can be placed to the class of giant EB effect materials Coercivity values of our samples, being 3 kOe at 10 K and 1 kOe at around 120 K, are also remarkable because the coercive field of unpinned Co layer
is 50 Oe at 2 K and 20 Oe at 295 K [15] The large magnitudes of both HE and HC are a clear evidence of very large unidirectional and uniaxial anisotropies induced during the field cooling process through the Ne´el tempera-ture in the AFM/FM system which possesses a strong exchange coupling of the AFM and FM moments at their interface and a very large volume magnetocrystalline anisotropy of the MnPd layer Our recent experiments on these samples on the dependence of JKas a function of the angle between the applied magnetic field and the initial field cooling direction showed that a huge value of
JKE10 erg/cm2
could be attained at an angle near 901[16] The AFM thickness dependence of both HE and HC
presented inFigs 1b and cshows that the onset of biasing appears at tMnPd less than 2 nm and the EB is fully established at around tMnPd¼12 nm In contrast the coercive force shows a peak at about 6 nm We note that the critical thickness for MnPd is approximately of the
-3000 -2000 -1000 0 1000 2000 3000 -0.2
-0.1 0.0 0.1 0.2
M10K M50K M125K M175K M225K
-0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8
-1500
-1000
-500 0 500 1000
H (Oe)
-3000 -2000 -1000 0 1000 2000 3000
H (Oe)
T (K)
0 50 100 150 200 250 300
T (K)
H E
0 200 400
800
600 1000
H E
H E1
H E2
T B
T B
M10K M75K M125K M175K M225K
(d) (c)
Fig 3 Temperature variation of hysteresis loops and the derived EB fields of the MnPd (6 nm)/Co (10 nm) bilayer which is annealed at T ann ¼ 240 1C (a and c) and T ann ¼ 320 1C (b and d).
Trang 4same size as that found in the IrMn/Co system and is much
lower than those observed in most of other AFM layers,
e.g FeMn[15] This can be considered as an indication of
the quite large value of the volume anisotropy of the MnPd
layer providing a large thermal stability for the AFM
domain structure and consequently resulting in a large
coercivity value mentioned above
Although in the limited range of Tann (from 240 to
320 1C) the influence of annealing regime seems not to be
critical to EB values, it has a stronger effect on the
coercivity and especially on the behavior of the hysteresis
curve (seeFig 2b) Indeed, in the sample of MnPd (12 nm)/
Co (10 nm), the appearance of SSL can only be observed
with Tann¼320 1C, below which the sample has DSL
behaviors It is thus interesting to note that the appearance
of the specific kind of loop behavior (either SSL or DSL)
can be tuned not only by the variation of the relative AFM
and FM thickness but also by annealing conditions
However, the mechanism for effect of the annealing regime
is not clear at the present
The origin of the DSL is not identical according to
different authors It may be due to the contribution of
uniaxial anisotropy apart from unidirectional anisotropy
that brings about exchange bias [17,18] or the pinning of
the FM spin in the direction perpendicular to the cooling
field by a mechanism like the Koon trapping [19] In this
paper we suggest that this anomaly may result from the
overlap of PEB and NEB[13,20] This suggestion is verified
by our results on temperature dependence of the EB in two
kinds of samples with either DSL or SSL behavior as presented inFigs 3a–d In case of sample with DSL, the overall loop is considered to consist of two sub-loops each
of which corresponds to one domain of exchange bias, i.e one sub-loop refers to one domain with NEB (character-ized by EB field HE1) and the other to PEB (HE2) We now note in Figs 3c and d that HE1 and HE2 decrease as temperature increases in the same manner as the decrease
of HEin the case of sample with SSL For the sample with DSL the blocking temperature, the temperature above which DSL disappears, is about 200 K, nearly the same as that of the sample with SSL This result shows a strong correlation between DSL and EB One may argue that the DSL can be caused by un-biased loops superimposed on negative biased loops, as often observed in spin-valve structure However, if this were the reason, the kinked points of the ascending and descending branch would be symmetric through the original point of the M–H loops, which is not seen in our results Therefore, this possibility can be ruled out
As demonstrated by Roshchin et al [20], the DSL phenomenon in AFM/FM bilayers can be ascribed to the modification of EB by changing the relative AFM and FM domain sizes When the AFM domain size is larger or comparable to that in the FM layer, sample can be split into independent subsystems with EB of opposite signs resulting in DSL upon cooling sample through the Ne´el temperature Although Roshchin et al [10] have studied the systems with FeF2as the AFM layers their model does not require the interface exchange coupling to be AFM Therefore, it can be applied as well for the MnPd/Co1xFex
case even when the nature of the interface exchange coupling in this system has not known yet The observed variation of the HE or the JK as a function of the AFM thickness in our samples can be understood in this line because the thinner AFM layer corresponds to the larger AFM domain So the decrease of the AFM layer thickness while keeping the FM layer fixed will lead to the appearance of the DSL below some critical MnPd thickness On the other hand, the change from DSL to SSL behavior with Tann increasing shown in Fig 2a can also be explained as caused by the improvement of the interface crystallographic morphology
Concerning the results of the last sample series where the
FM layer is Co1xFexalloy with different Fe content x, it should be noted that HEand HCdepend on the Fe content
in different ways (see Fig 4a) The concentration dependence of JK derived from HE curve is presented in
Fig 4b For a comparison the result of IrMn/Co1xFex
system reported by Tsunoda et al.[10]is also plotted It is clear that the trends of the composition dependence of JK
in the two systems are quite different Furthermore, they both show large discrepancy with the composition varia-tion of the saturavaria-tion magnetizavaria-tion of bulk Co–Fe alloy (see the dashed line in the figure) As analyzed by Tsunoda
et al [10] if the Heisenberg expression for the interface exchange coupling energy is valid for the AFM/FM bilayer
0
200
400
600
800
1000
1200
1400
0.0
0.5
1.0
1.5
2.0
2.5
0.0 0.5 1.0 1.5 2.0 2.5
x
J K
2 )
(a)
(b)
Fig 4 Variation of H E , H C (a) and J K (b) of MnPd (30 nm)/Co 1-x Fe x (20
nm) samples versus the Fe content x in the Co 1-x Fe x layer measured at
123 K J K of IrMn/Co 1-x Fe x system and the saturation magnetization M S
of Co 1-x Fe x alloy are quoted from Ref [10] for comparison.
Trang 5systems, the derived JKvalues should follow the trend of
magnetization The observed results therefore can be
attributed to the electron transfer at the hetero-interface
giving rise to the change of the magnetic moments Our
results showed that this changing may be strongly depend
on the nature of the AFM layer in a system with a given
FM layer This problem certainly needs more effort to be
well understood
4 Concluding remarks
In summary, a systematic study of exchange bias in
MnPd/Co1xFex bilayers has been carried out with the
investigation of the dependence of exchange bias,
unidirec-tional anisotropy constant and coercivity on MnPd
thickness, temperature, annealing regime and the content
of Fe A very large unidirectional anisotropy has been
found and the DSL has been observed, the appearance of
which can be tuned by both changes of layer thickness and
the preparation conditions The study of Fe-composition
dependence of EB in MnPd/Co1xFex bilayers shows a
behaviour which is quite different from that observed in
IrMn/Co1xFexsystem[10]and thus needs further study
Acknowledgement
We express our sincere thanks to Dr N Dempsey and
Dr D Givord from the Laboratoire Louis Ne´el, CNRS,
Grenoble for their kind experimental help and stimulating discussion This work is partly supported by the State Program on Fundamental Research of Vietnam under the Grants 811604 and 811404
References
[1] W.H Meiklejohn, C.P Bean, Phys Rev 102 (1956) 1413.
[2] W.H Meiklejohn, J Appl Phys 33 (1962) 1328.
[3] J Nogues, I.K Schuller, J Magn Magn Mater 192 (1999) 203 [4] D Mauri, et al., J Appl Phys 62 (1987) 2929.
[5] A.P Malozemoff, J Appl Phys 63 (1988) 3874.
[6] A.J Devasahayam, et al., J Appl Phys 83 (1998) 7216.
[7] K Yagami, et al., J Appl Phys 89 (2001) 6609.
[8] M Saito, et al., J Magn Soc Japan 21 (1997) 505.
[9] S Araki, et al., IEEE Trans Magn 34 (1998) 387.
[10] M Tsunoda, et al., J Magn Magn Mater 239 (2002) 182 [11] I Imakita, et al., Appl Phys Lett 85 (2004) 3812.
[12] N.N Phuoc, et al., Physica B 327 (2003) 385.
[13] N.N Phuoc, et al., J Magn Magn Mater 298 (2006) 43.
[14] E Kren, G Kadar, Phys Lett A 29 (1969) 340.
[15] M Ali, et al., Phys Rev B 67 (2003) 172405.
[16] N.T Nam, et al., to be published.
[17] Y.J Tang, et al., J Appl Phys 88 (2000) 2054.
[18] C.H Lai, et al., Phys Rev B 64 (2001) 094420.
[19] N.C Koon, Phys Rev Lett 78 (1997) 4865.
[20] Roshchin, et al., Europhys Lett 71 (2005) 297.