nonlinear susceptibility and dynamic hysteresis loops of magnetic nanoparticles with biaxial...
TRANSCRIPT
Nonlinear susceptibility and dynamic hysteresis loops of magneticnanoparticles with biaxial anisotropyBachir Ouari, Serguey V. Titov, Halim El Mrabti, and Yuri P. Kalmykov Citation: J. Appl. Phys. 113, 053903 (2013); doi: 10.1063/1.4789848 View online: http://dx.doi.org/10.1063/1.4789848 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v113/i5 Published by the American Institute of Physics. Related ArticlesCarrier density dependence of the magnetic properties in iron-doped Bi2Se3 topological insulator J. Appl. Phys. 113, 043926 (2013) Structural variants and the modified Slater-Pauling curve for transition-metal-based half-Heusler alloys J. Appl. Phys. 113, 043709 (2013) d0 magnetism in semiconductors through confining delocalized atomic orbitals Appl. Phys. Lett. 102, 022422 (2013) Influence of Ga doping on the Cr valence state and ferromagnetism in Cr: ZnO films Appl. Phys. Lett. 102, 022414 (2013) Half-metallic ferromagnetism in wurtzite ScM (M=C, Si, Ge, and Sn): Ab initio calculations Appl. Phys. Lett. 102, 022404 (2013) Additional information on J. Appl. Phys.Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors
Nonlinear susceptibility and dynamic hysteresis loops of magneticnanoparticles with biaxial anisotropy
Bachir Ouari,1,2 Serguey V. Titov,3 Halim El Mrabti,1 and Yuri P. Kalmykov1
1Univ. Perpignan Via Domitia, Laboratoire de Math�ematiques et Physique, EA 4217, F-66860 Perpignan,France2Macromolecular Research Laboratory, Physics Department, Abou Bekr Belkaid University,BP 119 Chetouane, Tlemcen, Algeria3Kotelnikov’s Institute of Radio Engineering and Electronics of the Russian Academy of Sciences, VvedenskiiSquare 1, Fryazino, Moscow Region 141190, Russian Federation
(Received 29 August 2012; accepted 15 January 2013; published online 4 February 2013)
The nonlinear ac susceptibility and dynamic magnetic hysteresis (DMH) of a single domain
ferromagnetic particle with biaxial anisotropy subjected to both external ac and dc fields of
arbitrary strength and orientation are treated via Brown’s continuous diffusions model [W. F.
Brown, Jr., Phys. Rev. 130, 1677 (1963)] of magnetization orientations. The DMH loops and
nonlinear ac susceptibility strongly depend on the dc and ac field strengths, the polar angle between
the easy axis of the particle, the external field vectors, temperature, and damping. In contrast to
uniaxial particles, the nonlinear ac stationary response and DMH strongly depend on the azimuthal
direction of the ac field and the biaxiality parameter D. VC 2013 American Institute of Physics.
[http://dx.doi.org/10.1063/1.4789848]
I. INTRODUCTION
Fine Single domains ferromagnetic nanoparticles have
been an active focus of research for over 60 years prompted
by both fundamental aspects of their magnetization and
increasing interest by industry especially for data storage.1,2
Other important current applications include medical appli-
cations, e.g., in hyperthermia.3 Such nanoparticles are char-
acterized by instability of the magnetization M(t) due to
thermal agitation resulting in spontaneous change of its ori-
entation from one metastable state to another by surmounting
anisotropy-Zeeman energy barriers resulting in the phenom-
enon of superparamagnetism.4 Due to the large magnitude of
the magnetic dipole moment of single-domain nanoparticles
so that the Zeeman energy is relatively large even in weak
external magnetic fields, the magnetization reversal process
has a strong field dependence causing nonlinear effects in
the dynamic susceptibility, stochastic resonance, dynamic
magnetic hysteresis (DMH), etc.
The determination of the nonlinear magnetic susceptibil-
ity and DMH for arbitrary ac field strengths amounts to the
calculation of the nonlinear response of the magnetization in
the presence of the thermal agitation. This calculation is usu-
ally treated by the Langevin equation which is Gilbert’s5 (or
the Landau-Lifshitz6) equation augmented by a random field
h(t) with Gaussian white noise properties, accounting for
thermal fluctuations of the magnetization MðtÞ of an individ-
ual particle,
_MðtÞ ¼ cMðtÞ � � @V
@Mtð Þ � a
cMS
_MðtÞþh tð Þ� �
; (1)
where c is the gyromagnetic ratio, a is the dimensionless damp-
ing parameter, MS is the saturation magnetization, and V is
the free energy per unit volume comprising the non-separableHamiltonian of the magnetic anisotropy and Zeeman energy
densities. Here jMj ¼ MS is assumed to be constant so that the
only the orientation of M varies. Equation (1) is then used to
derive the accompanying Fokker-Planck equation7 governing
the time evolution of the probability density function Wðu; tÞof magnetization orientations on a unit sphere (u is a unit vec-
tor along M) and the relevant FPE is
2sN
@W
@t¼ b
au � ðrV �rWÞ þ r � ðrW þ bWrVÞ; (2)
where r ¼ @=@u is the gradient operator on the unit sphere,
sN ¼ s0ðaþ a�1Þ is the characteristic free diffusion time,
s0 ¼ bMS=ð2cÞ, b ¼ v=ðkTÞ, v is the volume of the particle,
k is Boltzmann’s constant, and T is the absolute temperature.
Various treatments of the nonlinear ac stationary
response have been effected using numerical solutions of the
governing dynamical equations (1) and/or (2). In particular,
efficient numerical algorithms for the calculation of the non-
linear ac stationary response of uniaxial superparamagnetic
nanoparticles have been proposed (see, e.g., Refs. 8–10) by
assuming that the ac driving field HðtÞ is directed along the
easy axis of the particle. However, in this axially symmetric
configuration, many interesting nonlinear effects cannot be
treated and understood because no dynamical coupling
between the longitudinal and transverse modes of motion
exists. Recently,11–13 these results have been generalized to
determine the nonlinear ac stationary response for a uniaxial
particle driven by a strong ac field applied at an angle to the
easy axis of the particle so that the axial symmetry is broken
by the Zeeman energy. In the above papers, only the axially
symmetric uniaxial anisotropy potential has been treated
considerably simplifying the analysis. However, the results
cannot be applied to particles with inherent nonaxially sym-
metric anisotropy, such as biaxial anisotropy,
bVbiaxð#;uÞ ¼ r sin2 #þ D sin2 # cos2 u: (3)
0021-8979/2013/113(5)/053903/9/$30.00 VC 2013 American Institute of Physics113, 053903-1
JOURNAL OF APPLIED PHYSICS 113, 053903 (2013)
Here, D and r are the biaxiality and barrier parameters,
respectively (D ¼ 0 corresponds to uniaxial anisotropy), #and u are polar and the azimuthal angles of the spherical
coordinate system. We treat this problem because biaxial an-
isotropy may yield an appreciable contribution to the free
energy density of magnetic nanoparticles.14 In particular, Eq.
(3) describes the magnetic anisotropy energy of a spheroidal
single-domain particle with easy anisotropy axis inclined at
a certain angle to the particle axis of symmetry15 as well that
of elongated particles, where easy- and hard-axis anisotropy
terms are present.16 Furthermore, biaxial anisotropy consti-
tutes a generic model for treatment of the thermally assisted
magnetization reversal in the presence of a spin-transfer tor-
que.17 Biaxial anisotropy generates azimuthally nonuniform
energy distributions with two saddle points that lead to new
effects, viz., strong intrinsic dependence of the magnetic
characteristics on the damping arising from coupling of the
longitudinal and transverse relaxation modes.18,19 Further-
more, in contrast to uniaxial particles, the nonlinear ac sta-
tionary response strongly depends on the azimuthal direction
of the ac field. A method of solution of the general nonlinear
problem [i.e., nonlinear ac stationary response for superpara-
magnets with arbitrary anisotropy and including the gyromag-
netic term in the Fokker-Planck equation (2)] has been given
by Titov et al.20 Here, using this approach, we calculate the
nonlinear magnetic susceptibilities and DMH of nanoparticles
with biaxial anisotropy subjected to a strong ac driving field
superimposed on a strong dc bias field. We shall demonstrate
that for biaxial anisotropy, the magnetization dynamics alter
substantially leading to new temperature and damping de-
pendent nonlinear effects at low, mid, and high frequencies.
II. BASIC EQUATIONS
The free-energy density V of a single-domain ferromag-
netic nanoparticle with biaxial anisotropy in superimposed
homogeneous external magnetic dc and ac fields H0 þH cos xt of arbitrary strengths and orientations is given by
bVð#;u; tÞ ¼ bVbiaxð#;uÞ � ðn0 þ n cos xtÞ� ðc1 sin# cos uþ c2 sin# sin uþ c3 cos#Þ;
(4)
where n0 ¼ bH0MS and n ¼ bHMS are the dc and ac external
field parameters, respectively, c1 ¼ sin w cos /; c2 ¼ sin wsin/; c3 ¼ cos w are the direction cosines of the ac external
field, w is the angle between H and the Z axis taken as the
easy axis of the particle, and / is the azimuthal angle (here
we have assumed that the vectors H0 and H are parallel). As
shown in Ref. 20, the solution of the Gilbert-Langevin equa-
tion (1) for an arbitrary anisotropy potential can be reduced
to that an infinite hierarchy of differential-recurrence equa-
tions for the statistical moments hYlmiðtÞ governing the dy-
namics of the magnetization. Here Ylmð#;uÞ is a spherical
harmonic of order l and rank m defined as
Ylmð#;uÞ ¼ ð�1Þmffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið2lþ 1Þðl� mÞ!
4pðlþ mÞ!
seimuPm
l ðcos#Þ;
where Pml ðxÞ are the associated Legendre functions. By using
the generalized differential-recurrence equations specialized
to biaxial nanoparticles, we have the ac stationary solution
for the magnetization,
MðtÞ ¼ MShcosHiðtÞ
¼ MS
ffiffiffiffiffiffi4p3
rfc3hY1;0iðtÞ �
ffiffiffi2p
Re½ðc1 � ic2ÞhY1;1iðtÞ�g;
in the direction of the ac driving field H, viz.,
MðtÞ ¼ MS
X1k¼�1
mk1ðxÞeikxt; (5)
where H is the angle between the vectors H and M,
mk1ðxÞ ¼
ffiffiffiffiffiffi4p3
r "c3ck
1;0ðxÞ
þðc1 þ ic2Þck
1;�1ðxÞ � ðc1 � ic2Þck1;1ðxÞffiffiffi
2p
#; (6)
the ckl;mðxÞ are the coefficients in the time Fourier series of
the averaged spherical harmonics, viz.,
hYlmiðtÞ ¼X1
k¼�1ck
l;mðxÞeikxt: (7)
The coefficients ckl;mðxÞ, thus, the magnetization MðtÞ, can
be evaluated using matrix continued fractions (see Appen-
dix). In calculating the nonlinear ac response, we shall con-
sider the parameter range r� 1 and n � 1 corresponding to
the low temperature limit and strongly nonlinear regime.
Henceforth, we shall assume that the vectors H0 and H lie in
the XZ (/ ¼ 0) or YZ (/ ¼ p=2) planes of the laboratory
coordinate system so that the direction cosines are
c1 ¼ sin w; c2 ¼ 0; c3 ¼ cos w
or
c1 ¼ 0; c2 ¼ sin w; c3 ¼ cos w;
respectively. Now, the DMH loop represents a parametric
plot of the steady-state time-dependent magnetization as a
function of the AC field, i.e., MðtÞ vs. HðtÞ ¼ H cos xt,where the time interval t coincides with the period of oscilla-
tion of that field for a given value of H. Here, we also calcu-
late the normalized area of the DMH loop An defined as9
An ¼1
4MSH
þMðtÞdHðtÞ ¼ � p
4Imðm1
1Þ; (8)
which is the energy loss per particle and per cycle of the AC
field.
III. NONLINEAR SUSCEPTIBILITY AND DMH LOOPS
For a weak ac field, n ! 0, vðxÞ ¼ 6m11ðxÞ=n defines
the normalized linear dynamic susceptibility and our results
053903-2 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
agree in all respects with the benchmark linear response so-
lution.18,19 The results indicate that a marked dependence of
vðxÞ on r (inverse temperature), D (biaxiality), n (ac field),
n0 (dc field), a (damping), and angles w and / exists and that
three distinct dispersion bands appear in the spectrum of
vðxÞ (see curve 1 in Fig. 1(b)). The low frequency behavior
of vðxÞ is dominated by the barrier-crossing mode. In addi-
tion, a weaker second relaxation peak appears at high fre-
quencies. This relaxation band is due to the “intrawell”
modes, which are virtually indistinguishable in the frequency
spectrum appearing as a single high-frequency band. At low
dc fields, the amplitude of this band is far weaker than that
of the first band. However, in a strong dc magnetic field
n0 � 1, this band can dominate the spectrum because as n0
increases, the magnitude of the low frequency band
decreases. Just as in the absence of the dc bias field, there is
an inherent geometric dependence of vðxÞ on the damping
parameter a arising from the coupling of the longitudinaland transverse relaxation modes. This coupling appears in
the dynamical equation of motion, where the longitudinal
component of the magnetization is entangled with the trans-
verse components resulting in the appearance of a third fer-
romagnetic resonance (FMR) peak in the spectrum of vðxÞdue to excitation of transverse modes with characteristic fre-
quencies close to the precession frequency of the magnetiza-
tion. The FMR peak appears only for low damping ða� 1Þ
and strongly manifests itself at high frequencies. As adecreases, the half-width of the FMR peak decreases.
In strong ac fields, n> 1, pronounced nonlinear effects
occur (see Figs. 1–3 illustrating the dependence of the non-
linear response on the biaxiality D, damping a, angle w, dc
bias field n0, and the ac field parameter n). In particular, the
low frequency band can no longer be approximated by a sin-
gle Lorentzian indicating that a variety of modes participate
in the magnetization reversal process. However, the charac-
teristic frequency xmax of the maximum magnetic loss may
still be used in order to estimate an effective reversal time of
the magnetization s, i.e., s x�1max. The behavior of xmax
(and, hence, s) as functions of the ac field amplitude depends
on whether or not a dc field is applied. For strong dc bias,
n0 > 1, the low frequency peak shifts to lower frequencies
reaching a maximum at n n0 thereafter xmax decreasing
with increasing n. In other words, as the dc field increases,
the reversal time of the magnetization initially increases and
then having attained its maximum at some critical value
n n0 then decreases (see Fig. 3). For weak dc bias field,
n0 � 1, the low frequency peak shifts monotonically to
higher frequencies. Furthermore, as seen in Fig. 1, with
increasing n, the magnitude of the FMR peak at the preces-
sion frequency xpr decreases and also broadens showing
pronounced nonlinear saturation effects characteristic of a
soft spring. Furthermore, for low damping, a second weak
FIG. 1. The imaginary part of the suscep-
tibility �Im½vðxÞ� vs. the dimensionless
frequency xs0 for barrier parameter
r ¼ 10, biaxiality parameter D ¼ 10, dc
field parameter n0 ¼ 5, angles w ¼ p=4
and / ¼ 0, and various values of damp-
ing a (a) and ac field amplitude n (b).
FIG. 2. �Im½vðxÞ� vs. xs0 for biaxiality
parameter D ¼ 5, dc field parameter
n0 ¼ 5, the ac field parameter n ¼ 5, angle
/ ¼ 0, and various angles w (a) and barrier
parameters r (b).
053903-3 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
resonance peak [owing to phase locking of the nonlinear os-
cillatory (precessional) motion of the magnetization MðtÞ]appears at frequencies xpr=2 (Fig. 2). This effect is similar
to that already existing in nonlinear parametric oscillator sys-
tems driven by an ac external force.21 We remark, however,
that the high-frequency behavior (x� xpr) remains virtually
unchanged. Indeed, the behavior of the imaginary part of sus-
ceptibility �Im½vðxÞ� for x� xpr is independent of n (see
Fig. 1). Moreover, in contrast to uniaxial particles, the FMRpeak always appears for low damping in the spectrum�Im½vðxÞ� at w ¼ 0 because the transverse modes alwaystake part in the longitudinal relaxation process. In Fig. 1(b),
the peak height of the first band firstly (at moderate fields)
increases with increasing the ac field amplitude and then
reduces at higher ac fields. This result implies that these
changes in the ac field magnitude may modulate the heat pro-
duction (specific power loss) in a magnetic nanoparticle
working as a hyperthermia source. In general, however, the
most effective way to tune the heat production is to vary the
magnitude of the dc field10 (see Fig. 3(b)). Our calculation
indicates that with increasing n0 the magnitude of the low fre-
quency band decreases due to the depletion of the population
in the shallowest potential well. Finally, the appearance of the
FMR band is a signature of the onset of dynamic hysteresis at
high frequencies, a feature which is absent for high damping.
In Figs. 4–8, normalized DMH loops, i.e.,
mðtÞ ¼ MðtÞ=MS vs: hðtÞ ¼ HðtÞ=H ¼ cos xt;
for various values of the ac field amplitude n, damping a,
and the dimensionless barrier parameter r / 1=T are pre-
sented. Clearly in Fig. 4, in contrast to uniaxial particles, the
DMH strongly depends on the azimuthal direction of the ac
field H and not only on the polar angle w between the easy
axis and H. Moreover, the coercivity, the remanent magnet-
ization, and the saturation magnetization strongly depend on
the barrier height (inverse temperature) parameter r, so that
considerable variation in the size and shape of loops exists
for different values of r at low frequencies. For xs0 ¼ 10�2,
a ¼ 10�1, n0 ¼ 0, and r � 5, all the hysteresis loops are
large, retaining a large fraction of the saturation field when
the driving field is removed. For r � 2, we have narrow hys-
teresis loops, which implies that a small amount of energy is
used up in repeatedly reversing the magnetization. Figures
5–10 show that the DMH loops and the normalized loop area
An depend strongly on the biaxial parameter D, the oblique
angle w, the damping a and the dimensionless frequency xs0
indicating that the relaxation of the magnetization is mostly
caused by thermal fluctuations. At high frequencies, e.g., at
xs0 ¼ 1 (see Fig. 8), the dynamic hysteresis is due to the
resonant absorption in the FMR band. Thus, the DMH aris-
ing from a high-frequency periodic signal may be evaluated
permitting quantitative analysis of ultrafast switching of the
magnetization. The rotation of the hysteresis loop in Fig.
8(c) is due to the sensitivity of the FMR frequency xpr to the
biaxially parameter (due to this dependence, the phase shift
between MðtÞ and HðtÞ may vary substantially in the vicinity
FIG. 3. �Im½vðxÞ� vs. xs0 for barrier pa-
rameter r ¼ 10, ac field parameter n ¼ 5,
damping a ¼ 0:05, angles w ¼ p=4 and
/ ¼ 0 and various biaxiality parameters
D (a) and dc field amplitudes n0 (b).
FIG. 4. DMH loops at the dimensionless
frequency xs0 ¼ 0:01 for biaxiality pa-
rameter D ¼ 10, damping a ¼ 0:1 and
various values of the barrier parameter r.
Figures (a) and (b) correspond to the ori-
entation of the external fields in the XZ(angle / ¼ 0) or YZ (/ ¼ p=2) planes,
respectively; the oblique angle w ¼ p=4
and the ac field parameter n ¼ 2r.
053903-4 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
of xpr). This phenomenon can be used to determine the biax-
iality factor D for magnetic materials. Clearly, at low fre-
quencies (Fig. 7) the DMH loop area increases as the
biaxiality parameter increases. This is also true when either
the damping or the angle w is decreased. This effect is shown
in Figs. 5–9. Although being relatively weak for r ¼ 10
(Fig. 5(c)), it can become quite pronounced in the superpara-
magnetic regime r ¼ 2 (Fig. 5(a)). To interpret this, two
hypotheses can be made. The first, as already discussed, is
that low damping implies a longer response time. The sec-
ond, which is required to explain the temperature depend-
ence, is that the precession actually takes place about the
direction of the total field, including thermal contributions.
Then the random nature of this noise-induced precession
may hinder the magnetic response, with the consequence
that the area is increased.
FIG. 5. DMH loops at xs0 ¼ 0:01 for
damping a ¼ 0:1, ac field parameter n¼ 10, dc field parameter n0 ¼ 0, angles
w ¼ p=4 and / ¼ 0 and various biaxial-
ity D and barrier r parameters.
FIG. 6. DMH loops at xs0 ¼ 0:01 for barrier parameter r ¼ 5, damping a ¼ 0:1, ac field parameter n ¼ 10, dc field parameter n0 ¼ 0 and various biaxiality
parameters D, / ¼ 0, and angle w ¼ 0; p=4; p=3; and p=2.
053903-5 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
FIG. 7. DMH loops at xs0 ¼ 0:01 for
barrier parameter r ¼ 5, ac field parame-
ter n ¼ 10, dc field parameter n0 ¼ 0,
angles w ¼ p=4 and / ¼ 0, and various
biaxiality parameters D and damping
a ¼ 0:05; 0:1; 1; 10.
FIG. 8. DMH loops for barrier parameter
r ¼ 5, damping a ¼ 0:1, ac field parame-
ter n ¼ 10, dc field parameter n0 ¼ 0,
angles w ¼ p=4 and / ¼ 0, and various
biaxiality parameters D and dimension-
less frequency xs0 ¼ 10�3, 10�2, 1, and
100.
FIG. 9. Normalized area of the DMH loop
An as functions of the biaxiality parameter
D for damping a ¼ 0:1, dc field parameter
n0 ¼ 0, frequency xs0 ¼ 10�2, / ¼ 0,
and various barrier parameters r and
angles w.
053903-6 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
IV. CONCLUSION
A rigorous treatment of the nonlinear susceptibility and
the dynamic hysteresis loops of a single domain particle with
biaxial anisotropy have been given using Brown’s model. In
carrying out all the calculations, the system was initially
assumed to possess a biaxial anisotropy and no approxima-
tions were used except those already inherent in Brown’s
model. The present calculation is the first attempt to treat an
anisotropy other than uniaxial in this manner. We remark
that a detailed treatment of the same problem restricted to
uniaxial anisotropy for both an ac field and dc field of arbi-
trary strength applied at angles to the easy axis of magnetiza-
tion was given by Titov et al.20 and El Mrabti et al.12 They
evaluated the dependence of both the nonlinear dynamic sus-
ceptibility and DMH on the ac and dc field strength, temper-
ature and field angles. In our biaxial anisotropy problem just
as the uniaxial problem above, the intrinsic a dependence of
the response represents a signature of the coupling between
the longitudinal and precessional modes of the magnetization
(Fig. 1). Moreover, again as in the nonlinear response of uni-
axial particles, the FMR peak appears only for low damping
ða� 1Þ and strongly manifests itself at high frequencies.
We emphasize that it should again be possible as in the pre-
ceding case to determine a from measurements of nonlinear
response characteristics, e.g., by fitting the theory to the ex-
perimental dependence of observables on the ac and dc bias
field angles and strengths. As before in such measurements,
the sole fitting parameter is a, which can be determined
at different temperatures T, so yielding its temperature de-
pendence. This is significant because a knowledge of a and
its T dependence allows separation of the various relaxation
mechanisms while in strong ac fields, n> 1, pronounced
nonlinear effects occur.
Numerical calculations of the nonlinear dynamic sus-
ceptibility and the DMH show that a marked dependence of
these quantities on both the biaxiality parameter D and the
azimuthal angle / exists. This phenomenon is a general fea-
ture of nonaxially symmetric systems arising as usual in
from coupling of the longitudinal and transverse relaxation
modes. It is especially important here because unlike uniax-
ial anisotropy the biaxial case has intrinsic azimuthal angle
dependence arising from the D term in Eq. (3). In other
words, the nonaxially symmetric anisotropy generates an azi-
muthally nonuniform energy distribution with saddle points
leading to strong intrinsic dependence of magnetic character-
istics on D arising from the mode coupling due to D. As far
as the DMH is concerned at low frequencies, the DMH loops
exhibit a pronounced damping dependence due to coupling
of the thermally activated magnetization reversal mode to
the precessional modes of the magnetization via the driving
ac field. Again the frequency dependence of the DMH for
fixed angles w and / and temperature is very significant. The
underlying reason is that the remagnetization time is highly
sensitive to the frequency of the applied field. For example,
under a strong ac driving field, the Arrhenius dependence of
the relaxation time on temperature s er, which accurately
accounts for the linear regime, is modified because the strong
ac field drastically reduces the effective response time so
facilitating new re-magnetization regimes, which can never
be attained for weak ac fields. The dc bias component under
the appropriate conditions efficiently tunes this effect either
by enhancing or blocking the action of the ac field. This
description of the dynamic magnetic hysteresis is valid in a
wide field-frequency-temperature range and is free of over-
simplifications such as linear response or high-barrier
approximations so rendering it a useful tool both for verifica-
tion of existing solutions and extension to novel applications
involving magnetic nanoparticles, e.g., low-frequency mag-
netic hyperthermia and precise focusing of the heat gener-
ated by them. Once again, we see that for small ac fields, the
relaxation of the magnetization is mostly caused by thermal
fluctuations, so that the magnetic response time retains a
strong dependence on temperature. Nonaxially symmetric
dynamical hysteresis has valuable potential applications in
magnetic recording because the phenomenon can be used by
industry especially in the field of data storage to characterize
the recording density, signal-to-noise ratio, etc. Clearly, as
the effect of thermally driven magnetization reversal (super-
paramagnetism) is fully accounted for in the discussions we
have given, one would then be able to model the switching
processes for any desirable field-temperature protocol, e.g.,
FIG. 10. Normalized area An of the DMH
loop vs. the biaxiality parameter D for
barrier parameter r ¼ 5, ac field parame-
ter n ¼ 10, dc field parameter n0 ¼ 0,
angles w ¼ p=4 and / ¼ 0, and various
frequencies xs0 (a) and damping a (b).
053903-7 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
heat-assisted or hybrid magnetic recording techniques. The
description of the angular dependence of the dynamic mag-
netic hysteresis which we have obtained could be used to an-
alyze the switching times in magnetic nanoparticles allowing
for both imperfect orientational distributions of the particle
easy axes and nonuniformity of the head field inside the re-
cording layer.
ACKNOWLEDGMENTS
We thank W. T. Coffey, P.-M. D�ejardin, L. M�echernene,
and A. Layadi for useful comments. One of the authors
(B.O.) acknowledges fruitful discussions with H. Aourag,
director of scientific research and technological development
in Algeria. The support of the work by the European Com-
munity (Programme FP7, project DMH, No. 295196) is
gratefully acknowledged.
APPENDIX: MATRIX CONTINUED FRACTIONSOLUTION OF EQ. (5)
As shown in Ref. 20, the stationary ac response can be
calculated from the formally exact matrix continued fraction
solution,
C1 ¼
�
c�21 ðxÞ
c�11 ðxÞc0
1ðxÞc1
1ðxÞc2
1ðxÞ�
0BBBBBBBBBBB@
1CCCCCCCCCCCA¼ �1ffiffiffiffiffiffi
4pp S1 �
�
0
p�1q�1p�10
�
0BBBBBBBBBBB@
1CCCCCCCCCCCA; (A1)
where the infinite matrix continued fraction S1 is defined by
the recurrence equation,
Sn ¼ �½Qn þQþn Snþ1Q�nþ1 ��1;
the elements of the three-diagonal supermatrixes Qn and Q6n
are defined as
½Qn�l;m ¼ dl�1ma2n b2n
d2n�1 a2n�1
� �þ dlm
X2n W2n
Y2n�1 X2n�1
� �
þ dlþ1ma2n b2n
d2n�1 a2n�1
� �;
½Qþn �l;m ¼ dl�1mo d2n
o o
� �þ dlm
Z2n Y2n
o Z2n�1
� �
þ dlþ1mo d2n
o o
� �;
½Q�n �l;m ¼ dl�1mo o
b2n�1 o
� �þ dlm
V2n o
W2n�1 V2n�1
� �
þ dlþ1mo o
b2n�1 o
� �;
(dlm is Kronecker’s delta) and the column vectors ck1ðxÞ, p�1 ,
and q�1 are defined as
ck1ðxÞ ¼
ck2;�2ðxÞ
ck2;�1ðxÞck
2;0ðxÞck
2;1ðxÞck
2;2ðxÞck
1;�1ðxÞck
1;0ðxÞck
1;1ðxÞ
0BBBBBBBBBBBBBBB@
1CCCCCCCCCCCCCCCA
; q�1 ¼
�ffiffiffiffiffiffiffiffiffiffi3=10
pD
02rþ Dffiffiffi
5p
0
�ffiffiffiffiffiffiffiffiffiffi3=10
pD
ðc1 � ic2Þn0ffiffiffi6p
c3n0ffiffiffi3p
�ðc1 þ ic2Þn0ffiffiffi6p
0BBBBBBBBBBBBBBBBBBBBB@
1CCCCCCCCCCCCCCCCCCCCCA
;
p�1 ¼
0
0
0
0
0ðc1 � ic2Þn
2ffiffiffi6p
c3n
2ffiffiffi3p
�ðc1 þ ic2Þn2ffiffiffi6p
0BBBBBBBBBBBBBBBBBBB@
1CCCCCCCCCCCCCCCCCCCA
:
Here o and 0 are zero matrices and vectors of appropriate
dimensions, respectively. The five submatrices Vl, Wl, Xl
Yl, and Zl also appear in the linear response of biaxial par-
ticles and are defined explicitly in Ref. 18. The tridiagonal
submatrices al ; bl, and dl coincide with those for uniaxial
particles and are given explicitly in Ref. 20. For D ¼ 0, the
above solution reduces to that for uniaxial particles.20 Here
ckl;mðxÞ are the Fourier coefficients in the Fourier time series
Eq. (7). Clearly Eq. (A1) contains all the Fourier amplitudes
required for the nonlinear stationary response. Having deter-
mined the amplitude ck1;0ðxÞ and ck
1;61ðxÞ from Eq. (A1), we
can evaluate the magnetization given by Eq. (5).
Thus the nonlinear ac stationary response of biaxial
superparamagnetic particles driven by a strong ac field
applied at an angle to the easy axis of the particle can be
evaluated using matrix continued fractions. The calculations,
since they are valid for ac fields of arbitrary strength, pro-
vide a rigorous theoretical basis for comparison with experi-
ments on nonlinear response in strong ac fields, where the
results of perturbation theory are no longer valid, e.g., non-
linear magnetic susceptibility, nonlinear stochastic reso-
nance, and dynamic hysteresis (these results will be
published elsewhere).
1W. F. Brown, Jr., Phys. Rev. 130, 1677 (1963); IEEE Trans. Magn. 15,
1196 (1979).2C. D. Mee, The Physics of Magnetic Recording (North Holland, Amster-
dam, 1986); Y. Liu, D. J. Sellmyer, and D. Shindo, in Handbook ofAdvanced Magnetic Materials, Volumes I and II (Springer, New York,
2006). A. P. Guimar~aes, Principles of Nanomagnetism (Springer, Berlin,
2009).
053903-8 Ouari et al. J. Appl. Phys. 113, 053903 (2013)
3A. Pankhurst, N. K. T. Thanh, S. K. Jones, and J. Dobson, J. Phys. D:
Appl. Phys. 42, 224001 (2009); L. M. Lacroix, R. Bel Malaki, J. Carrey,
S. Lachaize, M. Respaud, G. F. Goya, and B. Chaudret, ibid. 105, 023911
(2009); J. Carrey, B. Mehdaoui, and M. Respaud, ibid. 109, 083921
(2011); N. A. Usov and B. Ya. Liubimov, ibid. 112, 023901 (2012).4C. P. Bean and J. D. Livingston, J. Appl. Phys. Suppl. 30, 120S (1959).5T. L. Gilbert, Phys. Rev. 100, 1243 (1956); IEEE Trans. Magn. 40, 1243
(2004).6L. D. Landau and E. M. Lifshitz, Phys. Z. Sowjetunion 8, 153 (1935).7W. T. Coffey and Yu. P. Kalmykov, The Langevin Equation, 3rd ed.
(World Scientific, Singapore, 2012).8Y. L. Raikher and V. I. Stepanov, Adv. Chem. Phys. 129, 419 (2004).9Yu. L. Raikher, V. I. Stepanov, and P. C. Fannin, J. Magn. Magn. Mater.
258–259, 369 (2003); Yu. L. Raikher, V. I. Stepanov, and R. Perzynski,
Physica B 343, 262 (2004); Yu. L. Raikher and V. I. Stepanov, J. Magn.
Magn. Mater. 300, e311 (2006).10P. M. D�ejardin and Yu. P. Kalmykov, J. Appl. Phys. 106, 123908 (2009);
P. M. D�ejardin, Yu. P. Kalmykov, B. E. Kashevsky, H. El Mrabti, I. S.
Poperechny, Yu. L. Raikher, and S. V. Titov, ibid. 107, 073914 (2010).11I. S. Poperechny, Yu. L. Raikher, and V. I. Stepanov, Phys. Rev. B 82,
174423 (2010).
12H. El Mrabti, S. V. Titov, P. M. D�ejardin, and Yu. P. Kalmykov, J. Appl.
Phys. 110, 023901 (2011); H. El Mrabti, P. M. D�ejardin, S. V. Titov, and
Yu. P. Kalmykov, Phys. Rev. B 85, 094425 (2012).13G. T. Landi, J. Magn. Magn. Mater. 324, 466 (2012); J. Appl. Phys. 111,
043901 (2012).14W. Wernsdorfer, Adv. Chem. Phys. 118, 99 (2001); M. Jamet, W. Werns-
dorfer, C. Thirion, V. Dupuis, P. M�elinon, A. P�erez, and D. Mailly, Phys.
Rev. B 69, 024401 (2004); C. Thirion, W. Wernsdorfer, and D. Mailly,
Nature Mater. 2, 254 (2003).15N. A. Usov and Yu. B. Grebenshchikov, J. Appl. Phys. 105, 043904 (2009).16D. A. Smith and F. A. de Rosario, J. Magn. Magn. Mater. 3, 219 (1976);
H. B. Braun, J. Appl. Phys. 76, 6310 (1994).17G. Bertotti, I. Mayergoyz, and C. Serpico, Nonlinear Magnetization Dy-
namics in Nanosystems (Elsevier, Amsterdam, 2009), Chapters 9, 10, and
references therein.18Yu. P. Kalmykov and B. Ouari, Phys. Rev. B 71, 094410 (2005).19B. Ouari and Y. P. Kalmykov, J. Appl. Phys. 100, 123912 (2006).20S. V. Titov, P. M. D�ejardin, H. El Mrabti, and Yu. P. Kalmykov, Phys.
Rev. B 82, 100413 (2010).21L. D. Landau and E. M. Lifshitz, Mechanics, 3rd ed. (Pergamon, London,
2000).
053903-9 Ouari et al. J. Appl. Phys. 113, 053903 (2013)