The magnitude of this shift is termed the exchange bias field HE and in almost all cases, the magnetic hys-teresis loop is shifted in the negative field if one defines the direction of t
Trang 1p s s
applications and materials science
status solidi www.pss-a.com
Anomalous training effect
in exchange-biased MnPd/Co bilayers
1 International Training Institute for Materials Science, Hanoi University of Technology, Hanoi, Vietnam
2 Department of Physics, Faculty of Science, National University of Singapore, 117542 Singapore
3 Department of MEMS and Micro-systems Technology, Faculty of Electronics and Communications, College of Technology,
Vietnam National University, Hanoi, Vietnam
Received 27 May 2008, revised 8 September 2008, accepted 25 September 2008
Published online 13 November 2008
PACS 75.25.+z, 75.30.Gw, 75.70.Cn, 81.15.Cd
* Corresponding author: e-mail nnguyenphuoc@yahoo.com , Phone: + 65-6516-2816, Fax: + 65-6777-6126
© 2009 WILEY-VCH Verlag GmbH & Co KGaA, Weinheim
between an antiferromagnet (AF) and a ferromagnet (FM),
discovered in 1956 [1], results in a shift of the hysteresis
loop along the magnetic field axis called exchange bias
(EB) This phenomenon has been studied extensively due
to its widespread application in spin valves and magnetic
tunnel junctions as well as its intriguing physical origin [2]
Normally, exchange bias is described as an additional
uni-directional anisotropy induced by the AF into the FM via
exchange coupling at the interface, causing a single
mag-netic hysteresis loop shifted along the magmag-netic-field axis
after the field-cooling procedure through the Néel point of
the AF The magnitude of this shift is termed the exchange
bias field (HE) and in almost all cases, the magnetic
hys-teresis loop is shifted in the negative field if one defines
the direction of the cooling field (HFC) as the positive
direction This case is referred to as negative EB The
phenomenon of positive EB was first observed in 1996 by
Nogués et al [3] when studying the systems of Fe/FeF2
and Fe/MnF2 They found that the sign of the EB field
changes from negative to positive as the cooling field
in-creases Very recently, it was found that the state of
coex-istence of positive and negative EB could be achieved in
some specific systems [4 – 7] This state manifests itself as
a double hysteresis loop
In this paper, we report on the observation of double-shifted loops in MnPd/Co when the MnPd thickness is less than 18 nm, which can be ascribed to the superposition of positive and negative EB Moreover, we present additional evidences to firmly support this assumption, namely the experimental results of temperature dependence and the observed abnormal training effect
the present work with the structure of Si(100)/MnPd (x nm)/
Co (10 nm) (x = 1.2, 3.6, 6, 12, 18, 36 nm) were fabricated
at room temperature by an RF sputter-deposition system The MnPd layers were sputtered from a composite target constituting a Pd target with Mn chips placed on it The base pressure was about 10–6 mbar, whereas the working argon pressure during deposition was 10–3 mbar The com-position of the MnPd films, identified by energy-dispersive X-ray spectroscopy (EDS), is Mn30Pd70 The samples were then annealed in a high-vacuum oven (10–5 mbar) at the temperature of 570 K for 1 h The purpose of annealing the samples at such a high temperature is to enhance the crys-tallinity of MnPd layers to obtain the antiferromagnetic phase Subsequently, they were cooled in a magnetic field
of 5 kOe to room temperature to induce the exchange bias effect The magnetic properties of the annealed bilayers
Exchange bias has been studied for a series of MnPd/Co
bilayers sputtered onto Si(100) by an RF sputter-deposition
system The double-shifted loops with an anomalous training
effect have been observed The manifestation of
double-shifted loops is interpreted as the coexistence of positive ex-change bias and negative exex-change bias, which is in agree-ment with the temperature dependence and the observed anomalous training effect
Trang 2were characterized by a vibrating sample magnetometer
(VSM) in the temperature range from 10 K to 300 K It is
known that Co may be spontaneously oxidized, forming a
thin CoO layer on the top of the samples In order to check
this possibility, we made a similar sample with the Mo
layer on the top and found that the magnetic behavior of
that sample is identical with the sample without a Mo layer
Hence, the effect of oxidation of the Co layer, if any, may
be neglected in our investigation
3 Results and discussion Figure 1a and b show the
magnetic hysteresis loops measured at T = 120 K of
MnPd (6 nm)/Co (10 nm) and MnPd (36 nm)/Co (10 nm)
bilayers, respectively A double-shifted loop is observed
in the MnPd (6 nm)/Co (10 nm) bilayer, while the
MnPd (36 nm)/Co (10 nm) bilayer shows a single-shifted
loop, i.e normal EB In the series of MnPd(x nm)/
Co (10 nm) samples, double-shifted loops are seen in
sam-ples with the thickness of MnPd layers smaller than 18 nm
These double-shifted loops have also been observed in
FeF2/Ni [4, 5], and CrMn/Co [6] systems and have been
at-tributed to the coexistence of positive and negative EB In
these systems, an antiferromagnetic coupling at the
inter-face of AF/FM is favored, causing a competition between
the Zeeman and the AF/FM exchange energy, which
re-sults in the crossover from negative to positive EB as the
cooling field is increased When the sample is cooled in an
intermediate applied field, the state of coexistence of
posi-tive and negaposi-tive EB is realized, in which a fraction of AF
spins is aligned in the cooling field due to Zeeman energy, while the rest is aligned in the opposite direction due to AF exchange coupling In our case, if the cooling field is small (less than 4 kOe), only negative EB is observed and as the cooling field is beyond 4 kOe, double hysteresis loops are seen However, due to the limit of the field-cooling system (maximum 5 kOe), we cannot observe the state of complete positive EB One should note that the present state of superposition of positive and negative EB differs from other studies [8, 9] where a double loop can be found after demagnetizing the FM, since in our case the cooling field is strong enough to saturate the magnetization of
Co layer For the sake of convenience, HE1 and HE2 are denoted as the positive and negative exchange bias fields and MS1 and MS2 as the spontaneous positive-biased and negative-biased magnetizations, respectively as seen in Fig 1a Figure 1c and d show the dependences of the EB field, unidirectional anisotropy, and coercivity on the MnPd thicknesses It is noted that for the samples with double loops, the EB field of the whole sample is defined
as the average value of HE1, HE2 and HC is estimated as the average value of the coercivities of each subloop The ob-tained unidirectional anisotropy constant JK is rather large (up to 1.1 erg/cm2) compared to other exchange bias sys-tems [2]
Shown in Fig 2 are the magnetization curves for the MnPd (6 nm)/Co (10 nm) bilayer measured at T = 120 K with the applied magnetic field rotated in the plane of the film It is seen that as the angle θ between the applied field
Figure 1 Magnetic hysteresis loops of MnPd (6 nm)/Co (10 nm) and MnPd (36 nm)/Co (10 nm) bilayers measured at 120 K (Fig 1a
and b) The dependence of exchange bias ( H E ), unidirectional anisotropy constant ( J K ) and coercivity ( H C ) on MnPd thickness (Fig 1c and d) The definitions of H E1 , H E2 , M S1 and M S2 are shown in Fig 1a (See the text for more detail.)
Trang 3Figure 2 M–H loops of MnPd (6 nm)/Co (10 nm) at T = 120 K
measured at θ = 0°, 45° and 90° with respect to the field-cooling
direction as shown in the inset
and the cooling field is 45°, we still observe the
double-shifted loops, although the subloops are slanted, indicating
that the applied field is now deviated from the easy axis
As θ = 90°, i.e the applied magnetic field is now along the
hard axis, both subloops become hard-axis curves,
result-ing in a total sresult-ingle hard-axis loop as observed in Fig 2
The present result suggests that double-shift loops
ob-served in our system are quite different from that reported
earlier in the Refs [10 – 12], where double hysteresis loops
are found when the external magnetic field is applied along
the hard axis of the AF, attributed to an additional
biquad-ratic AF – FM interaction
For more evidences to support the idea of two
oppo-sitely oriented AF domains, we have carried out a study of
the temperature dependence of this effect Shown in
Fig 3a are some representative magnetic hysteresis loops
of the MnPd (12 nm)/Co (10 nm) bilayer measured at
various temperatures It is seen that the double-shifted
loops disappear as the temperature is beyond 220 K,
which is the same as the blocking temperature of the
MnPd (36 nm)/Co (10 nm) sample that exhibits a
single-shifted loop The blocking temperature of about 220 K is
very close to the Néel point of Mn30Pd70 bulk materials
[13] The temperature dependence of positive EB field
(de-noted as HE1) and negative EB field (denoted as HE2) is
shown in Fig 3b It is clearly seen that the value of
posi-tive and negaposi-tive EB fields, though different, are varied in
a similar manner This fact can be considered as additional
evidence that the double-shifted loop originates from the
coexistence of positive and negative EB It should be
men-tioned that this explanation is valid only in the case of an
atomically flat interface with uncompensated spins In
real-ity, the natural interface roughness may play a vital role in
the mechanism of exchange bias as argued by some
au-thors [16, 17] Hence, the AF layer may be split into
multi-domain structure, which causes the applied field of 5 kOe
to be subcritical and supercritical, corresponding to the
crossover from negative EB to positive EB
More interestingly, we have observed an abnormal
training effect in the samples that exhibit double-shifted
Figure 3 (a) M–H loops of MnPd (6 nm)/Co (10 nm) bilayer measured at different temperatures (b) Temperature dependence
of positive ( H E1 ) and negative ( H E2 ) exchange bias fields in MnPd (6 nm)/Co (10 nm) bilayer H E1 and H E2 are defined in Fig 1a
loops Figure 4a and b present some representative mag-netic hysteresis loops of a MnPd (12 nm)/Co (10 nm) film after some cycles of measurement and the corresponding values of positive (HE1) and negative (HE2) EB fields as a function of the number of measurements, respectively Since the magnitudes of MS1 and MS2 correspond to the ar-eas that make positive and negative EB respectively, we can denote the normalized values of mS1 and mS2, where
mS1 = MS1/(MS1 + MS2) and mS2 = MS2/(MS1 + MS2), as the fractions of positive- and negative-biased areas The de-pendence of these positive- and negative-biased area frac-tions on the cycle measurement is shown in Fig 4c At first, the negative EB decreases very rapidly from 840 Oe to
740 Oe and levels off after 6 cycles, while the positive EB decreases slower After 11 cycles, there is no change in mS1 and mS2 At cycle 12, there is a drastic jump of HE1 and HE2
as well as mS1 and mS2 The drastic change of mS1 and mS2, namely the increase of mS2 and the decrease of mS1, implies that there is an enlargement of the negative-biased area at the cost of reducing the positive-biased area A similar ab-normal training effect has also been observed in the system
of CrMn/Co bilayers [6] and has been explained as the movement of the AF domain wall toward the positive-
Trang 4Figure 4 (a) Representative of training-effect hysteresis loops of
MnPd (12 nm)/Co (10 nm) film M–H loops for the cycles 1, 6,
12, 24 and 48 are shown (b) Positive ( H E1 ) and negative ( H E2 )
ex-change bias fields as a function of cycles of measurement (c)
Fractions of positive ( m S1 ) and negative ( m S2 ) exchange bias areas
as a function of cycles of measurement
biased domain Of great interest is the reverse training
effect for negative EB as seen in Fig 4b HE2, which is
de-creased to 740 Oe after 11 cycles of measurement, regains
the previous value of 840 Oe at cycle 12 This surprising
effect implies that the movement of the AF domain wall,
causing the enlargement of the negative-biased area,
rein-duces the untrained state Recently, Brems et al [14]
re-ported on the reverse training effect in CoO/Co bilayers, in
which the EB field regained its untrained value after
carry-ing out a M–H loop measurement along the hard axis and
this effect can be interpreted as a change in the
magnetic-domain structure in the AF layer It should be noted that
recently the physical origin of exchange bias was described
as due to a fraction of uncompensated interfacial spins
(about 4%) that are locked to the AF lattice and do not
ro-tate in an external magnetic field, while most of the other interfacial spins are affected by the external field [15] The training effect can therefore be understood as the loss of the pinned uncompensated spins Taking into account this argument, we can state that after the movement of the AF domain wall causing the enlargement of the negative-biased area, the fraction of uncompensated interfacial spins increases to its original value that results in the reverse training effect observed in our system However, for the decrease of the positive-biased area, the positive exchange bias decreases from 400 Oe to 250 Oe after the movement
of the AF domain wall, i.e the reverse training effect has not been observed in the positive subloop It is well known that positive exchange bias is in a high-energy state so it is less stable than the negative exchange bias Therefore, we may expect that after the AF domain-wall movement, the fraction of pinned uncompensated spins will be reduced It
is also very interesting to see that as we carried out more measurement, namely cycle 12 upward, the positive EB is not changed while the change of negative EB from cycle
12 to cycle 17 is very similar to the change from cycle 1 to cycle 6 This implies that after the movement of the AF domain wall, the negative-biased area is reinduced to the untreated state It is noticed that previously, Nowak et al [17, 18] explained the training effect in terms of the rear-rangement of AF domain by using their domain state model Hauet et al [19] studied the mechanism of the training effect in hard/soft Tb12Fe88/Gd40/Fe60 bilayers with positive exchange bias and found that a partial magnetiza-tion reversal of soft layers generates new domain of the hard layers, which has been subjected to a training effect
It is evidenced from their experiments that the training ef-fect is due to an irreversible reorientation of the hard-layer magnetization Hence, the present explanation employing the idea of the movement of AF domain seems to be rea-sonable for the interpretation of the abnormal training ef-fect However, one should not rule out other possibilities, e.g the granular model for the AF with thermal activation [20, 21], as a potential explanation for this behavior Re-cently, Binek [22] considered the training effect in the framework of nonequilibrium thermodynamics and found that training of the exchange bias effect originates from spin configurational relaxation, which is activated through consecutive-cycled hysteresis loops Based on this idea, one can explain the breakdown of the power-law behavior when n = 1, which is quite similar to the nonmonotonic behavior in this work Further investigation may thus need
to be performed to get a better understanding of this in-triguing effect
4 Summary and conclusion To summarize, we
pre-sent an EB system exhibiting a double-shifted loop, which results from the overlap of the two oppositely biased loops The angular and temperature dependences are consistent with this argument An abnormal training effect is qualita-tively explained using this assumption We believe that the present results will be useful for understanding the
Trang 5mecha-nism of some peculiar effects associated with exchange
bias
AcknowledgementsWe thank Dr Givord and Dr
Demp-sey from the Laboratory Louis Néel for their kind experimental
supports and stimulating discussions This work is supported by
the State Programs on Fundamental Research of Vietnam under
the Grants 4.049.06 and 4.105.06
References
[1] W H Meiklejohn and C P Bean, Phys Rev 102, 1413
(1956)
[2] J Nogués and I K Schuller, J Magn Magn Mater 192,
203 (1999)
[3] J Nogués, D Lederman, T J Moran, and I K Schuller,
Phys Rev Lett 76, 4624 (1996)
[4] I V Roshchin, O Petracic, R Morales, Z P Li, X Batlle,
and I K Schuller, Europhys Lett 71, 297 (2005)
[5] O Petracic, Z P Li, I V Roshchin, M Viret, R Morales,
X Batlle, and I K Schuller, Appl Phys Lett 87, 222509
(2005)
[6] N N Phuoc, N P Thuy, N A Tuan, L T Hung, N T
Thanh, and N T Nam, J Magn Magn Mater 298, 43
(2006)
[7] N N Phuoc, N A Tuan, N P Thuy, D Babonneau, and
J Rabier, Physica B 327, 385 (2003)
[8] P Miltényi, M Gierlings, M Bamming, U May, G
Gün-therodt, J Nogués, M Gruyters, C Leighton, and I K
Schuller, Appl Phys Lett 75, 2304 (1999)
[9] C L Chien, V S Gornakov, V I Nikitenko, A J Shapiro, and R D Shull, Phys Rev B 68, 014418 (2003)
[10] C H Lai, Y H Wang, C R Chang, J S Yang, and Y D Yao, Phys Rev B 64, 094420 (2001)
[11] T Zhao, H Fujiwara, K Zhang, C Hou, and T Kai, Phys Rev B 65, 014431 (2001)
[12] H Shi and D Lederman, Phys Rev B 66, 094426 (2001)
[13] E Krén and G Kádár, Phys Lett A 29, 340 (1969)
[14] S Brems, D Buntinx, K Temst, C Van Haesendonck,
F Radu, and H Zabel, Phys Rev Lett 95, 157202 (2005)
[15] H Ohldag, A Scholl, F Nolting, E Arenholz, S Matt,
A T Young, M Carey, and J Stöhr, Phys Rev Lett 91,
017203 (2003)
[16] A P Malozemoff, Phys Rev B 35, 3679 (1987)
[17] P Miltényi, M Gierlings, J Keller, B Beschoten, G Gün-therodt, U Nowak, and K D Usadel, Phys Rev Lett 84,
4224 (2000)
[18] U Nowak, K D Usadel, J Keller, P Miltényi, B Bescho-ten, and G Güntherodt, Phys Rev B 66, 014430 (2002) [19] T Hauet, S Mangin, J McCord, F Montaigne, and E E Fullerton, Phys Rev B 76, 144423 (2007)
[20] E Fulcomer and S H Charap, J Appl Phys 43, 4190 (1972)
[21] B Craig, R Lamberton, A Johnston, U Nowak, R W Chantell, and K O’Grady, J Appl Phys 103, 07C102 (2008)
[22] C Binek, Phys Rev B 70, 014421 (2004)