Basics and Applications of Solid State Kinetics A Pharmaceutical Perspective

REVIEW
Basics and Applications of Solid-State Kinetics:

A Pharmaceutical Perspective
AMMAR KHAWAM, DOUGLAS R. FLANAGAN
Division of Pharmaceutics, College of Pharmacy, University of Iowa, Iowa City, Iowa
Received 2 June 2005; revised 20 September 2005; accepted 31 October 2005

Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20559
ABSTRACT: Most solid-state kinetic principles were derived from those for homogenous
phases in the past century. Rate laws describing solid-state degradation are more complex
than those in homogenous phases. Solid-state kinetic reactions can be mechanistically
classified as nucleation, geometrical contraction, diffusion, and reaction order models.
Experimentally, solid-state kinetics is studied either isothermally or nonisothermally.
Many mathematical methods have been developed to interpret experimental data for both
heating protocols. These methods generally fall into one of two categories: model-fitting
and model-free. Controversies have arisen with regard to interpreting solid-state kinetic
results, which include variable activation energy, calculation methods, and kinetic
compensation effects. Solid-state kinetic studies have appeared in the pharmaceutical
literature over many years; some of the more recent ones are discussed in this review.
ß 2006 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 95:472–498, 2006
Keywords: solid-state kinetics; homogenous kinetics; heterogeneous kinetics; stability;
degradation; desolvation; reaction-order; isothermal; nonisothermal; arrhenius equation
INTRODUCTION tions have many forms, however, those that

involve weight or enthalpic change are of high
Many transformations may occur when a solid interest as their kinetics can be studied by
sample is heated, such as: melting, sublimation, thermal analytical methods. Pharmaceutically,
polymorphic transformation, or degradation.1 many solid-state kinetic studies are either desol-
These solid-state reactions, are quite common in vation reactions or polymorphic transformations.
pharmaceutical sciences, especially polymorphic Interest in these reactions is increasing as many
transformations and degradation. Solid-state formulated drugs, including compendial drugs,
chemistry has recently gained much interest in are solvates; mainly hydrates. Physicochemical
pharmaceutical sciences, which renders the topic stability of solvates is a concern to pharmaceutical
of solid-state kinetics important. Solid-state reac- scientists since they may convert to an amorphous
form upon desolvation while others may become
The authors dedicate this review to the memory of
chemically labile. For example, cephradine dihy-
Dr. David J.W. Grant who passed away on December 9, drate dehydrates and produces an amorphous
2005. Dr. Grant was an internationally known authority at the form that is further oxidized. Other hydrates may
University of Minnesota on the solid-state properties of drugs.
He will be remembered as a kind, humble, and brilliant
change their state of hydration producing forms
scholar. with different solubility characteristics.2 Due to
Correspondence to: Ammar Khawam (Telephone: 319-335- the impact of solvates on the development process
8819; Fax: 319-335-9349;
E-mail: ammar-khawam@uiowa.edu)
and drug performance, it is important to know the
Journal of Pharmaceutical Sciences, Vol. 95, 472–498 (2006)
stability of these solvates. In addition, with many
ß 2006 Wiley-Liss, Inc. and the American Pharmacists Association crystal forms of a drug, stability of the marketed
472 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 3, MARCH 2006

BASICS AND APPLICATIONS OF SOLID-STATE KINETICS 473
crystal form should be well understood. One rately. However, applying these concepts was
recent example of polymorphic instability is justified in the solid-state because of similarities
ritonavir, which is Abbott’s protease inhibitor for to some homogenous reactions. For example, the
human immunodeficiency virus (HIV). It was Arrhenius equation was historically developed
temporarily withdrawn from the market because empirically, after which theoretical justification
of the transformation of the marketed crystal for its use was later introduced in gases through
form (Form I) to a more thermodynamically stable the collision theory and in solutions through the
and less soluble form (Form II).3,4 Another classic transition-state theory. A similar justification was
example is chloramphenicol palmitate, which was claimed for the use of this equation in solid-state
reported to exist in more than one crystalline form kinetics.6 Therefore, solid-state kinetics evolved
and one form was found to be as much as seven from homogenous kinetic principles. However, ap-
times more potent therapeutically than other plications of these kinetic principles are different
forms.5 Therefore, the solid-state stability of because of the differences between solids, solu-
polymorphic drugs and the kinetics of such tions, and gases. For example, particle size, inter-
transformations is vital information that needs face advance, and geometric shape are variables
to be evaluated in the development of such drugs unique to heterogeneous reactions and have no
or stable dosage forms. equivalent in homogenous reactions. Differences
Kinetics in the solid-state bear similarities to between homogenous and heterogeneous kinetics
those in homogenous phases like solution or gases. will be highlighted throughout this review.
In fact, many of the basic mathematical principles
are shared among all three phases. However, solid-
state reactions differ substantially from those in Rate Laws
the homogenous state. These differences include
There are many types of solid-state reactions; we
experimental procedures employed for their study
will focus on reactions that involve a single solid
and computation methods for analyzing data.
reactant. A reaction that has acquired great
A review of solid-state kinetics has not appeared
interest is that which follows the reaction scheme
in the pharmaceutical literature while many such
below:
studies have appeared. As several theories and
models for solid-state kinetics have been proposed AðsÞ ! BðsÞ þCðgÞ
and applied in the last century, it is important to
review them so that pharmaceutical scientists can Pharmaceutically, desolvation is a reaction that
learn the range of models and apply them. obeys the above scheme, it involves the removal of
Additionally, kinetic software is available from solvent molecules from the crystalline solvate
equipment manufacturers or other sources to below its melting point (with dehydration being
analyze solid-state kinetic data. There needs to the specific loss of water).7,8 In desolvation, A is
be an understanding of the underlying assump- the solvate or hydrate; B is the parent drug, and C
tions and limitations in such software before one is the solvent or water vapor. The rate9 of the
can accurately interpret the results generated. above reaction is often proportional to the con-
This work aims to review solid-state reaction centration of the reactant or products raised to an
kinetic concepts and applications for pharmaceu- integer or fractional power according to:
tical solids and to assist pharmaceutical scientists
in employing these concepts. Rate / ½An or / ð½A0 ½BÞn or / ð½A0 ½CÞn
ð1Þ
SOLID-STATE KINETICS:
Where, A0 is the initial concentration of A, and n,
FROM HOMOGENOUS TO
is the order of the reaction. The rate of a reaction
HETEROGENEOUS PROCESSES
is usually studied by following the decrease in re-
actant concentration or increase in product concen-
Chemical kinetic concepts were originally based on
tration. Therefore, the reaction rate law becomes:
generalizations from empirical studies of homo-
genous reactions first in the gas phase. These
d½A d½B d½C
concepts were later applied to solution phase Rate ¼ ¼ ¼ ¼ k½An
processes and eventually to solid-state reactions. dt dt dt ð2Þ
Solid-state kinetic concepts did not develop sepa- ¼ kð½A0 ½BÞn ¼ kð½A0 ½CÞn
DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 3, MARCH 2006
474 KHAWAM AND FLANAGAN
Where, k is the reaction rate constant. If the when foreign atoms or ions (e.g., impurities)
evolved gas (C) is efficiently flushed such that occupy normal lattice positions. Dislocations in
C & 0, the above equation reduces to: the lattice structure result during crystal growth
due to surface or internal stresses. A dislocation is
d½A d½B
Rate ¼ ¼ ¼ k½An ¼ kð½A0 ½BÞn a discontinuity in the regularity of the lattice that
dt dt exists in the bulk of a crystal.10 For example, a
ð3Þ group of parallel planes in the lattice could be
If the reaction is an elementary unimolecular shifted by a certain lattice spacing, which could be
(n ¼ 1) reaction, the rate law would be (following due to rapid crystal growth in which molecules (or
only the reactant concentration): ions or atoms) do not have time to reach their
lowest energy states. Imperfection sites are ener-
d½A gized sites (i.e., have a higher free energy) in which
Rate ¼ ¼ k½A ð4Þ
dt the activation energy for reaction is least, thus
Which upon integration becomes: explaining why these sites are highly reactive
(Figs. 1–8).
½A
ln ¼ kt ð5Þ Solid-state kinetics can be studied with thermal
½A0 analytical methods1,11 by measuring a sample
The above expressions use concentration ([A]), property as it is heated or held at a constant
which is usually measured in solution kinetics. temperature. If a reaction involves weight loss,
However, in solid-state kinetics, concentration then weight is followed throughout the reaction
has little meaning because the sample is not and the kinetics are usually studied by thermo-
homogenous, hence reactivity is not the same gravimetry (TGA). Heat (evolved or consumed) is
throughout the sample (nonisotropic) and ‘‘con- another measurable property that is used for
centration’’ does not relate to reactivity. Figure 1 kinetic evaluation using differential scanning
depicts the difference between homogenously calorimetry (DSC) or differential thermal analysis
(dark spots, Fig. 1a) and heterogeneously distrib- (DTA). Weight loss or heat flow data are converted
uted reaction sites (dark spots, Fig. 1b). In solids, to a normalized form called conversion fraction (a).
reactions occur or are initiated at defects in the The conversion fraction ranges from 0 to 1 and is a
crystal lattice or at crystal surfaces, edges, or measure of reaction progress as a function of time
corners.5 or temperature.
Ideally, a perfect crystal contains no imperfec- For isothermal thermogravimetric analysis, the
tions (defects) thus having minimal reactivity. conversion fraction at any time is:
However, in reality, perfect crystals are rare and m0 mt
most crystal lattices contain imperfections. Lattice ¼ ð6Þ
m0 m1
imperfections can be point defects and disloca-
tions. Point defects are defects where units are where, m0 is the initial sample weight, mt is the
missing from the lattice leaving ‘‘vacancies’’ in the sample weight at time, t, and m1 is the final
lattice. These missing units may be atoms, mole- sample weight. Nonisothermally, the conversion
cules, or ions. Point defects are also generated fraction at any temperature is:
Figure 1. Schematic representation of reactivity (a) homogenous system; (b) hetero-

geneous system. Black dots represent reaction sites.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 3, MARCH 2006 DOI 10.1002/jps
Figure 3. Isothermal a-time plots of different solid-

state decomposition models, data simulated for a rate
constant of 0.049/min and temperature of 320 K:
(a) acceleratory, (b) constant.
For an isothermal DSC/DTA analysis, the

conversion fraction12 at any time is:
AUCt0
¼ ð8Þ
AUC1
0
where, AUCt0 is the sample peak area from 0 to t

and AUC1 0 is the total sample peak area.
Nonisothermally, the conversion fraction at any
temperature can be calculated from:
Figure 2. Transformations of TGA and DSC curves to
conversion fraction curves: (a, b); Desolvation thermo- AUCT0
gram of a solid solvate by TGA (a), and DSC (b); (c) a T ¼ ð9Þ
plot for the desolvation process for both DSC and TGA
AUC1
0
results.
where, AUCT 0 is the sample peak area from 0 to T.
Generally, any analytical method that mea-
m0 mT sures reactant loss or product generation can be
¼ ð7Þ converted to a-time (or temperature) plots, trans-
m0 m1
formations of TGA and DSC plots are shown in
where, mT is the sample weight at temperature, T. Figure 2.
Figure 3. (Continued ) (c–e) deceleratory, (f) sigmoidal.
Figure 4. Methods for studying solid-state kinetics.
d
¼ k f ðÞ ð12Þ
dt
gðÞ ¼ k t ð13Þ
where, f(a) is the differential reaction model and
g(a) is the integral reaction model (in some
references, f(a) and g(a) definitions may be
reversed).
The temperature dependence of the rate con-
stant (k) is usually given by the Arrhenius
equation:16
Ea
k ¼ AeRT ð14Þ
where, A is the preexponential (frequency) factor,
Ea is activation energy, T is absolute temperature,
and R is the gas constant. Substitution of Eq. 14
into Eqs. 12 and 13 gives:
d Ea
¼ AeRT f ðÞ ð15Þ
dt
and
Ea
gðÞ ¼ AeRT t ð16Þ
MODELS AND MECHANISMS IN

SOLID-STATE KINETICS
A model is a theoretical and/or mathematical

description of what is observed experimentally. In
solid-state reactions, a model describes a reaction
and can usually be converted into a mathematical
expression (i.e., rate expression). Many models
Figure 5. TGA data for a simulated dehydration have been proposed for solid-state kinetics, which
reaction: (a); isothermal, (b); nonisothermal. have been developed based on certain mechanistic
assumptions. Therefore, different rate expres-
sions are derived from these models. The models
Using the conversion fraction, rate expressions are generally categorized by their underlying
defined in Eqs. 4 and 5 can be written as: mechanistic assumptions or a versus time (a–t)
d shapes.13
Rate ¼ ¼ kð1 Þ ð10Þ Based on the shape of isothermal plots (a–t),
dt
models can be classified as, sigmoidal, accelera-
ln ð1 Þ ¼ kt ð11Þ tory, linear, or deceleratory as seen in Figure 3.
Based on the mechanistic assumptions, models are
Unlike rate laws in homogenous kinetics, which classified as:
usually depend on reaction order (i.e., first,
second, etc.), a rate law for an elementary solid- A. Nucleation—The rate-limiting step is
state reaction could depend on factors such as rate assumed to be the formation and growth of
of nuclei formation, interface advance, diffusion, nuclei, which are finite quantities of product
and/or geometrical shape of solid particles. These in the reactant lattice. After formation, a
factors lead to several decomposition models13–15 nucleus grows and the nucleation rate is
that do not exist in homogenous kinetics and are different from that of nuclei growth. Nuclea-
summarized in Table 1. Eqs. 10 and 11 can be tion models (A and P models) account for
generally expressed as: both nucleation and nuclei growth rates.
Figure 6. Isothermal model fitting method (conventional method): (a) simulated a-

time curves with 0.25% random error in time at: ~, 340 K; }, 345 K; ^, 350 K; &, 355 K,
and &, 360 K. Inset shows simulation parameters, (b); tabulated values obtained from the
curve in addition to g(a) values for each model, (c); first data fit (g(a) vs. t) for each model
and temperature (only two temperature values are shown), (d); second data fit (Arrhenius
plot) from which A and Ea can be calculated for each model.
Figure 7. Isoconversional methods for evaluating solid-state kinetics: (a); Standard

method for a set of isothermal curves at: &, 340 K; &, 345 K; *, 350 K; *, 355 K and ^,
360 K, (b); Ozawa–Flynn–Wall (OFW) method for a set of nonisothermal curves at: &,
1 K/min; &, 2 K/min; *, 4 K/min; *, 8 K/min and ^, 16 K/min.
where, m, n, and p are constants for a particular

model. Therefore, by assigning particular values
to these three terms, any mathematical model can
be generated. Table 1 lists the most common
models.13–15
In homogenous kinetics, when a reaction is
studied, the aim of the kinetic study is to obtain
kinetic parameters (e.g., rate constants) that can
be used to predict product stability. In addition,
the mechanism of the reaction is usually investi-
gated. Mechanisms often refer to detailed chemical
steps by which a reactant is converted to pro-
duct(s). In heterogeneous kinetics, the term
mechanism usually involves identifying a reaction
model, because this information involves chemical
steps that are otherwise difficult to obtain experi-
mentally.11 The choice of a model is generally
based on statistical fits of mathematical models
to data. However, model selection should also
be supported, when possible, by complementary
procedures such as microscopy, spectroscopy,
X-ray diffraction, product analysis, evolved gas
analysis, etc. For example, if nucleation, growth,
or gas diffusion are visually observed microscopi-
cally, such an observation can support a nucleation
or diffusion model obtained from statistical fitting,
in addition, conclusions drawn from statistical
fitting can be further substantiated by these com-
Figure 8. (a). Kinetic compensation effect (KCE), (b). plementary methods, especially X-ray diffraction.18
Isokinetic Relationship (IKR). Each point in (a) results
from an Arrhenius plot shown in (b).
METHODS FOR STUDYING
SOLID-STATE KINETICS
B. Geometrical contraction—Nucleation is There are many methods used to study solid-state

assumed to be instantaneous throughout kinetics. These methods can be generally grouped
the surface and the rate-limiting step is into two categories–experimental and computa-
the progress of the product layer from the tional; the methods are summarized in Figure 4.
surface of the crystal inward and is different
for various crystal morphologies (cubic,
Experimental Methods
cylindrical, spherical, etc.).
C. Diffusion—The rate-limiting step is the There are two approaches utilized to obtain solid-
diffusion of reactants into reaction sites or state kinetic data–isothermal and nonisothermal
products away from reaction sites. methods. For isothermal methods, samples are
D. Reaction-order—The rate law is based on studied at several constant temperatures while
the reaction order, similar to the same rate nonisothermal (dynamic) methods involve heat-
expressions in homogenous kinetics. ing samples at one or more constant heating rates
(usually linear) and following the course of the
reaction. Isothermal methods are similar to those
Sestak and Berggren17 have suggested a gen- used in homogenous kinetics to produce a–time
eral form for g(a) applicable to all models: data compared to concentration–time data in
homogenous kinetics (Fig. 5). On the other hand,
nonisothermal analysis produces a–temperature
gðÞ ¼ m ð1 Þn ðln ð1 ÞÞp ð17Þ data (Fig. 5).
Table 1. Solid-State Rate Expressions for Different Reaction Models, Shapes of These Models Is Given in Figure 3
Model Differential Forma f ðÞ ¼ 1k d

dt Integral Forma g(a) ¼ kt
Nucleation models
Power law (P2) 2 a(1/2) a(1/2)
Power law (P3) 3 a(2/3) a(1/3)
Power law (P4) 4 a(3/4) a(1/4)
Avarami-Erofe’ev (A2) 2(1 a)[ln(1 a)]1/2 [ln(1 a)]1/2
Prout–Tompkins (B1) a (1 a) ln[a/(1 a)]
Geometrical Contraction models
Contracting area (R2) 2(1 a)1/2 [1(1 a)1/2]
Contracting volume (R3) 3(1 a)2/3 [1(1 a)1/3]
Diffusion models
1-D diffusion (D1) 1/2a a2
2-D diffusion (D2) [ln(1 a)]1 [(1 a)ln(1 a)]þa
3-D diffusion-Jander eqn.(D3) 3(1 a)2/3/2(1(1 a)1/3) [1(1 a)1/3]2
Ginstling–Brounshtein (D4) (3/2((1 a)1/31) 1(2a/3)(1 a)2/3
Reaction-order models
Zero-order (F0/R1) 1 a
First-order (F1) (1 a) ln(1 a)
Second-order (F2) (1 a)2 (1 a)11
Third-order (F3) (1 a)3 0.5 ((1 a)21)
a
In some references f(a) and g(a) have opposite designations.
Isothermal Method da/dt is the isothermal reaction rate and dt/dT

is the heating rate (b). Substituting Eq. 15 into
This method is based on maintaining samples at
Eq. 19 gives,
several constant temperatures (i.e., isothermal)
and as a result, a set of a–time points is produced da A Ea
¼ e RT f ðaÞ ð20Þ
at each temperature. These methods are based on dT b
the isothermal rate equations (Eqs. 13 and 16).
Equation 20 represents the differential form of
the nonisothermal rate law.
Nonisothermal Method
This method employs a heating rate, usually Temperature Integral. Integrating the differen-
linear (b), to raise the temperature. A linear tial nonisothermal rate law (Eq. 20) produces the
heating program follows: integral form of the nonisothermal rate law:
T ¼ T0 þ bt ð18Þ
ZT
A Ea
where, T0 is the starting temperature, b is the gðaÞ ¼ eRT dT ð21Þ
b
linear heating rate (K/min), and T is the tem- 0
perature at time, ‘‘t’’. Nonisothermal experiments
are usually seen in solid-state kinetics, however, This integral is called the temperature integral
they have been previously applied to the study of and has no analytic solution.11,21 It has been
homogenous kinetics.19,20 reported that using nonlinear heating pro-
The following relationship can be defined for grams such as hyperbolic or parabolic22 or non-
nonisothermal experiments, Arrhenius temperature functions of the rate
constant23 leads to exact analytical solutions of
d d dt
¼ ð19Þ the temperature integral. However, this approach
dT dt dT has not been widely implemented. To transform
where, da/dT is the nonisothermal reaction rate; the above integral to a more general form found in
mathematical handbooks, the integration vari- Doyle24–26 approximated values of p(x) using the
able can be redefined as, first three terms of the Schlömilch series expan-
sion and the observed linear relationship for
Ea x ¼ 28–50 to obtain by regression the following
x¼
RT approximation of the temperature integral:
and the temperature integral then becomes,

log pðxÞ 2:315 0:4567x ð24Þ
Z 1 x
AEa e
gðÞ ¼ dx ð22Þ We have determined that the quality of this
bR x2
R 1 ex
x
linear relationship is quite high with an r2 ¼
If pðxÞ ¼ x x2
dx, then Eq. 22 can be written as, 0.99999.
AEa
gðÞ ¼ pðxÞ ð23Þ Senum-Yang Approximation. Senum and Yang27
bR developed an accurate nonlinear approximation of
where, p(x) is the exponential integral. The main the temperature integral. If variables in Eq. 22 are
approaches used for evaluating the temperature/ transformed so that x ¼ zy, the integral becomes,
exponential integral are:11 Z 1 zy
AEa e
gðÞ ¼ dy
1. Calculating values of p(x) numerically. Rz 1 y2
2. Converting p(x) to an approximate form that
can be integrated. Which can be written as,
3. Approximating p(x) by a series expansion.
AEa
gðÞ ¼ E2 ðzÞ ð25Þ
The two series most used for approximating the Rz
temperature integral are:1 E2(z) or generally, Ev(z) (where v is an integer) is a
well known integral28,29 for z > 0 given by the
I. An asymptotic series expansion following continued fraction:30

ex 2! 3! 4!
pðxÞ ¼ 2 1 þ 2 3 0 1
x x x x
B C
ðn þ 1Þ! B C
þ . . . þ ð1Þn þ . . . B C
xn B C
B C
B C
II. The Schlömilch series expansion B C
B C
B C
B C
ex 1 2 B C
pðxÞ ¼ 1 þ B C
xðx þ 1Þ ðx þ 2Þ ðx þ 2Þðx þ 3Þ B C
B C
4 z B C
Ev ðzÞ e B 1 C
þ ... ¼ B C
ðx þ 2Þðx þ 3Þðx þ 4Þ z z B vþ0 C
Bz þ C
B 1 C
B 1 þ C
Many approximations have been reported for B v þ 1 C
B z þ C
the temperature integral,1,23 two of which will be B 2 C
B 1 þ C
B vþ2 C
covered–the Doyle and Senum-Yang approxima- B z þ C
B 3 C
tions. They are among the most frequent tempera- B 1þ C
B vþ3 C
ture/exponential integral approximations as each B C
@ zþ A
is the basis of a particular kinetic calculation 4
1þ
method. z þ ...
ð26Þ
Doyle Approximation. The Doyle24–26 approxima- Truncating the number of terms in the above
tion of the exponential integral (p(x)) is based on continued fraction gives the first (one term),
the observation that log p(x) is fairly linear second (two terms), third (three terms), and
with respect to x over a short range of x values. fourth (four terms) degree rational approximation
Table 2. Senum-Yang Approximations of the gives the reaction rate constant and this fitting is
Temperature Integral repeated for each model at each temperature (only
two temperatures are shown in Fig. 6c). For each
Degree p(x)
model, rate constants from all temperatures (five
1
ex 1
in our simulation) are used for the second fit
according to the Arrhenius equation (Eq. 14) as
x ðx þ 2Þ
shown in Figure 6d. The frequency factor and
2 activation energy are obtained from the intercept
ex ðx þ 4Þ
2 and slope, respectively, of this plot. It is interesting
x ðx þ 6x þ 6Þ
to note that Ea values calculated isothermally by
3 the conventional model-fitting method appear to
ex ðx2 þ 10x þ 18Þ
3 be equal regardless of the model (i.e., model
x ðx þ 12x2 þ 36x þ 24Þ
independent). This behavior does not occur in
4a homogenous kinetic studies where, for example,
ex ðx3 þ 18x2 þ 86x þ 96Þ
4 activation energies obtained from a zero-order fit
x ðx þ 20x3 þ 120x2 þ 240x þ 120Þ
are substantially different from those obtained by
a
In Ref. 27, the 4th degree approximation is incorrectly a first, second, or third-order fit. This unusual
calculated, It is 86 not 88. result has been previously addressed18,33 –35 but
without a complete explanation.
known as the Senum-Yang approximation as
given in Table 2.23,27,31,32 Nonisothermal Model-Fitting Methods. There are
many model fitting methods that extract the three
Calculation Methods kinetic parameters known as the kinetic triplet
(A, Ea, and model) from nonisothermal data. These
There are two groups of methods used to analyze methods were used extensively in the early days of
either isothermal or nonisothermal solid-state solid-state kinetic analysis and they continue to
kinetic data–model-fitting and model-free meth- appear. These methods have been critically eval-
ods (Fig. 4). uated36–40 and it is been shown that the sole use of
these methods is not recommended because:
Model-Fitting Methods
For these methods, different models are fit to the 1. They assume a constant kinetic triplet
data and the model giving the best statistical fit is (A, Ea, and model).
chosen as the model of choice from which the 2. They involve fitting three parameters (A, Ea,
activation energy (Ea) and frequency factor (A) and model), which are simultaneously deter-
can be calculated. mined from a single curve.
3. They involve a single heating rate, which is
Isothermal Model-Fitting Method (Conventional not always sufficient to determine reaction
Method). This method is identical to that in kinetics.
homogenous phase kinetics. It involves two fits:
the first, determines the rate constant (k) of the There are a myriad of nonisothermal model
model that best fits the data according to Eq. 13, fitting methods. However, only a few have been
while the second determines specific kinetic para- extensively used, which will be discussed below.
meters such as the activation energy (Ea) and
frequency factor (A) using the Arrhenius equation Direct differential method. This method41,42 uses
(Eq. 14). Figure 6 illustrates the application of the the differential form of the nonisothermal rate law
conventional method for isothermal model-fitting. by
d
numerically
D
calculating the differential
Kinetic data were simulated with Ea ¼ 100 kJ/mole dT DT . Taking the logarithm of the nonisother-
and A ¼ 1013/min and assuming an R3 model with mal rate law, Eq. 20 gives:
a 0.25% random error in time (Fig. 6a). The upper d=dT A Ea
table in Figure 6b tabulates data from Figure 6a ln ¼ ln ð27Þ
f ðÞ b RT
while the lower table calculates g(a) for each
reaction model in Table 1. Figure 6c shows the Plotting the left-hand side (including the model
first fit, which determines the model that best fits f(a)) versus 1/T gives the activation energy (Ea)
the data according to Eq. 13. The slope from this fit and frequency factor (A) from the slope and
intercept, respectively. The model that gives the Eq. 20 generating d2a/dT2. According to Kissinger,
best linear fit is usually chosen as the model. the maximum reaction rate occurs when the
second derivative is zero from which the following
Freeman–Carroll (Difference-Differential) Method. equation can be obtained:
The Freeman and Carroll method43,44 is a differ- Ea
Ea b n1
ential method that was originally developed 2
¼ A nð1 Þ m eRTm ð33Þ
assuming a reaction-order model (f(a) ¼ (1 a)n). RTm
Taking the natural logarithm of the differential where, Tm is the temperature of the maximum
form of the nonisothermal rate law (Eq. 20) gives, rate and am is the conversion value at that rate.
da A Ea The maximum reaction rate represents the peak
ln ¼ ln þ ln f ðaÞ ð28Þ (i.e., inflection point) of a DSC or DTG curve.
dT b RT
Taking the natural logarithm of Eq. 33 and
If incremental differences in the variables of rearranging gives,
Eq. 28 are taken, we obtain: 0 n1
1
AR nð1 Þ
b A Ea
m
d Ea 1 ln 2 ¼ ln@ ð34Þ
D ln ¼ D ln f ðÞ D ð29Þ Tm Ea RTm
dT R T
Which can be rearranged to, The activation energy (Ea) is obtained by plotting
d the left-hand side of the equation versus 1/Tm for
D ln dT D ln f ðÞ Ea
¼ ð30Þ a series of runs at different heating rates. Eq. 34
D 1=T D 1=T R has been generalized to any reaction model
or (f(a)).49
It is worth noting that the Kissinger method is a
d
D ln dT Ea D 1=T model-free method as it does not require any
¼ ð31Þ
D ln f ðÞ R D ln f ðÞ modelistic assumptions to calculate Ea. However,
it is not an isoconversional method (discussed
The activation energy can be obtained by plotting below) because it does not calculate Ea values at
ln f ðÞ
the left-hand side of Eqs. 30 and 31 versus DD1=T progressive a values but rather assumes a con-
and evaluating the intercept for Eq. 30 or stant Ea like methods that assume a single
versus DDln1=T
f ðÞ and evaluating the slope for Eq. 31. Ea value, this method can not detect reaction
complexities.50
Coats–Redfern Method. This method45,46 uses
the integral form of the nonisothermal rate law Model-Free/Isoconversional Methods
(Eq. 23). Coats and Redfern utilized the asymptotic
Model-free methods calculate the reaction activa-
series expansion for approximating the tempera-
tion energy (Ea) without modelistic assumptions,
ture integral (p(x)), producing:
which is usually done by grouping terms such as
the frequency factor (A) and model into the
gðÞ AR 2RTexp Ea
ln 2 ¼ ln 1 ð32Þ intercept of a linear equation and using the slope
T bEa Ea RT
of that equation to calculate the activation energy
where, Texp is the mean experimental temperature. (Ea). The frequency factor (A) can be calculated
Plotting the left-hand side (including the model, from the intercept of the linear equation but
g(a)) of Eq. 32 versus 1/T gives the activation requires modelistic assumptions for such a deter-
energy (Ea) and frequency factor (A) from the mination. Therefore, model-free methods usually
slope and intercept, respectively. The model that report only activation energies.
gives us the best linear fit is chosen as the model. Isoconversional methods are model-free meth-
The Coats–Redfern equation was originally de- ods that evaluate kinetic parameters, namely the
rived assuming a first-order model (g(a) ¼ activation energy (Ea) at progressive conversion
ln(1 a)) but has been generalized to any reac- values (a).51 These methods require several kinetic
tion model. curves to perform the analysis and have therefore
been called by some as ‘‘multicurve’’ methods52,53
Kissinger Method. Kissinger47,48 proposed a kine- as shown in Figure 7. Calculations from several
tic analysis method for reaction-order models curves are performed on the same value of con-
(f(a) ¼ (1 a)n) based on taking the derivative of version (a), thus the name isoconversional. As a
result, these methods calculate the activation fore, two common approximations of the tempera-
energy for each conversion point (Eaa), resulting ture integral have been widely used:
in an isoconversional plot (Ea vs. a) as seen in
Figure 7. 1. A linear approximation (less accurate) uti-
The terms, ‘‘model-free’’ and ‘‘isoconversional’’ lizing the Doyle approximation has been
are sometimes used interchangeably, however, not used in the Ozawa and Flynn–Wall
all model-free methods are isoconversional. For methods.
example, the Kissinger method (discussed above) 2. A nonlinear approximation (more accurate)
is a model-free method but is not isoconversional.34 utilizing the Senum-Yang approximation
Isoconversional approaches can be used to analyze has been used in the Vyazovkin method.
both isothermal and nonisothermal data as
described below. Ozawa, Flynn, and Wall (OFW) Method. Ozawa56
and Flynn–Wall57 independently developed an
isoconversional calculation method for nonisother-
Isothermal Isoconversional Methods. These meth-
mal data, which is commonly referred to as the
ods utilize the isothermal rate law (Eq. 16) and
OFW method. Taking the common logarithm of the
include the standard and Friedman’s isoconver-
nonisothermal rate law (Eq. 23) gives the following
sional methods.
equation,
Standard Isoconversional Method. This method38,54 AEa

log gðaÞ ¼ log þ log pðxÞ ð40Þ
can be derived by taking the logarithm of the bR
isothermal rate law (Eq. 16) to give: Substituting Doyle’s approximation (Eq. 24) in
Ea Eq. 40 gives,
ln gðÞ ¼ ln A þ ln t ð35Þ
RT AEa
log gðaÞ ¼ log 2:315 0:457x ð41Þ
which can be rearranged to, bR
Substituting Ea/RT for x and rearranging gives,
A Ea
ln t ¼ ln ð36Þ
gðÞ RT AEa Ea
log b ¼ log 2:315 0:457 ð42Þ
A plot of ln t versus 1/T for each a gives Ea from gðaÞR RT
the slope for that a regardless of the model
according to: A plot of ln b versus 1/T at each a yields Ea from
the slope for that a regardless of the model
A Ea according to:
ln t ¼ ln ð37Þ
gðÞ RT
Aa Eaa Eaa
log ba ¼ log 2:315 0:457 ð43Þ
55
gðaÞR RTa
Friedman’s Isoconversional Method. This method
is a differential method and was one of the first
isoconversional methods. The logarithm of the Modified Coats–Redfern Method. Burnham and
isothermal rate law (Eq. 15) gives, Braun58 have transformed the model-fitting
Coats–Redfern method to an isoconversional
d Ea method by rearranging Eq. 32 to:
ln ¼ ðln Af ðÞÞ ð38Þ
dt RT
b AR 2RTexp Ea
A plot of ln (da/dt) versus 1/T at each a gives Ea ln 2 ¼ ln 1 ð44Þ
T Ea gðaÞ Ea RT
from the slope for that a regardless of the model
according to:
A plot of ln b/T2 versus 1/T at each a yields Ea from

d Ea the slope for that a regardless of the model
ln ¼ ðln Af ðÞÞ ð39Þ according to:
dt RT

b Aa R 2RTexp Eaa
Nonisothermal Isoconversional Methods. Unlike ln 2 ¼ ln 1
T a Eaa gðaÞ Eaa RTa
isothermal data, nonisothermal data involve the
use of the temperature integral (Eq. 23). There- ð45Þ
Example of Isoconversional Calculations. Figure 7 A Ea A Ea

IðEa ; T1 Þ ¼ IðEa ; T2 Þ
depicts the application of two isoconversional b1 R b2 R
ð48Þ
methods to simulated kinetic data–isothermal A Ea
(Fig. 7a–standard method) and nonisothermal ¼ ... ¼ IðEa ; Tan Þ
bn R
(Fig. 7b–Ozawa–Flynn–Wall). The kinetic data
were simulated with Ea ¼ 100 kJ/mole and which reduces to:
A ¼ 1015/min and assuming an R3 model in IðEa ; T1 Þ IðEa ; T2 Þ IðEa ; Tn Þ
Figure 7a and an F1 model in Figure 7b. Figure 7a ¼ ¼ ... ¼ ¼s
b1 b2 bn
shows a-time (a–t) plots while Figure 7b shows a-
temperature (a–T) plots. Calculations for each plot ð49Þ
(a–t or a–T) were performed for a single conversion where, s is a constant.
value (i.e., isoconversional). This is shown as an For a two heating rate study, using two terms in
isoconversional line with a ¼ 0.8 in Figure 7a and Eq. 49 we get:
a ¼ 0.4 for Figure 7b. Values of time (Fig. 7a) or
temperature (Fig. 7b) from each isoconversional IðEa ; T1 Þ IðEa ; T2 Þ
line are tabulated below each a–t or a–T plots. ¼ ¼s ð50Þ
1 2
Plotting the last two rows (ln t vs. 103/T or log b vs.
103/T) of these tables according to Eqs. 37 or 43, If both sides are divided by either the right-hand
respectively, as shown in Figure 7 give Arrhenius- term or left-hand term, we get either:
like plots for both isoconversional methods. The
activation energy (Ea) is obtained from the slopes of 2 IðEa ; T1 Þ s
¼ ¼1 ð51Þ
these plots according to Eqs. 37 or 43. The 1 IðEa ; T2 Þ s
calculated Ea represents a single point (a ¼ 0.8 or
0.4) in an isoconversional (Ea –a) plot (circled Ea or
values in Fig. 7). Repeating this analysis for 1 IðEa ; T2 Þ s
different a values gives completed isoconversional ¼ ¼1 ð52Þ
2 IðEa ; T1 Þ s
plots in Figure 7.
Adding Eq. 51 and Eq. 52 gives:
Vyazovkin (VYZ) Method. The temperature inte-
2 IðEa ; T1 Þ 1 IðEa ; T2 Þ
gral (p(x)) in the nonisothermal rate law (Eq. 23) is þ ¼2 ð53Þ
a function of Ea and temperature. Therefore Eq. 23 1 IðEa ; T2 Þ 2 IðEa ; T1 Þ
can be written as, For three heating rates a similar equation can be
AEa obtained, as shown below:
gðaÞ ¼ IðEa ; TÞ ð46Þ
bR
2 IðEa ; T1 Þ 3 IðEa ; T1 Þ 1 IðEa ; T2 Þ
where, I(Ea, T) ¼ p(x). The general assumption þ þ
1 IðEa ; T2 Þ 1 IðEa ; T3 Þ 2 IðEa ; T1 Þ
used in Vyazovkin’s59 method (or any other iso-
IðEa ; T2 Þ 1 IðEa ; T3 Þ
conversional method) is that the reaction model is þ 3 þ
independent of the heating rate (i.e., g(a) will be 2 IðEa ; T3 Þ 3 IðEa ; T1 Þ
the same for any heating rate). Therefore, for a IðEa ; T3 Þ
þ 2 ¼6
conversion value (a), the relationship below could 3 IðEa ; T2 Þ
be defined if two heating rates are applied: ð54Þ
Aa Eaa Aa Eaa For ‘‘n’’ heating rates, Eqs. 53 and 54 can be
gðaÞ ¼ IðEaa ; Ta1 Þ ¼ IðEaa ; Ta2 Þ ð47Þ
b1 R b2 R generalized as,
where, b1 is the first heating rate, b2 is the second X

n X
n
j IðEa ; Ti Þ
heating rate, Ta1 is the temperature for a ¼ nðn 1Þ ð55Þ
i¼1 j6¼i
i IðEa ; Tj Þ
particular a using the first heating rate, Ta2 is
the temperature at the same a using the second or
heating rate, Eaa is the activation energy at that a !
and Aa is the frequency factor at that a. For an X
n X
n
j IðEa ; Ti Þ
experiment having ‘‘n’’ heating rates, the relation- nðn 1Þ ¼0 ð56Þ
i¼1 j6¼i
i IðEa ; Taj Þ
ship would be,
For experimental data, Eq. 56 might not converge, This method allows for use of linear (Eq. 18) and
but an Eaa which minimizes the left-hand side can nonlinear heating programs and also can be used
be found if the following form is used: for isothermal analysis (b ¼ 0, T(t) ¼ Ti according
to Eq. 18). For any heating program, the integral
X n X n
j IðEa ; Ti Þ
can be numerically evaluated using the trapezoi-
nðn 1Þ ¼O ð57Þ
IðEa ; Taj Þ
i¼1 j6¼i i dal method.
Using the same procedures that were employed
Minimizing Eq. 57 is equivalent to minimizing the for obtaining Eq. 58, it can be shown that all values
following: of g(a) from Eq. 61 are equal, therefore, all J(Eaa,

X n X n
j IðEa ; Tai Þ T(ta)) values are equal. Thus, equating J(Eaa, T(ta))

O¼ ð58Þ values and using the same logic as for Eqs. 47–58,
i¼1 j6¼i i IðEa ; Taj Þ
we obtain:

Minimization of Eq. 58 is equivalent to minimiz- X n X n
JðEa ; Ti ðt ÞÞ

ing Eq. 57 because summations contain pairs of O¼ ð62Þ
i¼1 j6¼i JðEa ; Tj ðt ÞÞ
inverse ratios. These inverse ratios can be easily
seen in Eqs. 53 and 54 which forces each ratio to a The activation energy at each a (Eaa) is the value
value of ‘‘1’’ during minimization. Vyazovkin used that minimizes O.
the 3rd or 4th degree Senum-Yang approximation
of the temperature integral. According to this Vyazovkin’s Advanced Isoconversional (AIC)
method, the activation energy (Eaa) at each a is Method. Vyazovkin60 introduced a further mod-
the value that minimizes O. ification to his isoconversional method. This modi-
fication involved integration over smaller time
Vyazovkin’s Modified Isoconversional Method. intervals. Therefore, Eq. 60 was altered to give:
Vyazovkin22 modified his isoconversional method
to account for random temperature variation. This Zta Eaa
modified method is not limited to linear heating ðgðaÞÞa ¼ Aa eRTðtÞ dt ð63Þ
programs and can be used to analyze kinetics from taDa
a nonlinear heating program and also isothermal
experiments. A modification was introduced to the where, Da ¼ (1/m) and m is the number of
nonisothermal rate law (Eq. 21) where the heating segments (typically 10–50) into which the inte-
rate (b) has been included in the integral to gration is divided. Eq. 63 can be generally
represent a heating function rather than a single expressed as,
temperature (T(t) vs. T). Since the heating func- ðgðÞÞ ¼ A J0 bEa ; Tðt Þc ð64Þ
tion is a time dependent function, the integral was
changed from a temperature integral (Eq. 21) to a As with the methods for obtaining Eqs. 58 and 62,
time integral as shown below: we can obtain:
Zt
Ea X n X n
J 0 ðEa ; Ti ðt ÞÞ
gðÞ ¼ A exp dt ð59Þ
RTðtÞ O¼ ð65Þ
i¼1 j6¼i J 0 ðEa ; Tj ðt ÞÞ
0
where, T(t) is the heating program used. For each As in the previous methods, the activation energy
a, Eq. 59 becomes, (Eaa) at each a is the value that minimizes O in the
Zt above equation. The advanced isoconversional
Ea method (AIC) is claimed to be superior to other
ðgðÞÞ ¼ A exp dt ð60Þ
RTðtÞ isoconversional methods60–62 because integration
0 over smaller time segments can better account for
which can be generally expressed as, systematic Ea variations.
ðgðÞÞ ¼ A JðEa ; Tðt ÞÞ ð61Þ
Complementary Model-Free/Modelistic Approach
where,
Khawam and Flanagan18 have proposed a com-
Zt
Ea plementary approach that uses both model-free
JðEa ; Tðt ÞÞ ¼ exp dt and model-fitting methods for kinetic data analy-
RTðtÞ
0 sis. This approach utilizes an isoconversional
method (Vyazovkin’s) to obtain Ea values, which nature of the solid sample or due to a complex
are compared to values obtained by a model- reaction mechanism.
fitting method (Coats–Redfern). The most accu-
rate model is assumed to be the one, which Elementary Reactions. If an elementary reaction
produces an activation energy closest to that shows variable activation energy during its pro-
from the isoconversional analysis. This approach gress, it may be attributed to a systematic change
allows one to select models that might otherwise in the reaction kinetics. This is not usual for
be indistinguishable based on quality of regres- homogeneous reactions which occur between
sion fit alone. Therefore, the strengths of both freely moving, identical reactant molecules with
methods are used in the evaluation of solid-state random collisional encounters that are usually
kinetics to obtain A and Ea values as well as the unaffected by product formation. However, react-
best model. ing entities in a solid sample are not isolated but
interact strongly with neighboring molecules or
CONTROVERSIES IN SOLID-STATE KINETICS particles. Therefore, during such a reaction,
reactivity may change due to product formation,
Discussions over solid-state kinetic studies have crystal defect formation, intracrystalline strain, or
caused numerous debates and controversies.63 other similar effects.73
Disagreements include questioning whether such Solid-state reactivity could also be affected by
kinetics have a good theoretical framework,64,65 experimental variables that would change the
as well as critiques of approximations or assump- reaction kinetics by affecting heat or mass transfer
tions used.66–68 Some of the controversies will be at a reaction interface. For example, temperature
discussed below. changes could affect the kinetics not only through
the rate constant but also by mechanistic changes.
Varying Activation Energy in Solid-State Kinetics Elementary reaction kinetics at one temperature
could be different from that at another. Complex
Solid-state kinetics was developed from reaction reaction kinetics (described below) could vary with
kinetics in homogenous systems (i.e., gases and temperature due to changes in the contribution of
liquids). The Arrhenius equation (Eq. 14) relates each elementary step73 or change in the rate-
the rate constant of a simple one-step reaction to determining step. Purge gas flow rate, is another
the temperature through the activation energy experimental variable that could affect reactivity
(Ea) and preexponential factor (A). It has been when reactions produce or consume gaseous
generally assumed that activation energy (Ea) components. A low purge flow rate may not
and frequency factor (A) remain constant, how- preclude the reversibility of a reaction compared
ever, it has been shown69–71 in solid-state reac- to a higher flow rate, which could reduce the
tions that these kinetic parameters may vary with reversibility and cause variability in the apparent
the reaction progress (a). This variation can be activation energy and/or introduce errors in
detected by isoconversional methods. While this calculated reaction rates.
variation appears to be in conflict with basic
chemical kinetic principles, in reality, it may not Complex Reactions. If two or more elementary
be. steps, each having a unique activation energy,
Khawam and Flanagan54,72 have shown that affect the rate of product formation, the reaction is
activation energy variation is of two types—a true usually considered complex.9 In such a reaction, a
variation that results from the complex nature of change in the activation energy as the reaction
the solid-state reaction or an artifactual one progresses would be observed. This change will
resulting from the use of some isoconversional depend on the contribution of each elementary
methods. step, which gives an ‘‘effective’’ activation energy
that varies with reaction progress. The effective
True Variation in Activation Energy activation energy can be mathematically derived
from the nonisothermal degradation rate law (Eq.
Many explanations have been suggested for the
15), by taking the natural logarithm followed by
occurrence of a true variation in activation
differentiation to give:
energy, both in homogenous phases70 and hetero-
geneous phases.73 In the solid-state, a variation in
d ln d
dt Ea
activation energy could be observed for an ¼ ð66Þ
elementary reaction due to the heterogeneous d T1 R
From this expression, the activation energy at b. Imperfection distribution—Different sam-

each conversion (Eaa) can be obtained. If a react- ples of the same material may have different
ion is composed of two parallel steps according to imperfection distributions. Therefore, no
the following scheme, two solid samples are identical, although
they may be similar.74 This could change the
degradation kinetic profiles of each sample.
c. Sublimation along with other reaction pro-
cesses.
d. Surface adsorption—desorption processes
on the reactants/products.
The overall reaction rate is, e. Diffusion of a gaseous product through the
sample.
da Ea1 Ea2 f. Rate of growth may vary along each crystal-
¼ k1 f1 ðaÞ þ k2 f2 ðaÞ ¼ A1 e RT f1 ðaÞ þ A2 e RT f2 ðaÞ
dt lographic axis of a nucleus.71
ð67Þ g. Particle size—If a solid-state reaction occurs
at surfaces or defect points, larger particles,
Taking the natural logarithm of both sides of which have a lower specific surface area,
Eq. 67 and differentiating gives: will be less reactive than smaller particles.
Ea
1
Ea
2
Various particle sizes could have different
d ln da
dt A1 Ea1 e RT f1 ðaÞ þ A2 Ea2 e RT f2 ðaÞ kinetic behavior, therefore, a variable par-
¼ ð68Þ
d T1 Ea
1
Ea
2
R A e RT f ðaÞ þ A e RT f ðaÞ ticle-sized sample could show complex reac-
1 1 2 2
tion behavior.
Since the left-hand side of Eq. 68 is equal to h. Particle or solid morphology—Degradation
(Ea)apparent/R (which is also (Ea)a/R from Eq. 66), kinetics of a spherical particle or spherical
Eq. 68 becomes: compact could differ from that of a cylind-
rical one. A nonhomogenous sample that
Ea Ea
1 2
A1 Ea1 e RT f1 ðaÞ þ A2 Ea2 e RT f2 ðaÞ contains several solid shapes may show
Eaapparent ¼ Eaa ¼ Ea Ea complex reaction behavior.
1 2
A1 e RT f1 ðaÞ þ A2 e RT f2 ðaÞ i. Localized melting—Melt degradation rates
ð69Þ usually differ from that of the solid produ-
cing variable reactivity throughout the
which can be generalized as: sample.73
Ea1 k1 f1 ðÞ þ Ea2 k2 f2 ðÞ
E a ¼ ð70Þ Artifactual Variation in Activation Energy
k1 f1 ðÞ þ k2 f2 ðÞ
Isoconversional methods, use several TGA or DSC
Eq. 70 shows that the effective activation energy data sets for kinetic analysis. Some of these
(Eaa) is a function of each Ea and a. methods are sensitive to experimental variables
Kinetic complexities are not limited to multiple such that changes in these variables produces
chemical steps. They may also include physical errors in calculated kinetic parameters like
processes that have different activation energies activation energy. When performing isothermal
such as: experiments, care should be taken to insure that
every run is done under the same experimental
a. Nucleation and growth—The energy barrier
conditions (i.e., sample weight, purge rate, sample
for nucleation could be relatively large
size distribution, particle morphology, etc.) so
compared to growth. Once a nucleus is
that temperature is the only variable for each run.
established, the rate of interface advance
Experimental variation can be minimized, but not
can be much lower than that for nucleation.
totally eliminated. For example, sample mass
There is no sharp demarcation where
may vary from one run to the next and affect a
nucleation stops and growth starts, since
reaction because,75
the two are interdependent. The contribu-
tion of nucleation may diminish as the 1. Larger masses cause larger endothermic or
reaction progresses, leading to an effective exothermic (self-heating or self-cooling)
activation energy that varies with the effects, producing larger deviations from
reaction progress. the programmed linear heating rate.
2. Diffusion rates through the sample will (i.e., independent variables in these equations)
change; gaseous diffusion is faster through linearly related to one another. As kinetics
a lower mass compared to a higher mass. developed, most of these equations except, for
3. Thermal gradients through the sample that of Arrhenius, disappeared because they were
might vary, especially when a powder has theoretically unsound.16 As a result, the contro-
a low thermal conductivity; larger samples versy over temperature dependency was finally
could contain regions where the tempera- put to rest, but the controversy and confusion
ture differs significantly from other regions. surrounding reaction rate temperature depen-
dence still affect researchers in heterogeneous
Similarly, sample packing could affect solid kinetics.67,76–81
reaction kinetics where loosely packed powders Galwey and Brown6 have shown that use of the
contain air pockets that can reduce thermal Arrhenius equation in heterogeneous kinetics
conductivity or trap evolved gasses compared to a is conceptually sound and theoretically well
more densely packed powder which would mini- founded. However, use of the Arrhenius equation
mize these problems. If any of the above effects in nonisothermal experiments is problematic
occur, a thermogram can be altered such that it because the temperature integral has no analyti-
falls above or below the expected thermogram for cal solution. Use of the temperature integral can be
isothermal studies. This would introduce errors in avoided by performing kinetic studies isother-
the calculated kinetic parameters obtained from mally, or using the differential form of the rate
some isoconversional methods. law.23 However, with our current computational
tools, approximating the temperature integral is
no longer a serious problem because the approx-
Temperature Dependence of the Rate Constant
imations can be as exact as the kinetic data
The temperature dependence of the rate constant demands.
is almost universally expressed by the Arrhenius
equation. However, historically, there was con-
Kinetic Compensation Effect
troversy surrounding the temperature depen-
dence of rate constants with many workers A kinetic compensation effect (KCE)82–84, is a
proposing several forms for rate constant tem- relationship between the activation energy (Ea)
perature dependency, as summarized in Table 3. and frequency factor (A) according to:
These equations were empirically derived based ln A ¼ bEa þ c ð71Þ
on quality of fit. Selecting an equation because it
gives a reasonable fit to the data is not a sufficient where, b and c are constants. This relationship is
reason for its acceptance, as most of the cited called a ‘‘compensation’’ because a change in the
equations will reasonably represent the same activation energy (Ea) is partially or completely
experimental data. This occurs because kinetic compensated by a change in the frequency factor
studies are most often conducted in a narrow (A). KCEs have been classified into three
temperature range, which makes 1/T, T, and ln T types:84,85
a. Type-1—Occurs from a difference in the

Table 3. Summary of Temperature Dependencies of
Rate Constants16
physicochemical properties of the sample,
which includes groups of different but rela-
Equationa Year Referenceb ted reactions. For example, groups of reac-
tants that have different substitutions on the
k ¼ A0 F T ð1 þ G0 TÞ 1850 Wilhelmy
same parent molecule, or different crys-
k ¼ AeDT 1862 Berthelot
k ¼ a0 þðBDT
b0 T22Þ 1881 Warder
talline reactants of the same compound con-
k ¼ Ae T 1883 Schwab taining different defects and impurities.85
k ¼ AeT
B
1889 Arrhenius b. Type-2—Occurs from a difference in the
C B
k ¼ AT e T 1893 Kooij experimental conditions applied to a parti-
k ¼ ATC ðBDT2 Þ 1895 Harcourt and Esson cular reactant’s kinetic studies which inclu-
k ¼ ATC e T 1898 Van’t Hoff des different atmospheres, sample masses,
a heating rate, etc.
A, A0 , B, C, D, F, G0 , a0 , and b0 are temperature independent
constants. c. Type-3—Occurs by using different computa-
b
Ref. 16 cites the original articles for these expressions. tional methods for kinetic analysis of the
same data set. The significance of this type and nonisothermal experiments usually are not
of compensation has been questioned and in agreement. On the other hand, nonisothermal
is considered a mathematical artifact.84,85 studies are considered more convenient than
Garn76–80 considered that this effect results isothermal studies because a sample is not
from using the Arrhenius equation. subjected to a rapid temperature rise to a reaction
temperature (i.e., heat-up time)50 in which reac-
Graphically, each Arrhenius plot (ln k vs. 1/T) tion could occur but not be measured, thus
gives a pair of A and Ea values. Several Arrhenius introducing errors in the analysis. This is espe-
plots for a series of reaction studies will give an cially true if the isothermal temperature is high
equal number of A, Ea pairs that can be used to because some decomposition probably occurs
construct a KCE plot (Fig. 8a). If a KCE exists, before the fixed temperature study is initiated.
then overlaying the Arrhenius curves reveals an It was noted34,37,50 that the disagreement in
isokinetic relationship (IKR)84 as seen in results from isothermal and nonisothermal
Figure 8b. The IKR is characterized by a point experiments should be expected because each
called the ‘‘isokinetic point’’ where all Arrhenius covers a different temperature range and due to
curves intersect. The rate constant at this point the complex nature of solid-state reactions,
(kiso) is equal for all Arrhenius plots and the kinetics at different temperatures could vary.
temperature at this point is called the ‘‘isokinetic The choice of isothermal or nonisothermal experi-
temperature’’ (Tiso). ments is governed by needs of the study, whether
From the Arrhenius equation (Eq. 14) and it is desired to study reaction kinetics over a wide
compensation effect (Eq. 71), the isokinetic point temperature range (i.e., up to melting) or if a
can be determined by, narrow range is sufficient. Also, the application of
1 such results to predicting solid-state stability of a
Tiso ¼ ð72Þ drug affects the choice of most appropriate
Rb
reaction conditions.
ln kiso ¼ c ð73Þ Calculation methods have also raised many
controversies because there are many methods
Vyazovkin and Lesniovich86 have used an IKR to and their range of application and validity is
calculate the frequency factor (A), which cannot unclear. Results obtained by various calculation
be obtained directly from isoconversional meth- methods have often been different, even when
ods. applied to the same data set. A critical evaluation
The compensation effect has been widely of these methods was necessary and was initiated
reported, both in homogenous and heterogeneous in the ICTAC ‘‘kinetic project’’36–40 (described
reactions.82 However, it remains an empirical below). However, more work needs to be done to
observation and has little theoretical justification. standardize the field of solid-state kinetics experi-
Galwey and Brown83 have summarized the two mentally, computationally, and conceptually.
extreme positions about the compensation effect—
either it is an artifact or it has a real chemical
significance. If it has real chemical significance, ICTAC KINETIC PROJECT
the Arrhenius relationship’s meaning is weakened
and some basic principles of chemical kinetics may Solid-state kinetics has been associated with
need to be reformulated. controversies and these issues needed to be
addressed scientifically. One such approach was
through the establishment of a ‘‘kinetic project’’
Analytical Methods
by several researchers. A kinetics workshop was
Many questions have been raised about analysis held during the 11th International Congress on
and calculation methods used to study solid-state Thermal Analysis and Calorimetry (ICTAC) in
kinetics1. The use of nonisothermal experiments Philadelphia, in August, 1996. One of the sugges-
has been criticized in favor of isothermal experi- tions of that workshop was to evaluate the various
ments for two reasons—firstly, temperature is an calculation methods in a consistent and scientific
experimental variable in nonisothermal analyses fashion by establishing a global kinetic project.
while it is fixed in isothermal analyses, which The project distributed kinetic data sets to
reduces the total number of variables; secondly, volunteer participants to perform their data
kinetic parameters obtained by both isothermal analysis.87,88 Eight sets of kinetic data consisting
of real and simulated isothermal and nonisother- reviews were those by Garrett90,91, Lachman92,
mal experiments were distributed. Experimental and Carstensen93. Interest has recently increased
data were distributed in six sets for the decom- in pharmaceutical solids, their properties and, in
position of calcium carbonate and ammonium many cases, their degradation, dehydration or
perchlorate under nitrogen and vacuum both transformation kinetics.94 Several kinetic studies
isothermally and nonisothermally. Two data sets of pharmaceutical solids, which have appeared
were simulated isothermally and nonisothermally are described below.
using two equally-weighted, parallel, first-order
reactions (A1 ¼ 1010/min, Ea1 ¼ 80 kJ/mole; A2 ¼ Desolvation Kinetics
1015/min, Ea2 ¼ 120 kJ/mole).
The purpose of the project was to evaluate the Wyandt and Flanagan95 studied the desolvation
same data with different calculation methods and kinetics of sulfonamide-ammonia adducts. Sev-
make judgments based on the kinetic analysis eral nonisothermal model-fitting methods were
results. Participants were not limited to any evaluated96, which eventually resulted in the use
particular calculation method. Many such meth- of the direct differential method. A correlation
ods were used to analyze the kinetic data sets was found between calculated desolvation activa-
including: Ozawa–Flynn–Wall, Kissinger, Fried- tion energies of the solid sulfonamide and its pKa.
man, Coats–Redfern, direct differential, and Drugs with lower pKa values where found to have
many others. Some of these methods were incor- higher activation energies and vice versa. This
porated in software packages such as: TA-KIN1, finding was attributed to an acid-base-type inter-
NETZSCH1 thermokinetics, KINETICS1, and action between the sulfonamide (acid) and ammo-
AKTS-TA1.36 nia (base) in the solid-state. The pKa of the drug
Results for kinetic calculations showed that was found to inversely affect the strength of the
similar computational methods were in agreement ammonia-drug interaction, which affected desol-
among different laboratories. Analysis of simu- vation activation energy.
lated data showed that isoconversional methods Zhu and Grant97 studied the dehydration
produced results that were in general agreement kinetics of nedocromil magnesium pentahy-
with Ea values ranging from 80 to 120 kJ/mole; drate,98–100 a drug used in the treatment of
model-fitting results of simulated data gave single asthma. Dehydration kinetics was studied iso-
Ea values intermediate between the two Ea values thermally by TGA and nonisothermally using
used thus did not reveal the kinetic complexity of DSC. The conventional model-fitting method was
two parallel pathways.38 Also the kinetic results used to analyze isothermal data while the Kis-
for the same reaction under various experimental singer method was used for nonisothermal kinetic
conditions were different (i.e., comparing nitrogen analysis. Dehydration occurred in two steps where
and vacuum atmospheres for calcium carbonate the first step involved loss of four water molecules
and ammonium perchlorate). while the second step involved the loss of a single
The main conclusions from this project were water molecule. Dehydration kinetic results from
that the kinetic description of a process strongly isothermal TGA (conventional method) agreed
depends on the experimental conditions.89 In with those from nonisothermal DSC (Kissinger
addition, multiheating rate methods should be method). The first dehydration step followed the
employed to obtain reliable kinetic descriptions A2 model with Ea ¼ 70 kJ/mole (TGA) and 63 kJ/
and any kinetic process must be described by the mole (DSC) while the second dehydration step
complete kinetic triplet.37,89 followed the A3 model with Ea ¼ 121 kJ/mole
The project succeeded in bridging differences of (TGA) and 112 kJ/mole (DSC). The higher Ea of
data analysis in solid-state kinetics. The project the second step suggested that the fifth water
was a first step in ‘‘standardizing’’ solid-state molecule was more tightly bound in the crystal.
kinetic analysis methods. This project should be They also investigated the effect of particle size,
further expanded to cover other controversial sample weight and vapor pressure on the stability
areas in solid-state kinetics. of the drug hydrates. Dehydration Ea values from
DSC increased with increasing particle size and
PHARMACEUTICAL APPLICATIONS decreasing sample weight for both dehydration
steps. Changes in Ea with particle size were
Solid-state kinetics has not been recently attributed to changing the surface area/volume
reviewed in the pharmaceutical literature. Early ratio, which could affect dehydration kinetics.
Changes in Ea with sample weight were attributed model-fitting methods. Dehydration kinetics was
to self-cooling effects of larger sample masses for found to follow the phase boundary model (R2)
the endothermic reaction in addition to the vary- with an activation energy of about 68 kJ/mole.
ing sample geometry. When water vapor pressure was varied, the degra-
Nonisothermal dehydration kinetics of nedo- dation model changed from R2 to the A3 model, in
cromil sodium trihydrate was investigated by some cases, while no model could be fitted to other
Zhou et al.101. They reported that dehydration cases. With the aid of PXRD and DSC, they con-
occurred in two steps, the first step involved the cluded that dehydration under different water
loss of two water molecules while the second step vapor pressures alter the crystallographic form of
involved the loss of the third water molecule. They the drug. They claimed that water vapor pressure
studied dehydration kinetics by DSC and saw by can have opposing effects on the rate of dehydra-
thermomicroscopy that water escaped along the tion. Increased water vapor pressure can decrease
needle axis in the first dehydration step, while this the driving force (i.e., shift the reaction towards
directionality was not observed for the second the hydrate), reducing the reaction rate. On the
dehydration step. From isoconversional plots other hand, water can function as a plasticizer in
using Vyazovkin’s advanced isoconversional the solid and increase molecular mobility, increas-
(AIC) method, they obtained a conversion range ing nucleation and reaction rates. The overall
(0.2 < a < 0.8) having a relatively constant activa- dehydration rate is the sum of these two opposing
tion energy (70 kJ/mole). They then applied a effects of moisture.
model-fitting method (Coats–Redfern) to analyze
data within this conversion range. Their results
Degradation Kinetics
showed the R1 model to be the best fit for the
first dehydration step, however, no model gave Long et al.103 have studied the nonisothermal
a satisfactory fit for the second dehydration step, kinetics of anhydrous aspirin degradation in the
which had an Ea that varied between 130 and melt by TGA. They reported that degradation
140 kJ/mole. They suggested that a model- occurs in two steps, first; by formation of linear
fitting approach alone was insufficient for such oligomers of acetylsalicylic acid, which are con-
kinetic analysis. They also studied the effects of verted to cyclic oligomers in the second degrada-
particle size and purge gas flow rate on the tion step. Kinetics were analyzed using
dehydration kinetics and found that decreasing Vyazovkin’s AIC method and showed that the
particle size had little effect on the Ea value of apparent Ea decreases from 120 kJ/mole
the first dehydration step while it raised the Ea of (a ¼ 0.01) to 25 kJ/mole (a ¼ 0.5) and increases
the second dehydration step and increased its in the second degradation step to 130 kJ/mole
variability. This abnormal behavior was attribu- (a ¼ 0.93). From the calculated activation energy,
ted to formation of coherent particle aggregates they were able to make isothermal kinetic predic-
from which water escape was impeded. Changing tions of aspirin’s stability at different tempera-
the purge gas flow rate had similar effects on tures. Their predicted time for 5% decomposition
calculated Ea values where lower flow rates (10 at 308C (874 days) was similar to the labeled
mL/min, compared to 70 and 150 mL/min) raised shelf-life determined by the typical stability
the Ea value and increased Ea variability, espe- protocols. By combining thermal analysis and
cially for the second dehydration step. This mass spectrometry, they were able to propose a
behavior was attributed to slower water escape degradation scheme in which aspirin produces
from surface of the solid compared to its center. acetic acid and linear oligomers of acetylsalicylic
Finally, they were able to rationize dehydration acid, which were then converted into cyclic
behavior based on crystal structure where chan- oligomers.104
nels were observed through which water mole- Rodante et al.105,106 have studied decomposition
cules could move. kinetics of structurally related penicillins by
Han and Suryanarayanan102 studied the influ- simultaneous TGA-DSC analysis. They conducted
ence of temperature and water vapor pressure on both isothermal and nonisothermal experiments
the dehydration kinetics of carbamazepine dihy- and extracted kinetic parameters from each by
drate, an antiepileptic drug. Dehydration was model-fitting and isoconversional methods. Their
studied isothermally by TGA, DSC, and powder results showed that the decomposition of oxacillin
X-ray diffraction (PXRD). Dehydration kinetic and cloxacillin occurs in a single step while that
parameters were obtained from TGA data with for dicloxacillin, benzylpenicillin, and ampicillin
occur in two steps. Carbenicillin decomposes in SUMMARY

three steps. All decompositions were preceded by a
dehydration step. They have shown that solid- We have demonstrated in this review that solid-
state penicillin decomposition processes were state reaction kinetics is both a unique and
complex, and such complexity is not revealed complex area of research.
isothermally, even when isoconversional methods It is unique because of its significant deviation
are employed. from homogenous phase kinetic processes. Solid-
state kinetics is affected by particle size, crystal
defects, crystal strain, and other solid properties
Crystallization Kinetics
not relevant to liquid or gas phase processes. One
Zhou et al.33 used DSC to study crystallization must carefully interpret solid-state kinetic results
kinetics of amorphous nifedipine, a calcium and incorporate interpretive caveats that reflect
channel blocker under both isothermal and non- these perturbations. Generalization of such
isothermal conditions. For each condition, kinetic results is at the same time easy and difficult. It is
analysis was performed by model-fitting (conven- easy because the investigator can use the straight
tional and Coats–Redfern) and isoconversional forwarded extrapolation of activation energy,
(Vyazovkin’s AIC) methods. Based on the results frequency factor and model to other temperatures
of isoconversional methods, model-fitting meth- or conditions. It is difficult because kinetic results
ods were performed in conversion ranges where can depend upon a myriad of solid-state character-
the activation energy was relatively constant istics making it risky to make such extrapolations
(0.05 < a < 0.8). Isothermal and nonisothermal without knowing how these factors interact to
results from model-free analysis were comparable affect the reaction(s) of interest.
(Ea 130 kJ/mole) and model-fitting methods This area is complex because of the mathema-
showed that nifedipine crystallization followed tical tools and models used to interpret solid-state
the Avarami-Erofe’ev nucleation (A4) model kinetic data. There is a rich array of kinetic models
which transforms to A3 then A2 models as that arise from the nonisotropic nature of the solid-
crystallization proceeds. state. Also, many kinetic studies are carried out
Umeda et al.107 studied the polymorphic transi- under nonisothermal conditions, which further
tion kinetics of tolbutamide (form B to A) and complicates an already complex kinetic picture.
mefenamic acid (Form I to II) isothermally by DSC. We have attempted to summarize the range of
Kinetic analysis was performed by a model-fitting experimental methods (i.e., isothermal and non-
method described by Hancock and Sharp.108 They isothermal) and the attendant mathematical
showed that the polymorphic transition from approaches used to analyze such data. This review
forms B to A for tolbutamide followed a three- has not been exhaustive but rather representative
dimensional diffusion model (D3) with an activa- of the common tools and models used. Even though
tion energy of about 37 kcal/mole (156 kJ/mol). we have focused on reactions involving weight loss
Whereas, the polymorphic transition from Forms I (i.e., TGA), many of the tools are applicable to other
to II for mefenamic acid followed a zero-order thermal methods such as DSC. The only require-
model (F0) with an activation energy about 86 ment is the ability to convert collected data to
kcal/mole (360 kJ/mol). degree of reaction (a) versus time or temperature.
Kitamura et al.109 studied the effect of grinding For the pharmaceutical scientist, our review
on solid-state stability of cefixime trihydrate an will hopefully serve as an introduction into the
oral cephalosporin antibacterial, using DSC, TGA, realm of solid-state reaction kinetics. The use of
and powder X-ray diffraction. Kinetic analysis was such results to make extrapolations or conclusions
performed nonisothermally using the Kissinger about solid drug stability under ambient condi-
method. The dehydration Ea for intact cefixime tions awaits further development. The question
trihydrate was 72 kcal/mol (301 kJ/mole) which ‘‘what can I do with these results?’’ is presently
was reduced to 68 kcal/mole (285 kJ/mole) after difficult to answer in a general sense. For specific
a 4-h grinding. They concluded that grinding cases, answers may be generated for narrow
alters the bonding force between water and applications. Generality in solid-state kinetics of
cefixime molecules, which was reflected in a drugs awaits further investigation of the factors
decrease in the calculated dehydration activation affecting such processes and the successful extra-
energy for cefixime trihydrate and therefore, a polation to predicting API or formulation stability
decrease in its stability. characteristics.
We hope that this review has narrowed the gap ortho-diynylarenes to polynaphthalene networks.
between the thermochemical kinetics and phar- A comparison of calorimetric methods. Polymer
maceutical science worlds. We have cited exam- 41:4415–4422.
ples of solid-state kinetic investigations on 13. Jacobs PWM, Tompkins FC. 1955. Classification
pharmaceutical solids. Also, we have demon- and theory of solid reactions. In: Garner WE,
editor. Chemistry of the solid state. New York:
strated in a pedagogic fashion how collected
Academic Press, pp 184–212.
kinetic data (isothermal or nonisothermal) can be 14. Brown ME, Dollimore D, Galwey AK. 1980.
transformed or analyzed to obtain modelistic or Theory of solid state reaction kinetics. In: Bam-
model-free results. Finally, pharmaceutical inves- ford CH, Tipper CFH, editors. Chemical kinetics.
tigators are challenged to have a better under- Amsterdam: Elsevier, pp 41–72.
standing of the models, mathematical tools 15. Galwey AK, Brown ME. 1999. Thermal decomposi-
and software they apply to solid-state kinetic tion of ionic solids: chemical properties and
reactions. reactivities of ionic crystalline phases. Amsterdam:
Elsevier, pp 75–115.
16. Laidler KJ. 1984. The development of the Arrhe-
REFERENCES nius equation. J Chem Educ 61:494–498.
17. Sestak J, Berggren G. 1971. Kinetics of the
1. Brown ME. 1988. Introduction to thermal analy- mechanism of solid-state reactions at increasing
sis: techniques and applications. London: Chap- temperatures. Thermochim Acta 3:1–12.
man and Hall, pp 127–151. 18. Khawam A, Flanagan DR. 2005. Complementary
2. Morris KR. 1999. Structural aspects of hydrates use of model-free and modelistic methods in the
and solvates. In: Brittain HG, editor. Polymorph- analysis of solid-state kinetics. J Phys Chem B
ism in pharmaceutical solids. New York: Marcel 109:10073–10080.
Dekker, pp 125–181. 19. Zoglio MA, Windheuser JJ, Vatti R, Maulding HV,
3. Bauer J, Spanton S, Henry R, Quick J, Dziki W, Jr., Kornblum SS, Jacobs AL, Hamot H. 1968.
Porter W, Morris J. 2001. Ritonavir: an extra- Linear nonisothermal stability studies. J Pharm
ordinary example of conformational polymorph- Sci 57:2080–2085.
ism. Pharm Res 18:859–866. 20. Maulding HV, Jr., Zoglio MA. 1970. Flexible
4. Chemburkar SR, Bauer J, Deming K, Spiwek H, nonisothermal stability studies. J Pharm Sci 59:
Patel K, Morris J, Henry R, Spanton S, Dziki W, 333–337.
Porter W, Quick J, Bauer P, Donaubauer J, 21. Galwey AK, Brown ME. 1999. Thermal decom-
Narayanan BA, Soldani M, Riley D, McFarland position of ionic solids: chemical properties and
K. 2000. Dealing with the impact of ritonavir reactivities of ionic phases. Amsterdam: Elsevier,
polymorphs on the late stages of bulk drug process pp 139–171.
development. Org Process Res Dev 4:413–417. 22. Vyazovkin S. 1997. Evaluation of activation
5. Paul IC, Curtin DY. 1973. Thermally induced energy of thermally stimulated solid-state reac-
organic reactions in the solid state. Acc Chem Res tions under arbitrary variation of temperature.
6:217–225. J Comput Chem 18:393–402.
6. Galwey AK, Brown ME. 2002. Application of the 23. Flynn JH. 1997. The ‘Temperature Integral’—Its
Arrhenius equation to solid state kinetics: can this use and abuse. Thermochim Acta 300:83–92.
be justified? Thermochim Acta 386:91–98. 24. Doyle CD. 1961. Kinetic analysis of thermogravi-
7. Byrn SR, Pfeiffer RR, Stowell JG. 1999. Solid- metric data. J Appl Polym Sci 5:285–292.
state chemistry of drugs, 2nd edn. West Lafayette: 25. Doyle CD. 1962. Estimating isothermal life from
SSCI, Inc., pp 279–304. thermogravimetric data. J Appl Polym Sci 6:639–
8. Galwey AK. 2000. Structure and order in thermal 642.
dehydrations of crystalline solids. Thermochim 26. Doyle CD. 1965. Series approximations to the
Acta 355:181–238. equation of thermogravimetric data. Nature 207:
9. Steinfeld JI, Francisco JS, Hase WL. 1999. 290–291.
Chemical kinetics and dynamics, 2nd edn. Upper 27. Senum GI, Yang RT. 1977. Rational approxima-
Saddle River, NJ: Prentice Hall. tions of the integral of the Arrhenius function. J
10. Mullin JW. 2001. Crystallization, 4th edn. Oxford: Therm Anal 11:445–447.
Butterworth-Heinemann, pp 1–31. 28. Gautschi W, Cahill WF. 1964. Exponential inte-
11. Brown ME. 2001. Introduction to thermal analy- gral and related functions. In: Abramowitz M,
sis: techniques and applications, 2nd edn. Amster- Stegun I, editors. Handbook of mathematical
dam: Kluwer, pp 181–214. functions with formulas and mathematical tables.
12. Shah HV, Babb DA, Smith DW, Jr. 2000. Washington, DC: National Bureau of Standards,
Bergman cyclopolymerization kinetics of bis- pp 227–237.
29. Weisstein EW. ‘‘Exponential Integral.’’ From 42. Achar BNN, Brindley GW, Sharp JH. 1966. Int
MathWorld–A Wolfram Web Resource. http:// Clay Conf, Jerusalem. 67 p.
mathworld.wolfram.com/ExponentialInte- 43. Freeman ES, Carroll B. 1958. The application of
gral.html thermoanalytical techniques to reaction kinetics.
30. Weisstein EW. ‘‘ExpIntegralE.’’ From MathWorld– The thermogravimetric evaluation of the kinetics
A Wolfram Web Resource. http://functions.wol- of the decomposition of calcium oxalate monohy-
fram.com/GammaBetaErf/ExpIntegralE/10/0001/ drate. J Phys Chem 62:394–397.
31. Perez-Maqueda LA, Criado JM. 2000. The accu- 44. Freeman ES, Carroll B. 1969. Interpretation of
racy of Senum and Yang’s approximations to the kinetics of thermogravimetric analysis. J Phys
the Arrhenius integral. J Therm Anal 60:909– Chem 73:751–752.
915. 45. Coats AW, Redfern JP. 1964. Kinetic parameters
32. Urbanovici E, Popescu C, Segal E. 1999. On the from thermogravimetric data. Nature 201:68–69.
accuracy of Senum and Yang’s fourth degree 46. Coats AW, Redfern JP. 1965. Kinetic parameters
rational approximation of the temperature inte- from thermogravimetric data II. J Polym Sci Part
gral. J Therm Anal 55:325–327. B: Polym Lett 3:917–920.
33. Zhou DL, Schmitt EA, Zhang GG, Law D, 47. Kissinger HE. 1956. Variation of peak tempera-
Vyazovkin S, Wight CA, Grant DJW. 2003. ture with heating rate in differential thermal
Crystallization kinetics of amorphous nifedipine analysis. J Res Nat Bur Stand 57:217–221.
studied by model-fitting and model-free 48. Kissinger HE. 1957. Reaction kinetics in differen-
approaches. J Pharm Sci 92:1779–1792. tial thermal analysis. Anal Chem 29:1702–1706.
34. Vyazovkin S, Wight CA. 1997. Isothermal and 49. Elder JP. 1985. The general applicability of the
nonisothermal reaction kinetics in solids: in Kissinger equation in thermal-analysis. J Therm
search of ways toward consensus. J Phys Chem Anal 30:657–669.
A 101:8279–8284. 50. Vyazovkin S, Wight CA. 1999. Model-free and
35. Zhou DL, Grant DJW. 2004. Model dependence model-fitting approaches to kinetic analysis of
of the activation energy derived from nonisother- isothermal and nonisothermal data. Thermochim
mal kinetic data. J Phys Chem A 108:4239– Acta 341:53–68.
4246. 51. Simon P. 2004. Isoconversional methods—funda-
36. Brown ME, Maciejewski M, Vyazovkin S, Nomen mentals, meaning and application. J Therm Anal
R, Sempere J, Burnham A, Opfermann J, Strey R, 76:123–132.
Anderson HL, Kemmler A, Keuleers R, Janssens 52. Zsako J. 1973. Kinetic analysis of thermogravi-
J, Desseyn HO, Li CR, Tang TB, Roduit B, Malek metric data. VI. Some problems of deriving kinetic
J, Mitsuhashi T. 2000. Computational aspects of parameters from TG curves. J Therm Anal 5:239–
kinetic analysis. Part A: the ICTAC kinetics 251.
project-data, methods and results. Thermochim 53. Zsako J. 1996. Kinetic analysis of thermogravi-
Acta 355:125–143. metric data. XXIX. Remarks on the ‘many curves’
37. Maciejewski M. 2000. Computational aspects of methods. J Therm Anal 46:1845–1864.
kinetic analysis. Part B: the ICTAC Kinetics 54. Khawam A, Flanagan DR. 2005. Role of isoconver-
Project—the decomposition kinetics of calcium sional methods in varying activation energies of
carbonate revisited, or some tips on survival in solid-state kinetics: I. Isothermal kinetic studies.
the kinetic minefield. Thermochim Acta 355:145– Thermochim Acta 429:93–102.
154. 55. Friedman HL. 1964. Kinetics of thermal degrada-
38. Vyazovkin S. 2000. Computational aspects of tion of char-forming plastics from thermogravi-
kinetic analysis. Part C. The ICTAC Kinetics metry-application to a phenolic resin. J Polym Sci
Project—the light at the end of the tunnel? Part C: Polym Lett 6:183–195.
Thermochim Acta 355:155–163. 56. Ozawa T. 1965. A new method of analyzing ther-
39. Burnham AK. 2000. Computational aspects mogravimetric data. Bull Chem Soc Jpn 38:1881–
of kinetic analysis. Part D: the ICTAC kine- 1886.
tics project—multi-thermal-history model-fitting 57. Flynn JH, Wall LA. 1966. A quick, direct method
methods and their relation to isoconversional for the determination of activation energy from
methods. Thermochim Acta 355:165–170. thermogravimetric data. J Polym Sci Part B:
40. Roduit B. 2000. Computational aspects of kinetic Polym Lett 4:323–328.
analysis. Part E: the ICTAC Kinetics Project- 58. Burnham AK, Braun RL. 1999. Global kinetic
numerical techniques and kinetics of solid state analysis of complex materials. Energy Fuels 13:
processes. Thermochim Acta 355:171–180. 1–22.
41. Sharp JH, Wentworth SA. 1969. Kinetic analysis 59. Vyazovkin S, Dollimore D. 1996. Linear and
of thermogravimetric data. Anal Chem 41:2060– nonlinear procedures in isoconversional computa-
2062. tions of the activation energy of nonisothermal
reactions in solids. J Chem Inf Comput Sci 36: 76. Garn PD. 1975. Kinetic compensation effect.
42–45. J Therm Anal 7:475–478.
60. Vyazovkin S. 2001. Modification of the integral 77. Garn PD. 1976. The kinetic compensation effect.
isoconversional method to account for variation in J Therm Anal 10:99–102.
the activation energy. J Comput Chem 22:178– 78. Garn PD. 1978. Kinetic parameters. J Therm Anal
183. 13:581–593.
61. Budrugeac P, Segal E. 2001. Some methodological 79. Garn PD. 1988. Kinetics of decomposition of the
problems concerning nonisothermal kinetic ana- solid state: is there really a dichotomy? Thermo-
lysis of heterogeneous solid-gas reactions. Int J chim Acta 135:71–77.
Chem Kinet 33:564–573. 80. Garn PD. 1990. Kinetics of thermal decomposition
62. Vyazovkin S. 2002. Some confusion concerning of the solid state. II. Delimiting the homogeneous-
integral isoconversional methods that may result reaction model. Thermochim Acta 160:135–145.
from the paper by Budrugeac and Segal ‘‘Some 81. Arnold M, Veress GE, Paulik J, Paulik F. 1981.
methodological problems concerning nonisother- The applicability of the Arrhenius model in
mal kinetic analysis of heterogeneous solid-gas thermal analysis. Anal Chim Acta 124:341–350.
reactions’’. Int J Chem Kinet 34:418–420. 82. Galwey AK. 1977. Compensation effect in hetero-
63. Brown ME. 1997. Steps in a minefield—some geneous catalysis. Adv Catal 26:247–322.
kinetic aspects of thermal analysis. J Therm Anal 83. Galwey AK, Brown ME. 1997. Arrhenius para-
49:17–32. meters and compensation behaviour in solid-state
64. Galwey AK. 2004. Is the science of thermal decompositions. Thermochim Acta 300:107–115.
analysis kinetics based on solid foundations? A 84. Koga N. 1994. A review of the mutual dependence
literature appraisal. Thermochim Acta 413:139– of Arrhenius parameters evaluated by the ther-
183. moanalytical study of solid-state reactions: the
65. Galwey AK, Brown ME. 2000. Solid-state decom- kinetic compensation effect. Thermochim Acta
positions—stagnation or progress? J Therm Anal 244:1–20.
60:863–877. 85. Brown ME, Galwey AK. 2002. The significance of
66. Galwey AK. 1999. Some misconceptions in the ‘‘compensation effects’’ appearing in data pub-
kinetic analyses of pharmaceutical systems. J lished in ‘‘computational aspects of kinetic analy-
Pharm Pharmacol 51:879–886. sis’’: ICTAC project, 2000. Thermochim Acta
67. Galwey AK. 2003. Eradicating erroneous Arrhe- 387:173–183.
nius arithmetic. Thermochim Acta 399:1–29. 86. Vyazovkin SV, Lesnikovich AI. 1988. Estimation
68. Galwey AK. 2003. Perennial problems and pro- of the pre-exponential factor in the isoconver-
mising prospects in the kinetic analysis of non- sional calculation of effective kinetic parameters.
isothermal rate data. Thermochim Acta 407:93– Thermochim Acta 128:297–300.
103. 87. Brown ME, Maciejewski M, Vyazovkin S. 1997.
69. Vyazovkin S, Wight CA. 1997. Kinetics in solids. Kinetics analysis project. Thermochim Acta
Annu Rev Phys Chem 48:125–149. 307:201–203.
70. Vyazovkin S. 2000. On the phenomenon of vari- 88. Brown ME, Maciejewski M, Vyazovkin S. 1998.
able activation energy for condensed phase reac- Kinetics analysis project. J Therm Anal 51:327–
tions. New J Chem 24:913–917. 332.
71. Vyazovkin S. 2000. Kinetic concepts of thermally 89. Maciejewski M, Reller A. 1987. How (un)reliable
stimulated reactions in solids: a view from a are kinetic data of reversible solid-state decom-
historical perspective. Int Rev Phys Chem 19: position processes. Thermochim Acta 110:145–152.
45–60. 90. Garrett ER. 1962. Prediction of stability of drugs
72. Khawam A, Flanagan DR. 2005. Role of isoconver- and pharmaceutical preparations. J Pharm Sci
sional methods in varying activation energies of 51:811–833.
solid-state kinetics: II. Nonisothermal kinetic 91. Garrett ER. 1967. Kinetics and mechanisms in
studies. Thermochim Acta 436:101–112. stability of drugs. Adv Pharm Sci 2:1–94.
73. Galwey AK. 2003. What is meant by the term 92. Lachman L. 1965. Physical and chemical stability
‘variable activation energy’ when applied in the testing of tablet dosage forms. J Pharm Sci 54:
kinetic analyses of solid state decompositions 1519–1526.
(crystolysis reactions)? Thermochim Acta 397: 93. Carstensen JT. 1974. Stability of solids and solid
249–268. dosage forms. J Pharm Sci 63:1–14.
74. Brown ME, Brown RE. 2000. Kinetic aspects of 94. Glass BD, Novak C, Brown ME. 2004. The
the thermal stability of ionic solids. Thermochim thermal and photostability of solid pharmaceuti-
Acta 357-358:133–140. cals—a review. J Therm Anal 77:1013–1036.
75. Wendlandt WW. 1986. Thermal analysis, 3rd edn. 95. Wyandt CM, Flanagan DR. 1992. Solid-state
New York: Wiley. 33 p. nonisothermal kinetics of sulfonamide ammonia
adduct desolvation. Thermochim Acta 196:379– 103. Long GT, Vyazovkin S, Gamble N, Wight CA.
389. 2002. Hard to swallow dry: kinetics and mechan-
96. Wyandt CM, Flanagan DR. 1992. A critical ism of the anhydrous thermal decomposition of
evaluation of three non-isothermal kinetic techni- acetylsalicylic acid. J Pharm Sci 91:800–809.
ques. Thermochim Acta 197:239–248. 104. Long GT, Vyazovkin S, Gamble N, Wight CA.
97. Zhu HJ, Grant DJW. 2001. Dehydration behavior 2001. Hard to swallow dry: formation of linear and
of nedocromil magnesium pentahydrate. Int J cyclic oligomers in the anhydrous thermal decom-
Pharm 215:251–262. position of acetylsalicylic acid. J Chem Soc, Perkin
98. Zhu HJ, Khankari RK, Padden BE, Munson EJ, Trans 2:436–437.
Gleason WB, Grant DJW. 1996. Physicochemical 105. Rodante F, Vecchio S, Tomassetti A. 2002. Multi-
characterization of nedocromil bivalent metal salt step decomposition processes for some antibiotics:
hydrates. 1. Nedocromil magnesium. J Pharm Sci a kinetic study. Thermochim Acta 394:7–18.
85:1026–1034. 106. Rodante F, Vecchio S, Tomassetti M. 2002. Kinetic
99. Zhu HJ, Padden BE, Munson EJ, Grant DJW. analysis of thermal decomposition for penicillin
1997. Physicochemical characterization of nedo- sodium salts—model-fitting and model-free meth-
cromil bivalent metal salt hydrates. 2. Nedocromil ods. J Pharm Biomed Anal 29:1031–1043.
zinc. J Pharm Sci 86:418–429. 107. Umeda T, Ohnishi N, Yokoyama T, Kuroda T,
100. Zhu HJ, Halfen JA, Young VG, Padden BE, Kita Y, Kuroda K, Tatsumi E, Matsuda Y. 1985.
Munson EJ, Menon V, Grant DJW. 1997. Physi- Studies on drug nonequivalence. 14. A kinetic-
cochemical characterization of nedocromil biva- study on the isothermal transition of polymorphic
lent metal salt hydrates. 3. Nedocromil calcium. J forms of tolbutamide and mefenamic-acid in the
Pharm Sci 86:1439–1447. solid-state at high-temperatures. Chem Pharm
101. Zhou DL, Schmitt EA, Zhang GGZ, Law D, Wight Bull 33:2073–2078.
CA, Vyazovkin S, Grant DJW. 2003. Model-free 108. Hancock JD, Sharp JH. 1972. Method of com-
treatment of the dehydration kinetics of nedocro- paring solid-state kinetic data and its application
mil sodium trihydrate. J Pharm Sci 92:1367– to the decomposition of kaolinite, brucite, and
1376. barium carbonate. J Am Ceram Soc 55:74–77.
102. Han J, Suryanarayanan R. 1998. Influence of 109. Kitamura S, Miyamae A, Koda S, Morimoto Y.
environmental conditions on the kinetics and 1989. Effect of grinding on the solid-state stabi-
mechanism of dehydration of carbamazepine lity of cefixime trihydrate. Int J Pharm 56:125–
dihydrate. Pharm Dev Technol 3:587–596. 134.

Basics and Applications of Solid State Kinetics A Pharmaceutical Perspective

Uploaded by

Copyright:

Available Formats

You might also like

Basics and Applications of Solid State Kinetics A Pharmaceutical Perspective

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Basics and Applications of Solid State Kinetics A Pharmaceutical Perspective

Uploaded by

Copyright:

Available Formats

REVIEW

Basics and Applications of Solid-State Kinetics:

Received 2 June 2005; revised 20 September 2005; accepted 31 October 2005

INTRODUCTION tions have many forms, however, those that

472 JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 95, NO. 3, MARCH 2006

Figure 1. Schematic representation of reactivity (a) homogenous system; (b) hetero-

Figure 3. Isothermal a-time plots of different solid-

For an isothermal DSC/DTA analysis, the

where, AUCt0 is the sample peak area from 0 to t

Figure 3. (Continued ) (c–e) deceleratory, (f) sigmoidal.

Figure 4. Methods for studying solid-state kinetics.

MODELS AND MECHANISMS IN

A model is a theoretical and/or mathematical

Figure 6. Isothermal model fitting method (conventional method): (a) simulated a-

Figure 7. Isoconversional methods for evaluating solid-state kinetics: (a); Standard

where, m, n, and p are constants for a particular

B. Geometrical contraction—Nucleation is There are many methods used to study solid-state

Model Differential Forma f ðÞ ¼ 1k d

Isothermal Method da/dt is the isothermal reaction rate and dt/dT

and the temperature integral then becomes,

Standard Isoconversional Method. This method38,54 AEa

Example of Isoconversional Calculations. Figure 7 A Ea A Ea

where, b1 is the first heating rate, b2 is the second X

From this expression, the activation energy at b. Imperfection distribution—Different sam-

a. Type-1—Occurs from a difference in the

occur in two steps. Carbenicillin decomposes in SUMMARY

You might also like