11. W. CHANTRELLand E. P.

WOHLFARTH:
The Field-Cooled Magnetisation 619

phys. stat. sol. (a) 91, 619 (1985)

Subject classification: 18; 22.8.1

School of Physics and Astronomy, Lancashire Polytechnic, Pyestonl) (u)

and Department of Mathematics, Imperial College of Science and Technology, London2)( b )

Rate Dependence of the Field-Cooled Magnetisation

of a Fine Particle System
BY
R. W. CHANTRELL
(a) and E. P. WOHLFARTH
(b)

Dedicated t o Prof. Dr. Dr. h. c. Dr. E. h. P. GORLICHon the occasion of his 80th birthday

A theory is developed of the field cooled magnetisation of a fine particle system. A cooling rate
dependent “blocking temperature” is derived, by means of which the magnetisation of a system
with a distribution of particle sizes can be calculated. Numerical calculations give good qualitative
agreement with experimental results on cobalt aluminosilicate glasses.
Eine Theorie fur die feldabgekuhlte Magnetisierung eines Iileinteilchensystems wird entwickelt.
Eine ,,Sperrtemperatur“ wird in Abhangigkeit von der Abkiihlrate berechnet und damit dic resul-
tierende Magnetisierung eines Systems mit Teilchengrol3everteilung abgeschiitzt. Numerische Be-
rechnungen ergeben eine gute qualitative Ubereinstimmung mit’ MeBergebnissen an Kobaltalu-
minosilikatglasern.

1. Introduction
It is well known that the magnetic properties of a fine particle system are time depend-
ent [l t o 31. This phenomenon arises from thermally activated transitions over the
energy barriers provided by the intrinsic magnetic anisotropy of the particles. The
response of particles of volume V is governed (in the absence of interparticle inter-
actions) by the Arrhenius-NBel relaxation time,

where K is the anisotropy constant and the frequency factor j o = lo9 s-l. By setting t
equal t o the experimental measurement time t a critical volume V , is determined, given
by :
= 14 tf, .
Because of the rapid variation of z with volume, V , defines the transition to thermal
equilibrium (superparamagnetic) behaviour as the particle size decreases.
Since all systems contain a distribution of V and hence t,a logarithmic time de-
pendence of the magnetisation is generally observed, resulting from the summation
of many exponential variations. The time dependence is usually observed by measuring
the vasiation in the magnetisation as the external parameters ( H , T,etc.) are held
constant after an abrupt change, usually in H . The field-cooled magnetisation (FCM)

1) Preston P R 1 2TQ, Great Britain.

2, Queen’s Gate, London SW7 2BZ, Great Britain.
620 R.W. CHANTRELLand E. P. WOHLFARTH

is obtained by a continuous reduction in the temperature with the system exposed

to a constant field H . Thus the PCM is achieved by a dynamic process, in which
therelaxational behaviour of the particles might be expected to play a significant role.
Consequently, the FCM should depend upon the rate a t which the temperature is
reduced. Experimentally such behaviour has been observed by Wenger and Mydosh [a]
in insulating spin glasses. These systems are a t least qualitatively similar to a fine
particle system in that the relaxational behaviour arises from thermally activated
transitions over local energy barriers.
Clearly the critical volume defined earlier is not applicable t o the calculation of the
FCM. In this paper we present a general expression for the FCM of a fine particle sys-
tem with a single volume. A cooling rate dependent critical temperature is then defined
and used in numerical calculations for a system with a distribution of particle volumes.
Finally, a qualitative comparison is made with the experimental data of Wenger and
Mydosh [4].

2. The FCM for a Single Particle Volume

We first calculate the FCM for a system with a single particle volunie V , assuming
the particles t o have a uniaxial anisotropy, with an anisotropy constant K . We
assume further that the particle easy axes are aligned parallel to the applied field H .
The anisotropy energy K V tends to constrain the magnetic moments t o lie within one
or other of the energy minimum parallel or antiparallel to H . Thus we can approximate
the particles by a two-energy-level system with populations n, and n2governed by the
differential equation

where z1 and z2 are the relaxation times for thermally activated transitions from
states 1 to 2 and 2 to 1, respectively, which in general will be different.
Solution of (3) leads to the following equation for the time variation of the magneti-
sation :

where r +
is the reduced magnetisation (relative t o saturation) and t - l = ti1 tgl;
-.
re
I, is the magnetisation a t time t = 0. The equilibrium inagnetisation is achieved
at time t = 03. Setting this value of t in (4)wre obtain

as the equilibrium magnetisation, and substitution into (4)gives

-~ - -
I = I, + ( I ~- I,lexp(-:). (5)

We now apply ( 5 ) to the calculation of the FCM. The field-cooling process is approxi-
mated by a stepwise reduction of the temperature in small increments A T from a n
initial temperature To t o a final temperature T,, the system remaining a t the given
temperature for a small time At. Within each temperature increment the magnetisation
is given by a n expression of the form of (5) and after i increments (corresponding to
a temperature Tf = To - i A T ) the reduced magnetisation is
Field-Cooled Magnetisation of a Fine Particle System 621

where ?,, is the equilibrium magnetisation corresponding to the temperature T,.

Applying (6) to each increment in turn gives the following general expression for I,:

- - -
where AI, = I,j - Equation (7) is valid as long as the system is in an equil-
ibriurli state a t the initial temperature To. We now write (7) as

For a continuous process, i tends to infinity so that At and AT tend to zero. In this
limit,
T. ’1’

(9)

where T = d T / d t is the rate a t which the temperature is reduced; d?/dT is the teni-
perature differential of the equilibrium magnetisation.
For a system with aligned easy axes the equilibrium magnetisation is given by
7 = tanhp, (10)
where ,B = pH/kT with p the magnetic moment per particle.
The resultant expression for the FCM is

f ( T , ) = tanh pf - -
;f l’f
sech2p exp (- T-l
i
2’1
t-l dT’) .

If the r;ysteni has the particle easy axes randomly oriented (10) is no longer applicable.
(11)

It can be shown that in small fields the magnetisation has the same value as the Lange-
vin function,
-
I = coth p - p-‘ .
As a first approximation we shall use the Langevin function to represent the equi-
libriiini niagnetisation of a system with randonily oriented easy axes.

3. Rate Dependent Critical Temperature

In order to simplify the calculation of the FCM of a system with a distribution of
particle volumes we now introduce the concept of a cooling-rate-dependent critical
volume, which leads to a considerable simplification of (9). We first write (9) in the
form

where
T
J = J t-l d T ’ .
T€
622 R. W. CHANTRELLand E. P. WOHLFARTH

Due to the rapid variation of T with T a useful approximation to J can be determined.

The relaxation time T is determined by the Arrhenius-Xkel law,

where A E l , are the energy barriers to rotation, which for uniaxial particles are given
by

where HI<= 2K/Ibbis the anisotropy field, I,I, being the saturation niagnetisation of
the bulk inaterial of which the particles are composed. We now make the simplifying
assumption that transitions out of the field direction are negligible (zl is small) so
that
7-l = f o exp [--n(l - h ) 2 ], (14)
where LY = KVIkT and h = H / H A . Therefore, the exponential factor in (14) will be
incorrect by a small factor in small fields. However, due to the magnitude of
f o (- lo9 s-l), itself a poorly known quantity, our results are insensitive t o this ap-
proxiniation.
Since f o is large we have the greatest interest in the evaluation of (13) for large
values of 01, since otherwise ~ - isl large, giving J 3 1 and (from (12)) y(T,) = Fe(Tf)
i.e. the system remains in thermal equilibrium. When AE > k T the decrease in t - l
as the temperature is decreased from T to T f is very rapid and most of the variation
can be considered as taking place a t temperatures close t o T . Thus we approximate
the integral by expanding a(1 - h ) 2 about T and retaining the first term in the ex-
pansion. This gives (with AT = T - T,)
AT

The variation of J will generally be doiriinated by the first exponential factor, and
since iiiost of the variation takes place close to T , J is insensitive to the actual value
of the upper limit AT. We thus obtain a useful approximation to J by setting AT =
= co. This gives
T
J=fo exp [--ol(l - (15)
a ( l - h)2
From (12)) I(T,) is governed by the behaviour of exp (-?-V) which will have
a very rapid variation with temperature. Thus we define a rate-dependent critical
temperature TB a t which T - l J = 1. Then, to a good approximation, for T > TB,
exp (-@-'J) = 0 and for T <
TB, exp ( - @ - l J ) = 1. Inspection of (12) then shows
>
that if T f TB, then ? ( T f )= Te(Tf),i.e. the system remains in thermal equilibrium.
However, if TB >T,,
TB
I(Tf)= Fe(Tf)+
-
J dl = je(TB)
Tt

Clearly, on passing through the critical temperature T Bthe equilibrium niagnetisation

is frozen in.
Field-Cooled Magnetisation of a Fine Particle System 623

We now make an estimate of T , in a manner which is consistent with the well-

known definition of the blocking temperature in the case of isothermal nieasurements.
Setting T - V = 1 in (15) gives

with ac = KV/kT,. The blocking temperature ( T I Cfor

) quasi-static isothermal meas-
urements in zero field is usually taken to be given by KV = 25kT,. Using this
definition (16) can be written in the form

I n order t o make our definition of T , similar to that of the isothernial blocking tem-
perature we choose a reference rate of reduction so that 2, = a,(l - h ) 2 = 25. This
reference rate is then equivalent to the assumed measurement time of t = 100 s for
quasi-static measurements in the isothermal ease. This criterion gives a value of
P I T K ( l- h ) 2 = 1.8 x and from (17) we then have

a,(l - h)2 = 25 - In
from which it follows that

I n zero field for TK = 100 K the reference rate of reduction corresponds to T =

== 0.05 K s-l, i.e. quasi-static nieasureiiients.
It is now possible to construct a FCM curve from (18) by assuming a transition t o
irreversible behaviour a t T,. Thus 7 = Fe(T)for T >
T,, and = &(T,) for T <
< T,. This approximation is shown in Fig. 1 (curve 2 ) for T T c l = 5.5 x 10-4 and
a reduced field of h = 0.01. Also shown in Fig. 1 (curve 1) is the exact solution,
calculated numerically from (16). The approximate curve can be seen t o be a good
representation of the exact relation. The effect of increasing T by a factor of 104 is
also shown in Fig. 1. Even a t such a rapid rate of reduction the approximation is
stsillin reasonable agreement with the exact calculahion.

52

36 Fig. 1. Comparison of the exact and approx-

3 imate calculations of the FCM for a single
particle volume. Cooling rates of 5.5 x 10-4
and 5.5 are shown. ( 1 ) T = 5.5 x (ex-
28
act), (2) 5.5 x (approx.), (3) 5.5 (exact),
(4)5.5 (approx.)
10 15
rjm - 20
624 R. W. CHANTRELLand E. P. WOHLFBRTH

Equation (18) can be used to calculate the PCM curve for a single paIticle size.
However, in general a system will contain a distribution of particle sizes and, conse-
quently, blocking temperatures. I n the following section we generalise the treatment
by the introduction of a particle size distribution.

4. Calculation of the Field-Cooled Magnetisation

In order to calculate realistic FCM curves, account must be taken of the distribution
-
of particle size. Since Tp V we here represent the size distribution in terms of
a distribution of T, which can be thought of as an intrinsic blocking temperature.
We assume here a lognormal distribution since this is often found to be appropriate
for fine particle systems. Thus the fraction of the (magnetic) volume having intrinsic
blocking temperatures between T, and T, + dT, is

where T,, is the median value of the distribution and u is the standard deviation. It
is convenient t o use a temperature parameter T, normalised with respect to T,,
which allows us t o write (18) in the form
T m T r (1 - h)2 = 1 -
T,
where T, = T,/T, and T r = d(T/T,,,)/dt.
The net reduced magnetisation is the sum (weighted by the distribution function)
of the equilibrium and non-equilibrium contribution, thus,

where T B , is the blocking temperature corresponding to a reduced temperature Tr

and T, = T,/T,,. We now illustrate the variation of the FCM by numerical calcula-
tions based on (21). Fig. 2 shows a family of FCM curves for different values of the
reduced field h. The effect of distribution of T , is t o smooth the FCM curve as can
be seen by comparison of Fig. 1 and 2. I n order t o make a qualitative comparison of
the theory with the experimental data of Wenger and Mydosh [4]we have also cal-
culated FCM curves for different rates of temperature reduction.Typica1 results are
shown in Pig. 3 for a range of T,.

Fig. 2 . The dependence of the field-cooled

magnetisation on the reduced temperature
for several values of the applied field. The
standard deviation is 0 = 0.3. (1)h = 0.01,
(2) 0.05, (3) 0.1, (4)0.2 (in u n i t s of H K )
Field-Ccloled Magnetisation of a Fine Particle System 825

Fig. 3. The rate dependence of the FCM cur-

ves for areducedfieldofh = 0.01 and a = 0.3.
( 1 ) T,= 5.5 x 10-4, (2) 10-3, (3) 5 x 10-3,
(4)5 x 10-2, ( 5 ) 1

The predicted theoretical curves are qualitatively similar to the experimental data
of Wenger and Mydosh [4].It is, however, not meaningful to attempt a fit between
theory and experiment a t this time due to the large number of adjustable parameters
involved. In this context it i s interesting to study the variation of magnetisation with
cooling rate a t a given temperature in a fixed field.
Consider the change in blocking temperature (and consequently the magnetisation)
when the cooling rate is increased from the “reference” rate given earlier to a value
of T . Erorn (18) it is straightforward t o show that as long as T is not too large, the
blocking temperature increases to

and T,(O) = TK(l- h)2is the blocking temperature corresponding to the reference
rate of hemperatwe reduction. The change in the cooling rate thus results in a higher
blocking temperature and a consequently smaller value of the frozen magnetisation
for particles with a given intrinsic blocking temperature T,. The new value of the
FCM i s

which can be written in the form

Td!h
where FCM(T, 0) is the value of the FCM corresponding to the reference rate of teni-
perature reduction. If the change in the blocking temperature is small the first two
integraltr of (24) will be negligible in which case

40 physica (a) 91/2

 626 and E. P. WOHLFARTH:
R. W. CHANTRELL The Field-Cooled Magnetisation

Thus we expect a logarithmic dependence of the PCM on the cooling rate, with

Te(ii.)

The negative sign is included explicitly, rernenibering that d f / d T is negative. The

logarithmic dependence on the cooling rate is consistent with the logarithmic time
dependence which is observed experimentally in fine particle systems. In each case
the logarithmic behaviour arises from the superposition of many exponential varia-
tions. It is interesting to note, however, that the rate dependence calculated here has
contributions fiom all particles with blocking temperatures greater than the measure-
ment temperature. This contrasts with the time dependence which arises froin those
particles having blocking temperatures in a narrow range slightly above the measure-
ment temperature. This is a consequence of the fact that the cooling process accesses
the whole range of blocking teniperatures whereas the time dependence arises predoni-
inantly from those particles with relaxation times of the order of the time of nieas-
ureinent .

5. Conclusions
We have investigated the rate dependence of the field-cooled niagnetisation of a fine
particle system. The rate dependence arises from the finite relaxation times of the
particles. An approximation, based on a rate-dependent “blocking temperature” has
been derived and used to calculate PCM curves for systems with a distribution of
blocking temperatures. The theoretical curves are qualitatively similar t o the ex-
periniental data of Wenger and Mydosh [4]on two insulating spin glasses; essentially
cobalt aluininosilicate glasses with different cobalt concentrations. It might be thought
inappropriate to apply our model to a spin glass. However, any system with a distri-
bution of energy barriers will have a corresponding spread in the relaxation times of
the magnetic moments. This will result in a distribution of T,, the basis of the present
model.
A t present it is not meaningful to attempt a qualitative comparison with the ex-
perimental data because of the large nuinher of adjustable parameters involved. It is
suggested that further nieasiirenients should be made on well-characterised fine par-
ticle systems for which the important parameters can be measured independently.
It would be of some interest to investigate the effects of interparticle interactions on
the FCM. We have recently [ 5 ] developed a model of the anhysteretic renianent mag-
netisation (ARM) which shows that in a strongly interacting system the dynaniics
of the process have an important bearing on the final state attained. Such an effect
should also be present in the FCM, perhaps manifested as a more complex rate depend-
ence than predicted here.

References
[l] R. STREETand J. C. WOOLLEY, Proc. Phys. Soc. A62, 562 (1949).
[2] A. T. CAYLESS, S. R. HOON,B. K . TANNER,R. W.CHANTRELL, and If. KILNER,J. Magnetism
magnetic Mater. 30, 303 (1983).
[3] E. P. WOHLFARTH, J. Phys. F 14, L155 (1984).
[4] L. E. WENGERand J. A. MYDOSH, J. appl. Phys. 66, 1717 (1984).
[5] A. LYBERATOS, E. P. WOHLFARTH, and R . W.CHANTRELL, IEEE Trans. Magnetics, 21, 1277
(1985).
(Receiced August 21, 1985)