ON STEPFUNCTION REACTION KINETICS MODEL IN THE

ABSENCE OF MATERIAL DIFFUSION

DMITRY GOLOVATY

Abstract. We propose a precise denition of the step-function kinetics suitable for approxi-
mating diuse propagating reaction fronts in one-dimensional gasless-combustion-type models when
a Lewis number is large. We investigate this kinetics in the context of free-radical frontal polymer-
ization (FP) in which a monomer-initiator mixture is converted into a polymer via a propagating,
self-sustaining reaction front. The notion of the step-function kinetics has been extensively used in
studies of the frontal dynamics both in FP and in combustion problems when the material diu-
sion is negligible. However, the models have always been eectively reduced to their point-source
approximations without dening exactly what the step-function kinetics is for the diuse reaction
fronts.
We demonstrate numerically that dynamics of diuse fronts in systems modeled with step-
function kinetics and in systems modeled with the Arrhenius kinetics are qualitatively the same
at time scales at which the bulk reaction ahead of the front can be ignored. We perform the stability
analysis for the traveling reaction wave and show that the stability threshold is in close agreement
with numerical simulations as well as with other existing kinetics approximations.
The benets of using the step-function kinetics are two-fold. The reaction dynamics predicted
by the step-function kinetics approximates the dynamics predicted by the Arrhenius kinetics over
a wider range of system parameters than the point-source approximation. Second, the systems
governed by the step-function kinetics can be analyzed both analytically and numerically within the
framework of a single model.
Key words. frontal polymerization, gasless combustion, Arrhenius kinetics, traveling wave,
reaction-diusion equations
AMS subject classications. 35
1. Introduction.
1.1. Physical Background and Existing Modeling Approaches. In this
paper we give a precise denition of the step-function kinetics and study it in the
context of frontal polymerization (FP).
Frontal polymerization is a mode of free-radical polymerization of a monomer
which in the presence of a thermally unstable initiator converts into a polymer via a
propagating, localized reaction zone [4], [5].
A typical frontal polymerization experiment is performed in a glass tube lled
with reagents. An external heat source, when applied at the top of the tube, initiates
a descending front that appears as a moving region of polymer formation. Depending
on the choice of reactants and the conditions of the experiment, the front either
may or may not propagate with a constant speed. Various non-uniform propagation
scenarios can occur, even if it is assumed that the front always remains at the
situation considered in this paper.
There are several conditions necessary for the existence of the frontal mode. First,
the ignition temperature must be high enough to generate and initially sustain the
reaction front. Further, the reaction rate must be extremely small at the initial
(ambient) temperature but very large at the front temperature. The high reaction
rate coupled with the exothermicity of the reaction must be sucient to overcome
heat losses into the reactants and product zones [5].

Department of Theoretical and Applied Mathematics, The University of Akron, Akron, OH

44325, USA (dmitry@math.uakron.edu). Supported in part by the NSF grant DMS-0305577
1
2 D. Golovaty
A more extensively studied chemical process with a similar reaction mechanism is
self-propagating high-temperature synthesis (SHS) a combustion process character-
ized by a heat release large enough to propagate a combustion front through a powder
compact, while consuming the reactant powders [6], [12]. The simplest models and
front propagation mechanisms for FP and SHS are essentially the same, except for
the magnitudes of the model parameters.
Both steady and unsteady front propagation have been observed in FP [17] as well
as SHS [13]. Unsteady front propagation is usually undesirable as it leads to the non-
uniform, layered structure of the nal product. One of the goals of the modeling is
to determine the range of material parameters within which the stability of uniformly
propagating polymerization front is guaranteed. The analysis of the full model is,
however, too complicated because it requires solving a system of coupled nonlinear
partial dierential equations describing multiple reactions and energy transport. In
order to make analytical predictions, numerous simplications are usually introduced
by employing asymptotics in terms of small parameters, considering eective kinetics,
etc.
In the presence of an appropriate small non-dimensional parameter, the reaction
zone can be replaced by a propagating front with the chemical reaction approximated
by a heat source attached to the front [11]. With the removal of a nonlinear reaction
term, the governing equations become signicantly simpler but, because the location
of the front is not known a priori, the reduced problem is of a freeboundarytype.
The approximate problem is easier to study analytically, especially from the point of
view of stability analysis for the traveling wave solutions.
Note that, although sharpfront approximation is not usually derived via a rig-
orous asymptotic method, it has been shown to be an eective tool to study SHS
and FP problems yielding qualitatively plausible results. On the negative side, the
problems with point-source kinetics are dicult to treat numerically [8]; further, for
certain regimes of FP, the reaction zone does not always remain narrow even in the
presence of a small parameter, resulting in non-negligible bulk eects [3] that are
ignored within a point-source kinetics approximation.
Another approach that has been successfully applied in a number of combustion
and polymerization studies is to introduce simplied distributed kinetics [1], [9] that
is usually combined with the moving front approximation [15], [19], [10]. Within this
approach, the Arrhenius temperature dependence is replaced by the step-function with
the height equal to the value of a certain pre-dened characteristic temperature within
the reaction zone. The exact choice of the characteristic temperature is determined
by the physics of the problem. Although the kinetics function in this setup is very
simple, the strong nonlinearity of the Arrhenius kinetics is inherited by making the
characteristic temperature dependent on the solution.
The advantage of the step-function kinetics is that the traveling wave solution can
generally be found analytically. The stability analysis for this solution is, however,
very tedious and the workers have resorted to additional simplications, in particular
via narrowreactionzonetype asymptotics that essentially lead back to point-source
kinetics.
Typically, the front is postulated to have a width that is determined by a small
non-dimensional parameter (cf. (1.11)) [15], [19], [10]. Then the characteristic tem-
perature is set as a limit of the temperature in an outer solution instead of prescribing
the explicit formula for the characteristic temperature for a diuse front and using
a rigorous asymptotic procedure. Since the characteristic temperature for the trav-
ON STEP-FUNCTION KINETICS 3
eling wave solution is indeed the same as the appropriate outer temperature limit at
the interface, this approach successfully captures the stability threshold for the fronts
propagating with a constant speed. On the other hand, narrow reaction zone approx-
imations of step-functions kinetics suer from the same deciencies as those for the
point-source kinetics.
In this paper, we introduce a version of a distributed step-function kinetics that
can be used to approximate a propagating reaction front in one-dimensional FP. The
benets of using the step-function kinetics are two-fold. The reaction dynamics pre-
dicted by the distributed step-function kinetics approximates the dynamics predicted
by the Arrhenius kinetics over a wider range of system parameters than the point-
source approximation. Second, the systems governed by the step-function kinetics
can be analyzed both analytically and numerically within the framework of a single
model.
We perform the stability analysis for the traveling reaction wave in a diuse
front setting and determine the stability boundary. In order to make the analysis
tractable, we assume that the non-dimensional parameter dened in (1.11) is small.
This assumption appears in a number of other works [15], [19], [10] as a justication
for considering sharp-front asymptotics of the step-function kinetics. The analysis of
[3] indicates that there is no clear relationship between the value of and the width
of the reaction zone, especially for pulsating fronts. Hence we do not consider the
reaction zone to be narrow in our calculations.
Because the algebra involved in handling perturbations of the ground state and
the resulting form of the dispersion relation are very complex, we handle some of the
symbolic calculations and solve the dispersion relation in Maple. We obtain the sta-
bility threshold and show that it is in excellent agreement with numerical predictions
and the existing sharp-front approximations. The combination of analytical compu-
tations and Maple has a clear advantage over the full numerical simulations in that it
does not require numerical solution of a system of partial dierential equations. Also
the former requires a signicantly shorter computational time (minutes vs. hours)
even for a single simulations run.
The computational costs are considerably lower for the step-function kinetics
model than for the point-source kinetics model since the position of the front is not
one of the unknowns in the problem. The numerically determined behavior of the
front for the step-function kinetics is qualitatively similar to that under the Arrhenius
kinetics [7], [17], as it shows a similar hierarchy of dynamics and similar solution
features, including those that result from the bulk mode of polymerization. The same
spectrum of system behaviors has also been demonstrated for the point-source kinetics
[8]; however, all bulk reactions are ignored in this setting.
1.2. Mathematical model. Although the mechanism of free-radical polymer-
ization involves three steps initiation, propagation, and termination and ve
reagents an initiator, an active initiator radical, an active polymer radical, a
monomer, and a complete polymer chain [17], we will make a number of simplify-
ing assumptions that reduce the complexity of the underlying mathematical model.
Hence we will assume [17], [18], [15] that
The rates of reactions between the initiator radicals and the monomer and
between the polymer radicals and the monomer are the same.
The rate of change of total radical concentration is much smaller than the
rates of their production and consumption.
The initial concentration of the initiator is so large that it is not appreciably
4 D. Golovaty
consumed during the polymerization process.
The material diusion is negligible compared to thermal diusion.
Both reagents and the nal product are viscous enough to ignore convective
eects and bubble formation.
The test tube is suciently thin with the adiabatic boundary conditions on
sidewalls so that the spatial dependence of the solution can be restricted to
the axial variable.
Suppose that a test tube containing the monomerinitiator mixture occupies a
region R
3
, and denote by M(x, t) the monomer concentration and by T(x, t)
the temperature of the mixture at the point x and the time t > 0 . Then the
process of free-radical polymerizations can be described [15] by what is known as a
singlestep, eective kinetics model of monomertopolymer conversion
M
t
= kMe
E
RgT
b

1
T
b
T

, (1.1)
T
t
= div (T) + kqMe
E
RgT
b

1
T
b
T

, (1.2)
where is a thermal diusivity of the mixture/nal product, k is the eective pre-
exponential factor in the Arrhenius kinetics, R
g
is the gas constant, E is the eective
activation energy, and T
b
is a reference temperature that will be specied below. The
constant parameter q is
H
c
, where H is the reaction enthalpy; c and are the
specic heat and the mixture density, respectively.
Throughout this paper we will assume that the test tube is onedimensional, =
[L, L] , and that the thermal diusivity is constant (we ignore possible dependence
of on temperature and degree of conversion 1 M/M
0
). Then the problem (1.1)
(1.2) reduces to
M
t
= kMe
E
RgT
b

1
T
b
T

, (1.3)
T
t
=

2
T
x
2
+ kqMe
E
RgT
b

1
T
b
T

. (1.4)
We will assume that T and M satisfy the constant initial conditions
T(x, 0) = T
0
, M(x, 0) = M
0
, x [L, L] . (1.5)
In order to initiate the reaction, heat must be supplied to the system; hence for the
rst t
0
seconds we will use the following boundary conditions
T
x
(L, t) = 0 , M
x
(L, t) = 0 , T(L, t) = T
b
, t (0 , t
0
) . (1.6)
During the front propagation regime, we will impose the adiabatic and impen-
etrability boundary conditions on the temperature and the monomer concentration,
respectively by setting
T
x
(L, t) = 0 , M
x
(L, t) = 0 , t t
0
. (1.7)
Multiplying (1.3) by q, adding the resulting equation to (1.4), integrating with
respect to x, applying the adiabatic boundary conditions in (1.7), and setting
H :=
_
L
L
(T + qM) dx, (1.8)
ON STEP-FUNCTION KINETICS 5
yield
dH
dt
= 0 , (1.9)
expressing conservation of enthalpy in the system when t > t
0
. Thermodynamics of
the problem dictates that the temperature of the reaction products away from the
front is given by
T
b
= T
0
+ qM
0
, (1.10)
where T
0
and M
0
are the initial temperature and concentration, respectively, in (1.5).
We introduce dimensionless parameters
=
R
g
T
b
E
, Z =
qM
0
E
R
g
T
2
b
, (1.11)

t =
kt
Z
, x =
_
k
Z
x,

M =
M
M
0
,

T =
T T
0
T
b
T
0
.
Here T
b
is as dened in (1.10) and the Zeldovich number Z is a non-dimensionalized
activation energy [14] constructed as a ratio of the diusion temperature scale T
b
T
0
to the reaction temperature scale
RgT
2
b
E
. Also, note that Z < 1 in order to insure
that the initial temperature of the mixture is greater than absolute zero. Then (after
dropping tildes) we obtain
M
t
= ZM exp
_
Z(T 1)
Z(T 1) + 1
_
, (1.12)
T
t
=

2
T
x
2
+ ZM exp
_
Z(T 1)
Z(T 1) + 1
_
. (1.13)
From (1.5)-(1.7), the non-dimensional temperature T and concentration M satisfy
the following initial and boundary conditions
T(x, 0) = 0, M(x, 0) = 1, x [l , l] , (1.14)
M
x
(l , t) = 0 , T
x
(l , t) = 0 , T(l , t) = 1 , t (0 ,
0
) , (1.15)
M
x
(l , t) = 0 , T
x
(l , t) = 0 , t
0
, (1.16)
where l =
_
k/ZL and
0
= k t
0
/Z.
2. Step-function kinetics. Stability analysis and numerical results.
2.1. Mathematical formulation and traveling wave solution. From now
on, we will assume that is small; then the system of equations (1.12)(1.13) reduces
to
M
t
= ZMe
Z(T1)
, (2.1)
T
t
=

2
T
x
2
+ ZMe
Z(T1)
. (2.2)
6 D. Golovaty
When the second non-dimensional parameter Z is large, this model has been approxi-
mated by using the step-function kinetics [2], [15] in a sharp-front limit as Z . In
this limit the model is identical to the the sharp-front model of solid combustion with
point-source-kinetics at the interface considered in [11]. Here we investigate the behav-
ior and the stability of solutions to the step-function kinetics model and compare the
results to both Arrhenius kinetics and the point-source kinetics. The diuse-interface,
step-function kinetics can be thought of as an intermediate approximation between
the point-source kinetics and Arrhenius kinetics in that it yields to relatively straight-
forward analytical, as well as numerical analyzes. As it is well-known, although the
systems modeled with Arrhenius kinetics can be studied numerically, the stability
analysis for such systems is very dicult as no analytical expressions are available for
traveling wave solutions. On the contrary, for a point-source kinetics (sharp-interface
approach) the stability analysis is straightforward while the numerical computations
are dicult as one has to track a free boundary.
A similar model was considered in [1] for a more general case when the Lewis
number Le (the ratio of thermal and material diusivities) is not necessarily large.
This work has served as a basis for numerous studies in frontal polymerization and
combustion. As it will be explained shortly, the straightforward reduction of the
treatment in [1] to (2.1)(2.2) cannot be considered a priori within an asymptotic
procedure as Le , as it requires additional assumptions. These assumptions will
be introduced below and studied both analytically and numerically in order to verify
their validity.
In the step-function-kinetics model, the Arrhenius term Z e
Z(T1)
in (2.1)(2.2)
is replaced by the stepfunction
K(T) :=
_
0 , T < T
p

1
Z
,
Z e
Z(Tp1)
, T T
p

1
Z
,
(2.3)
where T
p
is the temperature of the mixture immediately upon the completion of the
reaction (or, analogously, the temperature at the product end of the reaction zone).
Since this temperature is, generally, the highest temperature of the mixture, within
this model the reaction is assumed to occur in the temperature range
_
T
p

1
Z
, T
p

.
Unless the front is a steadily propagating wave, the maximum temperature inside the
test tube and, therefore, the shape of the kinetic function depend on time.
The justication for the particular form of K(T) in (2.3) is that the integral value
of the kinetic function over the temperature range within the reaction zone has to be
approximately the same as for the Arrhenius kinetics. Indeed,
_
Tp
0
K(T) dT = e
Z(Tp1)
,
while
_
Tp
0
Z e
Z(T1)
dT = e
Z(Tp1)
e
Z
.
For a suciently large Zeldovich number Z these expressions are essentially the same.
To x ideas, we need a rigorous denition of the temperature T
p
. In [1] this
quantity is dened as the temperature at the end of the reaction zone on the boundary
x
b
(t) separating the reacting mixture and the products of the reaction. Since the
reaction is negligible in the products zone, it is assumed that the concentration of the
ON STEP-FUNCTION KINETICS 7
reagent vanishes at x
b
(t) and, hence, T
p
(t) = T(x
b
(t), t). Both T
p
and x
b
are unknown
and are determined as a part of the solution procedure.
The spatial domain in [1] is assumed to consist of three zones: pre-heating zone,
reaction zone, and the products zone. When traveling-wave solutions are sought sub-
ject to boundary conditions at innity, the problem within each zone consists of two
second order ODEs with solutions that are glued together so that the temperature,
the reagent concentration, and their derivatives are continuous across all interfaces.
Consider now the same model when the inverse of the Lewis number is equal to
zero. Because monomer diusion is neglected, the concentration equation is becomes
a rst order ODE, reducing the number of boundary conditions within each zone we
either have to consider a singularly perturbed problem in terms of the Lewis number,
or we can no longer dene x
b
as a point at which the concentration of the monomer
will vanish. Indeed, it will either cause the concentration to vanish everywhere in the
domain, or x
b
as Le .
As it has been already noted, the stepfunction kinetics approach has been widely
applied in a number polymerization studies. Step-function kinetics was used to inves-
tigate polymerization waves for the two-species [10] and for the four-species models
[18] as well as for the two-step and one-step polymerization models in non-adiabatic
case [9]. The problem of dening the temperature of the reaction cuto has generally
been circumvented by assuming that the reaction zone remains narrow. In [9], for
example, it was assumed that the heat losses lead to a temperature prole with a
maximum T
m
within the reaction zone. The reaction is then cut o below the igni-
tion temperature T
f
which itself depends on T
m
. The value of T
m
is not known in
advance and has to be determined as a part of the solution procedure. Even in the
simplest case, the solution to this problem is quite complicated and requires additional
assumptions; in [9] it was assumed that the reaction zone has a small width of order
0 < 1.
Subsequently, the step-function kinetics was used in conjunction with narrow
reaction zone assumption to study the stability of uniformly propagating front in
various polymerization models in the presence [19] and in the absence [15] of heat
losses. The width of the front in both instances was assumed to be of order and T
f
was set to be equal to T
p
(1 ). Then T
f
= T
p
in the limit 0 thus reducing the
reaction term to a jump condition on the gradient of the temperature on the interface.
Here, instead of either considering asymptotics in terms of the Lewis number or
assuming that the reaction zone is narrow, we will set T
p
(t) = T(x
b
(t), t) , where x
b
(t)
is a point at which the monomer concentration falls below a prescribed threshold value
M(x
b
(t), t) = . Here the constant parameter > 0 is small.
Note that K(T) is actually a functional K[M, T] and the dependence of K on T
is nonlocal. The parameter was chosen through the numerical experiments in order
to guarantee that the reaction is cut o neither prematurely nor too late. The range
of reasonable values of beta appears to be robust in a sense that it is independent
of the other parameters of the problem. The numerically determined magnitude of
beta is of order 10
2
for the cases that we tested. Further, it turns out that for the
range of values of Z that we are interested in,
1
Z
.
Although, similar to e.g. [15], we imposed the condition that 1, the width
of reaction zone is determined by the inverse of the Zeldovich number 1/Z which we
have not assumed to be small.
Given our choice of K(T) in (2.3), the step-function approximation of the model
8 D. Golovaty
(2.1)(2.2) is
M
t
= MK(T) , (2.4)
T
t
=

2
T
x
2
+ MK(T) . (2.5)
Since we are assuming that the reaction begins when the temperature reaches the
threshold value of T
p

1
Z
, we will associate with this temperature the position (t)
of the reaction front by dening (t) implicitly through the relation
T((t), t) = T
p
(t)
1
Z
.
In order to obtain traveling-wave solutions and study their stability, we introduce
the front-attached spatial coordinate y = x (t). Then the system of equations
(2.4)(2.5) can be written as
M
t

(t)
M
y
= ZMe
Z(Tp1)
(y) , (2.6)
T
t

(t)
T
y
=

2
T
y
2
+ ZMe
Z(Tp1)
(y) , (2.7)
q where
(y) =
_
0 , if y < 0 ,
1 , if y 0 ,
is the Heaviside function.
We will assume that y R and that the following boundary conditions at innity
(cf. dimensional boundary conditions (1.5)(1.7) on a nite domain) are satised
T(, t) = 0 , T
y
(, t) = 0 , (2.8)
M(, t) = 1 , M(, t) = 0 . (2.9)
These must be supplemented by the conditions on M and T when y = 0. By the
denition of (t), we immediately have that
T(0, t) = T
p
(t)
1
Z
. (2.10)
We will require that both monomer concentration and the derivative of the tem-
perature are continuous across the polymerization front, that is
[T
y
]
y=0
= [M]
y=0
= 0 , (2.11)
where [f]
y=a
= f(a
+
) f(a

) denotes the jump of the function f at y = a.

Further, by the denition of T
p
, an additional condition
T (x
b
(t) , t) = T
p
, where M (x
b
(t), t) = , (2.12)
must be satised. Here we assume that the monomer concentration is a monotone
function of x for all t > 0. In general, as has been pointed out in [3], this assumption
may not be correct for non-uniformly propagating reacting front. This complication
ON STEP-FUNCTION KINETICS 9
can be easily circumvented, however, by assuming that x
b
(t) is the leftmost point in
the reaction zone satisfying the condition M (x
b
(t), t) = .
First, we seek traveling-wave solutions of (2.6)(2.7) propagating in the negative-
y direction. We set

(t) = v, where v is a positive constant and suppose that

M
t

T
t
0 . Then (2.6)(2.7) reduce to the system of the ordinary dierential
equations
v
d

M
dy
= Z

Me
Z(Tp1)
(y) , (2.13)
v
d

T
dy
=
d
2

T
dy
2
+ Z

Me
Z(Tp1)
(y) , (2.14)
where

M and

T are time-independent solutions of (2.6)-(2.7).
Fix Z and . Denote the temperature at the reaction front, as in (2.10), by
T
f
:=

T(0) = T
p

1
Z
, (2.15)
and set
A := Z e
Z(Tp1)
= Z e
Z(T
f
1)+1
, (2.16)
to be the strength of the kinetics term. The problem (2.13)(2.14), (2.8)(2.12) admits
the following set of solutions

M(y) =
_
1 , y < 0 ,
e

Ay
v
, y 0 ,

T(y) =
_

_
(1 )Z 1
(1 )Z
e
v y
, y < 0 ,
1
1
(1 )Z
e

Ay
v
, y 0 ,
(2.17)
where
A = Ze

1
, v =

Ze

1
Z(1 ) 1
,

T
f
= 1
1
(1 )Z
. (2.18)
Furthermore, because of the constraint (2.12), the reaction zone extends from y = 0
to y = y
b
, where
y
b
:=
v ln
A
= ln

1
Z (Z(1 ) 1)
, (2.19)
so that

M ( y
b
) = ,

T ( y
b
) =

T
f
+
1
Z
. (2.20)
2.2. Stability analysis. Next, we consider the following perturbations of the
base state (2.17)(2.19)
M(y , t)=

M(y) + e
t
(y) ,
T(y , t) =

T(y) + e
t
(y) ,
T
f
(t) =

T
f
+ e
t
,
(t) = vt + e
t
,
y
b
(t) = y
b
+ e
t
,
10 D. Golovaty
where 0 < 1 and C.
First, we linearize the condition (2.12). To the rst order in we have
= M (y
b
(t) , t) =

M (y
b
(t)) + e
t
(y
b
(t))

M ( y
b
) + e
t
_

M

( y
b
) + ( y
b
)
_
,
and
T
f
(t) = T (y
b
(t) , t) =

T (y
b
(t)) + e
t
(y
b
(t))

T ( y
b
) + e
t
_

( y
b
) + ( y
b
)
_
.
Then, using the perturbation of T
f
(t) and (2.19) we obtain

( y
b
) + ( y
b
) = 0 ,

T

( y
b
) + ( y
b
) = . (2.21)
It follows from (2.17), (2.20), and (2.21) that
= ( y
b
)

( y
b
)

( y
b
)
( y
b
) = ( y
b
) +
1
(1 )Z
( y
b
) . (2.22)
Linearization of (2.6)(2.11) yields the following problems
_
v

+ =

M

, y < 0 ,
v

+ ( + Z) =

M

Z
2

M , y 0 ,
(2.23)
and
_

=

T

, y < 0 ,

=

T

Z
_
+ Z

M
_
, y 0 ,
(2.24)
subject to the boundary conditions
() =

() = [

]
y=0
= 0 , (0) = , (2.25)
() = () = []
y=0
= 0 , (2.26)
and (2.22).
As it has already been pointed out,
1
Z
, however the value of ln that enters
into y
b
is large. We simplify the computations by keeping ln as a parameter in our
calculations and otherwise restrict our analysis to the O(1)-approximation of (2.22)
(2.26) in by setting = 0. Then the base state can be written as

M(y) =
_
1 , y < 0 ,
e

Z y
v
, y 0 ,

T(y) =
_

_
1
v
2
e
v y
, y < 0 ,
1
1
Z
e

Z y
v
, y 0 ,
(2.27)
where
v =
_
Z
Z 1
,

T
f
= 1
1
Z
, y
b
=
ln
_
Z(Z 1)
. (2.28)
The solution to (2.22)(2.26) and the dispersion relation satised by the param-
eters Z and is very complicated, in particular due to the coupling (2.22) and was
ON STEP-FUNCTION KINETICS 11
1 2 3 4 5 6
ln()
8.43
8.445
8.46
8.475
8.49
Z
point-source kinetics
step-function kinetics
1 2 3 4 5 6
ln()
26
28
30
32
34
36
P
e
r
i
o
d

o
f

o
s
c
i
l
l
a
t
i
o
n
s
,

s
Fig. 2.1. Dependence of the critical value of the Zeldovich number Z and the dimensional
period of velocity oscillations on .
obtained using Maple computer algebra system. In order to nd the stability bound-
ary, the real part of was set equal to zero, = i and the resulting system of
equations was solved in Maple for Z and .
The dependence on ln of the critical value of the Zeldovich number Z and the
dimensional period of front velocity oscillations =
2Z
k
at the critical Z are shown
on Figure 2.1. Note that the stability boundary is in remarkable agreement with the
sharp-front stability boundary obtained in [11] (Z = 2
_
2 +

5
_
8.47.) when the
reaction kinetics is approximated by the point source on the front. The same critical
value of the Zeldovich number Z as in [11] can be obtained by using the sharpfront
limit Z of the model with stepfunction kinetics; the analysis analogous to
the one in [15] leads to this conclusion once the jump condition for the balance of
heat on the front is imposed in place of the similar, one-sided condition employed in
[15]. The latter condition was derived on the basis of generalized matched asymptotic
expansions [14] and leads to the stability threshold of Z = 6. This value disagrees
with predictions of other models.
The value of the articial parameter can be tuned so that the stability thresh-
old coincides with that obtained in [11]. Further, the dependence of the period of
oscillations on is much stronger than that for the critical Zeldovich number.
Although the stability boundaries for the approximate models discussed here are
essentially the same (Z 8.47), the stability boundary for model with full Arrhenius
kinetics [16] and small is slightly higher (Z 9.1).
2.3. Numerical results. To verify these conclusions, we conducted numerical
experiments with the model (2.4)(2.5) with the Arrhenius kinetic function being
replaced by the step function.
The system of equations was solved numerically using a nite dierence method
with semi-implicit time integration. The physical model has no-ow homogeneous
Neumann boundary conditions at both ends of the domain. For some parameter
combinations, however, we found it necessary to apply Dirichlet boundary condition
T(L, t) = T
b
, where T
b
is dened in (1.10), at the ignition end of the domain for a short
period of time to initiate the reaction, and then switch to the homogeneous Neumann
boundary condition. Numerical experiments have demonstrated that the long-term
behavior of the reaction-diusion equation system studied in this paper is not aected
12 D. Golovaty
by the application of the Dirichlet boundary condition during the initiation stage.
At each time step, the location of the reaction front was dened as the rst gird
point, going from left to right, at which the concentration of the monomer drops below
50% of the initial value. The average velocity of the front was calculated by
v = a
x
t
where x is the distance between grid points, t is the size of the time step, and a is
the number of grid intervals traveled through by the front in t seconds. Note that it
may take multiple, say m, time steps for the front to travel through one grid interval.
In that case, we have a =
1
m
.
Since we do not use an adaptive scheme, we used uniform grid renement tech-
nique which clearly indicated numerical convergence and demonstrated that all sharp
features are resolved and grid independent.
We will assume that the parameters
q = 33.24 K

L/mol , = 0.0014 cm
2
/s , k = 1 s
1
, T
b
= 500 K

,
are xed; then the state of the system is completely determined once the values of Z
and are specied. The length of the spatial domain (test tube) in our computations
varies from 6 cm to 10 cm, depending on the characteristic time scale of the process
of interest.
Using this choice of system parameters we varied Z while keeping =1.E-3 and
=2.E-2 xed. Since ln 0.02 3.92, the analytical stability threshold is almost the
same (Figure 2.1) as the one predicted by the stability analysis for the point-source
kinetics [11].
Our simulations predict that the stability threshold is Z 8.5. The typical
velocity and the front position proles are presented on Figures (2.2)-(2.5). The front
propagates with the constant velocity when Z = 8.2; the velocity oscillations appear
when Z = 8.5 and become more pronounced once Z is increased (Z = 8.7). Further
increase in Zeldovich number shows that the behavior of the system is similar to
the behavior of a system reacting via Arrhenius kinetics for instance, the period
doubling can be observed for Z = 9.4.
Another similarity with Arrhenius kinetics was demonstrated in [3] where we
showed that pulsating fronts in systems governed by the step-function kinetics evolve
via a combination of bulk and frontal modes. Note that both bulk contribution and
the related solution features are always absent from the sharpfrontbased models,
since the reaction is limited to the front.
The numerically determined period of velocity oscillations for Z = 8.5 near the
threshold of instability is approximately 31 s. This value is almost identical
the value obtained through the stability analysis (Figure 2.1). We conclude that
the predictions of the stability analysis are in very close agreement with numerical
simulations.
3. Conclusions. In the context of frontal polymerization, we introduced a pre-
cise denition of distributed step-function kinetics without resorting to a sharp-front
approximation. This kinetics is appropriate for simulating the behavior of one-
dimensional, large-Lewis-number reaction systems governed by Arrhenius kinetics.
Among the interesting features of the distributed step-function kinetics is the numer-
ical and analytical tractability of the corresponding model that takes into account
possible bulk reactions behind the advancing front.
ON STEP-FUNCTION KINETICS 13
0 100 200 300
Time, s
0
0.005
0.01
0.015
0.02
0.025
0.03
F
r
o
n
t

v
e
l
o
c
i
t
y
,

c
m
/
s
=8.2, =0.001
0 100 200 300 400
Time, s
-4
-3
-2
-1
0
1
2
F
r
o
n
t

p
o
s
i
t
i
o
n
,

c
m
Z=8.2
Fig. 2.2. Reaction front velocity and position when Z = 8.2 and = 10
3
.
0 100 200 300 400
Time, s
0.005
0.01
0.015
0.02
0.025
V
e
l
o
c
i
t
y
,

c
m
/
s
=8.5, =0.001
0 100 200 300 400
Time, s
-4
-3
-2
-1
0
1
2
F
r
o
n
t

p
o
s
i
t
i
o
n
,

c
m
Z=8.5
Fig. 2.3. Reaction front velocity and position when Z = 8.5 and = 10
3
.
0 100 200 300 400 500 600 700
Time, s
0
0.01
0.02
0.03
0.04
F
r
o
n
t

v
e
l
o
c
i
t
y
,

c
m
/
s
Z=8.7
0 100 200 300 400 500 600 700
Time, s
-8
-6
-4
-2
0
2
F
r
o
n
t

p
o
s
i
t
i
o
n
,

c
m
Z=8.7
Fig. 2.4. Reaction front velocity and position when Z = 8.7 and = 10
3
.
14 D. Golovaty
0 100 200 300 400
Time, s
0
0.05
0.1
0.15
0.2
F
r
o
n
t

v
e
l
o
c
i
t
y
,

c
m
/
s
Z=9.4
0 100 200 300 400 500
Time, s
-4
-3
-2
-1
0
1
2
F
r
o
n
t

p
o
s
i
t
i
o
n
,

c
m
Z=9.4
Fig. 2.5. Reaction front velocity and position when Z = 9.4 and = 10
3
.
We demonstrated numerically that dynamics of fronts in systems modeled with
the distributed step-function kinetics and in systems modeled with the Arrhenius
kinetics are qualitatively the same for the time scales at which bulk reactions ahead
of the front can be neglected. Further, we showed that the stability threshold of the
traveling wave solution for the step-function kinetics is in excellent agreement with its
numerically determined value as well as with other existing kinetics approximations.
4. Acknowledgments. The author would like to express their gratitude to L. K.
Gross, V. A. Volpert, and J. Zhu for the valuable discussions.
REFERENCES
[1] A. P. Aldushin and S. G. Kasparyan, Thermodiusional instability of a stationary ame
wave, tech. report, Institute of Chemical Physics, Chernogolovka, 1978. preprint.
[2] , Thermodiusional instability of a combustion front, Sov. Phys. Dokl., 24 (1979), pp. 29
31.
[3] S. A. Cardarelli, D. Golovaty, L. K. Gross, V. Gyrya, and J. Zhu, A numerical study of
one-step models of polymerization: Frontal vs. bulk mode. To appear in Physica D, 2004.
[4] K. M. Chechilo, R. Y. A. Khvilivitskii, and N. S. Enikolopyan, Phenomenon of polymeri-
sation reaction spreading, Doklady Akademii Nauk SSSR, 205 (1972), pp. 11801181.
[5] Y. Chekanov, D. Arrington, G. Brust, and J. A. Pojman, Frontal curing of epoxy resins:
Comparison of mechanical and thermal properties to batch-cured materials, Journal of
Applied Polymer Science, 66 (1997), pp. 12091216.
[6] Y. Choi, J. K. Lee, and M. E. Mullins, Densication process of TiCx-Ni composites formed
by self-propagating high-temperature synthesis reaction, Journal of Materials Science, 32
(1997), pp. 17171724.
[7] P. Dimitriou, J. Puszynski, and V. Hlavacek, On the dynamics of equations describing
gasless combustion in condensed systems, Combust. Sci. and Tech., 68 (1989), pp. 101
111.
[8] M. Frankel, V. Roytburd, and G. Sivashinsky, Complex dynamics generated by a sharp in-
terface model of self-propagating high-temperature synthesis, Combust. Theory Modelling,
2 (1998), pp. 118.
[9] P. M. Goldfeder and V. A. Volpert, Nonadiabatic frontal polymerization, Journal of Engi-
neering Mathematics, 34 (1998), pp. 301318.
[10] P. M. Goldfeder, V. A. Volpert, V. M. Ilyashenko, A. M. Khan, J. A. Pojman, and
S. E. Solovyov, Mathematical modeling of free-radical polymerization fronts, Journal of
Physical Chemistry, 101 (1997), pp. 34743482.
ON STEP-FUNCTION KINETICS 15
[11] B. J. Matkowsky and G. Sivashinsky, Propagation of a pulsating front in solid fuel combus-
tion, SIAM J. Appl. Math., 35 (1978), pp. 465477.
[12] A. G. Merzhanov, A. K. Filonenko, and I. P. Borovinskaya, New phenomena in combus-
tion of condensed systems, Doklady Akademii Nauk SSSR, 208 (1973), pp. 122125.
[13] A. G. Merzhanov and B. I. Khaikin, Theory of combustion waves in homogeneous media,
Proc. Energy Combust. Sci., 14 (1988), pp. 198.
[14] Daniel A. Schult, Matched asymptotic expansions and the closure problem for combustion
waves, SIAM J. Appl. Math., 60 (1999), pp. 136155.
[15] D. A. Schult and V. A. Volpert, Linear stability analysis of thermal free radical polymer-
ization waves, International Journal of Self-Propagating High-Temperature Synthesis, 8
(1999), pp. 417440.
[16] K. G. Shkadinskii, B. I. Khaikin, and A. G. Merzhanov, Propagation of a pulsating exother-
mic reaction front in the condensed phase, Fizika Goreniya i Vzryva, 1 (1971), pp. 1928.
English Translation in Combust. Expl. Shock Waves, 7 (1971), pp. 1522.
[17] S. E. Solovyov, V. M. Ilyashenko, and J. A. Pojman, Numerical modeling of self-
propagating polymerization fronts: The role of kinetics on front stability, Chaos, 7 (1997),
pp. 331340.
[18] C. A. Spade and V. A. Volpert, On the steady-state approximation in thermal free radical
frontal polymerization, Chemical Engineering Science, 55 (2000), pp. 641654.
[19] , Linear stability analysis of no-adiabatic free-radical polymerization waves, Combust.
Theory Modelling, 5 (2001), pp. 2139.