IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 19 (2008) 185702 (8pp) doi:10.1088/0957-4484/19/18/185702

Surface spin-glass freezing in interacting

core–shell NiO nanoparticles
E Winkler1 , R D Zysler1,5 , M Vasquez Mansilla2 , D Fiorani2 ,
D Rinaldi3 , M Vasilakaki4 and K N Trohidou4
1
Centro Atómico Bariloche, CNEA-CONICET, 8400 S C de Bariloche, RN, Argentina
2
Istituto di Struttura della Materia, CNR, Area della Ricerca di Roma, CP 10,
I-00016 Monterotondo Stazione, Rome, Italy
3
Dipartimento di Fisica e Ingegneria dei Materiali e del Territorio, Università Politecnica delle
Marche via Brecce Bianche, I-60131 Ancona, Italy
4
Institute of Materials Science, NCSR ‘Democritos’, GR-15310 Athens, Greece

E-mail: [email protected]

Received 6 December 2007, in final form 19 February 2008

Published 2 April 2008
Online at stacks.iop.org/Nano/19/185702

Abstract
Magnetization and AC susceptibility measurements have been performed on ∼3 nm NiO
nanoparticles in powder form. The results indicate that the structure of the particles can be
considered as consisting of an antiferromagnetically ordered core, with an uncompensated
magnetic moment, and a magnetically disordered surface shell. The core magnetic moments
block progressively with decreasing temperature, according to the distribution of their
anisotropy energy barriers, as shown by a broad maximum of the low field zero-field-cooled
magnetization ( MZFC ) and in the in-phase component χ ’ of the AC susceptibility, centred at
∼70 K. On the other hand, surface spins thermally fluctuate and freeze in a disordered
spin-glass-like state at much lower temperature, as shown by a peak in MZFC (at 17 K, for
H = 50 Oe) and in χ  . The temperature of the high temperature χ  peak changes with
frequency according to the Arrhenius law; instead, for the low temperature maximum a power
law dependence of the relaxation time was found, τ = τ0 (Tg /(T (ν) − Tg ))α , where α = 8, like
in spin glasses, τ0 = 10−12 s and Tg = 15.9 K. The low temperature surface spin freezing is
accompanied by a strong enhancement of magnetic anisotropy, as shown by the rapid increase
of coercivity and high field susceptibility. Monte Carlo simulations for core/shell
antiferromagnetic particles, with an antiferromagnetic core and a disordered shell, reproduce
the qualitative behaviour of the temperature dependence of the coercivity. Interparticle
interactions lead to a shift to a high temperature of the distribution of the core moment blocking
temperature and to a reduction of magnetization dynamics.
(Some figures in this article are in colour only in the electronic version)

1. Introduction exchange bias properties that stabilize the magnetic moment of

FM nanoparticles in magnetic recording [7]. The remarkable
The magnetic properties of nanoparticle systems are a magnetic properties of nanoparticles, which are very different
subject of continuously growing interest, driven by both from those of their parent massive materials, are due to a
fundamental research [1] and technological applications, complex interplay between finite size and surface effects. The
e.g. in magnetorecording [2] and biomedicine [3]. In particular first are related to the reduced number of exchange coupled
the study of antiferromagnetic nanoparticles is a topical spins within the particle core. The latter are related to the
challenge due to their novel application as, for example, in local symmetry breaking of the lattice, resulting in site-specific
spin valves [4, 5], magnetic random access memory [6] and the surface anisotropy and to fluctuations in the number of atomic
neighbours and interatomic distances, due to broken bonds and
5 Author to whom any correspondence should be addressed.
defects, resulting in magnetic frustration and spin disorder.

0957-4484/08/185702+08$30.00 1 © 2008 IOP Publishing Ltd Printed in the UK

Nanotechnology 19 (2008) 185702 E Winkler et al

Thus, the magnetization is not uniform throughout the particle

and the magnetic properties are strongly affected by the nature
of the interface exchange coupling between the magnetically
ordered core (with reduced ordering temperature because of
the size confinement) and the magnetically and structurally
disordered surface shell. As a matter of fact, particle core
and surface effects are not easily distinguishable, as they are
intimately mixed in determining the actual anisotropy and the
overall magnetic behaviour of nanoparticles.
Interparticle interactions, when present, represent a further
important effect to be taken into account to understand
the magnetic properties of nanoparticle assemblies [8–10].
Interparticle interactions provide a further contribution to the
particle anisotropy energy and they determine the relative
orientation of particle moments, competing with the single- Figure 1. X-ray (Cu Kα ) diffraction pattern of the NiO nanoparticle
particle anisotropy. For strong enough interactions, an powder.
assembly of interacting nanoparticles represents a randomly
oriented correlated moments system, giving rise to a collective
type spin dynamics at low temperature [11, 12], like in In the present paper, we report an investigation, by means
spin-glasses, with a broad distribution of interrelated energy of magnetization and AC susceptibility measurements, of
barriers, separating multiminima energy states. the magnetic properties of 3 nm NiO particles in powder
Surface effects are expected to play a growing role with form. The aim of this work is to investigate the effect of
decreasing particle size, due to the increase of the surface interparticle interactions on the static and dynamic properties
to volume ratio of particles. These effects are enhanced in of such particles, with a special interest in their influence on
antiferromagnetic nanoparticles [13, 14], due to the much the surface anisotropy, through comparison with those of the
lower core moment with respect to ferromagnetic particles polymer dispersed particles, where interparticle interactions
of the same size. According to the Néel model [15], are weak. The results show that, in the powder system,
antiferromagnetic nanoparticles have a net magnetic moment, interparticle interactions lead to a shift to higher temperature
due to an uncompensated number of spins in the two of the distribution of the blocking temperatures of the core
sublattices and depending on the crystal structure and particle particle moments, modifying the magnetization relaxation.
morphology. Indeed quite large magnetic moments have The low temperature surface spin frozen state is mainly due
been observed in nanoparticles of antiferromagnetic materials to the intraparticle interactions. However, the interparticle
such as NiO. Bulk NiO has a rhombohedral structure and interaction increases the degree of frustration of surface spins
is antiferromagnetic, with Néel temperature TN = 523 K, affecting the low temperature dynamical behaviour.
above which it has cubic structure [16]. NiO nanoparticles
were found to exhibit large coercive fields and high field
susceptibility at low temperature, attributed to surface 2. Results
anisotropy [17–19]. Such behaviour was explained through
The sample preparation, based on chemical precipitation, by
numerical modelling by a multisublattice spin configuration
mixing Ni(NO3 )2 and NaOH aqueous solutions, is reported
and reduced coordination of surface spins. Exchange
in [21]. The x-ray diffraction pattern (figure 1) of the powder
bias effects, which manifest themselves through shifted
hysteresis loops after field cooling, were also observed in sample corresponds to that of the NiO Fm 3m phase without
5–8 nm NiO particles, due to interface exchange coupling any trace of other phases. The size distribution is lognormal
between the antiferromagnetically ordered core and the frozen with a mean diameter d = 3 nm and a standard deviation σ =
magnetically disordered shell, below the surface spin freezing 0.2, as shown by transmission electron microscopy (TEM)
temperature [20]. Such effects were found to be strongly images (figure 2). The value determined from x-ray diffraction
particle-size-dependent, disappearing below a given particle by using the Scherrer equation [22], φ = (3 ± 1) nm, is in
size which depends on its microstructure [18]. agreement with TEM observations.
Recently, we have investigated the magnetic properties Figure 3 shows the temperature dependence of the
of NiO nanoparticles of d = 3 nm diameter, dispersed magnetization, measured, by a SQUID magnetometer,
in a polymer (polyvinyl-pyrrolidone) [21]. In this case, applying a field of 50 Oe for the powder and the dispersed
interparticle interactions are weak, being the average particles. The two ZFC curves show the same basic features:
interparticle distance ∼28 nm. The high temperature magnetic (i) a very broad maximum centred at T1 , ∼70 K and ∼60 K
behaviour is determined by the progressive blocking process for the powder and dispersed particle sample, respectively;
of core particle moments, whereas surface spins thermally (ii) a much sharper maximum at low temperature at T2 ,
fluctuate freely. At low temperature (below ∼15 K), the slightly different for the two samples, ∼17 K and ∼15 K
magnetic behaviour is dominated by the random freezing of for the powder and dispersed particles sample, respectively.
surface spins, accompanied by a large increase of coercivity Differences are observed in the low temperature behaviour
and remanence. of the FC magnetization: for the powder sample, it tends

2
Nanotechnology 19 (2008) 185702 E Winkler et al

Figure 2. TEM image of the NiO powder sample. Inset: detail of an

hexagonal NiO particle shape (right); size distribution histogram
Figure 4. Temperature dependence of the magnetization for the
obtained from different TEM micrographs (left). The solid line
dispersed and powder NiO samples measured at H = 50 kOe.
corresponds to a lognormal fitting of the distribution with
d = 3 nm and σ = 0.2.

Figure 3. Zero-field-cooled ZFC (full symbols) and field-cooled

(open symbols) magnetization measurements at H = 50 Oe for the
powder sample (circles). The results for the dispersed particles,
from [21], is also shown for comparison (triangles). The inset shows Figure 5. Temperature dependence of the real, χ  (solid symbols),
a detail of the high temperature region. and imaginary component, χ  (open symbols), of the AC
susceptibility for the powder sample, applying an AC field of
H = 10 Oe, measured at different frequencies (circle 10 Hz, square
270 Hz, diamond 1250 Hz, up triangle 3430, stars 7290 Hz and down
to saturate below T2 , whereas for the dispersed sample triangle 10 kHz). The inset shows the high temperature region.
it continues to increase with decreasing temperature, with
a downward curvature below 15 K. Differences are also
observed in the temperature dependence of the ZFC high AC susceptibility is measured at 80 Hz. The peaks shift toward
field magnetization (figure 4): it shows a Curie–Weiss-like higher temperature with increasing frequency. The inset of
behaviour for the dispersed particle sample, whereas for the figure 6 shows the plot of ln(1/ν) as a function of 1/TP ,
powder the magnetization shows a downward curvature below where TP is the χ  peak position obtained from the fit of the
∼15 K. high temperature maximum with a Curie plus a Lorentzian
The AC susceptibility measurements for the powder and function. From this plot it can be observed that the relaxation
dispersed sample were performed at different frequencies ν , time follows an Arrhenius law with an activation energy of
between 10 Hz and 10 kHz, applying an AC field of 10 Oe. E disp /kB = 1500 K for the dispersed nanoparticles and
Figure 5 shows the temperature behaviour of the real (χ  ) and E powder /kB = 5400 K for the powder sample. From the
imaginary component (χ  ) of the AC susceptibility for the fitting we obtain the characteristic relaxation time of the system
powder sample. From these measurements it is remarkable τ0 = 10−12 s and τ0 = 10−30 s for the dispersed and powder
the presence of two peaks located at 15 and 64 K for the sample, respectively. As can be observed, while the τ0 value
dispersed sample; and 17 and 85 K for the powder when the is reasonable for the AFM dispersed sample the value reached

3
Nanotechnology 19 (2008) 185702 E Winkler et al

Figure 6. Frequency dependence of the low temperature maxima of

χ  for the powder (circles) and dispersed particles (triangles)
samples. Note that the relaxation time follows a power law:
τ = τ0 (T ∗ /(T (ν) − T ∗ ))α . The inset show the Arrhenius plot for the
frequency dependence of the high temperature maximum. Figure 7. Hysteresis loops of the NiO powder sample measured at
different temperatures. The right insets show the completed cycles.
The upper inset shows, for comparison, a detail of the hysteresis
cycle measured at 5 K of the dispersed particle sample.
for the powder system is not physically acceptable. The large
value obtained for the activation energy and the relaxation
time for the powder sample implies that the interparticle
interactions play an important role in this system. The
frequency dependence of the χ  (T ) low temperature maximum
does not follow an Arrhenius law; instead a power law for the
relaxation time (τ = 1/ν ) is found: τ = τ0 (T ∗ /(T (ν) − T ∗ ))α
(figure 6), with τ0 = 10−12 s, α = 8 and T ∗ = 15.9 K. The
same dependence is followed by the dispersed particle sample
with τ0 = 10−10 s and T ∗ = 14.0. This frequency dependence
of the high peak temperature relaxation time is in agreement
with a spin-glass behaviour, as we are going to discuss in the
next section.
Figure 7 shows hysteresis cycles at different temperatures
for the powder sample. The observed behaviour suggests that
the magnetization consists of two components: a component
tending to saturate at low fields and a non-saturating
Figure 8. Saturation magnetization of the irreversible component
component, responsible for the almost linear variation at high associated with the non-compensated magnetic moment of the
fields. Figure 8 shows the temperature dependence of the nanoparticles. The inset shows the initial magnetization curve of the
magnetization value of the saturating component obtained from NiO powder sample for different temperatures.
the extrapolation to H = 0 of the high field linear component.
Both the total magnetization and the saturating magnetization the virgin curve of the hysteresis loop shows an ‘S’ shape,
component are smaller for the powder, which shows a tendency whereas with increasing temperature it progressively evolves
to saturation below ∼15 K. At very low temperature, T = 5 K, towards a continuous downward curvature (see inset figure 8).

4
Nanotechnology 19 (2008) 185702 E Winkler et al

Figure 11. Viscosity, S , as a function of the temperature.

Figure 9. Temperature dependence of the coercive field for the
powder sample. The solid line is a guide for the eyes.

increase of HC at low temperature is much stronger, reaching at

5 K a value almost twice that of the powder sample. No shift of
the hysteresis cycle is observed after cooling the samples under
a field of 50 kOe. The lack of exchange bias effect is coherent
with the very small particle size, in agreement with the results
reported in [18] where large coercive and exchange fields were
obtained for particles with diameters between 20 and 31 nm
and both fields tend to zero for smaller sizes.
Figure 10 shows the normalized relaxation of the remanent
magnetization for the dispersed (figure 10(a)) particles and
powder (figure 10(b)) sample. The measurements were
performed as follows: the samples were cooled under an
applied field of H = 200 Oe from 200 K down to the
measuring temperature, then the field was turned off and
the remanent magnetization (thermoremanent magnetization)
decay was recorded. A logarithmic decay of the magnetization
was observed, as expected for an assembly of non-interacting
(or weakly interacting) nanoparticles with a distribution of
relaxation times:

M(t, T )/M0 = 1 − S(T ) ln(t/τ0 )

where M0 is the initial magnetization at T in the presence of

a magnetic field H < Hc , S(T ) is the so-called magnetic
viscosity and τ0 is the characteristic nanoparticle relaxation
time. Note that the above equation is valid for measurements
times t  τ0 [23]. The remanent magnetization values above
∼15 K superimpose in a T ln(t/τ0 ) dependence (for τ0 =
10−11 s), according to a thermally activated mechanism (inset
figure 10). For both samples, the temperature dependence of
Figure 10. Normalized magnetization relaxation as a function of
time for the dispersed particles (a) and powder sample (b). The insets the viscosity (figure 11) shows two relaxation regimes, above
show the magnetization as a function of T ln(t/τ0 ), with and below ∼15 K, where an abrupt change is observed. In
τ0 = 1 × 10−11 s. the whole temperature range, the viscosity is smaller for the
powder sample.

The temperature dependence of the coercive field, HC , shown 3. Discussion

in figure 9, exhibits with decreasing temperature a maximum at
T = 40 K, then a minimum at ∼15 K and finally a very rapid The ensemble of the above data suggests the following
increase. Similar behaviour was observed in the dispersed picture of magnetic behaviour. The high temperature
sample [21], for which the maximum is located at 30 K and the regime and the broad maximum at T1 are associated with

5
Nanotechnology 19 (2008) 185702 E Winkler et al

the progressive blocking, with decreasing temperature, of In canonical spin glasses α = zν , where z is a dynamic expo-
nanoparticle moments, i.e. uncompensated moments of the nent and ν is the critical exponent characterizing the divergence
particle core, with decreasing temperature, giving rise to a of the correlation length ξ (τ = ξ z = Tg /(T (ν) − Tg )) [28].
hysteretic behaviour. In this temperature range, the spins of Moreover, the higher T2 value for the powder sample
the disordered surface shell thermally fluctuate and then they and the different behaviour of the FC magnetization below
do not contribute to the observed hysteretic behaviour. The T2 indicates that the surface spin freezing is affected by
width of the maximum reflects the distribution of blocking interparticle interactions. The saturation tendency of the FC
temperatures, TB , proportional to the anisotropy energy barrier, magnetization of the powder sample, similar to that found
E , separating the easy magnetization directions (E = in canonical spin glasses below the freezing temperature,
K a V for uniaxial anisotropy, where K a is the anisotropy suggests that interparticle interactions introduce a further
constant and V the particle volume). According to the frustration in the particle system. The temperature dependence
Néel–Brown model [24], for non-interacting particles with of the magnetization measured at 50 kOe (figure 4) for the
uniaxial anisotropy, the blocking temperature is TB = two samples is coherent with the above description. For
K a V /[kB ln(τ0 /τ )], where kB is the Boltzmann constant, τ0 is the dispersed particle sample, the temperature dependence
the characteristic relaxation time (10−11 –10−12 s) and τ is the of the magnetization resembles a Curie–Weiss behaviour,
relaxation time, equal to the measuring time at T = TB . The coherent with a Zeeman interaction between individual particle
anisotropy constant calculated from the blocked temperature moments and a strong enough magnetic field. On the
in the dispersed nanoparticle system K a ∼7 × 106 erg cm−3 is other hand, for the powder sample, a progressive downward
in agreement with that obtained from the AC magnetization deviation from this behaviour is observed below ∼30 K,
measurements. The higher value of T1 observed for the powder indicating a more frozen magnetic state, as the system follows
sample (it moves from ∼60 to ∼70 K) is due to interparticle the magnetic field less and less, although it is as high as
interactions, which produce an increase of the effective particle 50 kOe. This is also consistent with a smaller viscosity S(T ),
anisotropy energy, leading to an increase of TB , as theoretically i.e. slower dynamics, observed for the powder sample in the
predicted [25, 26]. From the AC susceptibility measurements all temperature range. Similar features were observed in other
we have obtained large values for the activation energy and interacting nanoparticle systems, like ferrimagnetic Mn3 O4
the relaxation time in the powder sample which show the and Fe3 O4 nanoparticles [29, 30].
crucial role played by the interparticle interactions. In this The features of the hysteresis cycles at different
temperature range mainly dipole–dipole interactions between temperature support the above picture. The easily saturating
core moments should be effective, as surface spins thermally component is associated with the uncompensated moment of
fluctuate. the particle core, whereas the non-saturating component is
The low temperature maximum (at 15–17 K) is weakly associated with the spins of the disordered surface. The
dependent on the particle aggregation state, the difference change of shape of the virgin curve with temperature (from
in T2 being quite small, and this indicates that it is mainly an ‘S’ shape to a continuous downward curvature, with
associated with intraparticle effects. The magnetization increasing temperature) is consistent with a highly anisotropic
behaviour below ∼30 K, where most of the core particle disordered frozen state of surface spins at low temperature
moments are blocked, is mainly determined by the surface (figures 5, 7). The ‘S’ shape of the virgin curve, with a
spins, which thermally fluctuate under an effective field, due very small value of the low field susceptibility (see the curve
to the applied field and that produced by the blocked core at 5 K), similar to what is observed in spin-glasses, reveals
moments. With decreasing temperature, thermal fluctuations a strong increase of anisotropy below T2 . The virgin curve
slow down, correlation between surface spins grows, leading exhibits the same behaviour in the dispersed sample and this
to spin cluster formation, and finally the surface spins freeze confirms that the low temperature frozen state is basically
in random directions. The low temperature maximum at T2 an intraparticle phenomenon, originating from the surface
should correspond to a spin-glass-like freezing temperature disorder and frustration. Above T2 , surface spins are unfrozen,
of the surface spins randomly coupled between them and thermally fluctuate and tend to follow the core particle moment
the adjacent core spins [21]. Such a maximum is weakly under the orienting action of the magnetic field, giving rise to
field-dependent, unlike the large temperature maximum at T1 , the continuous downward curvature in the field dependence of
confirming that it is associated with the onset of a highly the magnetization.
anisotropic magnetic state. The change of the magnetic state It is worth pointing out that the magnetization of the
below T2 is confirmed by the abrupt change in the temperature powder sample is much lower. This indicates a resulting
dependence of the viscosity (figure 11). demagnetizing character of interparticle interactions, since the
The existence of a spin-glass-like surface spin freezing at formation of correlated clusters of particles generates random
a well-defined temperature, identified by the sharp peak at T2 , interaction fields, as theoretically predicted when particles
is supported by divergence of the relaxation time according to coalesce, as in a powder [31]. The temperature dependence of
a power law, τ = τ0 (T ∗ /(T (ν) − T ∗ ))α , as in spin glasses, HC is coherent with what is reported above. With decreasing
where the α value (8.0 for the powder sample; 7 for the dis- temperature, the increase to 40 K is in agreement with what
persed particle sample) is within those reported in the literature is expected for the blocked state of core particle moments.
(7 < α < 9) for spin-glasses [27]. Then, T ∗ actually has to be The subsequent decrease is determined by the coexistence of
considered the glass temperature Tg for the surface spin state. the increasing paramagnetic type contribution from surface

6
Nanotechnology 19 (2008) 185702 E Winkler et al

constant K c = 0.1. The surface thickness is taken equal to one

lattice spacing, and if the spin i lies in the outer layer of the
particle then K i = K S = 0.2. In our calculations, random
axis anisotropy at the surface to model the disordered state of
the surface is considered. The third term corresponds to the
Zeeman energy.
We have used the Monte Carlo method for a Heisenberg
system [13, 32] to study the magnetic behaviour of the
nanoparticle. Typically, 104 Monte Carlo steps per spin
were found to be adequate in the simulations. The
results were averaged over 40 different samples (namely,
independent random number sequences that give different spin
configurations) cooled down under the same conditions. The
coercive field is defined as the magnetic field that reverses
the magnetization of the particle in the simulation time, so
that the z component of the magnetization vanishes. Note
Figure 12. Calculated coercive field, HC (T ), from the Monte Carlo that H and HC are given in units of J/gμB , T in units of
simulations of antiferromagnetic nanoparticles by using the energy
defined in equation (1) (see text). The solid line is a guide for the
J/kB and the anisotropy coupling constants K in units of J .
eyes. In the simulations, the field cooling procedure starts with an
magnetically disordered nanoparticle at temperature T = 3
(which is above the critical temperature of the nanoparticle).
spins and clusters, determining an actual narrowing of the Consequently the nanoparticle is cooled in the presence of a
hysteresis cycle. Finally, the large and rapid increase below magnetic field along the z axis. The results of the simulations
T2 is determined by the increase of anisotropy associated with for the coercive field as a function of temperature are plotted in
the surface spin freezing. The much smaller HC value at figure 12. It is noticeable that in the curve there is a minimum
low temperature for the powder sample with respect to non- at T = 0.2 in our units, a maximum at T = 0.6, and then the
interacting particles is coherent with the presence of much curve goes down. The observed behaviour of the HC (T ) curve
stronger dipolar interparticle interactions, where the weaker is qualitatively similar to that experimentally observed for the
anisotropy of the cluster of particles with respect to isolated NiO nanoparticles.
particles is the outcome of the coexistence of many local
easy axis directions originating from the correlated individual 4. Conclusions
particles [31].
In order to verify the core–shell picture of the The effect of interparticle interactions on ∼3 nm NiO nanopar-
nanoparticles Monte Carlo simulations [32] for the coercive ticles, considered as consisting of an uncompensated antiferro-
behaviour of the non-interacting nanoparticles have been magnetically ordered core and a magnetically disordered sur-
performed. For the MC simulations, simple cubic (sc) systems face shell, has been investigated by comparing the magnetic
with classical Heisenberg exchange interactions between the properties of a powder particle sample (interacting particles)
spins were considered which at each crystal site experience a and of a polymer dispersed particle sample (weakly interacting
uniaxial anisotropy. A spherical particle of radius R is defined particles). The results of magnetization and AC susceptibil-
by fixing the origin at a certain spin and including all spins ity measurements indicate that interparticle interactions, lead
within a distance of R lattice spacings. R is taken equal to to an increase of the effective anisotropy energy of the mag-
5.1 lattice spacings, so the volume of the nanoparticles, is netic core moments, shifting the distribution of their block-
similar to the experimental one. In the presence of an external ing temperatures (T1 ) to higher temperatures. A much lower
magnetic field, the total energy of the system is [13] magnetization was found for the powder sample, implying a
demagnetizing character of interparticle interactions, coherent

N  
N 
N with dipolar fields generated within the assembly of particles.
E =− Ji j Si · S j − K i ( Si · êi )2 − H · Si . (1) Such interactions are responsible for the slowing down of the
i j =i i i particle moments dynamics in the powder sample. For both
samples, the surface spins fluctuate freely at high temperature
The exchange interaction is supposed as the nearest- and freeze in a disordered spin-glass-like state at low temper-
neighbour coupling with Ji j equal to −J ( J > 0). Here si is ature (at T2 T1 ), as shown by a peak in the low field ZFC
the atomic spin at site i and êi is the unit vector in the direction magnetization and AC susceptibility, and by the associated di-
of the easy axis at site i . N is the total number of spins in the vergence of the relaxation time according to a power law. The
particle. The first term gives the exchange interaction between surface spin freezing is accompanied by a strong enhancement
the spins in the antiferromagnetic nanoparticles (the exchange of magnetic anisotropy, as shown by the rapid increase of coer-
coupling constant J is taken equal to one). The second term civity and high field susceptibility. Interparticle interactions in-
gives the anisotropy energy of the particle. The core anisotropy crease the degree of frustration in the nanoparticle system due
is considered uniaxial along the z axis with anisotropy coupling to random correlations between surface cluster freezing. The

7
Nanotechnology 19 (2008) 185702 E Winkler et al

qualitative behaviour of the temperature dependence of the co- [11] Mamiya H, Nakatani I and Furubayashi T 1998 Phys. Rev. Lett.
ercivity is quite well reproduced by temperature Monte Carlo 80 177
[12] Jonsson T, Norblad P and Svedlindh P 1998 Phys. Rev. 57 497
simulations for core/shell antiferromagnetic particles, with an
[13] Trohidou K N, Zianni X and Blackman J A 1998 J. Appl. Phys.
antiferromagnetic core and a disordered shell. 84 2795
[14] Zianni X and Trohidou K N 1999 J. Appl. Phys. 55 1050
[15] Néel L 1961 Comptes Rendus 252 4075
Acknowledgments Néel L 1961 Comptes Rendus 253 9
Néel L 1961 Comptes Rendus 253 1286
This work has been accomplished with partial support from [16] Smart J S and Greenwald S 1951 Phys. Rev. 82 1134
Argentina–Italy SECYT-MAEIT/PA03-EIII/085 cooperation [17] Richardson J T, Yiagas D I, Turk B, Forster K and Twigg M V
project, PICTs 3-13294 and 4-25317, PIP 5250/03, and 1991 J. Appl. Phys. 70 6977
[18] Kodama R H, Makhlouf S A and Berkowitz A E 1997 Phys.
Fundación Antorchas grants (Argentina). We thank Rev. Lett. 79 1393
H E Troiani for the TEM images. [19] Tiwari S D and Rajeev K P 2005 Phys. Rev. B 72 104433
[20] Kodama R H and Berkowitz A E 1999 Phys. Rev. B 59 6321
[21] Winkler E, Zysler R D, Vasquez Mansilla M and
References Fiorani D 2005 Phys. Rev. B 72 132409
[22] Jenkins R and Snyder R L (ed) 1996 Introduction to X-ray
[1] Dormann J L, Fiorani D and Tronc E 1997 Adv. Chem. Phys. Powder Diffractometry (New York: Wiley)
98 283 [23] Vincent E, Hammann J, Prené P and Tronc E 1994
[2] Hadjipanayis G C (ed) 2001 Magnetic Storage Systems Beyond J. Physique I 4 273
2000 (Dordrecht: Kluwer–Academic) [24] Brown W F Jr 1963 Phys. Rev. 130 1677
[3] Pankhurst Q A, Connolly J, Jones S K and Dobson J 2003 [25] Shtrikman S and Wolhfarth E P 1981 Phys. Lett. A 85 467
J. Phys. D: Appl. Phys. 36 R167–81 [26] Dormann J L, Bessais L and Fiorani D 1988 J. Phys. C: Solid
[4] Nogués J, Sort J, Langlais V, Skumryev V, Suriñach S, State Phys. 21 2015
Muñoz J S and Baró M D 2005 Phys. Rep. 422 65 [27] Bontemps N, Rajchenbach J, Chamberlin R V and
[5] Berkowitz A E and Takano K 1999 J. Magn. Magn. Mater. Orbach R 1986 J. Magn. Magn. Mater. 54–57 1
200 552 [28] Binder K and Young A P 1984 Phys. Rev. B 29 2864
[6] Comstock R L 2002 J. Mater. Sci. Mater. Electron. 13 509 [29] Winkler E, Zysler R D and Fiorani D 2004 Phys. Rev. B
[7] Skumryev V, Stoyanov S, Zhang Y, Hadjipanayis G, 70 174406
Givord D and Nogués J 2003 Nature 423 850 [30] Lima E Jr, Vargas J M and Zysler R D 2008 J. Nanosci.
[8] Dormann J L, Fiorani D and Tronc E 1999 J. Magn. Magn. Nanotechnol. at press
Mater. 202 251 [31] Kechrakos D and Trohidou K N 2003 J. Magn. Magn. Mater.
[9] Morup S and Tronc E 1994 Phys. Rev. Lett. 72 3278 262 107
[10] Blanco-Mantecon M and O’Grady K 2006 J. Magn. Magn. [32] Binder K 1984 Applications of Monte Carlo Methods in
Mater. 296 124 Statistical Physics (Berlin: Springer)