VOLUME 88, NUMBER 23 PHYSICAL REVIEW LETTERS 10 JUNE 2002

Scaling Approach to the Magnetic Phase Diagram of Nanosized Systems

J. d’Albuquerque e Castro*
Instituto de Física, Universidade Federal do Rio de Janeiro, Caixa Postal 68.528, RJ, 21945-970, Brazil

D. Altbir and J. C. Retamal

Departamento de Física, Universidad de Santiago de Chile, Casilla 307, Santiago2, Chile

P. Vargas
Departamento de Física, Universidad Técnica Federico Santa María, Casilla 110-V, Valparaíso, Chile
(Received 3 November 2001; published 23 May 2002)
Based on the discovery of a simple scaling relation, a new technique for the investigation of the
phase diagram of nanosized magnetic systems is proposed. By scaling the exchange interaction between
magnetic moments, the magnetic phase diagram of currently lithographically produced particles can be
obtained from those corresponding to much smaller systems. Such a technique reduces the computation
time by several orders of magnitude, and provides a new approach to the investigation of the relative
stability of distinct internal magnetic configurations of nanosized systems. The technique is illustrated
by the determination of the phase diagram of cylindrically shaped Co particles.

DOI: 10.1103/PhysRevLett.88.237202 PACS numbers: 75.75. +a, 75.60. –d

Magnetic particles with dimensions in the nanometer
range have become the subject of increasing interest in
the past few years, mostly because of their potential
applications in nonvolatile magnetic memory devices or
high-resolution magnetic field sensors [1]. Arrays of
discrete, lithographically patterned magnetic elements,
such as magnetic pillars, pyramids, and dots, have
been proposed as a new generation of ultrahigh density
patterned magnetic storage media [2]. Within those
elements or particles, different magnetic arrangements
may be observed: ferromagnetic states, in which the
magnetization reaches high values close to saturation,
and vortex configurations of the spins, in which the
magnetization of the system is very close to zero. In the
ferromagnetic state, the magnetic structure is close to that
of a single domain, which is of interest for the production
of magnetic memory chips, spin transistors, and hard-disk
read heads [3]. On the other hand, flux-closure states may
be relevant to technological applications, since in such
a case the magnetostatic interaction between individual
elements is significantly reduced [4].
Arrays of cylindrically shaped clusters have been pro-
duced and their magnetic properties have been studied
by several groups [5 –7]. For such geometry, the total
magnetization of individual clusters in the ferromagnetic
state may be either parallel to the basis of the cylinder
(ferro-in-plane) or perpendicular to it (ferro-out-of-plane).
From the hysteresis curves for these systems, attempts have
been made to determine their experimental magnetization
phase diagram in the D-H plane, where D is the diameter of
the basis and H is the height of the cylinder [8,9]. Since the
hysteresis curves for the ferromagnetic states are distinct
from those for the vortex state [8,9], interest has been fo-
cused on identifying the ranges of D and H corresponding
to three alternative configurations, namely, ferro-in-plane
(I), ferro-out-of-plane (II), and vortex state (III). The diffi-
culty in this approach, however, resides in the fact that the
transition between those configurations is not sharp, with
the cluster going through intermediate states characterized
by rather complex arrangements of the magnetic moments.
Such states are not clearly identifiable from the hysteresis
curves. Therefore, it would be highly desirable to obtain,
from theoretical considerations, a phase diagram based on
the relative stability of the three phases, (I), (II), and (III),
which could be used as a guide for the production of pat-
terned films with specific magnetic properties.
The purpose of the present work is precisely to obtain
such a phase diagram, using a simple but realistic model
for those systems. Our approach consists in evaluating,
for each point in the D-H plane, the total energy of the
three configurations in order to find the lowest-lying one.
However, as we make clear below, for cylinders with di-
mensions in the range of those experimentally investigated,
this is far from being a straightforward calculation. The
computational effort soon becomes prohibitively large, so
in order to accomplish this task we shall resort to scaling
techniques.
The magnetic structure within each cylinder is basically
determined by the internal interactions, namely, the
direct exchange between neighboring atoms, the clas-
sical dipolar coupling (which is responsible for the shape
anisotropy), and possibly crystalline anisotropy terms. It
has been found that the interaction energy between cylin-
ders is much weaker than the internal one [5], so it can be
safely disregarded in a first approach if the inter-cylinder
distance is bigger than D [10]. It follows that the internal
energy Etot of a single cylinder with N magnetic moments
in a given configuration 兵m !
i 其 can be written as

237202-1 0031-9007兾02兾 88(23)兾237202(4)$20.00 © 2002 The American Physical Society 237202-1

VOLUME 88, NUMBER 23 PHYSICAL REVIEW LETTERS
1 X
Etot 共兵m
!
i 其, 兵Jij 其, K兲 苷 关Eij 2 Jij m
!
i ? m j 兴 1 UK ,
!
2 ifij
! !
where Eij is the dipolar interaction between m i and m j ,
Jij is the exchange coupling, and UK is a crystalline
anisotropy term. The dipolar energy is given by
!
m ! !
bij 兲 共m
!
bij 兲
i ? m j 2 3共m i ? n j ? n
Eij 苷
rij3
! !
where rij is the distance between m bij is
i and m j , and n
the unit vector along the direction that connects the two
magnetic moments. Regarding the exchange interaction,
we assume that Jij 苷 J for nearest neighbor moments,
and zero otherwise. For the sake of simplicity, we also
assume that the magnetic moments occupy the sites of an
underlying lattice with cubic symmetry, which we take to
be face centered cubic. Thus, the crystalline anisotropy
term can be written in the form
X
N
UK 苷 K 关ai2 bi2 1 bi2 gi2 1 gi2 ai2 兴 ,
i苷1
where 共ai , bi , gi 兲 are the direction cosines of m
!
i referred
to the cube axis [11]. Since there is a bulk of experimental
ij 苷
!
results on granular Co systems, we have considered jm
m 苷 1.7mB , J 苷 2.9 mRy [12], K 苷 2 3 10 erg兾cm3 , 6
and the lattice parameter a0 苷 3.6 Å.
At this point, it is important to comment on the restric-
tions imposed upon our approach by the number of mag-
netic moments involved in the calculations. In fact, when
one deals with cylinders with dimensions comparable to
those experimentally investigated, N may be larger than
109 , which would require a computational effort way be-
yond present-day standard computational facilities (recall
that the computation time increases with N 2 ). On the other
hand, for cylinders small enough to be handled computa-
tionally within reasonable amounts of time, the exchange
interaction completely dominates the magnetic behavior,
and the lowest-lying state always turns out to be ferromag-
netic. The actual configuration can be either in-plane or
out-of-plane, depending on the aspect ratio R 苷 H兾D.
In order to circumvent this difficulty, we adopt the fol-
lowing strategy. We scale the exchange interaction by a
factor x, so as to reduce its strength. That is to say, we re-
place J with J 0 苷 xJ in the expression for the total energy,
and look at the corresponding phase diagram. As expected,
such a procedure allows the three configurations, (I), (II),
and (III), to show up even for much smaller systems (e.g.,
for x 苷 0.04, N may be smaller than 106 ). We minimize
the energy by exploring in a regular mesh the D-H plane.
The lower-lying state was determined by comparing the to-
tal energy Etot for the three configurations. The boundaries
between different phases have been determined by refining
the mesh in appropriate regions of the diagram. Figure 1
shows the phase diagrams one obtains for x 苷 0.04, 0.06,
237202-2
10 JUNE 2002
70
60
50
Height H (nm )
III
40
, II
30
20
10
I
0
0 10 20 30 40 50
Diameter D (nm )
FIG. 1. Phase diagram for x 苷 0.04 (circles), x 苷 0.06 (tri-
angles), x 苷 0.08 (squares), and x 苷 0.1 (stars) showing the
boundaries between the in-plane ferromagnetic (I), out-of-plane
ferromagnetic (II), and vortexlike (III) configurations.
0.08, and 0.1. The four diagrams have the same shape, with
a triple point that moves along a common straight line as x
is varied. The slope of such a line, which separates the two
ferromagnetic configurations, is given by the geometry of
the cluster. Indeed, the difference in total energy between
the two ferromagnetic configurations is determined by the
dipolar interaction, via the demagnetizing field, which de-
pends on the sample shape and magnetization orientation.
For cylinders, it can be shownpthat the line separating the
two phases is given by H 苷 p D兾2.
The remarkable result in Fig. 1 is the relation between
cluster height Ht (or diameter Dt ) at the triple point and
scaling factor x. This is illustrated in Fig. 2. We clearly
see that Ht scales with x according to Ht 苷 Ax h , with
A ⯝ 91.1 nm and h ⯝ 0.55. The uncertainty in the values
of both A and h is smaller than 1%. Such a relation
enables us to estimate the position (Dt , Ht ) of the triple
point for the full strength of the exchange coupling, i.e., for
x 苷 1, which for the case of Co would be Dt ⯝ 102.8 nm
and Ht 苷 91.1 nm. This corresponds to a cylinder with
N 艐 108 , which is already too large to be handled on a
computer. In fact, it is easy to show that the number of
atoms Nt at the triple point scales with x according to the
relation Nt 苷 N0 x 3h , with N0 ⯝ 6.5 3 107 .
Moreover, the scaling relation we have just obtained
leads to an even more important result, namely, the sought-
after phase diagram for the full strength of J. Figure 3
shows the four previous phase diagrams in which the D and
H axes have both been scaled by a factor 1兾x h . All four
diagrams collapse upon a single one, which corresponds
to x 苷 1. The diagram in Fig. 3 is in agreement with the
237202-2
VOLUME 88, NUMBER 23 PHYSICAL REVIEW LETTERS 10 JUNE 2002

100 300

0.55
H t = 91.1 * X

200

Height H (nm )
III
H t (nm )

II

100

I
10 0
0.0 0 .1 1.0 0 50 100 150 200 250
x Diam eter D (nm )

FIG. 2. Cylinder height at the triple point as a function of the FIG. 3. Phase diagram for the values of x as in Fig. 1, with
scaling factor x. the axes scaled by a factor 1兾x h (see text).

a scaling relation in the phase diagram of these systems.

available experimental data. It establishes a lower bound
Such a relation has allowed us to investigate systems of
to Dt , of about 100 nm, for the occurrence of vortex states,
the size of those experimentally studied, whose number
which agrees quantitatively with the experimental data of
of magnetic moments is so large that it would be prohibi-
Cowburn et al. [8]. In addition, our phase diagram entirely
tively time consuming to deal with them using present-day
supports the experimental finding of Lebib et al. [9]. that
standard computational facilities. The phase diagram we
for Co dots with sufficiently low H the magnetization re-
obtained may be of great interest for the production of new
versal is controlled by a single-domain state.
devices based on nanostructured magnetic materials.
Although our results have been obtained for cylindri-
We thank S. L. de Queiroz for helpful discussions. This
cally shaped clusters, one might plausibly expect to find
work has been partially supported by Fondo Nacional
similar scaling relations for clusters with different shapes,
de Investigaciones Científicas y Tecnológicas (FONDE-
which would be extremely useful to the study of the mag-
CYT, Chile) under Grants No. 1010127, No. 7010127, and
netism of nanostructured systems. It is worth commenting
No. 1990812, Grant No. P99-135F from the Millennium
on the role played by the exchange and dipolar interactions
project (Mideplan, Chile), DICYT from Universidad de
in determining the relative stability of the three phases. In
Santiago de Chile, and by CNPq, VITAE, FAPERJ, and the
all cases considered here, the exchange energy turned out
“Millennium Institute for Nanosciences/CNPq” of Brazil.
to be larger than the dipolar one. However, transitions
between configurations (or phases) do occur because, for
some geometry of the system, the difference in the ex-
change energy between two given configurations DEex re-
sults in smaller magnitude than the corresponding change *Electronic address: jcastro@if.ufrj.br
in the dipolar one DEdip . We remark that the change in the [1] J. N. Chapman, P. R. Aitchisom, K. J. Kirk, S. McVitie,
anisotropy energy was always smaller in magnitude than J. C. S. Kools, and M. F. Gillies, J. Appl. Phys. 83, 5321
both DEex and DEdip . This result has an interesting conse- (1998).
quence. Since the boundaries between the three phases are [2] C. A. Ross, Henry I. Smith, T. Savas, M. Schattenburg,
determined solely by the exchange and dipolar interactions M. Farhoud, M. Hwang, M. Walsh, M. C. Abraham, and
(both proportional to the square of the modulus of the mag- R. J. Ram, J. Vac. Sci. Technol. B 17, 3168 (1999).
[3] S. Y. Chou, Proc. IEEE 85, 652 (1997); G. Prinz, Science
netic moments), the scaling relation implies that the phase
282, 1660 (1998).
diagrams that one obtains for fixed shape and different ma- [4] J. Zhu, Y. Zheng, and G. Prinz, J. Appl. Phys. 87, 6668
terials, A and B, should scale as 共JA 兾JB 兲h , where JA and (2000).
JB are the respective exchange interactions. [5] M. Grimsditch, Y. Jaccard, and Ivan K. Schuller, Phys. Rev.
In conclusion, several interesting results have been ob- B 58, 11 539 (1998).
tained from our study of the phase diagram of nanostruc- [6] M. C. Abraham, H. Schmidt,, T. A. Savas, Henry I. Smith,
tured magnetic clusters. We have found the existence of C. A. Ross, and R. J. Ram, J. Appl. Phys. 89, 5667 (2001).

237202-3 237202-3
VOLUME 88, NUMBER 23 PHYSICAL REVIEW LETTERS 10 JUNE 2002

[7] M. Hwang, M. C. Abraham, T. Savas, Henry I. Smith, R. J.
Ram, and C. A. Ross, J. Appl. Phys. 87, 5108 (2000).
[8] R. P. Cowburn, D. K. Koltsov, A. O. Adeyeye, M. E.
Welland, and D. M. Tricker, Phys. Rev. Lett. 83, 1042
(1999).
[9] A. Lebib, S. P. Li, M. Natali, and Y. Chen, J. Appl. Phys.
89, 3892 (2001).
[10] K. Yu Guslienko, Sug-Bong Choe, and Sung-Chul Shin,
Appl. Phys. Lett. 76, 3609 (2000).
[11] C. Kittel, Introduction to Solid State Physics (Wiley,
New York, 1986), p. 566.
[12] N. Rosengaard and B. Johanson, Phys. Rev. B 55, 14 975
(1997).

237202-4 237202-4