Journal of Magnetism and Magnetic Materials ()

Contents lists available at ScienceDirect

Journal of Magnetism and Magnetic Materials

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

Current Perspectives

Specic absorption rate determination of magnetic nanoparticles

through hyperthermia measurements in non-adiabatic conditions
M. Cosson a,n, G. Barrera a,b, F. Celegato a, L. Martino a, F. Vinai a, P. Martino c, G. Ferraro d,
P. Tiberto a
a

INRIM, strada delle Cacce 91, 10135 Torino (TO), Italy

University of Torino, Chemistry Department, via P. Giuria 7, 10125 Torino (TO), Italy
c
Politronica srl, via Livorno 60, 10144 Torino (TO), Italy
d
Center for Space Human Robotics, Istituto Italiano di Tecnologia - IIT, corso Trento 21, 10129 Torino (TO), Italy
b

art ic l e i nf o

a b s t r a c t

Article history:
Received 11 September 2015
Received in revised form
12 October 2015
Accepted 13 November 2015

An experimental setup for magnetic hyperthermia operating in non-adiabatic conditions is described. A

thermodynamic model that takes into account the heat exchanged by the sample with the surrounding
environment is developed. A suitable calibration procedure is proposed that allows the experimental
validation of the model. Specic absorption rate can then be accurately determined just from the
measurement of the sample temperature at the equilibrium steady state. The setup and the measurement procedure represent a simplication with respect to other systems requiring calorimeters or crucial
corrections for heat ow. Two families of magnetic nanoparticles, one superparamagnetic and one
characterised by larger sizes and static hysteresis, have been characterised as a function of eld intensity,
and specic absorption rate and intrinsic loss power have been obtained.
& 2015 Elsevier B.V. All rights reserved.

Keywords:
Hyperthermia
Magnetic nanoparticles
SAR

1. Introduction
Hyperthermia [1] has been the subject of wide interdisciplinary
interest in the last years, as one of the several approaches that
could turn out to be viable for selective tumour cell destruction
[2,3]. In particular, magnetic particles hyperthermia (MPH) has
been regarded as particularly promising, and has therefore attracted signicant research efforts [4,5]. In spite of the several
advantages that magnetic hyperthermia could bring [6] (direct
thermal effects on cancer cells, heat assisted chemotherapy and
radiotherapy whose outcome is signicantly improved when
coupled with MPH [7]), several open questions still hinder the
application of this methodology, such as the complex relationship
between the intrinsic properties of the magnetic material and its
specic power losses [8,9], or the difculty of obtaining reliable
estimates of specic absorption rates (SAR) of different kinds of
magnetic particles in different media and under different measurement conditions [10]. Many recent studies have concentrated
on the optimisation of the materials' properties, by exploring different shapes and anisotropy values for the particles [11], by
deeply investigating crystalline properties of the materials [12], by
analysing how interparticle interactions may affect hyperthermia
losses [13], or by suggesting entirely new families of particles,
n

Corresponding author.

such as nanorings [14], microdisks [15,16] or nanodisks [17].

However, the lack of standards in MPH has also led to many
different experimental approaches and to results that are difcult
to compare: one of the main issues is the need of performing MPH
experiments, leading to the measurement of the SAR, in adiabatic
conditions [10], or under suitable, well-calibrated corrections for
compensating the non-ideal adiabatic conditions [18]. In this paper, we present an approach that does not require the development of an adiabatic or almost adiabatic experimental setup for
magnetic hyperthermia. On the contrary, a simpler system is used
that allows heat exchange with the surrounding environment. A
modelling of the thermodynamics of the system, together with a
suitable calibration procedure, allows for the accurate determination of the specic absorption rate (and its derived quantity
intrinsic loss power) of hyperthermia agents such as magnetic
particles. A detailed knowledge of the experimental details will be
the basis for future standardisation of the hyperthermia treatment.

2. Magnetic nanoparticles
Fe-oxides particles are currently the most frequently used
material for MPH [11]. In this work, Fe3O4 nanoparticles have been
obtained by two different suppliers: Alfa Aesar (sample A, nominal
particles diameter in the range 50100 nm) and Politronica s.r.l.
(sample P, nominal particles diameter in the range 1020 nm). For

http://dx.doi.org/10.1016/j.jmmm.2015.11.044
0304-8853/& 2015 Elsevier B.V. All rights reserved.

Please cite this article as: M. Cosson, et al., Journal of Magnetism and Magnetic Materials (2015), http://dx.doi.org/10.1016/j.
jmmm.2015.11.044i

M. Cosson et al. / Journal of Magnetism and Magnetic Materials ()

Fig. 1. Room temperature hysteresis loops of sample A (blue curve) and sample P
(red curve). Inset: magnication at low elds to better show the coercivity. (For
interpretation of the references to colour in this gure caption, the reader is referred to the web version of this paper.)

both samples, solutions containing 10 mg/ml of particles in deionised water have been prepared. Room temperature hysteresis
loops of as-received dried particles have been measured with a
LakeShore 7410 VSM: the results are reported in Fig. 1.
Sample A is constituted by larger particles, with a lower surface
to volume ratio with respect to sample P; therefore, as expected,
their saturation magnetisation is higher, and almost equal to the
saturation of bulk magnetite. Sample P, instead, is characterised by
smaller particles, with a higher surface to volume ratio leading to a
reduced saturation, that display a superparamagnetic behaviour,
with zero coercive eld and a much slower approach to saturation.
These two samples are representative of two different classes of
particles often studied for MPH applications.

3. Hyperthermia setup: calibration and modelling

The hyperthermia setup used in our experiments consists in a
water-cooled copper coil made of 4 turns, connected to a rf generator through a matching network, and able to generate electromagnetic elds with intensity up to 100 mT at a frequency of
100 kHz. Inside the coil, a PTFE sample holder hosts an Eppendorf
test tube containing 1 ml of water into which known concentrations of magnetic nanoparticles are dispersed. At equilibrium and
with no rf eld applied, the sample remains at a constant temperature Ta 23.5 C thanks to the water-cooling of the coil. In
order to measure the temperature of the water solution, a type-T
(wires diameter 0.127 mm) thermocouple is available, that does
not allow the measurement of the sample temperature during the
rf irradiation process. However, it can be inserted in the test tube
immediately after the rf eld is switched off, in order to measure
the time-dependent temperature decay of the sample towards Ta.
As we will show, if a suitable calibration of the experimental setup
is available, an accurate determination of the specic absorption
rate (SAR), and consequently of intrinsic loss power (ILP), can be
obtained by just knowing the two temperatures Ta and TM, the
latter being the equilibrium temperature reached by the nanoparticles solution at steady state during irradiation.
An accurate determination of SAR would require a hyperthermia setup operating in ideally adiabatic conditions [10,18],
i.e. a calorimeter would be the optimum choice for this kind of
measurements. However, such complex setups are often not

Fig. 2. Thermodynamic scheme of the hyperthermia system. S is the source subsystem (i.e. the magnetic nanoparticles). W is the water subsystem in which the
nanoparticles are dispersed. P is the exciting power. Q terms represent the heat
exchanged by the two subsystems with each other and with the environment.

available and SAR is therefore determined by measuring the initial

slope of the sample temperature vs. time when the rf eld is
switched on [11]. This procedure leads to approximate results that
may be affected by how the temperature vs. time curve is analysed
and by the accuracy of the adiabatic hypothesis in the actual experimental setup, and requires suitable corrections for compensating the heat exchange with the surrounding environment
[12,13].
To avoid these problems, we opted to keep simple the hyperthermia experiment, while taking into account the heat exchange of the sample with the surrounding environment in the
physical model that describes its thermodynamics. In this way,
after a suitable calibration procedure, the SAR can be accurately
measured even in non-adiabatic conditions.
The hyperthermia system can be modelled as described in
Fig. 2. A heat source S (e.g. the nanoparticles inside the solution)
provides energy Q s, in to the system (e.g. through the power losses
of the magnetic particles excited by the rf eld), and exchanges
heat with the water W, Q s, out and Q w, in representing the same
amount of heat, respectively, owing out of the S subsystem and
into the W one. Finally, the water exchanges heat Q w, out with the
environment (at temperature Ta). At each instant of time, the heat
exchange equations for the S and W subsystems are the following:

Q s, in
Q s, out
Q s
=
+
dt
dt
dt

(1a)

Q w, in
Q w, out
Q w
=
+
dt
dt
dt

(1b)

where each term is taken with its proper sign (positive for heat
entering the subsystem, negative for heat released by the
subsystem).
Q s, in
The rst term
is given by the power P that is provided
dt

through the source. The two terms

Q s, out
dt

and

sign and given by the following expression:

Q s, out
Q w, in
cm
=
= s s [Ts (t ) Tw (t )]
s
dt
dt

Q w, in
dt

are of opposite

(2)

where cs is the specic heat of the source, ms is its mass, s is the

time constant of the heat exchange process between the source
and the water, Ts is the source temperature and Tw is the temperature of the water. Both quantities Ts and Tw depend on time,
but in typical hyperthermia experiments involving magnetic

Please cite this article as: M. Cosson, et al., Journal of Magnetism and Magnetic Materials (2015), http://dx.doi.org/10.1016/j.
jmmm.2015.11.044i

M. Cosson et al. / Journal of Magnetism and Magnetic Materials ()

45
experimental data

40

Q s
cm
= P s s [Ts (t ) Tw (t )]
s
dt

(4a)

Q w
cm
( + w2 ) cw mw
= s s [Ts (t ) Tw (t )] w1
[Tw (t ) Ta ]
dt
s
w1w2

(4b)

The term P appearing in Eq. (4a) is the power that is released to

the system by the source. In a typical hyperthermia experiment,
this term is the power generated by the nanoparticles that are
submitted to the rf eld (this is not the power emitted by the rf
generator driving the coil, that is usually much higher). From the
rst principle of thermodynamics, U = Q L , with U being the
internal energy of the system, and Q and L the heat and work
exchanged by the system with the outside, respectively. If we assume that in the studied temperature range the water does not
appreciably change its volume, then L = 0. The internal energy of
both subsystems S and W can then be expressed in terms of
Q
temperature through the expressions T s, w = cs, w ms, w . As a con-

35

30

25

s, w

20
0

500

1000

1500

2000

time (s)
Fig. 3. Symbols: time dependence of the temperature of 1 ml of pre-heated water
cooling down in the hyperthermia setup. Red line: tting with the sum of two
exponential functions. (For interpretation of the references to colour in this gure
caption, the reader is referred to the web version of this paper.)

particles only Tw is accessible.

Finally, the term Q w, out needs to be evaluated. To this purpose, a
dt

rst calibration procedure has been carried on, consisting in putting in the test tube 1 ml of water that has been pre-heated on a
hot plate. The test tube is in its sample holder and the cooling
water is let circulate in the rf coil, even though no rf eld is generated. The thermocouple logs the time dependence of the temperature of water as it gradually recovers to the temperature Ta of
the surrounding environment. Fig. 3 reports the results of one of
the several tests that have been performed on water that has been
pre-heated at different temperatures. An exponential decay of the
temperature over time is expected. However, the experimental
data could not be tted with a single exponential function,
whereas two were required to reproduce the data. The time constants of the two exponential functions turned out to be fairly
reproducible within the whole set of measurements; an average
value has then been calculated, giving rise to the nal results:
w1 = 137.5 s and w2 = 498.1 s. The presence of two time constants
indicates that the dissipation of heat towards the external environment is constituted by two mechanisms: the rst possibly
involves heat transfer through the test tube walls and sample
holder, and the second through the top surface of the liquid in the
test tube.
Therefore, it is necessary to express Q w, out as the sum of two

sequence, the thermodynamic behaviour of the hyperthermia

setup can be modelled with the following set of coupled differential equations:

cs ms
dT (t )
[Ts (t ) Tw (t )] = cs ms s
s
dt

(5a)

( + w2 ) cw mw
cs ms
[Ts (t ) Tw (t )] w1
[Tw (t ) Ta ]
s
w1w2
dT (t )
= cw mw w
dt

(5b)

In order to solve the system (5) for the two unknowns Ts(t) and
Tw(t), values for cs, ms and s should be provided. Instead, cw is
known (4187 J kg  1 K  1), as well as mw, since the amount of water
put in the test tube can be measured. However, the specic heat,
mass and time constant of the source are not easily determined for
a typical hyperthermia sample consisting on magnetic nanoparticles. Therefore, we performed a second calibration step where
the source term is replaced by a known resistor having a small size
and a value of 10.2 . When a known current of constant intensity
is let pass through the resistor, it produces a known power P = RI2
that is used to heat the water. This is also the power term that
enters Eq. (4a) and following. To simplify the system and ease the
calibration process, the resistor has been immersed in a solution of
water and ice, that is therefore at temperature Tice 0 C ; this
temperature will not change when the resistor heats up due to the
current that ows through it, leading to a much simplied thermodynamic description of the system. The thermocouple has been
placed in direct contact with the resistor, in order to measure its
temperature as it heats up when the current ows through it. In
this way, the system (5) reduces to a single equation with the
temperature of the source (resistor) being the only unknown:

cs ms

dTs (t )
cm
= P s s [Ts (t ) Tice ]
dt
s

(6)

dt

terms in the form:

Q w, out
c m
c m
= w w [Tw (t ) Ta ] w w [Tw (t ) Ta ]
w1
w2
dt

(w1 + w2 ) cw mw
[Tw (t ) Ta ]
w1w2

(3a)

(3b)

where cw and mw are the specic heat and mass of the water,
respectively. w1 and w2 are the time constants of the two heat
transfer mechanisms to the surrounding environment.
Therefore, Eq. (1) can be written in the following way:

where the quantities with sufx s now refer to the resistor that
acts as heat source in this particular case. Eq. (6) can be rearranged
as:

P
dTs (t )
1
T
= Ts (t ) +
+ ice
cs ms
dt
s
s
that is in the form

y = ay b that, for the initial condition

y (0) = y0, has the analytical solution y =

Eq. (6) can be solved as:

Ts (t ) = Tice +

(7)

Ps
[1 et/ s ]
cs ms

b
a

+ [y0 a ] eat . Therefore,

(8)

Please cite this article as: M. Cosson, et al., Journal of Magnetism and Magnetic Materials (2015), http://dx.doi.org/10.1016/j.
jmmm.2015.11.044i

M. Cosson et al. / Journal of Magnetism and Magnetic Materials ()

10

40

35

experimental data

< s> = 89.0 s

<csms> = 7 J / K

thermodynamic model

30
4

P = 0.41 W

25

P = 0.10 W

experimental data
thermodynamic model

0
0

100

200

300

400

500

20
0

600

for the initial condition Ts (0) = Tice . The experimental results of the
time dependence of the temperature of the resistor immersed in
water and ice for a value of P 0.41 W are reported in Fig. 4, that is
representative of the several measurements that have been performed with different heating powers up to E 1 W (higher powers
would have led to an excessive temperature increase of the water,
well beyond the range that is interesting for MPH applications).
The data reported in Fig. 4, as well as all the other available
data, have been tted with Eq. (8) leaving s and the product cs ms
as free parameters. The obtained values have been averaged and
the nal results have been obtained: s = 89.0 s and cs ms = 7.0 J/K .
We will assume that these values describe the properties of the
resistor when it exchanges heat with water, both in the conditions
exploited for calibration purposes and in an actual hyperthermia
experiment represented in Fig. 2 and modelled by Eq. (5), when
the water will effectively heat up.
It is therefore now possible to validate the thermodynamic
model by performing a measurement of the temperature of the
water in the test tube in its sample holder with the cooling water
circulating in the rf coil but without the rf eld, as the heat source
will be the power provided by the calibration resistor. In practice,
an experiment in the same conditions as a normal hyperthermia
measurement is performed, with the difference that the heat
source is not the magnetic nanoparticles excited by the rf eld, but
the calibration resistor that heats the water by means of the Joule
effect with a known power. Since the rf eld is off, in this case it is
possible to use the thermocouple to measure Tw(t) during both
heating and cooling. The results, for heating powers of 0.41 W and
0.10 W (taken as examples) applied for 7200 s and 1800 s, respectively, are reported in Fig. 5 (black symbols).
Eq. (5) is solved numerically using the values for cs ms , s, w1
and w2 obtained in the two calibration steps, and with a volume
of 0.7 ml (mw 0.7 g) of water. The solutions, calculated for the
same heating powers and heating times adopted in the experiment, are plotted in Fig. 5 as the green lines. The agreement with
the experimental results is excellent. It is important to point out
that no free parameter has been adjusted to obtain this

10

t (103 s)

t (s)
Fig. 4. Symbols: time dependence of the temperature of the calibration resistor
immersed in iced water when an electrical current ows through it giving a heating
power of 0.41 W. Blue line: tting with Eq. (8) with s and the product cs ms left as
free parameters. (For interpretation of the references to colour in this gure caption, the reader is referred to the web version of this paper.)

Fig. 5. Symbols: time dependence of the temperature of the water heated by

means of the calibration resistor with a power of 0.41 W. and 0.10 W. Green lines:
numerical solutions of Eq. (5) with the values obtained in the course of the two
calibration processes, for the same heating powers. (For interpretation of the references to colour in this gure caption, the reader is referred to the web version of
this paper.)

agreement: all the quantities involved in Eq. (5) have been either
measured during the two calibration processes (time constants of
the heat transfer process from the test tube to the environment,
heat capacity of the calibration resistor, time constant of the heat
transfer process from the calibration resistor to water, power), or
are otherwise known (volume of water, its specic heat).
As a result, we can therefore consider the system of (5) as an
adequate thermodynamic representation of our hyperthermia
setup. We can then exploit them to determine the SAR of an unknown source (magnetic nanoparticles). If the system is brought to
thermal equilibrium during the heating process, the energy entering the system through the rf eld exciting the nanoparticles
and the energy released to the environment are equal and opposite in sign; as a consequence, both dTs (t ) and dTw (t ) are equal to
zero. Therefore, from Eq. (5):

P=

dt

dt

(w1 + w2 ) cw mw
[TM Ta ]
w1w2

(9)

where TM is the maximum (equilibrium) temperature reached

during the heating process. With this approach, it is not necessary
to measure the temperature of the water during the heating process and to correct its time derivative for the residual heat exchange with the surrounding environment. Instead, it is sufcient
to know TM, that can be measured in stationary conditions with an
optical thermometer or immediately after the rf eld is switched
off with just a thermocouple, to determine P, as all the other
quantities entering Eq. (9) are now known and the model Eq. (9)
belongs to has been validated. With P obtained in this way, by
knowing the mass of particles used in the experiment, their specic absorption rate is obtained accurately for the given rf eld
intensity and frequency.
4. SAR measurements on magnetic nanoparticles
Magnetic hyperthermia has been measured on both samples A

Please cite this article as: M. Cosson, et al., Journal of Magnetism and Magnetic Materials (2015), http://dx.doi.org/10.1016/j.
jmmm.2015.11.044i

M. Cosson et al. / Journal of Magnetism and Magnetic Materials ()

Fig. 6. Time dependence of the temperature of 1 ml of water containing 10 mg of

particles of samples A (top) and P (bottom), submitted to three values of rf eld for
1800 s. The origin of the abscissa is at the instant of switching on of the rf eld.
Insets: dependence of the equilibrium temperature TM as a function of the rf eld
intensity.

and P under three different values of rf eld (30, 60, and 90 mT)
applied for 1800 s. According to our preliminary investigations,
carried on by performing several measurements at different
heating times, 1800 s are sufcient to reach the equilibrium stable
state. The thermocouple is inserted in the test tube immediately
after switching off the rf eld, therefore the rst measured temperature is assumed to be equal to TM. The samples are then left
cool down to Ta. Fig. 6 shows the three temperature vs. time curves
for both samples A and P; for convenience, the abscissa is relative
to the switching on of the rf eld; as only the cooling part of the
curve is acquired, the horizontal axis in Fig. 6 starts at 1800 s.
The equilibrium temperatures TM for the three applied rf eld
intensities are shown in the inset: as expected, the system heats at
higher temperatures when higher rf elds are applied. With the
measured values of TM for both samples A and P it is then possible
to calculate SAR using Eq. (9):

SAR =

P
mparticles

(10)

where mparticles is the mass of the particles dispersed in the solution. The results are reported in Fig. 7a. As expected, SAR increases
with increasing eld value; for rf elds below 30 mT the power

Fig. 7. Specic absorption rate (SAR, panel (a)) and intrinsic loss power (ILP, panel
(b)) as a function of rf exciting eld of samples A and P.

losses of the studied particles become negligible.

However, the intrinsic loss power (ILP) has been proposed as a
quantity more suitable for comparing hyperthermia results obtained on different families of particles, since the power losses are
normalised not only to the sample mass, but also to eld intensity
and frequency [19]:

ILP =

SAR
H 2

(11)

where is the frequency of the rf exciting eld H. In Fig. 7b values of

intrinsic loss power are reported for both samples A and P. A clear
distinction between the two samples can be observed: sample A,
which is constituted by larger particles that display static hysteresis
(see Fig. 1), has a non-monotonous dependence of ILP with the rf
eld, indicating non-linear magnetisation dynamics giving rise to
power losses that depend on eld intensity. Conversely, sample P,
that is constituted by smaller particles having a superparamagnetic
behaviour, is characterised by an almost constant ILP as a function
of H, having a lower value with respect to sample A.

5. Conclusions
A thermodynamic model and a calibration procedure have been
proposed for a magnetic hyperthermia setup that does not need to

Please cite this article as: M. Cosson, et al., Journal of Magnetism and Magnetic Materials (2015), http://dx.doi.org/10.1016/j.
jmmm.2015.11.044i

M. Cosson et al. / Journal of Magnetism and Magnetic Materials ()

operate in adiabatic conditions. Accurate determination of the

specic absorption rate is possible without measuring the initial
slope of the temperature vs. time curve; instead, by measuring the
equilibrium temperature at the stationary state, SAR can be obtained. The thermodynamic model is validated on the actual experimental setup through a proper calibration procedure, and the
method is nally applied on a set of samples consisting in water
solutions of magnetic nanoparticles allowing the experimental
determination of SAR and ILP.

[8]
[9]
[10]
[11]
[12]

[13]

References
[14]
[1]
[2]
[3]
[4]

S. Dutz, R. Hergt, Nanotechnology 25 (2014) 452001.

M. Latorre, C. Rinaldi, P. R. Health Sci. J. 28 (2009) 227.
C.S.S.R. Kumar, F. Mohammad, Adv. Drug Deliv. Rev. 63 (2011) 789.
Nguyen T.K. Thanh (Ed.), Magnetic Nanoparticles from Fabrication to Clinical
Application, CRC Press, Taylor & Francis Group, FL, USA, 2012.
[5] A. Figuerola, R. Di Corato, L. Manna, T. Pellegrino, Pharmacol. Res. 62 (2010)
126.
[6] J. van der Zee, Ann. Oncol. 13 (2002) 1173.
[7] F. Cardoso, A. Costa, L. Norton, E. Senkus, M. Aapro, F. Andr, C.H. Barrios,
J. Bergh, L. Biganzoli, K.L. Blackwell, M.J. Cardoso, T. Cufer, N. El Saghir,
L. Falloweld, D. Fenech, P. Francis, K. Gelmon, S.H. Giordano, J. Gligorov,
A. Goldhirsch, N. Harbeck, N. Houssami, C. Hudis, B. Kaufman, I. Krop,

[15]
[16]
[17]
[18]

[19]

S. Kyriakides, U.N. Lin, M. Mayer, S.D. Merjaver, E.B. Nordstrm, O. Pagani,

A. Partridge, F. Penault-Llorca, M.J. Piccart, H. Rugo, G. Sledge, C. Thomssen,
L. vant Veer, D. Vorobiof, C. Vrieling, N. West, B. Xu, E. Winer, Ann. Oncol.
(2014) 1.
R. Hergt, S. Dutz, R. Mller, M. Zeisberger, J. Phys.: Condens. Matter 18 (2006)
S2919.
A.E. Deatsch, B.A. Evans, J. Magn. Magn. Mater. 354 (2014) 163.
E. Natividad, M. Castro, A. Mediano, J. Magn. Magn. Mater. 321 (2009) 1497.
R. Di Corato, A. Espinosa, L. Lartigue, M. Tharaud, S. Chat, T. Pellegrino,
C. Mnager, F. Gazeau, C. Wilhelm, Biomaterials 35 (2014) 6400.
K. Simeonidis, C. Martinez-Boubeta, Ll. Balcells, C. Monty, G. Stavropoulos,
M. Mitrakas, A. Matsakidou, G. Vourlias, M. Angelakeris, J. Appl. Phys. 114
(2013) 103904.
D. Sakellari, K. Brintakis, A. Kostopoulou, E. Myrovali, K. Simeonidis, A. Lappas,
M. Angelakersi, Mater. Sci. Eng. C 58 (2016) 187.
X.L. Liu, Y. Yang, C.T. Ng, L.Y. Zhao, Y. Zhang, B.H. Bay, H.M. Fan, J. Ding, Adv.
Mater. 27 (2015) 1939.
E.A. Rozhkova, V. Novosad, D.-H. Kim, J. Pearson, R. Divan, T. Rajh, S.D. Bader, J.
Appl. Phys. 105 (2009) 07B306.
D.-H. Kim, E.A. Rozhkova, I.V. Ulasov, S.D. Bader, T. Rajh, M.S. Lesniak,
V. Novosad, Nat. Mater. 9 (2010) 165.
P. Tiberto, G. Barrera, F. Celegato, G. Conta, M. Coisson, F. Vinai, F. Albertini, J.
Appl. Phys. 117 (2015) 17B304.
A. Chalkidou, K. Simeonidis, M. Angelakeris, T. Samaras, C. Martinez-Boubeta,
Ll. Balcells, K. Papazisis, C. Dendrinou-Samara, O. Kalogirou, J. Magn. Magn.
Mater. 323 (2011) 775.
M. Kallumadil, M. Tada, T. Nakagawa, M. Abe, P. Southern, Q.A. Pankhurst, J.
Magn. Magn. Mater. 321 (2009) 1509.

Please cite this article as: M. Cosson, et al., Journal of Magnetism and Magnetic Materials (2015), http://dx.doi.org/10.1016/j.
jmmm.2015.11.044i