Approximation Techniques in

Complex Reaction Kinetics

DANNY SUMMERS
Department of Mathematics & Statistics, Memorial University of Newfoundland,
St. John’s, Newfoundland, Canada A l C 5S7

JOHN M. W. SCOTT and SCOTT K. RALPH

Department of Chemistry, Memorial University of Newfoundland, St. John’s,
Newfoundland, Canada A l B 3x7

Abstract
We examine the kinetic model A + B
ki
k2
C-k3
D. The differential equations de-
scribing this reaction scheme are cast in a nondimensional form and analyzed in four
basic approximation regimes: a ‘pseudo-first order’ approximation valid for small values
of the ratio of the initial concentrations of the reactants; an asymptotic solution valid for
large values of k S ;the standard steady state (Bodenstein) approximation; and an approxi-
mation to a second order system without intermediate. Interconnecting relationships
between the various approximations derived are examined, and the approximations are
compared to numerical solutions to the full equations. The results are assessed from the
standpoint of the experimental kineticist, and it is suggested that the reaction studied.
and consequently many other more complex reactions, may under certain circumstances
be subject to non-unique interpretation.

Introduction
Recently, Blandamer, Scott and Summers [11studied the second order
reaction,
(1) A + B T Ck”-D
in the case when the back reaction rate k, is small. The purpose of the
present investigation is to develop a number of other analytical tech-
niques for studying reaction (11, that may be applied also to more
complex systems, that elucidate important features of the reaction, and
that may be profitably applied to the benefit of the experimental kinet-
icist. In the section entitled “Reaction Equations,” we set up the basic
equations governing (1)in nondimensional form, we give the simple
conservation integrals, and we deduce the qualitative behavior of the

International Journal of Chemical Kinetics, 1’01. 19, 553-570 (1987)

C 1987 John Wiley & Sons, Inc. CCC 0538-8066/87/060553-18$04.00
554 SUMMERS, SCOTT AND RALPH

time profiles of the concentrations. In the “Analysis” section we use

standard techniques of applied mathematics, in particular techniques
of regular perturbation analysis and boundary layer theory, to derive
approximate solutions to reaction (1) in the cases when the ratio of
the initial concentrations of the reactants is small, and when the reac-
t i m rate (k,) to the product is large. We also apply the steady state
zpproximation, and we approximate (1)to a second order system with-
out intermediate. The set of solutions developed in this paper is rela-
tivelv involved, and we attempt to clarify their relationships to the
original system and to each other in Figure 1. Finally, in “Numerical
Results and Discussion” we compare a representative selection of
approximate analytical solutions with solutions to the full equations
obtained by numerical integration, and we assess the impact of our
results in chemical terms.

Reaction Equations
The reaction (1)is described by the following nonlinear differential
system:

d[A
- - 1- - k l [ A ] [ B ] + k2[C]
dt

(3)

(4) d[C1 - k , [ A ] [ B ]- ( k 2 + k3)[ C ]

- -
dt

d[D1 - k3[C]
- -
dt
We shall consider solutions to equations (2)-(5) subject to the initial
conditions,
(6) [ A ] = [Ale, [ B ] = [BlO, [ C ] = 0 , [ D ] = 0 a t t = 0
and we shall measure the concentrations [ A ] ,[ B ] ,[C], [ D ] in the units
mol dm-3, the rate constant k , in dm3mol-’ s-’, and the rate constants
k 2 and k3 in s-’.
From equations (2)-(6), we may readily deduce the conservation
integrals,
 COMPLEX REACTION KINETICS 555

Following [l]we now introduce the nondimensional variables,

x = k,[A]ot
Then the system (2148) may be expressed in the form,

(10)

(11)

(12) Yl(0) = 1, Y2(0) = 0

(13) Y3 = 1 - y1 - y2
(14) y4 = y1 - (1 - a )
where we have also introduced the nondimensional parameters

(15)

Since the system (2145) is symmetrical with respect to the spe-

cies A and B, we may take [Bl0 5 [A]* without loss of generality. Thus,
the problem reduces to the solution of the nonlinear, autonomous
system (10)-(11) together with the initial conditions (121, where
0 < cy 5 1and p, y and positive parameters; y3 and y4then follow from
equations (13) and (14). We note that the parameter a denotes the ratio
of the initial concentrations of the species, and that the parameters P
and y respectively measure the rates of the back reaction and the
reaction to the product in terms of the forward reaction rate. It is not
possible to derive exact analytical solutions to the system (101-4 141,
even in the case a = 1. Nevertheless, it is possible to obtain several
useful approximate analytical solutions to (10)-(14), and we shall e-<-
amine these in the following section. The system (lO)-(l‘?) may bc
integrated numerically by means of one of the many algoritims : ‘-?t
have been developed for initial-value problems. We shall find thzt c-
comparison of approximate analytical solutions with ‘exact’ numei ical
solutions is a useful device for elucidating the nature of the solutions
to the original system (2)-(6). Of course, the qualitative behavior of the
solutions to (101414) may be deduced by elementary techniques (e.g. by
556 SUMMERS, SCOTT AND RALPH

using the method of isoclinals); in this regard we find that

(16) y1- 1- a , yz- 0, Y3 - a,
y4-0 as x - - x
which is the expected stoichiometric behavior of the reaction at large
times. Moreover, we find that the reactants ( y l and y4) are mono-
tonically decreasing functions of time, the intermediate ( y2) mono-
tonically increases to a maximum thereafter monotonically decreasing
to zero, and the product (y3)monotonically increases to its asymptotic
value. Thus, reaction (11, in common with many much more complex
reaction networks, exhibits kinetic behavior of only the most mundane
kind. Nevertheless, we shall find reaction (1)to be a useful vehicle for
exploiting perturbation methods of wide applicability in kinetics, and,
to a certain degree, we shall show that reaction (1) is a n archetypal
reaction representing omnipresent difficulties facing the experi-
mentalist concerned with the evaluation of rate constants.

Analysis
We shall derive four approximate solutions to the system (10)-(14),
and also analyze their inter-related properties.

(i) Approximation to a First Order System

Let us replace the second order reaction (1)by the first order reaction,
(17)
and the reaction equations:

(18)

(19) d'B1
- - - -h,[A]o[B]+ h,[C]
dt

(20)

d[D1
- - - k3[C]
dt
We shall now prove that this (completely soluble) first order system
approximates the reaction (1)in the limit as a = [B],/[A],-+ 0. Sys-
tem (18)-(21) may be expressed in the following form:

(22)
 COMPLEX REACTION KINETICS 557

where

-
and x , a , p, y are as defined above; denotes species in the first order
system (17)-(21).
Substituting the formal expansions,

(29)

Further, by substituting the expressions

(31)
(32)
in equations (29)-(30), we find that we precisely recover equa-
tions (22)-(23). Thus, from (281, (311, (32) we derive the results,

(33)

It also follows that

(34)

Results (33) and (34) formally establish the validity of the approxi-
mation of the reaction (1) by the first order system (17)-(21) as a -+ 0;
(33) and (34) also show that the approximation is uniformly valid for all
time t, and the results further provide numerical information concern-
558 SUMMERS, SCOTT AND RALPH

ing the introduced error. The results are also independent of the rate
constants kl,kz,and k,. As we shall demonstrate below, it is not actu-
ally necessary for the parameter LY to be vanishingly small in order for
the approximation to be of practical use. The approximate solutions f l ,
y z are readily deduced from (22)-(24) (by Laplace transform or other
methods), and are found to be
(35)
f1=1- A2)e-'It - (kl[AIO-

(36)

where A l and A z are the roots of the characteristic equation

(37) A' - (kl[A]o+ kz + k 3 ) A + k1k3[A]O 1 0
Solutions for y3 and y4 then follow from equations (25) and (26). The
roots A and A z are necessarily real, positive, and distinct for any (posi-
tive) values of the rate constants kl,k2 and k3,and any initial concen-
tration [A],. For completeness, we note that results (351-437) are also
valid if kz = 0, except that if kz = 0 and kl[AIO= k3 then (35) and (36)
must be replaced by
(38) y1 = 1 - (y + (ye-kdA10t

(39) fz = ( ~ k ~ [ A ] , t e - ~ 1 [ ~ ~ 0 ~
The type of approximation given in (i) is sometimes referred to as a
pseudo-first-order approximation. Pseudo-first-order reactions occur
typically in the method of experimental design known as flooding [21,
and also in catalytic reactions in which a species is regenerated almost
as fast as it is consumed. This experimental procedure, also known as
the isolation method, was originally introduced by Ostwald in 1902.
As far as the present authors are aware, the material presented in this
section constitutes a first rigorous analysis of the method.

(it, Asymptotic Solution for k, + ffi

We shall construct an approximate solution of system (10)-(14) that

is valid in the case that the rate of reaction (k,) of the intermediate
species to the product is large. Formally, we set y = k3/(k1[A],) = 1/c
in equation (ll),and we seek a solution to system (10)-(12) that is
uniformly valid for E + 0. If we re-write equation (11) in the form
dY2
(40) E-
dx
+ (1 + PE)YZ = -(1 - 4 E Y 1 + EY:
 COMPLEX REACTION KINETICS 559

we note that the ‘highest derivative’ is multiplied by the perturbation

parameter E . This is the classic indicator of the ‘boundary-layer’prob-
lem. We shall solve the present problem by constructing a solution in
the outer region away from the boundary layer (called the outer solu-
tion), by constructing a solution in the boundary layer, the inner region
(called the inner solution), and by matching the outer and inner solu-
tions by the technique of matched asymptotic expansions. In order to
construct the outer solution in the region x = O(1) we set

in equations (10) and (40).

We find that

O < f f < l

where c1 and c2 are constants of integration to be determined. We

deduce that the boundary layer region is x = O(E),so in this (inner)
region we set
(46) YI = (Yl)in(X), YZ = (Yz)in(X), X = x / c
in equations (10) and (40).The equations for the inner solution are then

We construct the inner solution in the region X = O(1) by setting

- Yp’(x)+ EY\l)(X) + . . ’
(49)
(Y1)Ln(x)
(YZ),,(X) - E Y y ( X ) + 2 Y y y X ) + . . .
in equations (47)and (48).
I
560 SUMMERS, SCOTT AND RALPH

The equations for the leading terms are found to be

d Y i0’
(50) - - -0
dX

the solutions of which may be written,

(52) Yi0]= c3

(53) YP’ = a ( 1 - c4e-X)

where c3 and c4 are further constants of integration to be found.
The principle of matched asymptotic expansions requires that
(54)

(55)

in order to ensure the matching of the outer and inner solutions. Con-
ditions (54) and (55) imply that c1 = c2 = c3 = 1, for 0 < a 5 1.
Further, the application of initial conditions (12) a t x = 0 to solu-
tions (52) and (53) gives c4 = 1, and, moreover, we see that the inner
solution we have constructed extends uniformly to x = 0. We also see
that the above outer solution for y1 extends uniformly to x = 0.
Finally, in order to construct a (single) uniform approximation valid
for E + 0, and all x 2 0, we combine the outer and inner solutions by
using the expression

for each of y1 and yz, where ymatcllis the value of y in the region of
overlap of the outer and inner regions (the ‘matching’ region). The
results are

O<a<l
Solutions (44), (45) and (52), (53) are leading order solutions, and,
correspondingly, we have carried out the matching process (541, (55) to
leading order. In order to improve the asymptotic approximations (571,
(58) further we could derive higher order terms and achieve higher
order matching, though we shall not pursue this approach here. We
note that the boundary layer coincides with the induction period €or the
 COMPLEX REACTION KINETICS 561

reaction, which is essentially the period during which the intermediate

builds up to its maximum concentration.

f i i i ) Steady-State Approximation
A formal application of the standard ‘steady-state’ method of ap-
proximation of system (2146)requires that we set d [ C ] / d t = 0 in equa-
tion (4).Then, from equations (10) and (11)we derive the solutions,

(60)
1 - a E a ( l- a ) 2 e r l n1r11 f i e ,
Iyl = 1- ae-il-OIX’I+i(el y2 = (1 + p E ) ( l- a e - ~ l - ~ r ~ r ~ 2l ’+ ~ c ~

O<a<l

where again we have set E = 1/y ; the corresponding solutions for y3

and y4 are then given by equations (13) and (14). In the derivation of
solutions (59) and (60) we have applied the condition y,(O) = 1,but, by
virtue of the short-comings of the steady-state approximation, the con-
dition yz(0) = 0 is violated. In fact, yz(0) = a / ( 1+ P E ) for 0 < a 5 1.
We now perceive a n intimate connection between the steady state ap-
proximation, and the above asymptotic solution for k3 -+ x, uiz., to
leading order, the solutions for all species in the outer region x = O(1)
for k3 -+x,are precisely those solutions resulting from the steady-state
approximation.

(iu) Approximation to a Second Order System without Intermediate

Suppose we replace reaction (1)by the reaction
k
(61) A +B-D
which is governed by the equations,

(62) d[A1
- - - -h[A][B]
dt

(63)

(64) dlD1
- - - k[A][B]
dt
with
(65) [ A ]= [A],,, [B] = [B],,, [D]= 3 a t t = 0
562 SUMMERS, SCOTT AND RALPH

where k is a n initially unspecified rate constant.

System (62)-(65) is readily solved and we find that

(66)

where 6 = k/kl,and y4 is correspondingly given by equation (14).

Solutions (66) and (67) are commonly available in the chemical litera-
ture [31. We note, for instance, that the solutions for yl in (66) and (67)
(and the corresponding solutions y4) precisely agree with the steady-
state solutions (59) and (60) (and corresponding solutions y4) if we set
6 = 1/(1+ PE);these solutions also agree with the asymptotic solu-
tions for yl, y4 for E -+ 0 given in (571, (58) and (14) if we let S 1. The-+

latter limit may be effected by letting E + 0 or p -+0. Thus, taking

additionally into account the solutions obtained by Blandamer, Scott
and Summers (1985), we conclude that the approximate solutions for

-
both reactant concentrations [A] and [BI are common to four different
regimes: the exact system (1)with E -+ 0 (i.e. h3 x ,for any h, and k,);
-+

the exact system (1)with p -+ 0 (i.e. k, 0, for any k , and k& the
steady state regime with either E -+ 0 or p -+ 0 or both; and the second
order regime (61) with h = h,. Moreover, this conclusion regarding the
commonality of the solutions y1 and y4 (given by (571, (58) and (14)) is
valid for any value of a i.e. for any initial concentrations.

(u) Interconnection of Approximations fi)-fiv)

We have already examined close relationships connecting the solu-

tions given in the analysis section (ii)-(zu). Let us further examine the
relationships among the solutions given in this article. If we apply the
limit k, +. x to the solutions obtained (for a -+ 0) in 3 ( i ) , or, equiva-
lently, if we apply the limit a 0 to the solutions obtained (for h3 x)
--f -+

in 3(ii),then we obtain the following solutions that are valid for CY -+ 0,

h3 +. = (the limits taken in either order):

(68) y1 - 1 + a(-1 + e-"), y2 - aE(e-" - e-x'c),

y 3 - a(1 - e-"), y4 - ae-"

Likewise, if we apply the limits a -+ 0, k3 -+ = (in either order) to the

steady-state solutions given in Analysis (iii) we obtain the solutions,

(69) y1 - 1 + a ( - 1 + e-"), yz - ace-',

y3 - a ( l - e-"), y4 - ae-"
 COMPLEX REACTION KINETICS 563

We note that, as to be expected, the solutions for y,,y3 and y4 in (68)and

(69) agree; the solutions for the intermediate y2 in (68) and (69) agree
for x = O(1).
Finally, letting cr -+ 0 in the solutions obtained for the second order
system without intermediate in Analysis (iu), we obtain the solutions
(70) y1 - 1 + a(-1 + e-'"), y3 - a(1 - e-"), y4 -
Solutions (70) agree precisely with the solutions for yi, y3, and y 3 in
(68) and (69) if we set 6 = 1 (or k = k,)
The steady-state (or stationary state) hypothesis, originally intro-
duced by Bodenstein 141, has indeed achieved widespread, and not
always justified, use in chemical kinetics. We note the critical studies
by Aris [ 5 ] and Murray [61, and references contained therein. These
authors also give detailed accounts of singular perturbation techniques
in kinetics, and Aris (1969) points out that the original discovery of the
relationship between the steady-state solution and the (outer) asymp-
totic solution is reported by Bowen, Acrivos, and Oppenheim [71.
In Figure 1 we give a diagrammatic representation of the solutions
obtained in this article and their inter-relationships via the appro-
priate limits.

A + B A D a*O
System (62) - (65). ISolutions: (70) !
Solutions: ( 6 6 ) , ( 6 7 ) , ( 1 4 ) I
1

omit

I
intermediate

ICOl I c n \
li

k3 * =

. --
Solutions: ( 6 9 ) System (1s) - (21),(6).
S o l u t i o n s : ( 3 5 ) , ( 3 6 ) ,(25) , ( 2 6 ) /'
,
-- --
-1c

.-________- ----- /---'

Figure 1. Approximation schemes used in this article; a = IBI,/[A],. Solu-
tions (68), (691, and (70) are identical for [A], IB],and ID].
564 SUMMERS, SCOTT AND RALPH

Numerical Results and Discussion

In the Analysis section, four types of approximate solution to the
ordinary differential equations which describe the second-order
reaction (1)are given, namely, a pseudo-first-order approximation, an
asymptotic solution for k, + m, the steady-state approximation, and a n
approximation to a second-order system without intermediate. It is not
our purpose here to carry out a numerical study of equation (1)for a
comprehensive set of values of the rate constants k l , k 2 , k, and the ratio
of the initial concentrations [ B l , / [ A ] ,= a. However, we will compare
the above approximate solutions to numerical (Runge-Kutta) solutions
to the full equations (10)-(12) for a representative selection of parame-
ter values. In Figures 2-4 we compare the pseudo-first-order approxi-
mations with Runge-Kutta solutions for a given set of rate constants
( k l = 1,k2 = k3 = 0.5) in each of the three cases a = 1, a = 0.6, and
a = 0.1. Results (33)-(34), which are uniformly valid for all time t and
for all values of the rate constants, indicate that the pseudo-first-order
approximation to [ A ] ,the dominant species, is more accurate than the
approximations to the other species, for a given value of a. It is inter-
esting to note that in the practical implementation of the flooding
technique species B , the dominated species, is the species usually mea-
sured. For a = 0.1 the maximum relative errors between the pseudo-
first-order solution and the Runge-Kutta solution for the species A and
B are respectively 0.17% and 2.21%. Thus, (Y need not be vanishingly
small in order for the pseudo-first-order approximations to be useful.
When a = 0.6 these two errors are 11.63%and 15.62%respectively. For
completeness, and also to exhibit a typical worst case, we have also
shown the concentration profiles for a = 1 in Figure 2. Although the
pseudo-first-order approximations (33) and (34) are uniformly valid for
all time t, the approximations for ya and y4 become less useful for large
values of time because yz and y4 tend to zero exponentially as t + =.
In Figures 5-7 we have taken cy = 1, and k l = k2 = 1, k3 = 10 (i.e.
E = 0.1); the asymptotic solutions, the steady-state solutions and the
second-order solutions without intermediate are shown together with
Runge-Kutta solutions. The inter-related solution properties derived
in the Analysis section and exhibited in Figure 1are also clearly dem-
onstrated in Figures 5-7. As E ( = k l [ A ] , / k , )becomes smaller the con-
centration profiles of A, B , and D in the three approximation regimes
shown become indistinguishable from the Runge-Kutta solutions. In
Figure 9, in which a = 1, k l = k 2 = 1 and kj = 100, i.e., E = 0.01,
these approximate solutions for [ A ] ,[BI, and [Dl are essentially coin-
cident. Figures 6 and 8 illustrate the superiority of the asymptotic
solutions for E -+ 0 over the steady-state solutions in the sense that the
former are applicable throughout the reaction while the latter are
applicable only beyond the induction period.
 COMPLEX REACTION KINETICS 565

A0'l.O BO'l.0

Time ( Seconds 1
Figure 2. Comparison between pseudo-first-order solutions (351, (361, (251, (261
and Runge-Kutta solutions for a = 1;in this figure and throughout the paper the
units of k , , k , , k , and [A], or [BI0 are respectively dm3 rnol-ls-', s-', s ', and
mol dm-'. ------Pseudo-first-order solution, ___ Runge-Kutta solution.

Time (Seconds 1
Figure 3. As Figure 2 except that a = 0.6. ----- Pseudo-first-order solution,
___ Runge-Kutta solution.
566 SUMMERS, SCOTT AND RALPH

u)
0.6

c
e
c

Time (Seconds 1
Figure 4. As Figure 2 except that a = 0.1.----- Pseudo-first-order solution,
_ _ Runge-Kutta solution.

I .o I I I I

k l 51.0 k2=1.0 k3=10.0

m A,= 1.0 Bo = 1.0

b
a
-I-
0
c
._
c
0

c
e
c
a
t
V
0
0
0.2 -

. I I I I A
O.OO.0 2.o 4.O 6.0 0.0 10.0

...
Time ( Seconds 1
Figure 5. ___ Runge-Kutta solution, ----- Asymptotic solution,
Steady-state solution, Second-order solution without intermediate.
..
 COMPLEX REACTION KINETICS 567

Time (Seconds)
Figure 6. ___ Runge-Kutta solution, ----- Asymptotic solution, 0.
Steady-state solution.

1.0 I I I I

Time ( Seconds )
Figure 5-7. Comparisons between asymptotic solutions (57), (13), (141, steady-
state solutions (59), (131, (141, second-order solutions without intermediate (66),
(14) with 6 = 1, and Runge-Kutta solutions, for the case LI = 1. _ _ Runge-
Kutta solution, ----- Asymptotic solution, 0 Steady-state solution,
Second-order solution without intermediate.
568 SUMMERS, SCOTT AND RALPH

.OK I I I I

k, z1.0 k, = 1.0 k, = 100

0 A, =LO eo=i.o
a .OOE
c
._
0
U
a,

5
-
2
c
\c
0
.oot

--e
c
._
0
.OOL

c
a,
u
c
u" .ooi

I I I I
.OO(
) 0. I 0.2 0.3 0.4 0.5

Time (Seconds 1
Figure 8. -Runge-Kutta solution, ----- Asymptotic solution, 0.0
Steady-state solution.

Time (Seconds)
Figures 8-9. As Figures 5-7 except that k~ = 100.___ Runge-Kutta solution,
A A A Asymptotic solution, Steady-state solution, # Second-order solution
without intermediate.

Experimental determination of the rate constants in such reactions

as (1)is beset with a number of technical problems including the mea-
surement of the concentration-time profile of the intermediate species
 COMPLEX REACTION KINETICS 569

Time (Seconds)
Figure 10. Least squares linear fit to the (Runge-Kutta) second-order function of
A and B occurring in solution (67); 8 = A,, - A = BO B . A A A Runge-Kutta
~

solution, ___ Least squares linear fit.

(especially when the concentration of the intermediate is small). In-

deed, we are not aware of a study that determines the rate constants,
k l ,k, and k , in the simpler (first-order) reaction
A e B - - % k C

from experimental concentration-time data for all the species; see also
[S], [9], and [lo]. From Figures 5, 7 and 9 we see that measurements
of the reactants and product only might lead an experimentalist to
conclude erroneously that the reaction is of second order without inter-
mediate. Such difficulties are compounded by the presence of experi-
mental errors. To show that the rate constant k3 need not be large for
such problems to arise we have set k , = k 2 = k3 = 1 and a = 0.5 in
Figure 10. Here, we plot, against time, Runge-Kutta values for the
second order quantity ln([Blo([Alo- 8)/[Alo([Blo- 8))/([Alo - [BM,
where 8 = [A], - A = [BI, - B , which occurs in solution (671, and we
have performed a least-squares linear fit to 24 points of the Runge-
Kutta solution, giving a second order rate constant of 0.42 dm3 mol-'
sec-'. Again, on this evidence alone an experimentalist might conclude
erroneously that the slight apparent departure from second order
behavior is due to experimental error in concentration-time mea-
surements and that the reaction is a simple second order process.
Aside from practical experimental difficulties in the study of complex
reactions, there remains a need to develop analytical methods appro-
priate to these nonlinear systems. The analysis and solutions presented
570 SUMMERS, SCOTT AND RALPH

in this article usefully develop existing theory and provide both tools
and caveats to the experimentalist in the determination of rate con-
stants from concentration-time profiles.

Acknowledgments
This work is supported by the Natural Sciences and Engineering
Research Council of Canada through the award No. 3-30862 to D.S.
and J.M.W.S. from the General Grant to the President of Memorial
University of Newfoundland. We are grateful to Professor Peter Gray
of the University of Leeds for helpful comments.

Bibliography
111 M. J . Blandamer, J.M. W. Scott, and D. Summers, Can. J . Chem., 1985, 63, 3294.
I21 J. H. Espenson, 1981 Chemical kinetics and reaction mechanisms. New York:
McGraw Hill.
[31 W. J . Moore, 1972 Physical chemistry. New Jersey: Prentice-Hall.
[41 M. Bodenstein, 2. Physik. Chem., 1913, 85, 329.
[5] R. Aris, Znd. Eng. Chem., 1969, 61, 17.
[6] J . D. Murray, 1977 Lectures on nonlinear-differential-equationmodels in biology.
Oxford: Clarendon Press.
[71 J . R . Bowen, A. Acrivos, and A. K. Oppenheim, Chem. Eng. Sci., 1963, 18, 177.
[81 D. D. Perrin, and I. H. Pitman, J . Chem. SOC.,1965, 7071.
[91 L. P. Hammett, 1970 Physical organic chemistry. New York: McGraw Hill.
[lo] M.J. Blandamer, R.E. Robertson, E. Ralph, and J.M.W. Scott, J . Chem. SOC.
Furuduy Trans., 1983, 1, 79, 1289.

Received September 22, 1986

Accepted November 3, 1986