Phy Revb PDF
Phy Revb PDF
Phy Revb PDF
Institutional Repository
by the/an author.
Additional Information:
http://link.aps.org/abstract/PRB/v72/e014433.
Publisher:
c American Physical Society
We have found that in a system consisting of small magnetic particles a phenomenon related to the formation
of fractal structures may arise. The fractal features may arise not only in the distribution of magnetic moments
but also in their energy spectrum. The magnetization and the susceptibility of the system also display fractal
characteristics. The multiple structures are associated with exponentially many locally stable minima in a
highly complex energy landscape. The signature of these fractal structures can be experimentally detected by
various methods.
of domain structures in a linear chain and in other small coupling constant J of the exchange or dipolar interaction
multiparticle clusters. We investigate this model in detail between particles, which was probed in experiments by
considering arrays of a few and many nanoparticles, and Cowburn.17
shown that they have a strong potential for sensor applica- These assumptions are plausible and will strongly sim-
tions and data storage. In particular we have shown that such plify the considered model for a system made of these small
systems have a very complex nontrivial magnetic behavior particles. This will allow us to find exact analytic nontrivial
and therewith different nontrivial structures displaying frac- solutions describing all states, magnetic properties, and en-
tal features may be formed. Specifically the system behavior ergy spectra of the system under consideration. The total
共with increasing particle number兲, the values of magnetic distribution of moments between the particles are also deter-
moments, the energy spectrum, coercive forces, hysteresis mined by an exchange and dipole-dipole interactions be-
loops, and all such properties may display the features of a tween the magnetic particles. The chain of magnetic particles
fractal. The formation of these fractal structures is mostly can be described by the Hamiltonian
related to the discrete nature of the systems made of small
particles and does not depend on the specific models which A B
we have been considering. Therefore the phenomenon of the F = − J 兺 Sជ iSជ j + 兺 Sជ 2i + 兺 Sជ 4i − 兺 H
ជ Sជ + K 兺 共S2 + S2 兲,
i ix iy
具i,j典 2 i 4 i i i
fractal creation has a very general character and must be
taken into account in the design of any system made of small 共2兲
particles and having a potential for applications.
For an illustration of the fractal formation in the present where we assume that particles are interacting ferromagneti-
work we consider one of these systems made of small par- cally with the exchange constant J and the value of the con-
ticles, namely, a linear chain. Such system can be and has figurational anisotropy constant K / J → ⬁ and −K / A → ⬁. Let
already been produced from small ferromagnetic particles us make an expansion of the free energy with respect to these
made, for example, of Fe 共see, also, Ref. 16兲. If all spins very large parameters. The stationary configurations of the
within a single particle are ferromagnetically ordered, then magnetic moments are determined with the use of equations
each particle may be considered as having a single classical obtained by the minimisation of F with respect to the com-
spin Sជ . The value of Sជ may be well described by a Ginzburg- ponents Six , Siy, and Siz. In zero approximation with respect
Landau model applied to the single particle to the very large parameters, K / J Ⰷ 1 and −K / A Ⰷ 1 the
equations obtained for the first two components have a very
A ជ2 B ជ4 ជ ជ simple form: Six = 0 and Siy = 0 and obvious solutions. Taking
Fsingle = S + S − HSi , 共1兲 into account these solutions, the form of the vectors describ-
2 i 4 i
ing the magnetic moments, Sជ i, is simplified to the form Sជ i
where Sជ = 共Sx , Sy , Sz兲; here A = a共T − Tc兲 and B are phenom- = 共0 , 0 , Szi兲. After the substitution of this form for Sជ i the
enological constants and H ជ is an external magnetic field. The terms with large parameters in the expression for the free
relative orientation of the total magnetic moment of each energy F disappear and it is simplified to the form
冉 冊兺
single particle is determined by anisotropy constants. The
JS2 A B 4
shape of the particles gives rise to the so-called configura-
tional anisotropy, which was not included yet. For small par-
F= 兺 共zi − z j兲2 + 2 S2 − JS2
2 具i,j典 i
z2i +
4
S 兺 z4i
i
ticles the value of the configurational anisotropy is the larg-
est. This situation with the largest configurational anisotropy − HS 兺 zi . 共3兲
can be readily produced for all magnetic particles having, for i
example, a needle shape, so that each individual magnetic Let us make the substitution zi = 冑共␣J / BS2兲xi. After this
moment associated with each single particles can be oriented substitution the expression for the free energy, Eq. 共3兲 will be
perpendicular to or along the chain of these particles.15,16 Let transformed into a new form. Then, after rescaling and using
us assume that all particles have such a shape which gives dimensionless units we obtain a standard model as
the strongest configurational anisotropy. For example, sup-
冋 册
pose all particles have ellipsoidal shape prolongated along N
1 ␣
the z direction. Then due to the configurational anisotropy all F共x1,x2,…,xN兲 = 兺 共xi − xi−1兲2 + 共x2i − 1兲2 − h 兺 xi ,
magnetic moments associated with single individual particles i=1 2 4 i
are oriented along the z direction. Therefore, since the mag-
共4兲
netic moments of all particles will be collinear, we can intro-
duce the ansatz Sជ i = 共0 , 0 , Szi兲, where the magnitude zi de- where ␣ = 2 − A / J and h = H冑B / ␣J3. Here the new energy
scribes the value of the magnetic moment in the ith particle. scale is F0 = ␣J2 / B, i.e., F = F / F0. The physical variable xi
Below we show that this ansatz arises exactly as a result of specifies the value of the the magnetic moment of the i-th
perturbation theory expansion in the limit of a large aniso- particle oriented in the z direction while the total magnetiza-
tropy constant K in a comparison with a interparticle inter- tion of the system M = ⌺ixi.18–20 It is specifically interesting
action J. We assume also that the magnetic field is oriented to investigate the situation when the particles form a ring or
along the axis of the anisotropy as H ជ = 共0 , 0 , H兲 and the con- a loop geometry. Similar kind of loops were recently widely
figurational anisotropy constant K is much larger than the discussed as a candidate for magnetic memory elements.21
014433-2
FRACTAL STRUCTURES IN SYSTEMS MADE OF SMALL… PHYSICAL REVIEW B 72, 014433 共2005兲
冉 冊
N four minima are well separated in the space of individual
x*i 共4xi−1
* *
− xi−2 兲−3
F = 兺 1 − xi−1
*
x*i + +¯ . 共8兲 magnetic moments, their separation in the energy space is
i=1 2␣ very small, ie significantly smaller than the barrier heights
associated with the saddle points 共see, Fig. 1兲. Thus due to
Introducing the quantities l and l⬘ by N − 2l : = ⌺i=1
N
xixi−1 and such an energy landscape the system could be locked in one
N − 2l⬘ : = ⌺i=1xixi−2, respectively, the total free energy per
N
of these minima with decreasing temperature, even if this
particle can be approximated by the simple expression minimum does not correspond to the ground state.25
014433-3
K. E. KÜRTEN AND F. V. KUSMARTSEV PHYSICAL REVIEW B 72, 014433 共2005兲
FIG. 1. 共Color online兲 The energy landscape for the system FIG. 2. A representative series of absolute values of magnetic
consisting of two particles, i.e., the dependence of the total energy moments of the individual particles are presented on the horizontal
on the values of magnetic moments 共x1 , x2兲. The picture illustrates axes. The scale of the vertical axes is meaningless. Top: The distri-
that the difference between energy related to absolute minima aris- bution of magnetic moments for a chain consisting of N = 50 000
ing at the values of moments ±共1 , 1兲 and the energy related to particles for the parameter ␣ = 8. Bottom: The zoomed fine structure
metastable minima arising at the points ±共冑1 − 4 / ␣ , −冑1 − 4 / ␣兲 is of the distribution of magnetic moments in the range of moment
much smaller than the barrier height defined by the saddle points values from 0.6705 to 0.6745. This distribution displays all features
presented on this figure. of the Cantor set. The total set of real values for magnetic moments
is obtained from the set of absolute values presented in the figure
Therefore, the formation of a particular structure will de- with the use of the symmetry of Eq. 共5兲, by simple substitution xi
pend strongly on the history of the system, for example at → −xi.
which rate and from which temperatures and at which mag-
netic field the material was cooled or heated. The pair of
* ken. The different 2N / N locally stable distinct configurations
domain walls created and described by the configuration xD
* of the domains may be analytically described up to arbitrary
has a higher energy than the soliton xS. This indicates that
order with the parameter ␣ Ⰷ 1. The distribution of magnetic
there is a decoupling energy which is needed to decouple
moments and the energy spectrum will be presented below.
domains having different topological charges 共kink from an-
tikink兲. That is there is the energy cost in decoupling of the
domain walls forming the soliton. In general domain walls IV. DISTRIBUTION OF MAGNETIC MOMENTS
are always created in pairs. One always needs the energy to
perform a transition from the state with l domain walls to the The calculated values of the magnetic moments are shown
state with l + 2 domain walls. Note that the highest energy in Fig. 2. Pronounced discrete structures appear marking for-
F = 2共1 − 2 / ␣兲N is achieved for the antiferromagnetic con- bidden regions. It is reminiscent of band chaos often found in
figuration xi = 共−1兲i冑1 − 4 / ␣ with l = N and l⬘ = 0 or in other nonlinear dynamical system theory. A physical origin of
words for maximal number of domains. This structure exists these bands and gaps may be related to a bound state of the
only for ␣ 艌 4 and is physically stable only for ␣ 艌 6. Simi- stable topological excitations like kinks and antikinks. An
larly, if the magnetization of one particle is zero 共xi = 0兲 the example is the formation of a soliton. The fine structure is
energy cost due to this “node” would be of the order of ␣ due to the kink-antikink interaction. This interaction modifies
Ⰷ 1. The configurations with nodes are physically unstable the value of the magnetic moment xi associated with each
saddle points and correspond to barriers separating the states single ith particle 共see, Fig. 2兲.
with domain walls and solitons. With increasing and decreas- The fine structures reveal some self-similarity reminiscent
ing magnetic fields there arises multiple instability. At each of those for the standard logistic map xn+1 → rxn共1 − xn兲 with
such instability the saddle point associated with one of the growth parameter r. When it is close to the critical constant
barriers will coincide with the local minimum, and the lo- rc = 3.5699… it is specifying the onset of chaos, where the
cally stable configuration, which was associated with that attractor is a Cantor set. In both cases the origin of the fractal
minimum, ceases to exist. These metastable states and mul- structures is in the discreetness of the system and in the
tiple instability created with increasing and decreasing mag- nonlinearity. In our system the latter is related to the interac-
netic field lead to a formation of multiple hysteresis loops, tion between domain walls. A crude estimate for the fractal
see below. dimension with the aid of a box counting algorithm gives the
The classification of the metastable states is simple. The value D = 0.555 compared to D = 0.538 for the logistic map.26
first is the two domain walls 共bound state of kink + antikink兲. In order to get an insight into the evolution of the fractal
The next one consists of four domain walls 共two kinks and set, we calculate systematically all values of magnetic mo-
two antikinks兲, etc. At zero magnetic field the energy differ- ments for all possible configurations. We consider small clus-
ence between successive minima is of the order of 1 while ters 共rings兲 with N = 1,3,5,7,… and impose periodic boundary
the height of the barrier is always of the order of ␣ Ⰷ 1. conditions 共x0 = xN , xN+1 = x1兲. For a single particle, N = 1, we
When the coupling constant ␣ Ⰷ 1 we may expect that the only have the two degenerate ground states. These two states
system has some kind of glassy behavior where an ideal correspond to the two values magnetic moments, 共x = ± 1兲.
ferromagnetic ordering of the ferromagnetic particles is bro- For N = 3 we find the two ground states and six soliton struc-
014433-4
FRACTAL STRUCTURES IN SYSTEMS MADE OF SMALL… PHYSICAL REVIEW B 72, 014433 共2005兲
014433-5
K. E. KÜRTEN AND F. V. KUSMARTSEV PHYSICAL REVIEW B 72, 014433 共2005兲
014433-6
FRACTAL STRUCTURES IN SYSTEMS MADE OF SMALL… PHYSICAL REVIEW B 72, 014433 共2005兲
014433-7
K. E. KÜRTEN AND F. V. KUSMARTSEV PHYSICAL REVIEW B 72, 014433 共2005兲
014433-8
FRACTAL STRUCTURES IN SYSTEMS MADE OF SMALL… PHYSICAL REVIEW B 72, 014433 共2005兲
not correspond to a specific deep minimum. Then all types of induce the transition of the system from this minimum into
thermal fluctuations exist and all types of configurations will another neighboring local minimum that has the closest mag-
be created. With fast cooling these configurations could be netic configuration. Such a transition must occur when a cer-
frozen into one of the configurations of the multivalley en- tain branch of the hysteresis set presented in Figs. 6 and 7 is
ergy landscape, which will be revealed by the fractal struc- terminated and such transition will be reflected in the
ture of the measured values of the total magnetization. Barkhausen jump of the total magnetization M. This will be
The complicated structure of the hysteresis loop presented seen in the experiment measuring the total magnetization of
in Figs. 6 and 7 may be observed in a series of the field the system. With the next continuous change of the field h,
cooled and zero field cooled experiments. Let us note that first the magnetization will be changed continuously until
this structure is a result of the complicated energy landscape this branch of the hysteresis set will be terminated. Then
consisting of the many locally stable minima separated by another transition into a new minimum and another
large barriers. The formation of these minima and the barri-
Barkhausen jump in the total magnetization will occur. Thus
ers is beginning with decreasing temperature starting from
if the system will be trapped originally in one of these locally
the critical temperature Tc of the bulk ferromagnetic transi-
tion. With the next decreasing temperature the minima be- stable 共not absolute兲 minima, due to the fast cooling or some
come deeper and deeper while the barrier height increases. other way, then with continuous change of the field h we will
At all temperatures the energy positions of the bottom of the recover a series of Barkhausen jumps in the total magnetiza-
minima are not very different from each other. Such differ- tion M. Such a Barkhausen 共devil兲 staircase may be different
ence is determined by the weak dipole-dipole interaction be- from one experiment to another, from one measurement to
tween particles which is characterized by the coupling con- another reflecting a very rich and complicated energy land-
stant J. Therefore, with decreasing temperature, when Tc scape associated with the system consisting of these small
− T Ⰷ J, we will have a system where there are many, practi- particles.
cally equivalent, minima separated by large barriers. A
simple estimation gives that for a particle consisting of 103
magnetic atoms the value of J must be significantly smaller VIII. SUMMARY
than 10 K. The latter, of course, depends on the distance Thus, our studies led us to amazing results, namely the
between particles r and ceases as 1 / r3. Therefore, it is clear spatial structures of domains in a chain made of small mag-
that the condition when Tc − T Ⰷ J can be easily realized. Ob-
netic particles show fractal self-similarity. Such a spatial dis-
viously that when the temperature decreases very fast the
tribution of magnetic moments associated with different
system may be easily trapped in any one of these minima. In
chains or clusters of particles is a difficult task to measure on
which of these minima the trapping will occur does depend
on the cooling rate and on initial configuration created by a experiments, although modern techniques, such as STM and
thermal fluctuation as well as on the strength of the applied AFM as well as the Kerr rotation, may allow such a task.
magnetic field. This property of the system may serve as a However, from our point of view, the most important result
basis of the following experiments. obtained is the fractal structure of the energy spectrum. The
First, if we cool the system in a strong magnetic field and energy landscape, which gives rise to such a fractal spec-
then continuously change the field h we will recover the trum, is associated with the creation of domains and fractal
main features of the bulk hysteresis loop 共see the dashed values of the total magnetization. This landscape energy sur-
lines in Fig. 6兲. In this case the system will be trapped in a face consists of many locally stable minima separated by
fully ferromagnetic state associated with the absolute mini- large barriers. Each of these minima corresponds to the state
mum. Moreover, the system will even be trapped in this state with some fixed number of domains. Even if such a number
when the minimum becomes not absolute, which happens is fixed the states associated with different configurations or
with the next continuous change of the field h. However, if rearrangement of these domains will correspond to different
we cool the system in a weak or zero field the resulting or the same minima. This is precisely the situation arising in
minimum in which the system will be trapped will depend on a glassy system. The shape of the energy landscape leads us
the cooling rate and on the initial high temperature state. to the conclusion that the system formed from magnetic par-
With the very slow cooling rate the system will have time to ticles is some kind of magnetic glass associated with the
termalize into the bottom of absolute minimum. However, creation of domains 共see also Ref. 27兲. We propose to make
due to the barriers of much height that surround these a detailed experimental investigation of systems made of
minima, when the cooling rate is not slow the system will be small magnetic particles 共chains or clusters兲 to identify this
trapped in other locally stable minima, which is different glassy character and the fractal features of their domain
from the absolute mininum and has different values of the structure. In this respect it might be useful to measure the
total magnetization. With the very fast cooling rate in a weak magnetization at zero field as well as in cooled regimes as
or zero field the system may be trapped in any minima hav- commonly practiced in experiments on spin glasses. Due to
ing locally stable configurations and characterized by differ- these energy landscapes described above the corresponding
ent values of magnetization. Such a dramatic difference be- magnetic structures at very low temperatures are very stable
tween the field cooled and zero field cooled magnetizations with respect to thermal as well as to quantum fluctuations. To
reminds us of the situation that is usually arising in spin reveal these fractals the experiments associated with fast
glasses. cooling should be setup. The repetition of the fast cooling
Then, when the system is trapped in one of these locally from high temperatures may drive the system to settle in a
stable minima, the continuous change of the field h may different valley of the energy landscape. Measurement of the
014433-9
K. E. KÜRTEN AND F. V. KUSMARTSEV PHYSICAL REVIEW B 72, 014433 共2005兲
total magnetization at each lap of cooling with the same or details Ref. 33兲. Such orbital moments give rise to a para-
different cooling rates may provide a set of numbers remi- magnetic Meissner effect33 observed in cuprate
niscent of parts of a fractal. The latter will depend on the superconductors.34,35 A chain or a planar array of electrically
shape and the number of particles of which the nanostructure isolated rings could be treated as a set of magnetic mo-
is formed. Since different clusters will be associated with ments oriented perpendicular to the plane and interacting via
different fractals, then in general these studies may lead to magnetic dipole forces. This dipole-dipole interaction modi-
the development of a type of spectroscopy where, with the fies the values of the orbital magnetic moments and leads to
aid of fast cooling magnetization measurements, the struc- the formation of the fractal in the way decribed in this paper.
ture of small clusters may be identified. These findings could
also open areas of applications for devices having very con-
trollable and stable giant magnetoresistance properties.28 ACKNOWLEDGMENTS
There is another interesting example of the systems,
where the described fractal structures can arise as well.30 We are grateful to F. Aliev, J. W. Clark, D. Edwards, G.
These are one-dimensional chains and two-dimensional ar- Gehring, C. Krattenthaler, and John Samson for useful dis-
rays of rings.31,32 A single ring is a superconducting loop cussions. The work has been supported by European Science
consisting of Josephson junctions where there is at least one Foundation in the framework of the network program: Ar-
junction.33 The phase shift by in such a junction results rays of Quantum Dots and Josephson Junctions as well as the
in the formation of an orbital moment on the ring 共see for EPSRC Grant No. GR/S05052/01.
1 M. A. Howson, Contemp. Phys. 35, 共5兲 347 共1994兲. Math. Phys. 8, 38 共2001兲.
2
C. Binns, Surf. Sci. Rep. 44, 1 共2001兲. 20
K. E. Kürten, in Condensed Matter Theories, 共Nova Science,
3 C. Binns, S. Louch, S. H. Baker, K. W. Edmonds, M. J. Maher,
New York, 2000兲 Vols. 14 and 15.
and S. E. Thornton, IEEE Trans. Magn. 38, 141 共2002兲. 21
P. Vavassori, M. Grimsditch, V. Novosad, V. Metlushko, and B.
4 H. Ardhuin, J. N. Chapman, P. R. Aitchison, M. F. Gillies, K. J.
Illic, Phys. Rev. B 67, 134429 共2003兲.
Kirk, and C. D. W. Wilkinson, J. Appl. Phys. 88, 2760, 共2000兲. 22 K. E. Kürten and C. Krattenthaler, in Condensed Matter Theories,
5 W. Schepper, A. Hütten, and G. Reiss, J. Appl. Phys. 88, 993
共Nova Science New York, 2004兲 Vol. 18.
共2000兲. 23
P. Bak and V. L. Pokrovsky, Phys. Rev. Lett. 47, 958 共1981兲.
6 M. V. Gvozdikova and A. S. Kovalev, Low Temp. Phys. 25, 1295 24 D. K. Campbell, S. Flach, and Y. S. Kivshar, Phys. Today 57, 43
共1999兲. 共2004兲.
7 25
F. V. Kusmartsev, H. S. Dhillon, and M. D. Crapper, J. Magn. D. Feinberg and F. V. Kusmartsev, Physics in Local Lattice Dis-
Magn. Mater. 198-199, 743 共1999兲. tortions, edited by Oyanagi and A. Bianconi 共AIP, New York,
8
A. Moser, K. Takano, D. T. Margulies, M. Albrecht, Y. Sonobe, 2001兲, p. 262.
Y. Ikeda, S. Sun, and E. Fullerton, J. Phys. D 35, R157 共2002兲. 26 P. Grassberger, J. Stat. Phys. 26, 173 共1981兲
9 27
S. P. Parkin, C. Kaiser, A. Panchula, P. Rice, M. Samant, and P. E. Jönsson, S. Felton, P. Svedlindh, P. Nordblad, and M. F.
S.-H. Yang, Nat. Mater. 3, 862 共2004兲. Hansen, Phys. Rev. B 64, 212402 共2001兲.
10 S. P. Parkin, X. Jiang, C. Kaiser, A. Panchula, K. Roche, and M. 28 P. Grünberg, Phys. Today 54, 31 共2001兲.
Samant, Proc. IEEE 91, 661 共2003兲. 29 H. Barkhausen, Z. Phys. 20, 401 共1919兲.
11 D. R. Lee, J. W. Freeland, G. Srajer, V. Metlushko, cond-mat/ 30 F. V. Kusmartsev, D. M. Forrester, and M. S. Garelli, in Physics
1997兲, Vol. 284 Kock, B. Roden, D. Khomskii, and D. Wohlleben, Phys. Rev.
19 H. S. Dhillon, F. V. Kusmartsev, and K. E. Kürten, J. Nonlinear
Lett. 68, 1908 共1992兲.
014433-10