Phys Chem Minerals (1999) 26: 591±600 Ó Springer-Verlag 1999

ORIGINAL PAPER

H.J. Bray á S.A.T. Redfern

Kinetics of dehydration of Ca-montmorillonite

Received: 28 August 1998 / Revised, accepted: 27 January 1999

Abstract Isothermal thermogravimetric experiments important uses as a physical seal and chemical absorbent
have been carried out to determine the reaction kinetics in chemical and nuclear waste management. In this
of the dehydration processes in fuller's earth, a natural context it is important to understand the characteristics
Ca-montmorillonite. Dehydration in swelling clays is a and controlling reaction mechanism of high-temperature
complex reaction, and analysis of the thermogravimetric transformations in these swelling clays.
data using empirical rate equations and time-transfor- Fuller's earth is the term used to describe a Ca-
mation analysis reveals that the nature of the rate montmorillonite with signi®cant swelling capactities.
controlling mechanism is dependent upon both the The structure comprises a dioctahedral 2:1 layer, readily
temperature regime of the sample as well as the extent of taking up water and organic molecules between the
reaction. For fuller's earth, we ®nd that the dehydration aluminosilicate sheets. Substitution in the central octa-
kinetics are dominated by a nucleation and growth hedral sheet (Fe2+ and/or Mg2+ for Al3+) and the outer
mechanism at low temperatures and fractions trans- tetrahedral sheet (Al3+ for Si4+) causes a negative
formed (stage I), but above 90 °C the last stages of the charge on the outer basal oxygens. Interlayer cations
reaction are di€usion controlled (stage II). The activa- balance this charge, forming hydration spheres and
tion energy for dehydration during stage I is around (together with interlayer molecular H2O) force the
35 kJ á mol)1, whereas the removal of water during crystalline layers apart. Dehydration involves the re-
stage II requires an activation energy of around moval of interlayer water up to 350 °C, followed (at
50 kJ á mol)1. These two stages of dehydration are as- higher temperatures) by dehydroxylation as water mol-
sociated with primary collapse of the interlayer (stage I) ecules are formed from hydroxyls in the octahedral
and movement of water that is hydrated to cations sheet.
within the interlayer (stage II). The hydration and dehydration reactions of mon-
tmorillonite are important in many geological processes
Key words Montmorillonite á Fuller's earth á and have been linked to phenomena as diverse as sedi-
Dehydration á Kinetics á Thermogravimetry ment overpressuring (Powers 1967; Barker 1972; Ma-
gara 1975; Plumley 1980; Bruce 1984), the migration of
petroleum (Powers 1967; Burst 1969), listric faulting
Introduction (Bruce 1984) and the well-known smectite to illite tran-
sition (Powers 1967; Perry and Hower 1972; Weaver and
The high-temperature dehydration transformations of Beck 1971; Hower et al. 1976; Freed and Peacor 1989;
fuller's earth are of great importance in ceramic tech- Pytte and Reynolds 1988; Velde and Vasseur 1992;
nology, since this and other related clays are typically Ransom and Helgeson 1995).
major components of commercial clay bodies. Temper- The temperatures and characteristics of dehydration
atures encountered during the production of ®red ce- reactions in the swelling clays provide important infor-
ramics can reach as high as 1100 °C, and the drying mation about the interlayer con®guration, since the de-
process of the clay beforehand greatly a€ects the phases hydration process depends on the composition of the
formed on ®ring. In addition, this natural material ®nds counterbalancing cations in the interlayer and charge on
the tetrahedral sheet. Experimental studies by Koster
van Groos and Guggenheim (1984, 1986, 1987) have
H.J. Bray á S.A.T. Redfern (&)
demonstrated that montmorillonite dehydrates in two
Department of Earth Sciences, University of Cambridge, stages. The two dehydration steps were interpreted as
Downing Street, Cambridge, CB2 3EQ, UK dehydration of voluminous, but weakly bonded, water
592

from the outer hydration shell of the interlayer cation, octahedral substitution (Mg2+ and/or Fe2+ for Al3+),
followed by dehydration of water from a more strongly- is more di€usely spread over the surface oxygens.
bonded inner hydration shell environment. It is well Finally, the end of dehydration is accompanied by
known that dehydration depends on interlayer cation collapse of the interlayer region, but the 2:1 tetrahe-
chemistry. For example, the radius, charge and hydra- dral:octahedral sheets remain intact. At this stage, the
tion energy of counterions can control dehydration interlayer cation is thought to either reside in the ditri-
temperatures. Dehydration may also be in¯uenced by gonal cavities of the outer tetrahedral sheet or migrate
other factors, for example the location of the layer- into the octahedral sheets (Heller-Kallai and Rozenson
charge and the clay particle size (Annabli-Bergaya et al. 1980).
1996). Many kinetic studies assume that the dominating
The kinetics of the smectite to illite transition have reaction processes is a single rate-controlling process
been investigated extensively in earlier studies, with the throughout all stages of the reaction. Furthermore, it is
estimated activation energy lying between 12.6 kJ á often accepted that the associated activation energy
mol)1 and 125.6 kJ á mol)1 (Eberl and Hower 1976; remains constant throughout the transition. This inves-
Robertson and Lahann 1981; Howard and Roy 1985; tigation seeks to test these assumptions, and shows how
Huang et al. 1994). There have been fewer attempts the rate-controlling process changes throughout the
to probe thoroughly the kinetics of dehydration of reaction and the temperature regime chosen. The ln
montmorillonites. Girgis et al. (1986) conducted non- `time to' method of kinetic analysis of isothermal data,
isothermal experiments on a series of naturally occurring described by Burke (1965), has been employed in order
predominantly Ca-smectites, they concluded that dehy- to test the validity of these assumptions and determine
dration follows a ®rst order reaction with values ranging any variations in the energetics of the reaction.
from 39.8 kJ á mol)1 to 52.3 kJ á mol)1. No discrimina-
tion was made between interlayer water that was
physically adsorbed and water attached in hydration Experimental procedures
shells of the exchangeable cations. Furthermore, GuÈler
and Sarier (1990) calculated a value of 14.72 kJ á mol)1, The sample selected for study was a naturally ocurring Ca-mon-
assuming that a ®rst order reaction occurs, following the tmorillonite from Redhill, UK, kindly provided by Dr. S. Drach-
man of RBB Ltd. X-ray di€raction of the sample revealed it to be a
work of Murray and White (1949, 1955a, b, c, d). smectite-rich clay with an initial 001 d-spacing of approximately
However, it should be noted that Murray and White 15 AÊ. The structural formula of the <2 lm fraction was deter-
studied dehydroxylation, rather than interlayer mined as Ca0.59K0.38Na0.13(Si7.72Al0.28) (Al2.71Fe2‡
0:50 Mg0.57Ti0.08)
dehydration. Waclawska (1984) described a zero-stage O20(OH)4 by X-ray ¯uorescence spectroscopy.
The sample was initially ground by hand in a pestle and mortar
process with activation energies ranging from 15.1 to and dry sieved below 53 lm, to further purify the clay fraction.
26.4 kJ á mol)1, but failed to provide any details of the Preliminary mass loss and di€erential scanning calorimetry (DSC)
physical nature of the dominating reaction mechanism. were carried out on a loosely packed sample placed in a platinum
More recently, computer simulations of the molecu- crucible in an atmosphere of air. A Netzsch STA 409 simultaneous
lar structure of water and the distribution of the thermal analysis instrument was used, employing alumina as a
reference. The data were obtained at a constant heating rate of
counterions in the interlayer region have extended our 10 °C/min, giving clear DSC data that can be correlated to mass
understanding of the structure of these clays (Chang loss events collected from the thermobalance (Fig. 1).
et al. 1995; Skipper et al. 1989, 1991, 1995a, b; Bridge- The <2 lm fraction was separated following a method similar
man et al. 1996; Karaborni et al. 1996). Good agreement to that of Tang and Sparks (1993): 50 g of clay was dispersed in 2 l
was found between the enthalpies of swelling clays deter-
mined from Monte Carlo and molecular dynamic sim-
ulations (De Siquieira et al. 1997) and enthalpies derived
from experimental thermodynamic data (Greene-Kelly
1955; Low, 1979; Moore and Hower 1986; Huang et al.
1994; Wu et al. 1997). These simulations also indicate
that the interlayer water structure deviates signi®cantly
from the structure of bulk liquid water, showing order-
ing akin to that of a smectic liquid crystal, but then
becoming increasingly disordered with increasing inter-
layer water and interlayer repeat. The degree of hydra-
tion and the location of charge de®ciency depends upon
the distribution of the interlayer cations within the
tetrhedral and octahedral sheets of the 2:1 layer struc-
ture. The simulations con®rm the earlier ®ndings of
Farmer and Russell (1967), that negative charge in the
tetrahedral sheet due to the substitution of Al3+ for Si4+ Fig. 1 Preliminary thermal analysis, TG and DSC experiment at
is localised on the three surface oxygens attached to 10 °C/min, for the <53 lm fraction of fuller's earth (Ca-montmo-
the Al3+ substitutant. Negative charge arising from rillonite)
 593

of water for 24 h. The clay was then left to settle out, and the upper the ®rst 500 seconds of heating, followed by a period of
portion syphoned o€ to collect the fraction below 2 lm as deter- slower more constant mass loss. A room temperature
mined by Stoke's law. The clay solution was then centrifuged for
one hour at 2500 rpm, and the supernatant ¯uid was decanted o€. experiment was performed to show the e€ect of constant
Finally, the clay fraction was freeze-dried. nitrogen gas ¯ow on water loss: a value of around 4
The kinetics of dehydration were investigated by isothermal wt.% can be attributed to humidity-dependent dehy-
mass loss experiments carried out using a Stanton Redcroft TG 761 dration due to the dry nitrogen atmosphere. The lowest
thermobalance. The atmosphere could be controlled by a constant
¯ow of dry nitrogen. The starting mass of all samples was between
temperature runs showed the smallest mass loss at the
9.9 and 10.1 mg to ensure there were no spurious e€ects due to the end of the hour long experiment: they retain a greater
build up of local vapour pressure. The mass loss experiments were amount of water than the samples run at higher tem-
carried out for one hour each, between room temperature and peratures. It seems that, up to a temperature of 100 °C,
150 °C. the mass of the sample at the end point of the reaction
decreases with increasing temperature. Above this tem-
Results perature the percentage mass lost at the end point of
each isothermal run was identical to within 0.2 wt.%.
Preliminary thermal analysis (TG and DSC) at 10 °C/
min indicate there are many stages of mass loss in the
temperature range of 25±1000 °C (Fig. 1). Interlayer Kinetic analysis
water loss dominates the behaviour below 350 °C. An
initial rapid mass loss (up to 150 °C) is followed by The thermal transformations for Ca-montmorillonite
slower mass loss approaching a constant value towards appear complex, re¯ected by the many endotherms re-
the end of dehydration. The reaction appears to occur in corded by DSC (Fig. 1). The dehydration rate is rapid,
two stages, revealed by two endotherms observed in the and can begin before the target temperature has been
DSC experiments. A total mass loss of over 10 wt.% can reached in the furnace (Fig. 2). It can be seen from the
be assigned to the dehydration of interlayer water. mass change versus time curves that complete dehydra-
Further mass loss, after the removal of interlayer tion was not attained in the experiments. Equilibration
water, occurs at higher temperatures (430±665 °C), the to a constant mass occurred at each temperature, even
region over which dehydroxylation becomes important. though the atmosphere was a constant dry nitrogen
Hydroxyls combine in the central octahedral sheet to purge, and this constant mass implies ®nite H2O con-
form water molecules. The reaction appears to proceed tents remain after long time intervals. This behaviour
over a wide range of temperatures and three endotherms could indicate that a local H2O vapour pressure is set up
are observed in the DSC run. These occur at 430, 520 in the sample, and that the gas purge was not e€ective in
and 665 °C, the two higher temperature endotherms removing the complete H2O layer (Line et al. 1995).
representing the main dehydroxylation events. A mass On the other hand, it seems more likely that the clay has
loss of 2 wt.% is observed for each of these reactions. At reached a stable water content as a function of the
very high temperatures, at around 800 °C, there appears conditions of the sample.
to be a slight mass loss (of around 1 wt.%) and a broad The ®nal mass reached at each temperature is de-
exotherm. A constant mass loss is observed above pendent on the amount of sample and the manner of
925 °C, at which temperature a further sharper exoth- packing of the sample into the platinum crucible. As
erm is seen. the initial sample mass was kept constant between 9.9±
Isothermal mass loss experiments (20±150 °C) for the 10.1 mg, e€ects due to packing must be responsible for
dehydration reaction (Fig. 2) show a rapid mass loss in di€erence in the ®nal mass change observed. To elimi-
nate any e€ects due to the random packing of the
sample, for each individual temperature the experiment
was repeated four times to ensure a reproducible and
reliable data set. It is important to note that even though
the ®nal mass change shown by the sample was some-
times anomalous, the rate of mass change as a function
of time remains unchanged for each data set. This shows
that the TG technique is a reliable method for measuring
the rate of reaction, as a function of time at a speci®c
temperature.
To determine the reaction coordinate, or fraction
transformed (a), one has to know the end point of the
reaction (the state corresponding to a = 1.0). A simple
®rst-order exponential function was ®tted to the data for
times greater than 1000 s, so the data could be expressed
in terms of a. The fraction transformed was then accu-
Fig. 2 Isothermal mass change against time for fuller's earth, rately determined, as a function of time for each iso-
collected from 20 °C to 150 °C thermal experiment (Fig. 3). The lower temperature
594

The reduced time plots shown in Fig. 4 use t/t0.3 to

test for the reaction mechanism involved. Frequently,
t/t0.66 has been employed, but by scaling the time de-
pendent behaviour to that at 30% of the reaction one is
better able to discriminate di€erences at higher values of
a (Mohamed and Sharp 1997). It can be seen that our
experimental data is not isokinetic between 20 °C and
150 °C: the reaction cannot be described by a single rate
equation over the enire temperature range. The experi-
ment at 50 °C lies between a ®rst order reaction and a
di€usion controlled reaction. At 100 °C the initial re-
action appears to be dominated by a phase-boundary
mechanism, and at 150 °C the initial reaction could be
described by the Avrami equation. Therefore, on in-
creasing temperature the rate-determining step appears
Fig. 3 Time evolution of the fraction transformed (a), calculated
from the mass change as a function of time
to be changing, and there is evidence that the reaction
could be multistage, or at least a mixture of two reaction
mechanisms.
experiments did not always go to completion, and
clearly react at a slower rate.
Empirical rate equations
In order gain some insight into the reaction mecha-
nism on an atomic scale, attempts were made to ®t the
Rather than ®tting the data to one of a number of
data to a series of solid-state rate equations. A well-
possible rate equations, and attempting to discern the
established method of data analysis assumes the general
rate controlling process by judging the goodness of ®t
rate Eq. (1), which expresses the rate of the reaction as
a product of a function of the fraction transformed (a)
and a rate constant (k):
da
ˆ kf …a† …1†
dt
which can be expressed in integrated form as
g…a† ˆ kt …2†
The function g(a) is dependent on the mechanism of the
reaction, and this dependence is often used to infer an
atomic reaction mechanism.
Sharp et al. (1966) have shown that reduced time
scale plots facilitate the comparison of experimental
data. Isokinetic data from a number of experimental
runs should line on a single curve. This curve can then be
compared to that expected for any one of a number Fig. 4 Reduced time plots (t/t0.3) for the dehydration reaction of Ca-
of di€erent theoretical rate equations (Table 1). The rate montmorillonite at 50, 100 and 150 °C. The expected evolutions of the
equations can be divided into three general groups: (1) fraction transformed for random nucleation (A2), phase boundary
(R2), di€usion (D2) and nucleation (F1) controlled reactions are
phase-boundary and ®rst order reactions, (2) di€usion shown by the lines. The data for Ca-montmorillonite dehydration to
controlled reactions, (3) reactions described by the Av- lie between a mixture of processes at low temperature, changing in
rami equation (Avrami, 1939). character with temperature and fraction transformed (a)

Table 1 Integrated rate equa-

tions, showing the correspond- Mechanism Function g(a) Slope (m) Slope (mA)
ing calculated values of the
exponent m for the Avrami and Zero order a 1.17 1.36
Austin-Rickett empirical rate R2 ± phase boundary reaction, cylindrical 1 ÿ …1 ÿ a†1=2 1.10 1.31
laws R3 ± phase boundary reaction, spherical 1 ÿ …1 ÿ a†1=3 1.06 1.27
F1 ± random nucleation, one nucleus/particle ÿ ln…1 ÿ a† 1.00 1.19
A2 ± random nucleation, Avrami equation I ‰ÿ ln…1 ÿ a†Š1=2 2.00 2.39
A3 ± random nucleation, Avrami equation II ‰ÿ ln…1 ÿ a†Š1=3 3.00 3.58
D1 ± one-dimensional di€usion a2 0.60 0.72
D2 ± two-dimensional di€usion, cylindrical …1 ÿ a† ln…1 ÿ a† ‡ a 0.57 0.67
D3 ± three-dimensional di€usion, spherical ‰1 ÿ …1 ÿ a†1=3 Š2 0.53 0.63
D4 ± three-dimensional di€usion, spherical, …1 ÿ 23 a† ÿ …1 ÿ a†2=3 0.55 0.66
Ginstling-Broughtein
 595

between the data and each rate equation, it is possible to

employ a general rate equation in which the function,
f(a), (in Eq. 1) is expressed in some general manner with
a variable parameter that re¯ects the rate controlling
mechanism. Such rate equations are sometimes termed
``empirical rate equations''. Two such equations are
commonly used, that developed by Avrami (1939) and
that developed by Austin and Rickett (1939). We have
used each of these to analyse our data for smectite de-
hydration. The Johnson-Mehl or Avrami method ex-
presses the general rate Eq. (1) such that the rate
constant (k) and the controlling reaction mechanism
(related to a parameter, m) can be determined simulta-
neously (Hancock and Sharp 1972):
da
ˆ k m
tmÿ1 …1 ÿ a† …3†
dt
Following integration of Eq. (3), the fraction trans-
formed can be written as the linearised form of the
Avrami equation:
ln…ÿ ln…1 ÿ a†† ˆ m ln k ‡ m ln t …4†
The values of k and m can be calculated from the in-
tercept and gradient of a `lnln' plot of the data. If a
single reaction mechanism operates through the tem-
perature range and several data sets are isokinetic, they
will plot as a set of parallel lines with a constant value of
m. Values of m (in Eq. 4) for a range of established rate
equations were ®rst presented by Hancock and Sharp Fig. 5 a Avrami `lnln' plot and b Austin-Rickett plot, for
(1972). We have recalculated these for 0.15 < a > 0.5 0.1 < a > 0.9. Deviations from linearity are observed towards the
for both the Avrami equation and for the Austin-Rickett end of the reaction at higher temperatures
equation (Table 1).
Our data for 0.1 £ a ³ 0.9 are shown as an Avrami
`lnln' plot in Fig. 5a. The gradient m increases with these temperatures, a second stage in the reaction be-
temperature, and at higher temperatures, is also a comes apparent (Table 2). The initial rate limiting pro-
function of the fraction transformed, a (Table 2). There cess (with m = 1.20) could be the advancement of a
appear to be three di€erent temperature regimes. phase boundary. At higher values of a (>0.70), the
Firstly, from 20 to 80 °C, the reaction appears isoki- gradient changes (m = 0.41). Therefore, we conclude
netic, with a single dominant rate-limiting step. The av- that during dehydration at these temperatures the
erage gradient, m, over this temperature range is 0.84. rate-controlling mechanism changes from ®rst-order to
This value does not correspond to any one of the nine di€usion controlled. The initial water loss is very rapid,
rate equations shown in Table 1. Rather, it is interme- and the ®rst-order control is most likely associated
diate between the value expected for a ®rst order reaction with a strain-related interface, as the interlayer space
and that expected for di€usion control. It seems likely, collapses upon dehydration of the ®rst layer of water
therefore, that the reaction is ®rst order at these tem- lost. The water removed during the second stage is more
peratures, slowed down by the di€usional control of the ®rmly held (probably in the hydration sphere around the
movement of molecular water from the interlayer space. interlayer cation) and is involved in water-cation inter-
Secondly, between 80 and 130 °C, there appears to be actions. This water does not start to leave the structure
a change in behaviour. The Avrami plots show a more until dehydration of the ®rst type is complete. The dif-
pronounced negative curvature and there appears to be fusion of this more ®rmly held water is therefore the
more than one linear section. Criado et al. (1984), Goss rate determining step during the ®nal stages of the
(1987) and Ruan and Gilkes (1996) found (in other dehydration process.
systems) that `lnln' plots may often be split into several Finally, at the highest temperatures (140 °C and
linear sections, indicating that several reaction processes 150 °C), the `lnln' plots once more show only one linear
are operating, their relative in¯uence on the kinetics of section (Fig. 5a). It is likely that this is because of the
the reaction changing as a function of a. Very rapid very rapid reaction of the sample, which begins even as
changes are still observed at the early stages of the re- the sample starts to heat up within the apparatus, but
action, as well as an increase in m with temperature. before it has reached the set soak temperature. The
Towards the end of the reaction, in isothermal runs at system is far from equilibrium upon heating to such high
596

Table 2 Values of the rate constant (k) and exponent (m) for the ®rst and second stages of dehydration of fuller's earth determined using
the Avrami method and Austin-Rickett method

T °C Avrami Austin-Rickett

k m k m

Stage I
20 0.00080(4) 0.9940(7) 0.00010(9) 1.37(2)
30 0.0033(3) 0.859(6) 0.00077(6) 1.19(1)
40 0.0069(3) 0.786(6) 0.00157(8) 1.137(8)
50 0.0068(2) 0.844(6) 0.0015(1) 1.21(1)
60 0.0112(7) 0.81(1) 0.0027(1) 1.194(8)
70 0.0120(5) 0.834(8) 0.0050(3) 1.18(7)
80 0.0104(6) 0.96(1) 0.00372(3) 1.30(4)
90 0.0085(6) 1.06(2) 0.0043(2) 1.32(8)
100 0.0056(4) 1.20(2) 0.0030(2) 1.46(1)
110 0.0042(4) 1.31(2) 0.0015(1) 1.68(2)
120 0.0027(2) 1.47(3) 0.00099(7) 1.85(3)
130 0.0035(4) 1.42(3) 0.0014(2) 1.79(3)
140 0.0026(3) 1.45(3) 0.00035(5) 2.12(4)
150 0.0025(2) 1.48(2) 0.00035(6) 2.15(5)
T °C Avrami Austin-Rickett

a k m a k m
Stage II
80 0.78 0.058(9) 0.62(3)
90 0.70 0.13(1) 0.48(2) 0.8 0.04(1) 0.87(5)
100 0.70 0.21(2) 0.41(1) 0.8 0.09(1) 0.76(2)
110 0.76 0.31(2) 0.34(1) 0.8 0.14(2) 0.71(3)
120 0.73 0.25(3) 0.42(2) 0.84 0.13(8) 0.80(9)
130 0.75 0.14(2) 0.57(3)

temperatures, and this is re¯ected in the increased value respectively. At intermediate temperatures, the latter can
of the gradient (m = 1.5). The driving force for the re- be split into two separate linear sections. The second
action is high, and no information about the second linear section occurs for a > 0.8, with an average value
stage of reaction can be extracted because the reaction of mA = 0.79 (Table 2). The break in slope in the
runs to exhaustion by a ®rst-order mechanism before Austin-Rickett plots occurs at a higher a than when
reaching the second stage seen at lower temperatures. plotted according to Eq. (4). This is because this method
Two reaction stages are recorded in the Avrami `lnln' removes the negative curvature seen in the Avrami plots,
plots. The speci®c point at which the gradient changes and better di€erentiates the two linear sections. The
could be a€ected by the slight negative curvature ob- values of mA determined for the second stage suggest a
served in the higher temperature Avrami `lnln'plots. If di€usion controlled process, con®rming the conclusions
the Avrami  lnln' plot show a slight negative curvature it from the Avrami `lnln' plots. It is very dicult to dif-
is often useful to replace (1)a) in Eq. (3) with (1)a)2 ferentiate between the di€erent di€usion-controlled rate
(Burke 1965), and the empirical rate equation becomes equations (Table 1). Considering the structure of mon-
equivalent to that of Austin and Rickett (1939): tmorillonite, a two dimensional di€usion mechanism is
expected, corresponding to movement of water parallel
da to (001).
ˆ k mA
tmAÿ1 …1 ÿ a†2 …5†
dt
or, in linearised integrated form:
 a  Determination of the activation energy (Ea)
ln ˆ mA ln kA ‡ mA ln t …6†
1ÿa From rate constants, k
This de®nes another exponent mA and rate constant kA,
which di€er from m and k in the Avrami equation. We have determined values of rate constant, k, for the
Analysis of the data using Eq. (6) (rather than Eq. 4) Avrami equation and kA for the Austin-Rickett equation
con®rms the results of the Avrami analysis for the data for isothermal runs between 20 °C and 150 °C (Table 2).
below 80 °C (dominantly ®rst order with nucleation and An estimate of the activation energy (Ea) can obtained
growth of a reaction front controlling the initial stage of using the Arrhenius equation, k ˆ A exp…ÿEa =RT †.
dehydration) and above 130 °C. At intermediate tem- As the data in the lowest temperature regime were
peratures the gradient mA increases with temperature in found to be isokinetic, those from the region from 30±
the same way as has been seen in the Avrami `lnln' plots. 80 °C have been used to determine the energetics for the
Figure 5a, b shows the Avrami and Austin-Rickett plots ®rst stage of the dehydration process (Fig. 6). In addi-
 597

Values of rate constant (k) can also be calculated

using the integrated rate equations in Table 1. First
order, phase boundary and di€usion rate equations
were used (whether they were deemed physically
appropriate or not) to calculate an activation energy.
An average value of 35.5 ‹ 0.5 kJ á mol±1 was ob-
tained (Table 3). Gualtieri et al. (1995) employed a rate
equation based on instantaneous nucleation and
growth model (®rst order), but with an experimental
coecient related to di€usion, which contributes to a
decrease in the reaction rate (0.75[)ln(1)a)4/3] = kt).
Using this rate equation we obtain an activation energy
of 35.4 kJ á mol)1, which is in agreement with the av-
erage value from all the rate equations and the value
obtained using Eq. (7). The spread of activation ener-
gies obtained using this diverse suite of rate equations
is quite small. The phase boundary and ®rst order rate
Fig. 6 Arrhenius plot of ln k against 1/T (K) for the initial stage of laws tend to give lower values of the activation energy,
dehydration, determined from the integrated rate equations, Eq. 7, whilst the di€usion controlled rate laws give values
and both the Avrami and Austin-Rickett rate laws
higher than the average. The values calculated from
Eq. 7 and from Gaultieri et al.'s (1995) rate equation
tion to using the empirical rate equations to determine both lie close to the average.
the temperature dependence of k, we have followed the The activation energy for the second stage of the
work of Miletich et al. (1997) to accurately calculate reaction is more dicult to calculate by this method, as
the reaction rate k for each individual temperature run. this subsequent di€usion controlled reaction was only
The mass change versus time data was ®tted by a least observed across a small temperature and a regime. It is
squares procedure to a simple exponential function therefore dicult to resolve accurate values of the in-
n
tercept in Avrami or Austin-Rickett plots, since this is
mt ˆ meqm ‡ Dm
eÿfk…tÿto †g …7† strongly dependent on choice of starting point for the
second stage of the reaction at each temperature. In fact,
where mt is initial mass at time t, meqm is the mass at
the values of k quoted for stage II of the dehydration
t = ¥, and Dm = mo ) meqm. The exponent n was ®xed
reaction in Table 2 do not seem to describe the rates of
at 0.84, in accordance with the value implied by the
reaction implied by the raw data.
Avrami equation. Using, therefore, values of k obtained
at each temperature Eqs. (4), (6), and (7), we have ob-
tained three possible values for the activation energy of
dehydration. These three methods of calculating acti- Using the ``time to a given fraction'' method
vation energies render values of 28.6 ‹ 0.9 kJ á mol)1
(Avrami method), 34.2 kJ á mol)1 (Eq. 7), and a higher A potential pitfall of the method described is the as-
value of 40.6 ‹ 2.2 kJ á mol)1 from the Austin-Rickett sumption that activation energy (Ea) remains constant
equation (Table 3). throughout the reaction process, and is correctly deter-
mined from the rate laws. The Avrami and Austin-
Rickett plots have already indicated that, at lower
Table 3 Activation energies (Ea), for the ®rst stage of dehydration, temperatures, the reaction is probably dominated by a
calculated from the rate equations in Table 1, by ®tting the data to combination of two processes. In addition, at higher
Eq. 7, and by both the Avrami and Austin-Rickett methods
temperatures (80±120 °C), the dominant mechanism
Rate equation Activation Energy appears to change through the reaction, and therefore
(kJ
mol )1) likely there is a change in activation energy as the
reaction proceeds. Redfern (1987) suggested that the
Avrami 28.6 ‹ 0.9
Austin-Rickett 40.5 ‹ 2.2
activation energy must be determined independently of
Gualtieri et al. (1995) 35.4 ‹ 1.6 the empirical function g(a), if any change in activation
First order 32.3 ‹ 1.9 energy is to be observed. This is often referred to as the
R2 31.3 ‹ 2.2 ``time to a given fraction'' method, and relies on the fact
R3 31.6 ‹ 2.1 that the fraction transformed (a) and the time t are
D1 37.5 ‹ 2.1
D2 38.1 ‹ 2.0 functionally related. The time t can be made the de-
D3 38.0 ‹ 1.9 pendent variable by rewriting Eq. (1):
D4 38.4 ‹ 2.0
Miletich et al. (1997) 34.2 ‹ 1.6 dt ˆ k ÿ1 f ÿ1 …a†da …8†
Average 35.5 ‹ 0.5
The time tY to transform a given fraction Y is then
598

Z
aˆY time (hence lines are horizontal): the sample has dehy-
tY ˆ k ÿ1
f ÿ1 …a†da …9† drated fully during heating to the isothermal soak tem-
perature. Furthermore, between 100 and 130 °C the
aˆ0
activation energy increases with fraction transformed at
and if the function ¦)1 (a) does not change over the values of a > 0.8: this represents the second stage of the
temperature range studied the integral has a constant dehydration process, only seen at these temperatures
numeric value. Therefore tY / k ÿ1 or tY / Aÿ1 exp and fractions transformed.
…Ea =RT †. Thus tY for a certain fraction transformed can Previously, it was noted that the reaction changes
be determined from the set of isothermal experiments, above values of a > 0.8 at these temperatures, therefore
using ln tY ˆ const ÿ ln A ‡ Ea =RT . A plot of ln tY we have attempted to determine the energetics of the last
against 1/T is linear if the activation energy does not stage of dehydration. Fig. 7b shows a ln ``time to a given
change within the range of Y chosen. fraction'' plot, for 0.8 < a > 0.9, from which the acti-
The data for 20 < T < 80 °C reveal a linear de- vation energy is found to rise to a value of 47 ‹
pendence at all values of fraction transformed between 6 kJ á mol)1.
a = 0.1 and a = 0.9, giving an average activation
energy, Ea, of 34.7 ‹ 0.9 kJ á mol)1 (Fig. 7a). At
T > 90 °C, the plots starts to tend towards horizontal Conclusions
at lower values of a. This is due to the fact that the
sample is reacting before reaching the isothermal soak The dehydration of Ca-montmorillonite is clearly a
temperature for that experiment, so that small fractions complex reaction, occurring rapidly under the condi-
of transformation occur at temperatures lower than the tions of this investigation. From the DSC measurements
notional set temperature as the sample heats in the ap- two endotherms are observed below 300 °C. The pres-
paratus. Only on reaching higher fractions transformed ence of two such endotherms is predominantly due to
is there some di€erentiation between the times required the hydration of calcium in the interlayer, since this re-
to reach certain points in the reaction. This e€ect is sults in the process of interlayer water removal occurring
particularly pronounced above 130 °C where the sam- several stages (Waclawska 1984).
ples reach all given fractions transformed at an identical Thermogravimetry shows that, as the sample
equilibrates over a range of isothermal temperatures, the
end point of dehydration (shown by the total mass loss)
is a function of temperature. It is essential to recognise
this if an accurate kinetic analysis is to be carried out for
the ®rst stage of dehydration, since the data must be
normalised against the expected end point for each
temperature. When this is done, the reduced time plots
(t/t0.3) show that, at 50 °C, the dehydration reaction
follows a kinetic path that lies between the curves cal-
culated for di€usion (D2) and random nucleation (F1)
controlled reactions. The Avrami plots ranging from 20
to 80 °C, reinforce this observation, indicating that the
reaction control is intermediate between nucleation and
growth process and di€usion. The Austin-Rickett (1939)
method implies the dominating reaction mechanism is
®rst order in nature, but this probably re¯ects that fact
that this method over-corrects for a negative curvature
in the `lnln' plots, not observed in the original Avrami
plots.
In conclusion, the rate-limiting step for the ®rst stage
of dehydration is well described by the nucleation and
growth of a reaction front, which is slightly impeded by
the strong electric ®elds of the Ca2+. The powerful
hydrating properties of the interlayer cations are able to
slightly hinder the migration of the water molecules
from their outer hydration spheres.
It is observed that at higher temperatures (above
90 °C) the reaction is no longer isokinetic as a function
of temperature: both the m and mA values (in the Av-
rami and Austin-Rickett plots, respectively) increase
Fig. 7 a ln ``time to a given fraction'' for 0.1 < a > 0.9, b for the with temperature. This is because the sample is far from
second stage of the reaction (0.8 < a > 0.9, 100 < T < 130 °C) equilibrium at the start of the isothermal shock heating
 599

experiment. We can not infer that the nature of water Avrami M (1939) Kinetics of phase change. I. General theory.
release is changing, this trend being recognised in both J Chem Phys 7: 1103±1112
Barker C (1972) Aquathermal pressuring ± role of temperature in
methods. development of abnormal-pressure zones. Am Assoc Petrol
The second stage of the reaction process is observed Geol Bull 56: 2068±2071
towards the end of the reaction at high temperatures, at Bridgeman CH, Buckingham AD, Skipper NT, Payne MC (1996)
a > 0.7 and a > 0.8 for the Avrami and Austin-Rickett Ab-initio total energy study of uncharged 2:1 clays and their
interaction with water. Mol Phys 89: 879±888
methods, respectively. A change in gradient of the `lnln' Bray HJ, Redfern SAT, Clark SM (1998) The kinetics of dehy-
plots is observed, suggesting that this second stage of the dration in Ca-montmorillonite: an in situ X-ray synchrotron
dehydration reaction is di€usion controlled, and could study. Mineral Mag 62: 647±656
be linked to the removal of water from a di€erent en- Bruce CH (1984) Smectite dehydration ± Its relation to structural
vironment. The separation of the two stages is better development and hydrocarbon accumulation in northern Gulf
of Mexico. Am Assoc Petrol Geol Bull 68: 673±683
seen using the Austin-Rickett method, and occurs at Burke J (1965) The kinetics of phase transformations in metals.
a > 0.8. Pergamon Press, Glasgow
Previous in situ X-ray synchrotron work on the de- Burst JF (1969) Diagenesis of Gulf Coast clayey sediments and its
hydration of the same fuller's earth sample, showed that, possible relation to petroleum migration Am Assoc Petrol Geol
Bull 53: 73±93
up to 80 °C, a one-layer hydrate is stable (Bray et al. Chang FRC, Skipper NT, Sposito G (1995) Computer simulation
1998). Waclawska (1984) found evidence of a second of interlayer molecular structure in sodium montmorillonite
stage of dehydration. Firstly the water that forms the hydrates. Langmuir 11: 2734±2741
hydration layers is loosely bound in the structure of Criado JM, Ortega A, Real C, Torres de Torres E (1984) Re-
examination of the kinetics of the thermal dehydroxylation of
montmorillonite is removed, whereupon water forming kaolinite. Clay Miner 19: 653±661
the hydration sheaths of cations is released. This occurs De Siqueira AVC, Skipper NT, Coveney PV, Boek ES (1997)
when a = 0.8, the second stage of dehydration ac- Computer simulation evidence for enthalpy driven dehydration
counting for loss of 20% of interlayer water. However, of smectite clays at elevated pressures and temperatures. Mol
no mention has been made of reaction mechanisms and Phys 92: 1±6
Eberl DD, Hower J (1976) Kinetics of illite formation. Bull Geol
no calculations for the energetics of the ®nal stage have Soc Am 87: 1326±1330
been made (Waclawska 1984). Farmer VC, Russell JD, (1967) IR absorption spectrometry in clay
From the `lnln' plots of the second stage of dehy- studies. Clays Clay Miner 15: 121±142
dration a di€usional process could describe the rate- Freed RL, Peacor DR (1989) Geopressured shale and sealing e€ect
of smectite to illite transition. Am Assoc Petrol Geol Bull 73:
limiting step. It is easy to envisage a two-dimension 1223±1232
di€usion dominating mechanism, since this would des- Girgis BS, El-Barawy KA, Feli NS (1986) Dehydration kinetics of
cribe the movement of water parallel to (001) from the some smectites: a thermogravimetric study. Thermochimica
interlayer to the reaction front. Acta 98: 181±189
Goss J (1987) The kinetics and reaction mechanism of the goethite
Activation energies (Ea) determined for the ®rst stage to hematite transformation. Mineral Mag 51: 437±451
of dehydration using the Avrami, Austin-Rickett, Eq. (7) Greene-Kelly R (1955) Dehydration of montmorillonite minerals.
and integrated rate equations generate a range of values, Clay Minerals Bull 5: 604±615
from 28.6 kJ á mol)1 to 40.6 kJ á mol)1 (Table 3). The Gualtieri A, Bellotto M, Artioli G, Clark SM (1995) Kinetic study
activation energy (Ea) determined from the ``ln time to a of the kaolinite-mullite reaction sequence. Part II: mullite for-
mation. Phys Chem Minerals 22: 215±222
give fraction method'' (34.7 ‹ 0.9 kJ á mol)1), ®tting to GuÈler C, Sarier N (1990) Kinetics of thermal dehydration of acid-
Eq. (9) (34.2 kJ á mol)1), and the Gualtieri integrated activated montmorillonite by the rising temperature technique.
rate equation (35.4 kJ á mol)1), are all very similar. This Thermochimica Acta 159: 29±33
is a good indication that the activation energy for re- Hancock JD, Sharp JH (1972) Method of comparing solid-state
kinetic data and its application to the decomposition of kaoli-
moval of the ®rst interlayer water by a nucleation and nite, brucite and BaCO3. J Am Ceram Soc 55: 74±77
growth process is around 35 kJ á mol)1. The second, Heller-Kallai L, Rozenson I (1980) Dehydroxylation of diocta-
slower, di€usion controlled reaction process, at a > 0.8, hedral phyllosilicates. Clays Clay Miner 28: 355±368
has a higher activation energy (Ea) of 47 ‹ 6 kJ á mol)1 Howard J, Roy DM (1985) Development of layer charge and ki-
netics of experimental smectite alteration. Clays Clay Miner 33:
as determined from the ``ln time to a given fraction'' 81±88
method. Hower J, Eslinger E, Hower ME, Perry EA (1976) Mechanism of
burial metamorphism of argillaceous sediment 1. Mineralogical
Acknowledgements The authors are grateful for support from RBB and chemical evidence. Geol Soc Am Bull 86: 725±737
R & D, who provided the sample of fuller's earth, and in particular Huang WL, Bassett WA, Wu T-C (1994) Dehydration and hydra-
to Dr Simon Drachman for his interest in the project. This work tion of montmorillonite at elevated temperatures and pressures
was supported by NERC in the form of studentship GT19/95/CS/2. monitored using synchrotron radiation. Am Mineral 79: 683±691
Karaborni S, Smit B, Urah J, van Oort E (1996) The swelling of
clays: molecular simulations of the hydration of montmorillo-
References nite. Science 271: 1102±1104
Koster van Groos AF, Guggenheim S (1984) The e€ect of pressure
Annabli-Bergaya F, Estrade-Szwarckopf H, Van Damme H (1996) on the dehydration reaction of interlayer water in Na-mon-
Dehydration of Cu-hectorite: water isotherm, XRD, and EPR tmorillonite. Am Mineral 69: 872±879
studies. J Phys Chem 100: 4120±4126 Koster van Groos AF, Guggenheim S (1986) Dehydratoin of K-
Austin JB, Rickett RL (1939) Kinetics of the decomposition of exchanged montmorillonite at elevated temperatures and pres-
austenite at constant temperature. Trans AIME 135: 396±443 sures. Clays Clay Miner 34: 281±286
600

Koster van Groos AF, Guggenheim S (1987) Dehydration of a Ca- histories of sedimentary basins. Springer Verlag, New York
and a Mg-exchanged montmorillonite (SWy-1) at elevated Berlin Heidelberg pp 133±140
pressures. Am Mineral 72: 292±298 Ransom B, Helgeson HC (1995) A chemical and thermodynamic
Line CMB, Putnis A, Putnis C, Giampaolo C (1995) The dehy- model of dioctahedral 2:1 layer clay minerals in diagenetic
dration kinetics and microtexture of analcime from two para- processes: dehydration of dioctahedral aluminous smectite as a
geneses. Am Mineral 80: 268±279 function of temperature and depth in sedimentary basins. Am J
Low PF (1979) Nature and properties of water in montmorillonite- Sci 295: 245±281
water systems. Soil Sci Am J 45: 651±658 Redfern SAT (1987) The kinetics of dehydroxylation of kaolinite.
Magara K (1975) Reevaluation of montmorillonite dehydration as Clay Miner 22: 447±456
cause of abnormal pressure and hydrocarbon migration. Am Robertson HE, Lahann RW (1981) Smectite to illite conversion
Assoc Petrol Geol Bull 59: 292±302 rates: e€ect of solution chemistry. Clays Clay Miner 29: 401±411
Miletich R, Zemann J, Nowak M (1997) Reversible hydration in Ruan HD, Gilkes RJ (1996) Kinetics of thermal dehydroxylation
synthetic mixite, BiCu6(OH)6(AsO4)3.nH20 (n < 3): hydration of aluminous goethite. J Thermal Anal 46: 1223±1238
kinetics and crystal chemistry. Phys Chem Minerals 24: 411±422 Sharp JH, Brindley GW, Narahari Achar BN (1966) Numerical
Mohamed BM, Sharp JH (1997) Kinetics and mechanisms of for- data for some commonly used solid state reaction equations.
mation of monocalcium aluminate, CaAl2O4. J Mat Sci 7: J Am Ceram Soc 49: 379±382
1595±1599 Skipper NT, Refson K, McConnell JDC (1989) Computer calcu-
Moore DM, Hower J (1986) Ordered interstrati®cation of dehy- lation of water-clay interactions using atomic pair potentials.
drated and hydrated Na-smectite. Clays Clay Miner 34: 379± Clay Miner 24: 411±425
384 Skipper NT, Refson K, McConnell JDC (1991) Computer simu-
Murray P, White J (1949) Kinetics of the thermal dehydration of lation of interlayer water in 2:1 clays. J Chem Phys 91: 7434±
clays. Trans Brit Ceram Soc 48: 187±206 7445
Murray P, White J (1955a) Kinetics of the thermal dehydration of Skipper NT, Chang FRC, Sposito G (1995a) Monte Carlo simu-
clays. Part 1. Dehydration characteristics of the clay minerals. lation of interlayer molecular-structure in swelling clay miner-
Trans Brit Ceram Soc 54: 137±149 als. 1. Methodology. Clays Clay Miner 43: 285±293
Murray P, White J (1955b) Kinetics of the thermal dehydration of Skipper NT, Sposito G, Chang FRC (1995b) Monte Carlo simu-
clays. Part II. Isothermal decomposition of the clay minerals. lations of interlayer molecular-structure in swelling clay-min-
Trans Brit Ceram Soc 54: 151±187 erals. 2. Monolayer hydrates. Clays Clay Miner 43: 294±303
Murray P, White J (1955c) Kinetics of the thermal dehydration of Tang L, Sparks DL (1993) Cation-exchange kinetics on montmo-
clays. Part III. Kinetics analysis of mixtures of the clay min- rillonite using pressure-jump relaxation. Soil Sci Am J 57: 42±46
erals. Trans Brit Ceram Soc 54: 189±203 Velde B, Vasseur G (1992) Estimation of the diagenetic smectite to
Murray P, White J (1955d) Kinetics of the thermal dehydration of illite transformation in time-temperature space. Am Mineral 77:
clays. Part IV. Interpretation of the di€erential thermal analysis 967±976
of the clay minerals. Trans Brit Ceram Soc 54: 204±238 Waclawska I (1984) Dehydration and dehydroxylation of smectites
Perry EA, Hower J (1972) Late-stage dehydration in deeply buried 1. Dehydration and dehydroxylation kinetics. Mineral Polonica
sediments. Am Assoc Petrol Geol Bull 56: 2013±2021 158: 91±107
Plumley WJ (1980) Abnormally high ¯uid pressure. Survey of some Weaver CE, Beck KC (1971) Clay water diagenesis during burial:
basic principles. Am Assoc Petrol Geol Bull 64: 414±430 how mud becomes gneiss. The Geological Society of America,
Powers MC (1967) Fluid-release mechanisms in compacting marine Boulder, Colorado
mudrocks and their importance in oil exploration. Am Assoc Wu T-C, Bassett WA, Huang W-L, Guggenheim S, Koster van
Petrol Geol Bull 51: 1240±1254 Groos AF (1997) Montmorillonte under high H2O pressure:
Pytte AM, Reynolds RC (1988) The thermal transformations of stability of hydrate phases, rehydration hysteresis, and the e€ect
smectite to illite. In: Vaeser ND, McCulloh TH, (eds.) Thermal of interlayer cation. Am Mineral 82: 69±78