J. Phys. Chem.

B 2006, 110, 17315-17328 17315

REVIEW ARTICLES

Solid-State Kinetic Models: Basics and Mathematical Fundamentals

Ammar Khawam and Douglas R. Flanagan*

DiVision of Pharmaceutics, College of Pharmacy, UniVersity of Iowa, Iowa City, Iowa 52242
ReceiVed: May 5, 2006; In Final Form: June 25, 2006

Many solid-state kinetic models have been developed in the past century. Some models were based on
mechanistic grounds while others lacked theoretical justification and some were theoretically incorrect. Models
currently used in solid-state kinetic studies are classified according to their mechanistic basis as nucleation,
geometrical contraction, diffusion, and reaction order. This work summarizes commonly employed models
and presents their mathematical development.

1. Introduction fraction. For a gravimetric measurement, R is defined by

Fifty years have passed since the publication of Chemistry m0 - mt

of the Solid State, edited by W. E. Garner,1 which covered R) (2)
m 0 - m∞
theories of the solid state, including solid-state reaction kinetics.
Jacobs and Tompkins2 covered theories and derivations of some where, m0 is initial weight, mt is weight at time t, and m∞ is
solid-state reaction models in Garner’s text, specifically, final weight.
nucleation and nuclei growth models. The derivation of these Kinetic parameters (model, A, Ea) can be obtained from
and other models has also appeared in Volume 22 of the isothermal kinetic data by applying the above rate law (eq 1).
Chemical Kinetics Series entitled, “Reactions in the Solid State” Alternatively, eq 1 can be transformed into a nonisothermal rate
by Brown et al.3 Later, Galwey and Brown4 presented many of expression describing reaction rate as a function of temperature
these same models. These and other older references are at a constant heating rate by utilizing the following:
becoming less accessible because they have been out of print
for many years. Researchers who seek the basis for these models dR dR dt
) (3)
and their mathematical foundations must find such old texts or dT dt dT
access even older journal articles. Unfortunately, no single
reference comprehensively presents the basics and mathematical where, dR/dT is the nonisothermal reaction rate, dR/dt is the
development of these models. The lack of such a source causes isothermal reaction rate, and dT/dt is the heating rate (β).
authors to redundantly present reaction models in tabular Substituting eq 1 into eq 3 gives the differential form of the
form5-10 as shown in Table 1. It is rare to find a solid-state nonisothermal rate law,
kinetic report that does not list such reaction models because
of the lack of a general source to which reference can be made.
dR A -(Ea/RT)
) e f(R) (4)
dT β
This review is intended to provide a summary and math-
ematical basis for commonly used reaction models in solid- Separating variables and integrating eqs 1 and 4 gives the
state kinetics. integral forms of the isothermal and nonisothermal rate laws,
respectively:
2. Reaction Rate Laws
g(R) ) Ae-(Ea/RT) t (5)
The rate of a solid-state reaction can be generally described
by and

A
dR
) Ae-(Ea/RT) f(R) (1) g(R) )
β
∫0T e-(E /RT) dT
a
(6)
dt
where, g(R) is the integral reaction model, defined by
where, A is the preexponential (frequency) factor, Ea is the
activation energy, T is absolute temperature, R is the gas dR
constant, f(R) is the reaction model, and R is the conversion g(R) ) ∫0R f(R)
* Corresponding author: E-mail address: douglas-flanagan@uiowa.edu. It should be noted that there is no consensus on the symbols
Fax: (319) 335-9349. used to represent these models, so the definitions of f(R) and
10.1021/jp062746a CCC: $33.50 © 2006 American Chemical Society
Published on Web 08/15/2006
17316 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan

Ammar M. N. Khawam received his B.S. in Pharmacy from the Jordan Douglas R. Flanagan received his B.S. in Pharmacy and M.S. and
University of Science and Technology in 1997. After working in the Ph.D. degrees in Pharmaceutical Chemistry, all from the University of
pharmaceutical industry, he joined the graduate program at the Michigan. He has been on the faculty at the University of Connecticut
University of Iowa and is completing his Ph.D. degree in Pharmaceutics. and is presently Professor of Pharmaceutics at the University of Iowa.
His research interests include physical properties of solids, drug stability, Dr. Flanagan’s research interests include solid-state properties of drugs,
and thermal analysis applications to pharmaceutical development. diffusion and dissolution phenomena, controlled release formulation,
and biodegradable polymeric systems.

g(R) may be reversed in some publications. Several reaction

3. Models and Mechanisms in Solid-State Kinetics
models (f(R) and g(R)) are listed in Table 1.
Reaction kinetics in the solid-state are often studied by A model is a theoretical, mathematical description of what
thermogravimetry, but they can also be studied by other occurs experimentally. In solid-state reactions, a model can
analytical methods such as differential scanning calorimetry describe a particular reaction type and translate that mathemati-
(DSC),11,12 powder X-ray diffraction (PXRD),13 and nuclear cally into a rate equation. Many models have been proposed in
magnetic resonance (NMR).14,15 For any analytical method, the solid-state kinetics, and these models have been developed based
measured parameter must be able to be transformed into a on certain mechanistic assumptions. Other models are more
conversion fraction (R) that can be used in the kinetic equations. empirically based, and their mathematics facilitates data analysis
Kinetic analysis (isothermal or nonisothermal) can be per- with little mechanistic meaning. Therefore, different rate expres-
formed by either model-fitting or model-free (isoconversional) sions are produced from these models.
methods.11 The use of isoconversional or model-free methods In homogeneous kinetics (e.g., gas or solution phases), kinetic
has increased recently12,16-21 due to the ability of these methods studies are usually directed toward obtaining rate constants that
to calculate Ea values without modelistic assumptions. However, can be used to describe the progress of a reaction. Additionally,
such methods have some disadvantages,22,23 and a reaction the reaction mechanism is typically investigated and rate
model is usually needed for a complete kinetic description of constant changes with temperature, pressure, or reactant/product
any solid-state reaction.24 Different solid-state kinetic analysis concentrations are often helpful in uncovering mechanisms.
methods have been recently reviewed.11 These mechanisms involve to varying degrees the detailed

TABLE 1: Solid-State Rate and Integral Expressions for Different Reaction Models
differential form integral form
model f(R) ) 1/k dR/dt g(R) ) kt
nucleation models
power law (P2) 2R1/2 R1/2
power law (P3) 3R2/3 R1/3
power law (P4) 4R3/4 R1/4
Avrami-Erofeyev (A2) 2(1 - R)[-ln(1 - R)]1/2 [-ln(1 - R)]1/2
Avrami-Erofeyev (A3) 3(1 - R)[-ln(1 - R)]2/3 [-ln(1 - R)]1/3
Avrami-Erofeyev (A4) 4(1 - R)[-ln(1 - R)]3/4 [-ln(1 - R)]1/4
Prout-Tompkins (B1) R(1 - R) ln[R/(1 - R)] + ca
geometrical contraction models
contracting area (R2) 2(1 - R)1/2 1 - (1 - R)1/2
contracting volume (R3) 3(1 - R)2/3 1 - (1 - R)1/3
diffusion models
1-D diffusion (D1) 1/(2R) R2
2-D diffusion (D2) -[1/ln(1 - R)] ((1 - R)ln(1 - R)) + R
3-D diffusion-Jander (D3) [3(1 - R)2/3]/[2(1 - (1 - R)1/3)] (1 - (1 - R)1/3)2
Ginstling-Brounshtein (D4) 3/[2((1 - R)-1/3 - 1)] 1 - (2/3)R - (1 - R)2/3
reaction-order models
zero-order (F0/R1) 1 R
first-order (F1) (1 - R) -ln(1 - R)
second-order (F2) (1 - R)2 [1/(1 - R)] - 1
third-order (F3) (1 - R)3 (1/2)[(1 - R)-2 - 1]
a
Constant of integration.
Review Article J. Phys. Chem. B, Vol. 110, No. 35, 2006 17317

chemical steps by which a reactant is converted to product. Nucleation is the formation of a new product phase (B) at
However, in solid-state kinetics, mechanistic interpretations reactive points (nucleation sites) in the lattice of the reactant
usually involve identifying a reasonable reaction model25 (A). Nucleation rates have been derived based on one of two
because information about individual reaction steps is often assumptions:2 nucleation is single- or multistepped (Table 2).
difficult to obtain. However, the choice of a reaction model Single-step nucleation assumes that nucleation and nuclei
should ideally be supported by other complementary techniques growth occur in a single step. For N0 potential nucleation sites
such as microscopy, spectroscopy, X-ray diffraction, etc.26 (having equal nucleation probability), once the nuclei (N) are
formed, they grow and the rate of nucleation is a simple first-
4. Model Classification
order process according to
Models are generally classified based on the graphical shape
of their isothermal curves (R vs t or dR/dt vs R) or on their dN
mechanistic assumptions. Based on their shape, kinetic models ) kN(N0 - N) (8)
dt
can be grouped into acceleratory, deceleratory, linear, or
sigmoidal models (Figure 1). Acceleratory models are those in where, N is the number of growth nuclei present at time, t, and
which the reaction rate (dR/dt) is increasing (e.g., accelerating) kN is the nucleation rate constant. Separating variables and
as the reaction proceeds (Figure 1a); similarly, deceleratory integrating eq 8 gives
reaction rates decrease with reaction progress (Figure 1b-d)
while the rate remains constant for linear models (Figure 1e),
and sigmoidal models show a bell-shaped relationship between N ) N0(1 - e-kNt) (9)
rate and R (Figure 1f). Nonisothermally, R-temperature plots
are not as distinctive in their shapes as they are isothermally. Differentiation of eq 9 gives the exponential rate of nucleation:
Figure 2 shows the nonisothermal R-temperature and dR/dT vs
R plots generated using eqs 4 and 6. Models in Figures 1 and dN
) kNN0e-kNt (10)
2 can be displayed on a single graph using a reduced-time plot dt
(isothermal data)27 or a master plot method (for nonisothermal
data).7 These methods for graphical presentation are easy means When kN is small, the exponential term in eq 10 is ∼1 and the
of visually determining the most appropriate model for a rate of nucleation is approximately constant, producing a linear
particular data set. rate of nucleation
Based on mechanistic assumptions, models are divided into
nucleation, geometrical contraction, diffusion, or reaction-order dN
(Table 1). ) kN N 0 (11)
dt
5. Model Derivation
However, when kN is very large, the rate of nucleation is very
Model derivation is based on several proposed reaction high, indicating that all nucleation sites are rapidly or instantly
mechanisms which include nucleation, geometric shape, diffu- nucleated producing an instantaneous rate of nucleation:
sion, and reaction order. Sestak and Berggren28 have suggested
a mathematical form that represents all models in a single dN
general expression: )∞ (12)
dt
g(R) ) Rm(1 - R)n(-ln(1 - R))p (7)
On the other hand, multistep nucleation assumes that several
where m, n, and p are constants. By assigning values for these
distinct steps are required to generate a growth nucleus.40
three variables, any model can be expressed. Derivations and
Accordingly, formation of product B will induce strain within
theoretical implications of specific models are discussed below.
the lattice of A, rendering small aggregates of B unstable and
5.1. Nucleation and Nuclei Growth Models. The kinetics
of many solid-state reactions have been described by nucleation causing them to revert back to reactant A. Strain can be
models, specifically, the Avrami models. These reactions include overcome if a critical number (mc) of B nuclei are formed.
crystallization,29-31 crystallographic transition,32 decomposi- Therefore, two types of nuclei can be defined: germ and growth
tion,33,34 adsorption,35,36 hydration,37 and desolvation.24 Skrdla nuclei. A germ nucleus is submicroscopic with B particles below
and Robertson38 have recently suggested a model that describes the critical number (m < mc), which will either revert back to
sigmoidal R-time curves based on the Maxwell-Boltzmann reactant A or grow to a growth nucleus, which is a nucleus
energy distribution and incorporates two rate constants: one with B particles exceeding the critical number (m > mc) of
for the acceleratory and one for the deceleratory region of the particles allowing further reaction (i.e., nuclei growth). There-
R-time curve. fore, a germ nucleus must accumulate a number of product
5.1.1. Nucleation. Crystals have fluctuating local energies molecules, “p”, before it is converted to a growth nucleus. The
from imperfections due to impurities, surfaces, edges, disloca- rate constant (ki) for addition of individual molecules in a
tions, cracks, and point defects.39 Such imperfections are sites nucleus up to p molecules (e.g., n < p) is assumed to be constant
for reaction nucleation since the reaction activation energy is or k0 ) k1 ) k2 ) k3 ) ... ) kp-1 ) ki (i.e., rate constant for
minimized at these points. Thus, they are called, nucleation addition of each molecule). After p molecules have been
sites.2,32 accumulated (e.g., n g p), the rate constant (kg) for further
A common reaction in solid-state kinetics follows the scheme: nucleus growth by addition of further molecules (>p) becomes,
kp ) kp+1 ) kp+2 ) kp+4 ) ... ) kg. It is assumed that the rate
A(s) f B(s) + C(g)
of nucleus growth is more than that of nucleus formation (i.e.,
where, a solid “A” decomposes thermally to produce a solid kg > ki). Therefore, according to Bagdassarian,40 if β successive
“B” and gas “C”. events are necessary to form the growth nucleus, and each event
17318 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan
Review Article J. Phys. Chem. B, Vol. 110, No. 35, 2006 17319

Figure 1. Isothermal dR/dt time and R time plots for solid-state reaction models (Table 1); data simulated with a rate constant of 0.049 min-1:
(a) acceleratory; (b-d) deceleratory; (e) constant; (f) sigmoidal.

has a probability equal to ki, then the number of nuclei formed 2). Equation 14 was also derived by Allnatt and Jacobs,41 but
at time, t, is they assumed unequal rate constants for addition of successive
molecules to the growth nuclei before reaching the critical size
N0(kit)β (n ) p).
N) ) Dtβ (13)
β! 5.1.2. Nuclei Growth. The nuclei growth rate (G(x)) can be
represented by the nuclei radius formed from growth. Nuclei
where, D ) N0(kit)β/β!. After differentiation, eq 13 becomes growth rates usually vary with size.2,4 For example, growth rates
dN of small nuclei (often submicroscopic) would be different from
) Dβtβ-1 (14) that of large nuclei. Low growth rates are due to the instability
dt
of very small nuclei (germ nuclei), which revert to reactants.
Equation 14 represents the power law of nucleation2,40 (Table The radius of a stable nucleus (growth nucleus) at time, t,
17320 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan
Review Article J. Phys. Chem. B, Vol. 110, No. 35, 2006 17321

Figure 2. Nonisothermal dR/dT and R temperature plots for solid-state reaction models (Table 1); data simulated with a heating rate of 10 K/min,
frequency factor of 1 × 1015 min-1, and activation energy of 80 kJ/mol. (a) P-models; (b) D-models; (c-e) F and R models; (f) A-models.

(r(t,t0)), is V(t) ) σ[r(t,t0)]λ (16)

r(t,t0) ) ∫t t G(x) dx
0
(15) where, λ is the number of growth dimensions (i.e., λ ) 1, 2, or
3), σ is the shape factor (i.e., 4π/3 for a sphere), and r is the
where, G(x) is the rate of nuclei growth and t0 is the formation radius of a nucleus at time t. Equation 16 gives the volume
time of a growth nucleus. occupied by a single nucleus; the total volume occupied by all
In addition to nucleus radius, two important considerations nuclei (V(t)) can be calculated by combining nucleation rate
in nuclei growth are also considered: nucleus shape (σ) and (dN/dt) and growth rate (V(t,t0)) equations while accounting for
growth dimension (λ). When these are considered, they describe different initial times of nucleus growth (t0):
nuclei growth rate through the volume occupied by individual
nuclei (V(t)). Therefore, a stable nucleus formed at time (t0) V(t) ) ∫0t V(t)(dN
dt )t)t
dt0 (17)
occupies a volume V(t) at time t according to 0
17322 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan

TABLE 2: Mathematical Expressions for Nucleation Rates

differential form integral form
nucleation rate law dN/dT N
exponentiala kNN0e-kNt N0(1 - e-kNt)
lineara kNN0 kNN0t
instantaneousa ∞ N0
powerb Dβtβ-1 Dtβ
a
Single-step nucleation. b Multistep nucleation.

where, V(t) is the volume of all growth nuclei and dN/dt is the Figure 3. Two types of nuclei growth restrictions: black dots are
nucleation rate. Substituting eq 15 into 16 and eq 16 into 17 nucleation sites; shaded areas are nuclei growth regions.
gives
models are usually applied to the analysis of the acceleratory
V(t) ) ∫0t σ(∫t tG(x) dx)λ (dN
0 dt )t)t
dt0 (18) period of a curve.
0 5.1.4. The AVrami-ErofeyeV (A) Models. In any solid-state
decomposition, there are certain restrictions on nuclei growth.
The above equation may be integrated for any combination of Two such restrictions have been identified4 (Figure 3). (a)
nucleation and/or growth rate laws to give a rate expression of Ingestion - elimination of a potential nucleation site by growth
the form (g(R) ) kt) as listed in Table 1. However, this is not of an existing nucleus; ingested sites never produce a growth
always possible since there is no functional relationship4 nucleus due to their inclusion in a growth nucleus. Ingested
between the nucleation and growth terms. Therefore, assump- nuclei are called “phantom” nuclei. (b) Coalescence - loss of
tions about nucleation (dN/dt) and growth (V(t)) rate equations reactant/product interface when reaction zones of two or more
must be made as described below. growing nuclei merge.
5.1.3. Power Law (P) Models. For a simple case where An expression relating the number of nuclei sites is42
nucleation rate follows the power law (eq 14) and nuclei growth
is assumed constant (G(x) ) kG), eq 18 becomes N1(t) ) N0 - N(t) - N2(t) (27)

V(t) ) ∫0t σ(kG(t - t0))λ(Dβt0β-1) dt0 (19) where N0 is the total number of possible nuclei-forming sites,
N1(t) is the actual number of nuclei at time t, N2(t) is the number
Evaluating the integral in eq 19 gives2 of nuclei ingested, and N(t) is the number of nuclei activated
(i.e., developed into growth nuclei). From eq 27, a nucleation

V(t) ) σkGλDβtβ+λ 1 - ( λβ
β+1
+
λ(λ - 1) β
2! β+2
... , λ e3 ) rate (dN/dt) known as the modified exponential law can be
developed.2 However, if this nucleation rate is substituted into
eq 17, the resulting expression does not have an analytical
(20)
solution.2 To deal with this issue, an extended conversion
If D′ ) Dβ(1 - λβ/(β + 1) + λ(λ - 1)/2! β/(β + 2)...) and n fraction (R′) was proposed,43 which is the conversion fraction
) β+λ, eq 20 becomes previously defined in eq 25 (R ) [kt]n) that neglects ingestion
(i.e., accounts for active and phantom nuclei) and nuclei
V(t) ) σkGλD′tn (21) coalescence. Therefore, R′ g R. Values of R can be evaluated
by determining their relation to values of R′.
Since, V(t) is directly proportional to the reaction progress (R), The extended conversion fraction (R′) was related to the actual
R can be represented as conversion fraction (R) by Avrami44 who obtained

R ) V(t) × C (22) dR
dR′ ) (28)
(1 - R)
where C is a constant equal to 1/V0 (V0 initial volume). From
eqs 21 and 22 we obtain which, upon integration, gives

R ) σkGλCD′tn (23) R′ ) - ln(1 - R) (29)

Substituting the value of (R′) from eq 25 into eq 29 gives
which can be rewritten as
(kt)n ) -ln(1 - R) (30)
R ) ((σkGλCD′)1/n)ntn (24)
which can be rearranged to
If k ) (σkGλCD′)1/n, eq 24 can be written as
[-ln(1 - R)]1/n ) kt (31)
n
R ) (kt) (25)
Erofeyev (Erofe’ev or Erofeev)45 followed a different ap-
Equation 25 can be rearranged to proach to derive a special case of eq 31 for n ) 3. Therefore,
eq 31 was attributed to both Avrami and Erofeyev and represents
(R)1/n ) kt (26) different Avrami-Erofeyev (A) models (Table 1) for different
values of n. These “A” models are also called the JMAEK
Equation 26 represents the various power law (P) models (Table models, which stands for Johnson, Mehl, Avrami, Erofeyev,
1). Since these models assume constant nuclei growth without and Kholmogorov, in recognition of the researchers that have
any consideration to growth restrictions (explained below), these contributed to their development.4
Review Article J. Phys. Chem. B, Vol. 110, No. 35, 2006 17323

5.1.5. Autocatalytic Models. In homogeneous kinetics, auto- Since Prout and Tompkins assumed that Ri ) 0.5, eq 39 reduces
catalysis occurs when the products catalyze the reaction; this to
occurs when the reactants are regenerated during a reaction in
what is called “branching”. The reactants will eventually be dR
) kBR(1 - R) (40)
consumed and the reaction will enter a “termination” stage dt
where it will cease. A similar observation can be seen in solid-
state kinetics. Autocatalysis occurs in solid-state kinetics if Separating variables and integrating eq 40 gives
nuclei growth promotes continued reaction due to the formation
R
of imperfections such as dislocations or cracks at the reaction ln ) kBt + c (41)
interface (i.e., branching). Termination occurs when the reaction 1-R
begins to spread into material that has decomposed.46 Prout and where c is the integration constant.
Tompkins47 derived an autocatalysis model (B1) for the thermal Equation 41 is the Prout-Tompkins (B1) model (Table 1)
decomposition of potassium permanganate which produced which well fits the thermal degradation of solid potassium
considerable crystal cracking during decomposition. permanganate. It should be noted that, unlike other models, the
In autocatalytic reactions, the nucleation rate can be defined integration of eq 40 was performed without limits, simply
by because with a lower limit (R ) 0), the value is negative infinity.
As a result, the integration constant appears in the Prout-
dN
) kNN0 + (kB - kT)N (32) Tompkins equation (eq 41). One of the limitations in some
dt literature on this equation is that it is reported without the
where, kB is the branching rate constant and kT is the termination constant term as
rate constant. If kNN0 is neglected, eq 32 becomes R
ln ) k Bt (42)
dN 1-R
) (kB - kT)N (33)
dt This causes confusion since eq 42 will give negative time values
for R <0.5 (Figure 4). To overcome this problem, an integration
This could occur in one of two cases: (1) kN is very large so
constant, c, in eq 41 is needed which shifts the curve toward
that initial nucleation sites are depleted rapidly and calculations
positive time values. There is no general criterion for what the
of dN/dt are valid for time intervals after N0 sites are depleted.
integration constant should be; however, Prout and Tompkins
(2) kN is very small so that kNN0 can be ignored.
used tmax, which is the time needed for the maximum rate (i.e.,
The reaction rate is related to number of nuclei by
the inflection point) and which is approximately the same as
dR t1/2 used by Carstensen.48 We have used a time equivalent to R
) k'N (34) ) 0.01 (30.21 min) for our simulation (Figure 4), but other
dt
values would be equally valid.
where, k′ is the reaction rate constant. Prout and Tompkins found The Prout-Tompkins model has been criticized because of
that the shape of R vs time plots for the degradation of potassium the assumptions required for its derivation; other forms of it
permanganate was sigmoidal. Therefore, an inflection point have been proposed.46,49-51 Skrdla52 considered nucleation and
exists (Ri,ti) at which dN/dt will change signs. From the branching as two separate processes (independent but coupled)
boundary conditions that need to be satisfied at that inflection having two different rate constants and proposed an autocatalytic
point (i.e., kB ) kT), the following can be defined: rate expression. The proposed expression gives the Prout-
Tompkins model if the nucleation and branching rate constants
k T ) kB ()
R
Ri
(35)
are equal.52 Guinesi et al.53 have shown that titanium(IV)-
EDTA decarboxylates in two steps, the first being the B1 model
while the second is the R3 model.
Substituting eq 35 into eq 33 gives 5.2. Geometrical Contraction (R) Models. These models
assume that nucleation occurs rapidly on the surface of the
dN
dt ( R
) kB 1 - N
Ri ) (36)
crystal. The rate of degradation is controlled by the resulting
reaction interface progress toward the center of the crystal.
Depending on crystal shape, different mathematical models may
Since dN/dR ) dN/dt‚dt/dR, eq 37 can be obtained: be derived. For any crystal particle the following relation is
applicable:
dN
dR (
) k'' 1 -
R
Ri ) (37) r ) r0 - kt (43)

where k′′ ) kB/k′. Assuming kB is independent of R, separating where r is the radius at time t, r0 is the radius at time t0, and k
variables in eq 37 and integrating gives is the reaction rate constant. If a solid particle is assumed to
have cylindrical or spherical/cubical shapes (Figure 5), the
contracting cylinder (contracting area) or contracting sphere/
(
N ) k′′ R -
R2
2Ri ) (38) cube (contracting volume) models,54 respectively, can be
derived. Dehydration of calcium oxalate monohydrate was
shown to follow geometrical contraction models.55-57
Substituting eq 38 into 34 gives 5.2.1. The Contracting Cylinder (Contracting Area) Model
- R2. For a cylindrical solid particle, the volume is hπr2, where
dR
dt (
) kB R -
R2
2Ri ) (39) h is the cylinder height and r is the cylinder radius. For “n”
particles, the volume is nhπr2. Since, weight ) volume ×
17324 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan

Therefore, for n reacting particles:

nFhπr02 - nFhπr2
R) (45)
nFhπr02

Equation 45 reduces to

(
R) 1-
r2
r02 ) (46)

Substituting the value of r from eq 43 gives

( ) r0 - kt 2
R)1- (47)
r0
Figure 4. Isothermal R time plots for the Prout-Tompkins reaction which can be rearranged to
model (Table 1); data simulated with a rate constant of 0.152 min-1:
(9) data simulated according to eq 42; (0) data simulated according
to eq 41 where c ) tmax (30.21 min).
k
1-R) 1- t
r0 ( )
2
(48)

If ko ) k/r0, eq 48 becomes the contracting cylinder model:

1 -(1 - R)1/2 ) kot (49)

5.2.2. The Contracting Sphere/Cube (Contracting Volume)

Model - R3. If a solid particle has a spherical or cubical shape,
a contracting sphere/cube model can be derived. A sphere has
a volume of 4πr3/3. For n particles, the volume is 4nπr3/3. Since
weight ) volume × density (F), the weight of n spherical
particles is
4
weight ) nFπr3 (50)
3
Equation 44, for a reaction involving n particles becomes
4 4
nFπr03 - nFπr3
3 3
R) (51)
4
nFπr03
3
which reduces to

R) 1-
( ) r3
r03
(52)

Substituting for r from eq 43 gives

( ) r0 - kt 3
R)1- (53)
r0

which can be rearranged to

k
1-R) 1- t
r0 ( )
3
(54)

Figure 5. Geometrical crystal shapes: (a) cylinder; (b) sphere; (c) If ko ) k/r0, eq 54 becomes the contracting sphere model as
cube.
1 -(1 - R)1/3 ) kot (55)
density (F), the weight of “n” cylindrical particles is nFhπr2.
From the earlier definition of conversion fraction (R, eq 2) and A similar approach for cubic crystals leads to the same general
assuming m∞ ≈ 0, we obtain expression. It should be emphasized that particle size is
incorporated in the rate constant (k) for these models and for
m 0 - mt other models where geometry of the solid crystal is part of the
R) (44) mathematical derivation (e.g., diffusion models). Therefore, a
m0 sample of varying particle size will have variable reaction rate
Review Article J. Phys. Chem. B, Vol. 110, No. 35, 2006 17325

Figure 7. Schematic representation of a spherical particle reaction.

Figure 6. One-dimensional diffusion through a flat plane.62 A and B
are reactants, AB is the product interface, l is the thickness of the Separating variables and integrating eq 57 gives
interface AB, and x is the distance measured from interface Q into
AB.
MAB(C2 - C1)
l2 ) 2D t (58)
constants, which will cause R-time or R-temperature curves to MBF
shift. This would produce a curved isoconversional plot if an
isoconversional method is used for kinetic analysis;22,23 sieving If k ) 2D[MAB(C2 - C1)]/MBF, eq 58 becomes
the solid sample will greatly reduce this effect. Particle size
effects on the shape of R-time/temperature plots has been l2 ) kt (59)
described by Koga and Criado.58,59
5.3. Diffusion (D) Models. One of the major differences Equation 59 is known as the parabolic law.62
The simplest
between homogeneous and heterogeneous kinetics is the mobil- rate equation is for an infinite flat plane that does not involve
ity of constituents in the system. While reactant molecules are a shape factor (e.g., one-dimensional), where the conversion
usually readily available to one another in homogeneous fraction (R) is directly proportional to product layer thickness,
systems, solid-state reactions often occur between crystal lattices “l”. Therefore, eq 59 becomes
or with molecules that must permeate into lattices where motion
R2 ) k′t (60)
is restricted and may depend on lattice defects.60 A product layer
may increase where the reaction rate is controlled by the where k′ is a constant. Equation 60 represents the one-
movement of the reactants to or products from the reaction dimensional diffusion (D1) model.
interface. Solid-state reactions are not usually controlled by mass The three-dimensional diffusional (D3) model is based on
transfer except for a few reversible reactions or when large the assumption of spherical solid particles (Figure 7). The
evolution or consumption of heat occurs. Diffusion usually plays conversion fraction for a reaction involving n spherical particles
a role in the rates of reaction between two reacting solids, when using eqs 44 and 50 is
reactants are in separate crystal lattices.3 Wyandt and Flanagan61
have shown that desolvation of sulfonamide-ammonia adducts 4 4
nFπR3 - nFπ(R - x)3
follows diffusion models. A correlation was found between 3 3
calculated desolvation activation energies of the ammonia R) (61)
4
adducts and the sulfonamide’s intrinsic acidity. This finding was nFπR3
3
attributed to an acid-base-type interaction between the sul-
fonamide (acid) and ammonia (base) in the solid state. The pKa where x is the thickness of the reaction zone. Upon simplifica-
of the drug was found to inversely relate to the strength of the tion, eq 61 becomes
ammonia-drug interaction, which in turn affected desolvation
3
activation energy.
In diffusion-controlled reactions, the rate of product formation
R)1- (R -R x) (62)
decreases proportionally with the thickness of the product barrier
Equation 62 can be rearranged to
layer. For metallic oxidation, this involves a moving boundary
and is considered a “tarnishing reaction”60,62 which is depicted
x ) R(1 -(1 - R)1/3) (63)
in Figure 6. According to Figure 6, the mass of B moving across
P (unit area) in time, dt, to form product AB is Jander63 used the parabolic law (eq 59) to define x. Therefore,
subsitituing eq 63 (after squaring x) into eq 59 gives
dl MAB dC
) -D (56)
dt MBF dx R2(1 - (1 - R)1/3)2 ) kt (64)

where, MAB and MB are the molecular weights of AB and B, Assuming k′ ) k/R2, eq 64 becomes the D3 (Jander) model:
respectively, D is the diffusion coefficient, F is the density of
the product (AB), l is the thickness of the product layer (AB), (1 -(1 - R)1/3)2 ) k′t (65)
C is the concentration of B in AB, and x is the distance from Ginstling-Brounshtein have shown64 that the Jander model
interface Q into AB. Assuming a linear concentration gradient (eq 65) which used the parabolic law (derived for a plane
of B in AB, dC/dx|x)l ) -(C2 - C1)/l, where C2 and C1 are surface) is oversimplified and holds only at low conversion
the concentrations of B at interfaces P and Q, respectively, eq values (i.e., low x/R values). The steady-state solution of Fick’s
56 becomes first law for radial diffusion in a sphere is65

dl MAB (C2 - C1) aC1(b - r) + bC2(r - a)

)D (57) C(r) ) (66)
dt MB F l r(b - a)
17326 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan

where C(r) is the reactant concentration at a particular value of

r (a < r < b), C1 is the concentration of the diffusing species
at surface r ) a, and C2 is the concentration of the diffusing
species at surface r ) b. The reaction at the interface is assumed
to occur at a much faster rate than diffusion, therefore, C1 ≈ 0.
Therefore, eq 66 becomes

bC2(r - a)
C(r) ) (67)
r(b - a)
Figure 8. Schematic representation of a cylindrical particle reaction.
Taking the derivative of the above equation with respect to r at
approach used for the D3 model. For a cylindrical particle, eq
r ) a gives
63 is defined as
dC (b - a)bC2
| ) (68) x ) R(1 - (1 - R)1/2) (75)
dr r)a a(b - a)2
If Jander’s approach is followed, the resulting equation is
According to Figure 7, a ) R-x and b ) R, so that eq 68
becomes (1 -(1 - R)1/2)2 ) k′t (76)

dC RC2 where, k′ ) k/R2. Equation 76 is not the D2 model usually cited

) (69) in the literature. The usual D2 model is derived following the
dr (R - x)x
Ginstling-Brounshtein approach. The steady-state solution of
The rate of reaction zone advance, dx/dt, can be related to dC/ Fick’s first law for radial diffusion in a cylinder is68
dr by64
C1ln(b/r) + C2ln(r/a)
dx D dC C(r) ) (77)
) (70) ln(b/a)
dt ǫ dr
where C(r) is the reactant concentration at a particular value of
where D is the diffusion coefficient, ǫ is a proportionality r (a < r < b), C1 is the concentration of the diffusing species
constant equal to Fn/µ (F and µ are the specific gravity and at surface r ) a, and C2 is the concentration of the diffusing
molecular weight of the product, respectively, and n is the species at surface r ) b. The reaction at the interface is assumed
stoichiometric coefficient of the reaction). Subsituting eq 69 to occur at a much faster rate than diffusion, making C1 ≈ 0.
into eq 70 gives Therefore, eq 77 becomes

dx D RC2 C2ln(r/a)
) (71) C(r) ) (78)
dt ǫ (R - x)x ln(b/a)
which can be rewritten as Taking the derivative of the above equation with respect to
r at r ) a gives
dx R
)k (72)
dt (Rx - x2) C2
dC
| ) (79)
where k ) DC2/ǫ. Separating variables and integrating eq 72
dr r)a aln(b/a)
gives
According to Figure 8, a ) R-x and b ) R, therefore, eq 79
becomes
x2 (21 - 3Rx ) ) kt (73)
dC C2
) (80)
Substituting for x in eq 73 with the value of x in eq 63 and dr (R - x) ln (R/(R - x))
rearranging gives
Subsituting eq 80 into 70 gives
2
1 - R - (1 - R)2/3 ) kt (74) dx k
3 ) (81)
dt (R - x) ln (R/(R - x))
Equation 74 is the Ginstling-Brounshtein (D4) model. The D4
model is another type of three-dimensional model. Buscaglia where k ) DC2/ǫ. Substituting for the value of x in eq 81 with
and Milanese66 have proposed a generalized form of the x from eq 75 and rearranging gives
Ginstling-Brounshtein model and have discussed limitations
related to the boundary conditions for this model. The reaction dx k
)- (82)
between manganese oxide (Mn3O4) and sodium carbonate was dt (1 - R) ln (1 - R)1/2
1/2

shown to follow the D4 model.67

If solid particles are assumed to be cylindrical, and diffusion The derivative of eq 75 is
occurs radially through a cylindrical shell with an increasing
reaction zone, a two-dimensional diffusion (D2) model can be R
dx ) dR (83)
derived. The D2 model can be derived using the same general 2(1 - R)1/2
Review Article J. Phys. Chem. B, Vol. 110, No. 35, 2006 17327

Substituting eq 83 into eq 82 and rearranging gives measured. We hope that this work has demonstrated that solid-
state kinetic models have a theoretical physical meaning and
dR k′ are not merely based on goodness of data fits to complex
)- (84)
dt ln(1 - R) mathematical expressions.
Finally, investigators are challenged to better understand the
where k′ ) 4k/R2. Equation 84 is the differential form of the models and mathematical tools they apply to solid-state kinetic
D2 model. The integral form (Table 1) of the D2 model can be reactions. We hope our work helps resolve some of the lack of
obtained by separating variables and integrating eq 84. understanding of solid-state kinetic models.
Finally, JMAEK models ((-ln [1 - R])1/n ) kt) have been
modified to account for diffusion69 where n becomes 1.5 (one-
dimension), 2 (two-dimensions), and 2.5 (three-dimensions); the References and Notes
diffusion coefficient (D) is included in the reaction rate constant
(1) Chemistry of the solid state; Garner, W. E., Ed.; Academic Press:
(k). New York, 1955.
5.4. Order-Based (F) Models. Order-based models are the (2) Jacobs, P. W. M.; Tompkins, F. C. Classification and theory of
simplest models as they are similar to those used in homoge- solid reactions. In Chemistry of the solid state; Garner, W. E., Ed.; Academic
Press: New York, 1955; Chapter 7.
neous kinetics. In these models, the reaction rate is proportional (3) Brown, M. E.; Dollimore, D.; Galwey, A. K. Reactions in the solid
to concentration, amount or fraction remaining of reactant(s) state. In ComprehensiVe chemical kinetics; Bamford, C. H., Tipper, C. F.
raised to a particular power (integral or fractional) which is the H., Eds.; Elsevier: Amsterdam, 1980; Vol. 22, Chapter 3.
reaction order. Some kinetic analysis methods force data into (4) Galwey, A. K.; Brown, M. E. Thermal decomposition of ionic
solids: Chemical properties and reactiVities of ionic crystalline phases;
an order-based model that may not be appropriate.4,27 Order- Elsevier: Amsterdam, 1999; Chapter 3.
based models are derived from the following general equation (5) Zhou, D. L.; Grant, D. J. W. J. Phys. Chem. A 2004, 108, 4239.
(6) Perez-Maqueda, L. A.; Criado, J. M.; Gotor, F. J.; Malek, J. J.
dR Phys. Chem. A 2002, 106, 2862.
) k(1 - R)n (85) (7) Gotor, F. J.; Criado, J. M.; Malek, J.; Koga, N. J. Phys. Chem. A
dt 2000, 104, 10777.
(8) MacNeil, D. D.; Dahn, J. R. J. Phys. Chem. A 2001, 105, 4430.
where dR/dt is the rate of reaction, k is the rate constant, and n (9) Alkhamis, K. A.; Salem, M. S.; Obaidat, R. M. J. Pharm. Sci. 2006,
is the reaction order. 95, 859.
If n ) 0 in eq 85, the zero-order model (F0/R1) model is (10) Capart, R.; Khezami, L.; Burnham, A. K. Thermochim. Acta 2004,
417, 79.
obtained and eq 85 becomes (11) Khawam, A.; Flanagan, D. R. J. Pharm. Sci. 2006, 95, 472.
(12) Vyazovkin, S.; Dranca, I. J. Phys. Chem. B 2005, 109, 18637.
dR (13) Kirsch, B. L.; Richman, E. K.; Riley, A. E.; Tolbert, S. H. J. Phys.
)k (86) Chem. B 2004, 108, 12698.
dt (14) Jones, A. R.; Winter, R.; Florian, P.; Massiot, D. J. Phys. Chem. B
2005, 109, 4324.
After separating variables and integrating, eq 86 becomes (15) Bertmer, M.; Nieuwendaal, R. C.; Barnes, A. B.; Hayes, S. E. J.
Phys. Chem. B 2006, 110, 6270.
R ) kt (87) (16) Yu, Y. F.; Wang, M. H.; Gan, W. J.; Tao, Q. S.; Li, S. J. J. Phys.
Chem. B 2004, 108, 6208.
If n ) 1 in eq 85, the first-order model (F1) model is obtained (17) Premkumar, T.; Govindarajan, S.; Coles, A. E.; Wight, C. A. J.
Phys. Chem. B 2005, 109, 6126.
and eq 85 becomes (18) Bendall, J. S.; Ilie, A.; Welland, M. E.; Sloan, J.; Green, M. L. H.
J. Phys. Chem. B 2006, 110, 6569.
dR (19) Vyazovkin, S.; Dranca, I. J. Phys. Chem. B 2004, 108, 11981.
) k(1 - R) (88) (20) Vyazovkin, S.; Dranca, I.; Fan, X. W.; Advincula, R. J. Phys. Chem.
dt
B 2004, 108, 11672.
Separating variables and integrating eq 88 leads to the first- (21) Milev, A. S.; McCutcheon, A.; Kannangara, G. S. K.; Wilson, M.
A.; Bandara, T. Y. J. Phys. Chem. B 2005, 109, 17304.
order integral expression (22) Khawam, A.; Flanagan, D. R. Thermochim. Acta 2005, 436, 101.
(23) Khawam, A.; Flanagan, D. R. Thermochim. Acta 2005, 429, 93.
-ln(1 - R) ) kt (89) (24) Khawam, A.; Flanagan, D. R. J. Phys. Chem. B 2005, 109, 10073.
(25) Brown, M. E. J. Therm. Anal. Calorim. 2005, 82, 665.
The first-order model, also called the Mampel model,70,71
is a (26) Brown, M. E. Introduction to thermal analysis: Techniques and
applications, 2nd ed.; Kluwer: Dordrecht, 2001; Chapter 10.
special case of the Avrami-Erofeyev (A) models where n ) (27) Vyazovkin, S.; Wight, C. A. Annu. ReV. Phys. Chem. 1997, 48,
1. Similarly, second-order (n ) 2) and third-order (n ) 3) 125.
models can be obtained (Table 1). Lopes et al.72 have shown (28) Sestak, J.; Berggren, G. Thermochim. Acta 1971, 3, 1.
(29) Yang, J.; McCoy, B. J.; Madras, G. J. Phys. Chem. B 2005, 109,
that decomposition of gadolinium(III) complexes follows a zero- 18550.
order model. Thermal oxidation of porous silicon73 and desorp- (30) Yang, J.; McCoy, B. J.; Madras, G. J. Chem. Phys. 2005, 122.
tion of 2-phenylethylamine (PEA) from silica surfaces74 were (31) Liu, J.; Wang, J. J.; Li, H. H.; Shen, D. Y.; Zhang, J. M.; Ozaki,
shown to follow a first-order model. Y.; Yan, S. K. J. Phys. Chem. B 2006, 110, 738.
(32) Burnham, A. K.; Weese, R. K.; Weeks, B. L. J. Phys. Chem. B
2004, 108, 19432.
6. Summary (33) Graetz, J.; Reilly, J. J. J. Phys. Chem. B 2005, 109, 22181.
(34) Wang, S.; Gao, Q. Y.; Wang, J. C. J. Phys. Chem. B 2005, 109,
We have attempted to summarize the assumptions and 17281.
mathematical derivation of the most commonly used reaction (35) Hromadova, M.; Sokolova, R.; Pospisil, L.; Fanelli, N. J. Phys.
models in solid-state kinetics. This review has not been Chem. B 2006, 110, 4869.
(36) Wu, C. Z.; Wang, P.; Yao, X. D.; Liu, C.; Chen, D. M.; Lu, G. Q.;
exhaustive but rather representative of the common models used. Cheng, H. M. J. Phys. Chem. B 2005, 109, 22217.
Hopefully our presentation is a useful pedagogic tool for (37) Peterson, V. K.; Neumann, D. A.; Livingston, R. A. J. Phys. Chem.
understanding solid-state kinetic models. Even though we have B 2005, 109, 14449.
focused on reactions involving weight loss (A(s)fB(s)+C(g)), (38) Skrdla, P. J.; Robertson, R. T. J. Phys. Chem. B 2005, 109, 10611.
(39) Boldyrev, V. V. Thermochim. Acta 1986, 100, 315.
many of the models are applicable to other solid-state reactions (40) Bagdassarian, C. Acta Physicochim. U.S.S.R. 1945, 20, 441.
where, for example, evolution or consumption of heat is (41) Allnatt, A. R.; Jacobs, P. W. M. Can. J. Chem. 1968, 46, 111.
17328 J. Phys. Chem. B, Vol. 110, No. 35, 2006 Khawam and Flanagan

(42) Guo, L.; Radisic, A.; Searson, P. C. J. Phys. Chem. B 2005, 109, (61) Wyandt, C. M.; Flanagan, D. R. Thermochim. Acta 1992, 196, 379.
24008. (62) Booth, F. Trans. Faraday Soc. 1948, 44, 796.
(43) Avrami, M. J. Chem. Phys. 1939, 7, 1103. (63) Jander, W. Z. Anorg. Allg. Chem. 1927, 163, 1.
(44) Avrami, M. J. Chem. Phys. 1940, 8, 212. (64) Ginstling, A. M.; Brounshtein, B. I. J. Appl. Chem. USSR 1950,
(45) Erofeyev, B. V. Dokl. Akad. Nauk SSSR 1946, 52, 511. 23, 1327.
(46) Jacobs, P. W. M. J. Phys. Chem. B 1997, 101, 10086.
(47) Prout, E. G.; Tompkins, F. C. Trans. Faraday Soc. 1944, 40, 488. (65) Crank, J. The mathematics of diffusion, 2nd ed.; Clarendon Press:
(48) Carstensen, J. T. Drug stability: Principles and practices, 2nd Oxford, England, 1975; Chapter 6, p 89.
revised ed.; Marcel Dekker: New York, 1995; p. 237. (66) Buscaglia, V.; Milanese, C. J. Phys. Chem. B 2005, 109, 18475.
(49) Brown, M. E.; Glass, B. D. Int. J. Pharm. 1999, 190, 129. (67) Eames, D. J.; Empie, H. J. International Chemical Recovery
(50) Brown, M. E. Thermochim. Acta 1997, 300, 93. Conference: Changing Recovery Technology to Meet the Challenges of
(51) Prout, E. G.; Tompkins, F. C. Trans. Faraday Soc. 1946, 42, 468. the Pulp and Paper Industry, Whistler, BC, Canada, June 11-14, 2001.
(52) Skrdla, P. J. J. Phys. Chem. A 2004, 108, 6709. (68) Crank, J. The mathematics of diffusion, 2nd ed.; Clarendon Press:
(53) Guinesi, L. S.; Ribeiro, C. A.; Crespi, M. S.; Santos, A. F.; Capela, Oxford, England, 1975; Chapter 5, p 69.
M. V. J. Therm. Anal. Calorim. 2006, in press. (69) Perez-Maqueda, L. A.; Criado, J. M.; Malek, J. J. Non-Cryst. Solids
(54) Carstensen, J. T. J. Pharm. Sci. 1974, 63, 1. 2003, 320, 84.
(55) Chunxiu, G.; Yufang, S.; Donghua, C. J. Therm. Anal. Calorim. (70) Mampel, K. L. Z. Phys. Chem. 1940, A187, 43.
2004, 76, 203.
(56) Liqing, L.; Donghua, C. J. Therm. Anal. Calorim. 2004, 78, 283. (71) Mampel, K. L. Z. Phys. Chem. 1940, A187, 235.
(57) Gao, Z. M.; Amasaki, I.; Nakada, M. Thermochim. Acta 2002, 385, (72) Lopes, W. S.; Morais, C. R. D.; de Souza, A. G.; Leite, V. D. J.
95. Therm. Anal. Calorim. 2005, 79, 343.
(58) Koga, N.; Criado, J. M. J. Am. Ceram. Soc. 1998, 81, 2901. (73) Pap, A. E.; Kordas, K.; George, T. F.; Leppavuori, S. J. Phys. Chem.
(59) Koga, N.; Criado, J. M. J. Therm. Anal. 1997, 49, 1477. B 2004, 108, 12744.
(60) Welch, A. J. E. Solid-solid reactions. In Chemistry of the solid (74) Carniti, P.; Gervasini, A.; Bennici, S. J. Phys. Chem. B 2005, 109,
state; Garner, W. E., Ed.; Academic Press: New York, 1955; Chapter 12. 1528.