Journal of Magnetism and Magnetic Materials 252 (2002) 370374

Heating magnetic uid with alternating magnetic eld

R.E. Rosensweig*
Exxon Research and Engineering Co. (ret.), 34 Gloucester Rd., Summit, NJ 07901, USA

Abstract

This study develops analytical relationships and computations of power dissipation in magnetic uid (ferrouid)
subjected to alternating magnetic eld. The dissipation results from the orientational relaxation of particles having
thermal uctuations in a viscous medium.
r 2002 Elsevier Science B.V. All rights reserved.

Keywords: Magnetic uid; Ferrouid; Heating; Medical hyperthermia

1. Introduction result is
dU H  dB; 1
It appears that heating of colloidal magnetic uid
1
(ferrouid) due to time-varying magnetic induction has where H (A m ) is the magnetic eld intensity and B (T)
not been signicantly studied. Heating ferrouid to the induction, both in the sample. Because the elds are
achieve hyperthermia in medical treatments is an colinear the relationship reduces to dU H dB where
emerging area of importance [1,2]. Heating effects in H and B are magnitudes. B m0 H M; where M
ferrouids may also be important in loudspeakers where (A m1) is the magnetization and m0 4p  107
temperature rise adversely affects performance [3]. Other (T m A1) is the permeability of free space. Substitution
applications exploit the ability to heat the uid into Eq. (1) followed with integration by parts shows that
magnetically, as in the thermal actuation of a polymer the cyclic increase of internal energy can be written as
I
gel in contact with a ferrouid.
This work develops dissipation relationships based on DU m0 M dH: 2
rotational relaxation of single domain magnetic particles
dispersed in a liquid matrix. Eddy current heating is When magnetization lags the eld, the integration yields
assumed negligible due to the small size of the particles a positive result indicating conversion of magnetic work
(o15 nm). The SI system of units is employed. to internal energy. It will be convenient to express the
magnetization in terms of the complex ferrouid
susceptibility w w0  iw00 : Then with an applied mag-
netic eld of the form
2. Power dissipation
Ht H0 cos ot ReH0 eiot
; 3
From the rst law of thermodynamics for a the resulting magnetization is
constant density system of unit volume dU dQ
dW ; where U is the internal energy, Q the heat Mt RewH0 eiot
H0 w0 cos ot w00 sin ot; 4
added and W the magnetic work done on the system. where it is seen that w is the in-phase component, and w00
0
For an adiabatic process dQ 0 with the differential the out-of-phase component of w. Substituting for M and
magnetic work, given in general [4] by dW H  dB; the H in Eq. (2) from Eq. (4) leads to
Z 2p=o
*Tel.: +1-908-277-2846; fax: +1-908-277-1084. DU 2m0 H02 w00 sin2 ot dt; 5
E-mail address: rerosen@comcast.net (R.E. Rosensweig). 0

0304-8853/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 8 8 5 3 ( 0 2 ) 0 0 7 0 6 - 0
 R.E. Rosensweig / Journal of Magnetism and Magnetic Materials 252 (2002) 370374 371

where only the component w00 survives, hence its name,

the loss component. Integrating and multiplying the
result by the cyclic frequency f o=2p gives the
following expression for volumetric power dissipation:
P f DU m0 pw00 fH02 : 6

To make use of this result w00 must be related to material

parameters of the ferrouid.

3. Relationship to material parameters

The relaxation equation of Shliomis [5] reduces to the

following form for motionless uid in an oscillatory
eld:
qMt 1
M0 t  Mt; 7
qt t 4:45 nm.
where t is the relaxation time, M0 w0 H0 cos ot
Rew0 H0 eiot is the equilibrium magnetization in the
applied eld whose value is given in Eq. (3), and w0 is the
equilibrium susceptibility. Substituting the complex
representations of M0 and Mt into Eq. (7) yields
w0
w ; 8 rotates as well. A second mechanism exists (N!eel
1 iot
which gives the dependence of complex susceptibility on
frequency and from which the components of suscept-
ibility are
w0
w0 ; 9a
1 ot2
ot
w00 w0 : 9b
1 ot2
These relationships are identical to the Debye spectra of
polar molecules in the absence of a constant eld [6].
The spectra for monodisperse particles are plotted in
Fig. 1 which shows that the real and imaginary
components have values that cross at the peak value
of w00 : The data of Fanin [7] display this trend. As it
follows from Eq. (9b), the predicted cross-over occurs at
ot 1: Also shown for comparison is the computed
spectra for a polydispersion of magnetic particles
assuming a log normal distribution of particle size with
standard deviation s 0:1 as discussed in Section 5. The
polydisperse curves are shifted greatly showing why
measurements of actual magnetic uids, which are
nearly all polydisperse, vary from the monodisperse
predictions.
Ferromagnets exhibit magnetic resonance at frequen-
cies B1081010 Hz yielding a change of sign to negative
value of w0 o and a sharp peak of w00 o [8]. Here tD Gt0 :
interest is limited to the megahertz frequency range and
below; hence, no further discussion is devoted to the
higher frequency aspects.
Fig. 1. Susceptibility components of monodisperse particles
compared to that of a polydispersion. B0 0:06 T; R
4. Time constants
With the Brownian mechanism of relaxation the
magnetic moment is locked to the crystal axis and when
the magnetic moment aligns with the eld, the particle
relaxation) in which the magnetic moment rotates within
the crystal. To achieve high heating rates the N!eel
relaxation must not be allowed to dominate.
The Brownian time constant is given by the following
relationship [9]:
3ZVH
tB ; 10
kT
where Z is the viscosity coefcient of the matrix uid, k
the Boltzmann constant (1.38  1023 J K1), and T the
absolute temperature (K). VH is taken as the hydro-
dynamic volume of the particle which is larger than the
magnetic volume VM 4pR3 =3 for a particle of radius
R: As a model for VH ; it is assumed that VH 1
d=R3 VM ; where d is the thickness of a sorbed surfactant
layer.
The N!eel relaxation time, denoted tN ; is given by the
following expression due to Brown [10],1 where G
KVM =kT with K as the anisotropy constant which may
be of magnetocrystalline or shape origin:
p
p exp G
tN t0 1=2 11
2 G
or in the equivalent form
p
p exp G
tN tD 3=2 12a
2 G
12b
1
The author thanks M.I. Shliomis for pointing out the
alternate expressions.
372 R.E. Rosensweig / Journal of Magnetism and Magnetic Materials 252 (2002) 370374
Because t0BK1 while tD BVM ; Eq. 11 is used when
determining the size dependence of tN ; where VM enters
only in G: Alternatively, when considering how tN
depends on K; Eq. (12) is used where K is contained
only in G:
A typical ferrouid has a broad distribution of
particle sizes with a mean size of B10 nm. Because the
Brownian and N!eel processes take place in parallel, the
effective relaxation time t is given by
1 1 1
: 13
t tB tN
The plot shown in Fig. 2 illustrates that the shorter time
constant tends to dominate in determining the effective
relaxation time for any given size of particle.
From Eqs. (6) and (9b) the power dissipation is
expressed as
2pf t
P pm0 w0 H02 f : 14
1 2pf t2
Eq. (14) expresses the power dissipation density for a
monodispersion, assuming susceptibility is constant.
Actual susceptibility w0 is magnetic eld dependent. As
a conservative estimate giving a lower bound of P; it is
assumed that w0 is the chord susceptibility correspond-
ing to the Langevin equation Lx M=Ms
coth xF1=x where x m0 Md HVM =kT: Ms fMd is
saturation magnetization of the ferrouid, Md the
domain magnetization of a suspended particle, and f
the volume fraction solids. Thus,
 
3 1
w0 wi coth x  ; 15
x x
where the initial susceptibility is wi qM=qHi
m0 fMd2 Vm =3kTas determined from differentiation of
Fig. 2. Time constants vs. particle size for magnetite particles.
the Langevin relationship. Temperature rise for the
monodispersion is computed as DT PDt=c; where c is
the ferrouid specic heat and Dt is the duration of the
heating. Specic heat is calculated as the volume average
of solid and liquid constituents.
5. Polydispersions
It is found that the log normal particle size distribu-
tion gR provides a reasonably good t to the measured
distribution for ferrouids [11]. gR is obtained by
substituting lnR for R in a Gaussian distribution
function, thereby circumventing the unphysical aspect of
contributions from negative particle sizes associated
with the Gaussian:
 
1 lnR=R0 2
gR p exp ; 16a
2psR 2s2
Z N
gR dR 1; 16b
0
where ln R0 is the median and s the standard deviation
of ln R: The most probable radius is given by
R0 exps2 =2 and the mean radius is R0 exps2 =2:
Numerical integration over the distribution function
transforms P of Eq. (14) to P; % the volumetric heat
release rate of a polydispersion:
Z N
P% PgR dR: 17
0
The corresponding temperature rise of an adiabatic
sample is DT% P% Dt=c:
6. Calculated results
The integrand of Eq. (17) is a huge expression when
the expressions for gR; P; w and t are substituted in it,
and analytical integration is not desirable. Instead,
numerical values of temperature rise are calculated as
examples. These are based on physical property values
listed in Table 1. A value of t0 109 s is employed
throughout.
Fig. 3a illustrates the heating rate for monodisperse
magnetite samples in a hydrocarbon carrier as a
function of particle radius and eld intensity
(B0 m0 H0 ). Very substantial heating rates are achiev-
able. An optimum particle size yielding maximum
heating exists that is nearly independent of the applied
eld intensity. Fig. 3b shows the inuence of carrier
viscosity on the heating rate with frequency as para-
meter for a sample with a xed particle size. It is seen
that heating rates peak at viscosity values lower than the
usually available ones.
 R.E. Rosensweig / Journal of Magnetism and Magnetic Materials 252 (2002) 370374 373

Table 1
Physical properties of magnetic solids

Magnetic solid Chemical formula Md (kA m1) K (kJ m3) c (J kg1 K1) r (kg m3)

Maghemite g-Fe2O3 414 4.6 B746 4600

Magnetite FeO  Fe2O3 446 2341 670 5180
Cobalt ferrite CoO  Fe2O3 425 180200 700 4907
Barium ferrite BaO  6Fe2O3 380 300330 B650 5280

Fig. 3. Temperature rise rate in monodisperse ferrouid based on magnetite with f 0:071: (a) Tetradecane carrier, frequency
300 kHz; (b) Magnetic induction 0.06 T, R 7 nm.

Fig. 4. Temperature rise rate in polydisperse ferrouid based on magnetite with f 0:071: (a) Tetradecane carrier, f 300 kHz;
(b) B0 0:06 T, f 900 kHz, R 7 nm.

The degradative inuence of polydispersity is illu-
strated in Fig. 4 for magnetite particles in hydrocarbon
carrier, showing the incentive to utilize highly mono-
disperse samples. The calculations are done for a
tetradecane carrier uid having specic heat of
2080 J kg1 K1, mass density 765 kg m3 and viscosity
0.00235 kg m1 s1. Surfactant layer thickness is set at
2.0 nm.
Fig. 5 compares heating rates for monodispersions of
the various magnetic solids listed in Table 1 assuming
t0 109 s. The carrier liquid is tetradecane in all cases.
The main differences between the performance of the
resultant ferrouids are due to the anisotropy constants
and the domain magnetizations. From the gure it can
be seen that barium ferrite and cobalt ferrite yield the
largest heating rates in the size range of typical
374 R.E. Rosensweig / Journal of Magnetism and Magnetic Materials 252 (2002) 370374
Fig. 5. Comparative heating rates for various magnetic solids.
1Barium ferrite, 2Cobalt ferrite, 3Magnetite, 4
Maghemite. (f 0:071; Z 0:00235 kg m1 s1, f 300 kHz,
B 0:09 T, d 2 nm).
ferrouids (RB425 nm). The highest heating rates are
provided by magnetite and maghemite; however, the size
range where this occurs is larger than that found in
stable ferrouids.
7. Conclusion
This study utilizes a collection of well-known princi-
ples in magnetic uid science to formulate and compute
heating rate in samples subjected to an alternating
magnetic eld. The methodology should be of interest to
investigators in numerous elds of technology.
Acknowledgements
The author thanks M. Zahn for helpful comments on
this work.
References
[1] A. Jordan, P. Wust, H. Fahling, W. Hohn, A. Hinz, R.
Felix, Int. J. Hyperthermia 9 (1) (1993) 51.
[2] R. Hiergeist, W. Andra, N. Buske, R. Hergt, I. Hilger, U.
Richter, W. Kaiser. J. Magn. Magn. Mater. 201 (1999) 420
(volume issued out of sequence).
[3] K. Raj, in: B. Berkovski (Ed.), Magnetic Fluids and
Applications Handbook, Begell House Inc., New York,
1996, pp. 716724.
[4] J.A. Stratton, Electromagnetic Theory, McGraw-Hill,
New York, 1941, p. 124.
[5] M. Shliomis, Sov. Phys. Uspekhi (Engl. transl.) 17 (2)
(1974) 153.
[6] P. Debye, Polar Molecules, The Chemical Catalog Co.,
Reprinted by Dover Publications, New York, 1929.
[7] P.C. Fanin, in: B. Berkovski (Ed.), Magnetic Fluids and
Applications Handbook, Begell House Inc., NewYork,
1996, p. 170.
[8] Y.L. Raikher, M.I. Shliomis, in: W. Coffey (Ed.),
Relaxation Phenomena in Condensed Matter, Advances
in Chemical Physics Series, Vol. LXXXVII, Wiley, New
York, 1994, pp. 596751.
[9] J. Frenkel, The Kinetic Theory of Liquids, Dover
Publications, New York, 1955.
[10] W.F. Brown, Jr., Phys. Rev. 130 (1963) 1677.
[11] R.W. Chantrell, J. Popplewell, S.W. Charles, IEEE Trans.
Magn. MAG-16 (1978) 975.