FLUID MECHANICS AND TRANSPORT PHENOMENA

Diffusing with Stefan and Maxwell

Neal R. Amundson
Dept. of Chemical Engineering and Dept. of Mathematics, University of Houston, Houston, TX 77204

Tsorng-Whay Pan and Vern I. Paulsen

Dept. of Mathematics, University of Houston, Houston, TX 77204

In most chemical engineering problems, diffusion is treated as an add-on to forced

ad®ection, and the boundary conditions are the Danckwerts conditions in order to
maintain conser®ation. If we treat problems in which there is no applied ad®ection or
chemical reaction for a steady-state situation and with an ideal gas, it is an equi-molar
process. Because it is considered the natural mo®ement of molecules mo®ing down a
concentration gradient, this requires the Stefan-Maxwell equations. In the standard 1-D
two-component boundary ®alue problem, analogous to the Dirichlet problem, the solu-
tion is direct and is probably the only one there is. In a ternary system, the solution is
already not direct and can only be obtained by numerical means, which is not se®ere. In
a quaternary system, it does not appear feasible to obtain a simple procedure like that
obtained for the ternary system. A different numerical scheme de®eloped is robust with
rapid con®ergence and works well with an arbitrary number of components, 60 ha®ing
been used in one problem. As the number of components increases, the solution profiles
tend to become linear and the dependence on particular diffusi®ities is less important.
This manifests itself when using diffusi®ities from a random collection. The problem
using a continuous distribution of components is sol®ed, and computationally and theo-
retically the profiles are probably linear and with a single pairwise diffusion coefficient.

Introduction
The purpose of this article is to present some manipula-
tions of the Stefan-Maxwell equations as applied to the sim-
ple one-dimensional Ž1-D. diffusion problem in the steady
state for an ideal gas with an arbitrary number of compo-
nents. Anyone who has examined these equations becomes
frustrated by their backward presentation which manifests it-
self in some way in almost every application of their use. In
addition, they are nonlinear in the two important sets of vari-
ables although linear in each set singly. The two sets of vari-
ables are the mol fraction of each component and the corre-
sponding flux of each. Manipulation of these equations re-
veals that their inversion to produce the fluxes in terms of
the gradient for a two-component system is trivial. For a
three-component system, it is doable and builds one’s confi-
dence, while for a four-component system the problem is a
character builder; for a five-component system, it seems im-
Correspondence concerning this article should be addressed to N. R. Amundson.
passable and impossible in an analytical way. Some of the
earliest attempts at using the equations were carried out by
Toor Ž1964. and Stewart and Prober Ž1964. who linearized
the equations to cast them into a usable form. This is a suc-
cessful procedure since the system is very benign, and, as will
be shown later, many of the solutions turn out to be close to
linear. Later workers developed many other schemes, and
these are discussed by Taylor and Krishna Ž1993. whose book
is required reading. The numerical scheme devised by
Krishna and Standart Ž1976. was and still is the method of
choice.
Our first effort will be to invert the Stefan-Maxwell equa-
tions for three and four component systems, and to attempt
to simplify these equations to make their use easier. For a
three-component system, this is relatively easy, but for a
four-component system, this is a trial and requires the intro-
duction of two sets of simple functions to present the inver-
sion in a relatively simple way. Some small surprises result
here and make its perusal worth the effort.

AIChE Journal April 2003 Vol. 49, No. 4 813

 A different numerical scheme for the 1-D diffusion prob- since motion of molecules must be balanced, that is, move-
lem is presented, and its application to several different ment of one molecule must be balanced by the motion of
problems is illustrated for a wide range of components. The another molecule. Thus, any one of the differential equations
equations for the diffusion problems with a continuous distri- can be replaced by one of the two conditions above.
bution of components are presented, and the 1-D diffusion For the ternary system, we will use the set of equations
problem is solved although the numerical solution is obtained ŽEq. 2.,
by a lumping procedure as for a finite distribution. For some
of these problems, it is shown by direct computation that the dx 1 x2 x3 x1 x1
final solutions for the mol fraction profiles are relatively in-
dependent of the choice of diffusion coefficients.
c
dz
sy
ž D 12
q
D 13 / N1 q
D 12
N2 q
D 13
N3 ,

dx 2 x2 x1 x3 x2
Ternary Inversion
c
dz
s
D 21
N1 y
ž D 21
q
D 23 / N2 q
D 23
N3 , Ž3.
The basic set of equations for 1-D diffusion in matrix form
0 s N1 q N2 q N3 .
is, for an n component system,

dx In order to invert the system we will treat this as a set of

c s Bx Ž1. linear simultaneous algebraic equations in N1, N2 , and N3
dz and simply solve using Cramer’s rule on Eq. 3. This proce-
dure gives solutions for N1, N2 , and N3 in terms of dx 1rdz
where c is the total molar local concentration, x is the vector
and dx 2rdz which are linear in these variables and clumsy,
of mol fractions, z is the axial space variable, Nj is the molar
and we must cast them into relations which are independent
flux of species j, and the Di j represent the pairwise diffusion
of the species numbering and also quasi symmetric as occurs
coefficients. With
in the original Stefan-Maxwell equations. To do this, we add
to the three relations quantities like
B s w ␮i j x Ž i row; j column .
3 dx i
then ␣ s0
Ý dz
is1
Ni n Nk
␮ i j sy , i/ j, ␮ ii s Ý
Di j k s1 Dik with ␣ appropriately chosen to each equation to produce the
k/i species independent property, and this is much easier to do
than it sounds. The Cramer denominator turns out to be
which shows the linearity in x i for known values of Nj . These
equations will also be used in the form x1 x2 x3
q q
dx D 13 D 12 D 12 D 23 D 13 D 23
c s␤N Ž2.
dz
which we will use in the form, on multiplying through
where D 12 D 13 D 23 , as

␤ s w ␯i j x ␾ Ž x . s D 23 x 1 q D 13 x 2 q D 12 x 3 . Ž4.
with
After some juggling, we obtain the matrix form of the inverse
xi n xk
␯i j s , i/ j, ␯ ii sy Ý . c
Di j k s1 Dik Nsy
k/i ␾Ž x.

These matrices are singular since each column will sum to Ž 1y x 1 . D13 D12 y x 1 D 23 D 12 y x 1 D 23 D 13
zero, and, hence, the sets of differential equations are also = y x 2 D 13 D 12 Ž 1y x 2 . D 23 D12 y x 2 D 23 D 13
singular. Since we are dealing with 1-D steady-state diffusion,
y x 3 D 13 D 12 y x 3 D 23 D 12 Ž 1y x 3 . D 23 D13
the singularity may be circumvented by using either the fact
that diffusion must be equimolar dx
. Ž5.
n dz
Ý Nj s 0,
js 1 In this form it is evident that the flux Ni depends on a sum of
terms with a net single diffusivity as it should.
or Since our main interest is in the solution of the 1-D diffu-
n sion equations, we consider diffusion in a 1-D conduit which
Ý x i s1, is adjoined at its ends by infinitely large reservoirs, each at a
is 1 fixed molar concentration and each at a fixed and the same

814 April 2003 Vol. 49, No. 4 AIChE Journal

temperature and same pressure. In one dimension the con- after dividing through by D 12 D 13 D 23. Integrating from 0 to z,
centration equations are, where l is the conduit length gives

dNi x 1 y x 10 x 2 y x 20 x 3 y x 30
s 0, 0 - z - l, is1, 2, 3 c q q
dz D 23 D 13 D 12

N1 N2 N3
and boundary conditions are, with x 0 and x l specified s
ž D 13 D 12
q
D 12 D 23
q
D 13 D 23 / z

x Ž0. s x 0 , x Ž l . s x l . so

This simple problem will be cast in dimensionless form later, 1 dx 1 1 dx 2 1 dx 3

since now the occurrence of the diffusivities is revealing. From q q
D 23 dz D 13 dz D 12 dz
the conservation equation, Ni is a constant for all x and
x 1l y x 10 x 2 l y x 20 x 3 l y x 30 1 K
n
s q q s Ž6.
D 23 D 13 D 12 l l
Ý Ni s 0
is 1
and also

since this diffusion movement of molecules only means that x 1 y x 10 x 2 y x 20 x 3 y x 30

they replace each other. q q
D 23 D 13 D 12
In an effort to simplify Eq. 5 by removing the unity in each
row, we multiply it by the diagonal matrix x 1l y x 10 x 2 l y x 20 x 3 l y x 30 z
s
ž D 23
q
D 13
q
D 12 / l
D 23 0 0
0 D 13 0 which is not a relation one might expect for three compo-
0 0 D 12 nents.
In order to use these equations, particularly Eq. 6, let us
consider the expression for N1, and write it in the form
and then add the rows to obtain
c
N1 sy Ž D12 D13 D 23 .
c dx 1 ␾Ž x.
y D 13 D 12 Ž y x 1 D 23 y x 2 D 13 y x 3 D 12 .
␾Ž x. dz 1 dx 1 1 dx 1 1 dx 2 1 dx 3
=
q D 23 D 12 Ž y x 1 D 23 y x 2 D 13 y x 3 D 12 .
dx 2 D 23 dz
y x1
ž D 23 dz
q
D 13 dz
q
D 12 dz /
dz
which reduces to an expression with a single derivative
dx 3
q D 23 D 13 Ž y x 1 D 23 y x 2 D 13 y x 3 D 12 .
dz c 1 dx 1 K
N1 sy Ž D12 D13 D 23 . y x1
s D 23 N1 q D 13 N2 q D 12 N3 . ␾Ž x. D 23 dz l

We see that, fortuitously, the three quantities within the If we now cast this in dimensionless form, using the Cramer
parentheses are ␾ Ž x . so that from Eq. 4 denominator ŽEq. 4., we obtain

dx 1
dx 1 dx 2 dx 3 y q x 1 K 23 s Z1 w c 23 x 1 q c13 x 2 q c12 x 3 x K 1
c D 13 D 12 q D 23 D 12 q D 23 D 13 d␰
dz dz dz

s D 23 N1 q D 13 N2 q D 12 N3 . with

z Di j
Integration between zero and l gives ␰ s , ci j s , K 23 s D 23 K , K 13 s D 13 K , K 12 s D 12 K ,
l D
x 1l y x 10 x 2 l y x 20 x 3 l y x 30 N1 l N2 l N3 l D2
c q q Z1 s , Z2 s , Z3 s , K 1s ,
D 23 D 13 D 12 Dc Dc Dc D 12 D 13

N1 N2 N3 D2 D2
s
ž D 13 D 12
q
D 12 D 23
q
D 13 D 23 / l K2s
D 12 D 23
, K3s
D 13 D 23

AIChE Journal April 2003 Vol. 49, No. 4 815

and in a similar way example, dx 3rd ␰ s 0 at a point, then it follows that

dx 2 dx 2 dx 1
y q x 2 K 13 s Z2 w c 23 x 1 q c13 x 2 q c12 x 3 x K 2 , q s0
d␰ d␰ d␰

dx 3
y q x 3 K 12 s Z3 w c 23 x 1 q c13 x 2 q c12 x 3 x K 3 . which is impossible since the functions defined by the corre-
d␰ sponding two SM equations are linearly independent. If we
differentiate the equation for x 1, we obtain
These last three equations should be simpler to apply to
the numerical problem than the original ones, but one needs d 2 x1 1 dx 2 1 dx 3 dx 1 N2 N3
to have a good estimate of Z1, Z2 , Z3 to initiate that solu-
tion. One can find such a solution by assuming that an ap-
c
d␰ 2
sy
ž D 12 d ␰
q
D 13 d ␰ / N1 q
d␰ ž D 12
q
D 13 / .

proximate solution is a straight line connecting the appropri-

ate boundary points and using the values of x 0 Žat z s 0. to If we assume this is zero, then, using
obtain
dx i
Ý s0
x 1l y x 10 i d␰
y q x 10 K 23 s Z10 Ž c 23 x 10 q c13 x 20 q c12 x 30 . K 1
l
x 2 l y x 20 we have
y q x 20 K 13 s Z20 Ž c 23 x 10 q c13 x 20 q c12 x 30 . K 2
l
1 1 dx 2 dx 1

from which a reasonable guess should ensue. ŽNote that

ž y
D 12
q
D 13 /ž N1
d␰
y N2
d␰ / s0

Ý Zi s 0..
We stress that the advantage to these equations is the fact which gives
that the lefthand sides are linear first-order operators since
K i j is a predetermined constant. If we write the original Ste- dx 2 dx 1
fan-Maxwell equations in the form N1 s N2
d␰ d␰

dx i Nj xj
yc q xi Ý s Ni Ý
dz Di j Di j
the left-hand side is a linear operator in x i , but its coefficient
contains the unknown Nj for all j except js i. It is evident
of course that these equations for constant values of Ni and
Ý x i s1 have an analytical solution which is not of great use
since the boundary conditions must be used to eventually de-
termine the Ni , which is a tortuous exercise, but doable.
It may be interesting to consider the structure of the solu-
tion profiles x i s x i Ž ␰ ., not because the solutions are rich
with pathology, but rather in order to show their simplicity.
Diffusional processes are normally very benign Žas long as
they are not coupled with exothermic chemical reactions.,
since molecular motion tends to smear things out.
If we consider a ternary system, there are three equations
of which only two are linearly independent in pairs. If, for
which is also impossible because of the linear independence.
Thus, the profiles of the components are strictly monotonic
between the boundaries and we shall illustrate some of these
later. The above may be generalized to higher-order equa-
tions.
Toor Ž1957. in the early 1950s solved the ternary diffusion
problem by obtaining the analytical solution for fixed fluxes,
and then solved for the fluxes by trial and error. Hsu and
Bird Ž1960. published a massive paper for three-component
chemical reaction problems in which the catalytic reaction
occurred at one of the boundaries. These were solved in an
analytical way with computations on an IBM 650.
Quaternary Systems
In order to realize the inversion we must solve the follow-
ing set of linear simultaneous equations in N1, N2 , N3 , and
N4 , where we use Eq. 2

dx 1 y x2 x3 x4 x1 x1 x1
c y y
dz D 12 D 13 D 14 D 12 D 13 D 14
dx 2 x2 y x1 x3 x4 x2 x2
c y y
dz s D 21 D 21 D 23 D 24 D 23 D 24 N.
dx 3 x3 x3 y x1 x2 x4 x3
c y y
dz D 31 D 32 D 31 D 32 D 34 D 34
0 1 1 1 1

816 April 2003 Vol. 49, No. 4 AIChE Journal

One could hope that the treatment of quaternary systems Table 1. Definition of the Functions in Eqs. 8.
might not be substantially different from ternary systems but, ␣ 12 ␣ 21 ␣ 13 ␣ 31 ␣ 14 ␣ 41
of course, the fourth-order manipulations are certainly more x1 x 2 x 2 x1 x1 x 3 x 3 x1 x1 x 4 x 4 x1
tedious in this Cramer solution. The challenging part of the
D 13 D 24 D 23 D 14 D 12 D 34 D 32 D 14 D 12 D43 D42 D 13
exercise, however, as before, is the casting of the resultant
w Di j s Dji x
solutions into forms which are independent of the numbering ␣ 23 ␣ 32 ␣ 24 ␣ 42 ␣ 34 ␣ 43
of the species and are quasi-symmetric in the component x2 x3 x3 x2 x2 x4 x4 x2 x3 x4 x4 x3
variables. The Cramer solution first obtained is linear in the
D 21 D 34 D 31 D 24 D 21 D43 D41 D 23 D 31 D42 D41 D 32
three derivatives dx 1rdz, dx 2rdz, and dx 3rdz since the last
Stefan-Maxwell equation is replaced by
4
and
Ý Ni s 0.
is 1 x2 x3 x4
X1 s q q ,
The Cramer denominator may be written as a fourth-order D 23 D 24 D 23 D 34 D 24 D 34
quadratic form. This is not readily apparent during the calcu- x1 x3 x4
lation, but it can be written in the form X2 s q q
D 13 D 14 D 13 D 34 D 14 D 34
␾Ž x. x1 x2 x4
X3 s q q ,
d12 d13 d14 d 24 d13 d12 d12 d13 d 34 d12 d14 d 34 D 14 D 12 D 12 D 24 D 14 D 24
d12 d14 d 23 d 24 d 23 d12 d12 d 23 d 34 d 24 d 34 d12 x1 x2 x3
s xT x; X4 s q q Ž9.
d13 d14 d 23 d 23 d13 d 24 d13 d 34 d 23 d 24 d 34 d13 D 12 D 13 D 23 D 12 D 13 D 23
d13 d14 d 24 d 24 d14 d 23 d 23 d14 d 34 d 24 d 34 d14
Žnote that these have the same gross form as the ternary
d i j s d ji Ž7. Cramer denominator., and

where d i j s Dy1 4
i j , that is, the elements in the matrix are re-
ciprocals of the pairwise diffusion coefficients. There are six ␣i s Ý Ž ␣ i s q ␣ si . Ž 10.
ss1
pairwise diffusion coefficients in a four component set, and s/ i
so there are 20 combinations of six things taken three at a
time. Clearly, four of these are missing in the above, and, which is the sum of the six terms in which x i appears in each
after some scrutiny, it develops that these four are those in term and ␣ i j is defined in Table 1.
which one of the four digits is missing; that is, there is no It is seen that X i contains no x i , and each x j has D l j D k j ,
subscript form like d12 d13 d 23. Physically, this makes sense l / k / i/ j in the denominator in Eq. 9. The X i’s are quasi-
since it means that every chemical component has an effect symmetric, as are the ␣ i j ’s. Note also that every ␣ i j has a
in each term. Since the quantity obtained as the Cramer de- 1234 subscript property in some order, and, therefore, there
terminant was not obviously a quadratic form a priori, the are only three different ones. If one examines the matrix so-
placement of the terms in the matrix in a quasi-symmetric lution carefully, one finds that it is quasi-symmetric, al-
form was somewhat arbitrary at the start and might need re- though, in some cases, one must go back to the definitions of
arrangement. the functions. It is readily apparent also that the matrix is
The solution obtained can be put into the matrix-vector singular since the sum of the rows is zero.
form given below, and the placement of the functions ␣ i j , The solution matrix can also be written in the form
defined below, is restricted by the form,
c dx
4 dx i N sy w wi j x
␾Ž x. dz
A Ý s0
is 1 dz
with
where A is chosen and the functions X i Ž x . and ␣ i j are cho- xl q xm
sen and are defined below wi j sy x i X j q x i ; i/ j/ l / m,
Di j D lm
c dx
N sy W Ž8. xl q xm
␾Ž x. dz wii s Ž 1y x i . X i q x i Ý ; s/ j/ l / m.
s Di s D lm
where the matrix W is

Ž 1y x 1 . X1 q ␣ 1 y x 1 X 2 q ␣ 13 q ␣ 14 y x 1 X 3 q ␣ 12 q ␣ 41 y x 1 X 4 q ␣ 21 q ␣ 31
y x 2 X 1 q ␣ 23 q ␣ 24 Ž 1y x 2 . X 2 q ␣ 2 y x 2 X 3 q ␣ 21 q ␣ 42 y x 2 X 4 q ␣ 12 q ␣ 32
y x 3 X 1 q ␣ 32 q ␣ 34 y x 3 X 2 q ␣ 31 q ␣ 43 Ž 1y x 3 . X 3 q ␣ 3 y x 3 X 4 q ␣ 13 q ␣ 23
y x 4 X 1 q ␣ 42 q ␣ 43 y x 4 X 2 q ␣ 41 q ␣ 34 y x 4 X 3 q ␣ 14 q ␣ 24 Ž 1y x 4 . X4 q ␣ 4

AIChE Journal April 2003 Vol. 49, No. 4 817

If we were to assume that all the diffusivities were the same, Using the expression for the Cramer denominator given by
then Eq. 7, we can produce some other interesting results. If we
examine ␾ Ž x ., in particular, the last three rows and the last
D 2 wi j sy x i Ý x s q x i Ý x s ; i/ j, three columns, we see that row two in the last three elements
s/ j s/ i may be written Ž x 2rD 12 . X 1, while the third and the fourth
s/ j
rows Žthe last three elements. become
sy x i Ž 1y x j . q x i Ž 1y x i y x j . sy x i2 , i/ j,
x3 x4
D 2 wii s Ž 1y x i . Ž 1y x j . q2 x i Ž 1y x i . s1y x i2 X1 q X1 .
D 13 D 14

and, hence, since the Cramer denominator equals Dy3 It then follows that ␾ Ž x . may be written

1y x 12 y x 12 y x 12 y x 12 x2 x3 x4 ␣ 12 q ␣ 21
␾Ž x.s ž q q
/ X1 q
y x 22 1y x 22 y x 22 y x 22 dx D 12 D 13 D 14 D 12
N sy cD ,
y x 32 y x 32 1y x 32 y x 32 dz
␣ 13 q ␣ 31 ␣ 14 q ␣ 41 x 12
y x 42 y x 42 y x 42 1y x 42 q q q . Ž 13.
D 13 D 14 D 12 D 13 D 14
dx
N sy cD , Now, to find a similar representation involving X 2 , we look
dz
at the first, third, and fourth rows, and first, third, and fourth
and so singularity is preserved. columns, and, proceeding as above, one obtains
Because we wish to simplify, if possible, the quaternary
problem we will spend some time now on the structure of x1 x3 x4 ␣ 21 q ␣ 12
what we have already established. Some surprising solutions ␾Ž x.s ž q q
/ X2 q
D 21 D 23 D 24 D 21
will surface. For example, if we examine the Cramer denomi-
nator ŽEq. 7., we discover that it has the form ␣ 23 q ␣ 32 ␣ 24 q ␣ 42 x 22
q q q . Ž 14.
D 23 D 24 D 21 D 23 D 24
␾Ž x.
x 12 x1 x 2 x1 x 3 x1 x 4 The expansion in terms of X 3 follows from the elements of
s q q q the first, second, and fourth columns, and first, second, and
D 12 D 13 D 14 D 12 D 13 D 24 D 12 D 13 D 34 D 12 D 14 D 34
fourth rows. One obtains
x 2 x1 x 22 x2 x3 x2 x4
q
D 14 D 21 D 32
q
D 12 D 24 D 23
q
D 12 D 23 D 34
q
D 12 D 24 D 34 x1 x2 x4 ␣ 13 q ␣ 31
␾Ž x.s ž q q
/ X3 q
D 31 D 32 D 34 D 31
x 3 x1 x3 x2 x 32 x3 x4
q
D 13 D 14 D 23
q
D 13 D 24 D 32
q
D 13 D 23 D 34
q
D 13 D 24 D 34 ␣ 23 q ␣ 32 ␣ 34 q ␣ 43 x 32
q q q . Ž 15.
D 32 D 34 D 31 D 32 D 34
x 4 x1 x4 x2 x4 x3 x 42
q q q q
D 13 D 14 D 24 D 14 D 23 D 24 D 14 D 23 D 34 D 14 D 24 D 34 The expansion in terms of X 4 follows from the elements of
the rows and columns in the upper lefthand corner and is
Ž 11 .

where we use the commonality of some species in each row x1 x2 x3 ␣ 14 q ␣ 41

␾Ž x.s ž q q
/ X4 q
and each column, and, if we look at this a little more closely, D41 D42 D43 D41
we can, using the ␣ i j functions defined before, write
␣ 24 q ␣ 42 ␣ 34 q ␣ 43 x 42
q q q . Ž 16.
x 12 ␣ 12 ␣ 13 ␣ 14 D42 D43 D41 D42 D43
␾Ž x.s q q q
D 12 D 13 D 14 D 12 D 13 D 14
We see also that there is a systematic scheme for the above
␣ 21 x 22 ␣ 23 ␣ 24 relations. The expansions in terms of X 1 and X 4 are made
q q q q
D 21 D 12 D 23 D 24 D 23 D 24 directly from ␾ Ž x .. The expansion in terms of X 2 and X 3
requires some minor symmetric juggling, more for X 3 than
␣ 31 ␣ 32 x 32 ␣ 34 for X 4 .
q q q q
D 31 D 32 D 31 D 32 D 34 D 34
Simple Quaternary Representation
␣ 41 ␣ 42 ␣ 43 x 42
q q q q . Ž 12. With the ternary system by some judicious juggling, we were
D41 D42 D43 D41 D42 D43 able to reduce the fundamental equations to equations which

818 April 2003 Vol. 49, No. 4 AIChE Journal

had a linear operator. This is desirable since the operator first the representation for N1
determines the form of the solution and should enable one to
produce a numerical solution, which is less likely to be unsta-
ble and to converge more rapidly. With this in mind, we will c dx 1 dx j dx
y N1 s X1 y x1 Ý X j q ␣1
try to do the same with the quaternary system written in the ␾Ž x. dz j dz dz
form

where ␣ 1 is the first row of ␣ . Then

c dx dx j dx
N sy X y Ý Xj ⴢ xq␣ Ž 17.
␾Ž x. dz dz dz
j y N1 ␾ Ž x . dx 1 dx 2
s w Ž 1y x 1 . X 1 q ␣ 1 x y w x 1 X 2 y ␣ 13 y ␣ 14 x
c dz dz
where X is the diagonal matrix with X i on the ith term of dx 3 dx 4
the diagonal and where we note that the quasi-symmetry in ␣ y w x 1 X 3 y ␣ 12 y ␣ 41 x y w x 1 X 4 y ␣ 21 y ␣ 31 x .
in the second matrix which contains the functions belonging dz dz
to ␣ i j
This equation is not as mean as it looks for all of the ␣ i j ’s
and ␣ ji’s contain an x 1. A little juggling with the formula
␣1 ␣ 13 q ␣ 14 ␣ 12 q ␣ 41 ␣ 21 q ␣ 31
␣ 23 q ␣ 24 ␣2 ␣ 21 q ␣ 42 ␣ 12 q ␣ 32
␣s dx j
␣ 32 q ␣ 34 ␣ 31 q ␣ 43 ␣3 ␣ 13 q ␣ 32 AÝ s0
dz
␣ 42 q ␣ 43 ␣ 41 q ␣ 34 ␣ 14 q ␣ 24 ␣4

␣1 will cast it into a form

␣2
s , Ž 18 . y N1 ␾ Ž x .
␣3 dx 1 dx 2
s w Ž 1y x 1 . X 1 x y Ž x 1 X 2 q ␤ 12 .
␣4 c dz dz

dx 3 dx 4
y Ž x 1 X 3 q ␤ 13 . y Ž x 1 X 4 q ␤ 14 . .
which is singular and dz dz

␣s Notice what one would call ␤ 11 s 0 and we have the follow-

ing for N1, N2 , N3 , and N4
x 1Ž x 3qx 4 . x 1Ž x 2qx 4 . x 1Ž x 2qx 3 .
␣1
1234 1324 1423 ␤ 12 s ␣ 12 q ␣ 21 q ␣ 31 q ␣ 41 , ␤ 21 s ␣ 21 q ␣ 12 q ␣ 32 q ␣ 42 ,
x 2 Ž x 3qx 4 . x 2 Ž x 1qx 4 . x 2 Ž x 1qx 3 .
␣2 ␤ 13 s ␣ 13 q ␣ 31 q ␣ 14 q ␣ 21 , ␤ 23 s ␣ 23 q ␣ 32 q ␣ 12 q ␣ 24 ,
1234 1423 1324
x 3 Ž x 2qx 4 . x 3 Ž x 1qx 4 . x 3 Ž x 1qx 2 . ␤ 14 s ␣ 14 q ␣ 41 q ␣ 12 q ␣ 13 , ␤ 24 s ␣ 24 q ␣ 42 q ␣ 23 q ␣ 21 ,
␣3
1324 1423 1234
␤ 11 s 0, ␤ 22 s 0,
x 4 Ž x 2qx 3 . x 4 Ž x 1qx 3 . x 4 Ž x 1qx 2 .
␣4
1423 1324 1234 ␤ 31 s ␣ 31 q ␣ 13 q ␣ 23 q ␣ 43 , ␤41 s ␣ 41 q ␣ 14 q ␣ 34 q ␣ 24 ,

Ž 19. ␤ 32 s ␣ 32 q ␣ 23 q ␣ 34 q ␣ 13 , ␤42 s ␣ 42 q ␣ 24 q ␣ 43 q ␣ 14 ,

␤ 34 s ␣ 34 q ␣ 43 q ␣ 31 q ␣ 32 , ␤43 s ␣ 43 q ␣ 34 q ␣ 41 q ␣ 42 ,
where in the denominator we have only written the subscripts
␤ 33 s 0, ␤44 s 0.
of Di j D lm.
In an effort to obtain a simpler representation of the inver-
sion given by Eq. 8 we will use Eq. 17 in a form by writing and also

Ž 1y x 1 . X1 y x 1 X 2 y ␤ 12 y x 1 X 3 y ␤ 13 y x 1 X 4 y ␤ 14
y N␾ Ž x . y x 2 X 1 y ␤ 21 Ž 1y x 2 . X 2 y x 2 X 3 y ␤ 23 y x 2 X 4 y ␤ 24 dx
s .
c y x 3 X 1 y ␤ 31 y x 3 X 2 y ␤ 32 Ž 1y x 3 . X 3 y x 3 X 4 y ␤ 34 dz
y x 4 X 1 y ␤41 y x 4 X 2 y ␤42 y x 4 X 3 y ␤43 Ž 1y x 4 . X4

AIChE Journal April 2003 Vol. 49, No. 4 819

The nondiagonal terms have a connection with the Cramer and we assume that there is equimolar diffusion. While the
determinant when multiplied by the symmetric matrix method is applicable to n species, our illustration will be for
four. The equations to be treated are Eq. 1

1 1 1
1 dx
D 12 D 13 D 14 c s Bx , 0 - z - l,
dz
1 1 1
1
D 12 D 23 D 24 x Ž0. s x 0 , x Ž l . s xl .
1 1 1
1
D 13 D 23 D 34 B is given in Eq. 1. We cast these equations in dimensionless
1 1 1 form using the standard notation; then, we have
1
D 14 D 24 D 34
dx i Z j x i y Zi X j
sÝ ; i/ j,
d␰ j ␤i j
However, taking the sum of the rows in the product does not
have the desired result which occurred with the ternary sys- Di j Nj l
tems because the unity terms on the diagonal do not disap- ␤i j s ; ␤i j s ␤ ji ; Zj s
D cD
pear and so a relation analogous to Eq. 6 does not result. In
addition, the above process, which looks so promising, intro-
duces a substantial number of other terms, and, while there where D is normally one of the Di j . We write these equa-
is some redundancy which could be removed, the overall re- tions in the form Žwhere A is defined below.
sult is not much simpler and does not reduce to a simple
linear operator as in the ternary system.
dx
s Ax , x Ž0. s x 0 , Ž 20.
d␰
Numerical Diffusion Problem
Solutions of the 1-D diffusion problem have been pre-
sented for the Stefan-Maxwell equations, the best known, as the formal solution of which is
mentioned earlier, being that developed by Krishna and
Standard Ž1976., which involve the solution of the set of x sexp Ž A ␰ . x 0 Ž 21.
equations treated as ordinary differential equations in the mol
fractions x i for fixed values of the fluxes. To determine the
fluxes, this set of equations then must be solved by some nu- In the matrix A there are n unknown values of Zi which are
merical procedure. This has apparently produced in some constant and Ý Zi s 0. Our problem is to determine the set
cases instability and some nonconvergence, and has been i
 Zi4 from the equation
studied by Taylor ŽTaylor and Webb, 1981; Taylor, 1982. who
has, in addition, developed some interesting and useful re-
sults in a series of papers referenced in his book. x l sexp Ž A . x 0 . Ž 22.
The purpose of this section is then to present a numerical
solution method for the standard 1-D multicomponent diffu-
sion problem with specified boundary values at z s 0 and z In order to solve this equation we will develop a different
s l for all species. The fluxes will be constant and unknown, representation for A, which follows

Z2 Z3 Z4 Z1 Z1 Z1
q q y y y
␤ 12 ␤ 13 ␤ 14 ␤ 12 ␤ 13 ␤ 14
Z2 Z1 Z3 Z4 Z2 Z2
y q q y y
␤ 21 ␤ 21 ␤ 23 ␤ 24 ␤ 23 ␤ 24
As Ž 23.
Z3 Z3 Z1 Z2 Z4 Z3
y y q q y
␤ 31 ␤ 32 ␤ 31 ␤ 32 ␤ 34 ␤ 34
Z4 Z4 Z4 Z1 Z2 Z3
y y y q q
␤41 ␤42 ␤43 ␤41 ␤42 ␤43

which is different from the form used earlier.

820 April 2003 Vol. 49, No. 4 AIChE Journal

 We shall decompose A into a sum of matrices which makes the exponential solution of Eq. 22, we obtain a vector ©
use of the linearity in the  Zi4; so
©s I q Ý Zi A i ␰ x 0 Ž 25.
1 1 1
0 y y y
␤ 12 ␤ 13 ␤ 14 To find the first set of Zi’s we set ©s x l and ␰ s1 so
1
0 0 0 x l y x 0 s Ý Zi A i x 0
␤ 12
As Z1
1 This is a singular set of algebraic equations in the Zi , so we
0 0 0
␤ 13 replace the last equation by Ý Zi s 0. Let us call the solution
1 so obtained  Zi0 4. We then consider the set of differential
0 0 0 equations
␤ 14
dx 1
1 s A0 x1 ; x 1Ž 0 . s x 0 Ž 26.
0 0 0 d␰
␤ 12
1 1 1 This set of equations is also singular so that we will substitute
y 0 y y Ý x i s1 for the last equation, giving the solution
␤ 12 ␤ 23 ␤ 24 i
q Z2
1
0 0 0 x 1 sexp Ž A 0␰ . x 0
␤ 23
1 which we rewrite as
0 0 0
␤ 24
x 0 sexp Ž y A 0␰ . x 1
1
0 0 0
␤ 13 and substitute this into the solution to obtain
1
0 0 0 x sexp Ž A ␰ . exp Ž y A 0␰ . x 1
␤ 23
q Z3
1 1 1 which we treat as before, keeping only the first-order term in
y y 0 y
␤ 13 ␤ 23 ␤ 34 the expansion of the exponential to give
1
0 0 0 ©s I q Ý A i Ž Zi y Zi0 . ␰ x 1
␤ 34
and at ␰ s1, we obtain
1
0 0 0
␤ 14 x l y x 1 s Ý A i Ž Zi y Zi0 . x 1 Ž 27.
1
0 0 0
␤ 24 which is a new set of linear simultaneous algebraic equations
q
1
Z4 Ž 24 . for  Zi4, which we treat as before. Call the solution  Zi14 which
0 0 0 we use in
␤ 34
1 1 1 dx 2
y y y 0 s A1 x 2 ; x 2 Ž0. s x 0
␤ 14 ␤ 24 ␤ 34 d␰

which has the solution

Such a decomposition exists for any n. We write the above as
x 2 sexp Ž A1␰ . x 0
As A 1 Z1 q A 2 Z2 q A 3 Z3 q A 4 Z4
and as before
Each of the A i is singular Žand each has very simple eigenval-
ues.. We consider now Eq. 20, 21, and 22 and the general x 0 sexp Ž y A1␰ . x 2
procedure will involve a sequence of paired steps. Approxi-
mate values for the fluxes Zi will be found by solving a set of Thus
linear simultaneous algebraic equations obtained by keeping
only the first term of the exponential solution. This step will x sexp Ž A ␰ . x 0 sexp Ž A ␰ . exp Ž y A1␰ . x 2 ,
be followed by solving the differential equation Eq. 21 with
the values of Zi so obtained. If we keep only the first term of ©y x 2 s Ý A i ␰ Ž Zi y Zi1 . x 2

AIChE Journal April 2003 Vol. 49, No. 4 821

and, with ␰ s1, ©s x l , we solve for a new set of  Zi2 4. The sure that the local truncation error is kept within a specified
procedure then repeats itself with equations bound. It consists of using a Runge-Kutta method with local
truncation error of order five to estimate the local error in a
A k s Ý A i Zik , Runge-Kutta method of order four and then adjusting the
step size to keep the local error within a specified bound; in
dx kq1 our examples the tolerance was 10y5. In the examples to fol-
s A k x kq1 ; x kq1 Ž 0 . s x 0 ,
d␰ low it took less than 20 iterations to have a convergent result.
The scheme is very robust.
x l y x kq1 s Ý A i Ž Zikq1 y Zik . x kq1
Numerical Examples
with the hope that Example 1
lim x k Ž 1 . s x l . This is taken directly from Taylor and Krishna Ž1993, p.
k ™⬁ 103.. Components are hydrogen Ž1., nitrogen Ž2., and carbon
dioxide Ž3..
The scheme described above can best be thought of as a Diffusion path length l s85.9 mm
quasi-Newton scheme for the nonlinear equation Temperature s 35.2⬚C

F Ž Z . sexp Ž AZ . x 0 ; F : ⺙R4 ™⺙Rs4 At z s 0, x 10 s 0.00000 At z s l, x 1l s 0.50121

x 20 s 0.50086 x 2 l s 0.49879
Returning to Eq. 27, we see that our wish is to solve F Ž Z⬁. s x 30 s 0.49914 x 3l s 0.00000
x l . If we apply Newton’s method with an initial estimate Z 0 s
0, we find that Pairwise diffusion coefficients
X
F Ž 0 . s w A1 x 0 A 2 x 0 A 3 x 0 A 4 x 0 x D 12 s83.3 m2rs =10y6 ␤ 12 s1
D 13 s 48.0 m rs =10
2 y6
␤ 13 s 0.8163
Forgetting for the moment the problem of invertability, we
see that the next approximation to Z⬁ is D 23 s16.8 m rs =10
2 y6
␤ 23 s 0.20168

Z1 s F Ž 0 .
y1
Ž xl y x0 . Total concentration cs 39.513 molrm3

Zi s 26.087 Ni
yielding x 1 s F Ž Z1 .. At the next step in Newton’s method,
one needs to compute F X Ž Z1 .. Our method approximates Figure 1 shows the composition profiles, and, as stated by
F X Ž Z n . by Taylor and Krishna, the nitrogen flux is substantially higher

w A1 x n A 2 x n A 3 x n A 4 x n x .

How well this latter approximates F X Ž Z n . can be shown to

depend on how well

exp Ž A Ž Z q Z n . . x 0

is approximated by

exp Ž AZ . exp Ž AZ n . x 0

for small Z. Since matrix exponentials do not commute un-

less the respective matrices commute, we would expect that
how well this approximation works depends on the size of
commutators of AZ nq1 and AZ n .
In the implementation of the quasi-Newton method we
have applied the Runge-Kutta-Fehlberg method, described in
Burden and Faires Ž2001., to solve the set of equations

dx kq1
s A k x kq1 , 0 - ␰ -1,
d␰
x kq1 Ž 0 . s x 0

for each k, k s1, 2, 3, . . . . This method adapts the number

and position of the nodes used in the approximation to en- Figure 1. Composition profiles with abnormal fluxes.

822 April 2003 Vol. 49, No. 4 AIChE Journal

 Table 2. Boundary Conditions in Examples 2a, 2b, and 2c.
Fig. x 10 x 20 x 30 x1 l x2 l x3l
2a 0.55 0.05 0.40 0.01 0.45 0.54
2b 0.55 0.05 0.40 0.10 0.72 0.18
2c 0.55 0.05 0.40 0.01 0.98 0.01

than one would anticipate from the gross gradient. In fact,

the flux of nitrogen approaches that of the flux of carbon
dioxide. The profiles show some curvature and are strictly
monotonic. Here we had to solve Eq. 26 reversely by starting
at ␰ s1.

Example 2
This is also chosen from Example 4.2.4 of Taylor and
Krishna and illustrates the effect on the solution of changes
in boundary compositions with hydrogen Ž1., nitrogen Ž2., and
carbon dichloride difluoride Ž3.. The pairwise diffusion coef-
ficients are

D 12 s 77.0 m2rs=10y6 ␤ 12 s 2.3293

D 13 s 33.1 m2rs=10y6 ␤ 13 s1.00000
D 23 s8.1 m2rs=10y6 ␤ 23 s 0.24482
The different cases considered are given in Table 2.
While Figure 2a is a normal case in the sense that the fluxes
are in the directions expected, they are somewhat different
than expected. Figure 2b is an abnormal case since it appears
that the flux of component three is the reverse of the direc-
tion. Figure 2c is a really abnormal case since it shows that
component three, in spite of having a large superficial gradi-
ent, has almost no flux and substantial profile curvature.
Figure 2b. Abnormal flux in component three.
Example 3
This is an eight component case in which the relative pair-
wise diffusivities vary by about a factor of seven and there are
28 of them. The boundary values are given in Table 3 and the
numerical results are in Figure 3.
This is a relatively normal case since the fluxes are in the
same direction as their superficial values. It may be said,

Figure 2a. Fluxes that are normal, but with surprising

values. Figure 2c. Very abnormal flux in component three.

AIChE Journal April 2003 Vol. 49, No. 4 823

Table 3. Boundary Conditions in the Eight Components in
Example 3 and Figure 3.
␰ , xi x1 x2 x3 x4 x5 x6 x7 x8
␰ s0 0.30 0.35 0.05 0.05 0.10 0.15 0.00 0.00
␰ s1 0.05 0.00 0.00 0.10 0.00 0.25 0.10 0.50

however, that while the superficial flux of x 8 is greater in

absolute value than that of x 1, the actual flux is less. The
profiles show some curvature, but most appear to be rela-
tively straight lines between boundary values, except that of
x 8 . The chemical species used in these calculations were H2
Ž1., O 2 Ž2., N2 Ž3., CO Ž4., CO 2 Ž5., CH4 Ž6., C2 H4 Ž7., and Figure 3. Eight-component case with tendency toward
C2 H6 Ž8.. Many of the pairwise coefficients were calculated straight line profiles.
using the Fuller-Schettler-Giddings ŽFuller et al., 1966.
scheme or were those listed in the same source.
if we assume equimolarity of diffusion. The pairwise diffu-
Example 4 sion coefficients will be DŽ p,q . where D is a function of two
This is a six component case with H2 Ž1., O 2 Ž2., N2 Ž3., variables p and q, pl F qF pg . If cŽ p . dp is the concentra-
CO Ž4., CO 2 Ž5., and CH4 Ž6., so there are 15 pairwise diffu- tion of component p, then
sion coefficients with a spread of values of over 11. There are
no real surprises here with the possible exception that the pg
flux of x 5 is negative, as seen in Figure 4. Otherwise, curva- Hp c Ž p . dps c
ture in minimal. l

Continuous Distribution of Components

and cŽ p . dprcs x Ž p . dp, and c is a constant. The Stefan-
We will now consider a slightly different approach in which Maxwell equations now become
we assume that the number of components is very large and,
in fact, that there is a continuous distribution of components.
In order to do this it is easier and, perhaps, even necessary, ⭸ x Ž p. pg xŽ q. pg NŽ q.
to change the procedure. With a large number of compo- yc s NŽ p. Hp dqy x Ž p . Hp dq.
⭸z l
D Ž p,q . l
D Ž p,q .
nents, say 60, in order to proceed as earlier, it could be nec-
essary to introduce 1,770 pairwise diffusion coefficients, all of
which probably must be estimated by the methods of Fuller
et al. Ž1966.. Let us now consider the problem in which there
is a continuous distribution of components, first, in which no
species is present in finite amount, and, second, in which some
species are present in the finite amount, but the remainder
as a continuous distribution.
To start this effort, we write the Stefan-Maxwell equations
in the form

dx i n xj n Nj
yc s Ni Ý y xi Ý , 0 - z - l, is1, . . ., N.
dz js 1 Di j js 1 Di j
j/ i j/ i

Ž 28.

If we suppose that p is a component designating variable and

x Ž p . dp is the mol fraction of that component in the mixture,
then
pg
Hp x Ž p . dps1
l

where pl is the lowest p and pg is the greatest p of the

continuous variable, pl F pF pg . The total fluxes of these
variables will be N Ž p . dp where
pg
Hp N Ž p . dps 0 Figure 4. Component five with a flat profile and a nega-
l tive flux.

824 April 2003 Vol. 49, No. 4 AIChE Journal

If we make the change of variables by normalizing DŽ p,q . as Now, with a little manipulation, one can discover a different
before, then form for this relation with which it is easier to compute. While
we will not show a number of possibilities, let us consider one
NŽ p. l D Ž p,q . z that seems like the best and easiest. If we write
s ZŽ p. ; ␤ Ž p,q . s ; ␰s ,
D Ž pl , pl . c D Ž pl , pl . l
1 1 1 1
then Eq. 28 becomes ( p
q
q
2
s
1
7r12 7r12
( p q 7r6 1r6
q
q p7r6 1r6

2
⭸ x Ž p. xŽ q. ZŽ q.
3 3 p q 1 1
y s ZŽ p. Hp
pg
dqy x Ž p .
pg
Hp dq, ž 'p q'q / ž p 1r3
q
q 1r3 /
⭸␰ l
␤ Ž p,q . l
␤ Ž p,q .
Ž 29. 1
s 7r12 7r12
w
p q
and, if we assume that we have 1-D steady-state diffusion at
constant pressure, we must specify the mol fraction distribu- and, if we compute w for values of p and q from ps qs 5
tion at z s 0 and z s l, or at ␰ s 0 and ␰ s1. At ␰ s 0 x Ž p . to ps qs 20 in various increments of p and q separately,
s x 0 Ž p . and at ␰ s1, x Ž p . s x l Ž p .. we will find that the average value of w will be 0.3567 with a
Having a continuous distribution of components implies standard deviation of 0.0053 and the value of w for ps q is
that we must have a scheme for determining the pairwise dif- '2 r4 f 0.3536. If in addition we normalize the DŽ p,q . by
fusion coefficients for that infinite distribution. One probably DŽ pl , pl . then Žwith pl s6.
would only be interested in such a problem when the mixture
8.09
has some organized system in its composition. One that comes ␤ pq s Ž 30.
7r12 7r12
to mind immediately, of course, is that of a continuous distri- p q
bution of hydrocarbons. The building blocks of hydrocarbons,
in general, vary from C to CH to CH2 with a few CH3’s In cases where we make calculations for a large number of
added on. Trying to compute the average CH= is a problem. chemical species we will use the above formula for the com-
For saturated hydrocarbons CH= is CH2 ". For unsaturated putation with the pairwise normalized diffusivities. The nor-
hydrocarbons CH= will be CH2 . or CH1 ". If we use the malization factor used above is specific to the numerical solu-
Fuller formula ŽTaylor and Krishna, 1993, p. 68. for pairwise tions which will follow where we have used D Ž6,6..
diffusion constants we have
More Examples
1 1

Di j s c
T 7r4 ( Mi
q
Mj
2
We are now in position to compute with the Stefan-Maxwell
equations for the continuous distribution case. We will use a
lumped constant model for all of these.
P 3 3
ž 'V q'V /i j
Example 5
where Mi and M j are the molecular weights and Vi and Vj This is a problem in which the continuous distribution is
are the partial molecular diffusion volumes. For carbon, the described at the two endpoints by
appropriate partial diffusion volume is 15.9 and, for hydro-
gen, it is 2.31. These values come from Taylor and Krishna °1 Ž py2.Ž py16. ; 2
6 F pF16,
Ž1993, p. 69.. For CH2 , the appropriate partial diffusion vol- x Ž p,0 . s~ 2,365
ume is 20.53. For the average, we would take 20 and for the
0
¢ 0; 16 F pF 20,
molecular weight 13.8 p and 20 p for the approximate Vi .
The Fuller formula for a hydrocarbon mixture then will be ° 1 Ž py12.Ž30y p . ;
~ 12 F pF 20,
x Ž p,l . s 444 Ž 31.
1 1
l
¢ 6 F pF12

D Ž pi , pj . s c
T 1.75
( pi
q
pj
2
1
0;

where we think of p as a continuous variable. We will use

P 3 3 '13.8 Ž20. 2r3 only integers in w6, 20x for p
ž 'p q'p / i j

x i s x Ž pi , ␰ . , is1,2,3,...,15,
and, if we choose T s 500 K, P s 500,000 Pa and with cs
p1 s6, p 2 s 7,..., pi s iq5,..., p15 s 20.
1.013=10y2 , D will be in meters squared per second Žmulti-
ply the above by 10y6 ., The normalization factors above are for fifteen components.
For the diffusivities, we use Eq. 30. The boundary values are
1 1

D Ž pi , pj . s 36.8
( pi
q
pj
2
=10y6 Ž m2rs .
listed in Table 4 and are computed from the definitions in
the problem. The calculations in Figure 5a show what appear
to be almost straight lines. If one calculates the difference
3 3
between these solution curves and the corresponding straight
ž 'p q'p /i j
lines, one obtains Figure 5b which shows that this difference

AIChE Journal April 2003 Vol. 49, No. 4 825

 Table 4. Boundary Conditions for the 15 Components in
Example 5 and Figure 5a.
i 1 2 3 4 5 6 7 8
x i Ž0. .1691 .1712 .1624 .1450 .1218 .0951 .0677 .0419
xiŽ l . 0 0 0 0 0 0 0 .0383
pi 6 7 8 9 10 11 12 13
i 9 10 11 12 13 14 15
x i Ž0. .0203 .0055 0 0 0 0 0
xiŽ l . .0721 .1014 .1261 .1464 .1622 .1734 .1802
pi 14 15 16 17 18 19 20

for curves with positive slopes are concave negative while

those with negative slopes are concave positive, but not by
much. It is interesting to determine what happens if one uses
random diffusivities between 0.25 and 1.25. The profiles are
not discernibly different from those in Figure 5a and the
comparison of the fluxes given in Figure 5c with the dots rep-
resenting the formula diffusivities and the asterisks the ran-
dom diffusivities case. The random diffusivities range is well
outside the physical range of ␤i j .
Example 6
This example is the same as Example 5. However, instead
of using 15 components as integers, we will now use 57, but
with the same distribution from ps6 to ps 20 so now
iy1
pi s6q , is1,2, . . . ,57.
4 2 c dx
We must make it known that the normalizations must now be
different since in Example 5 we had 15 rectangles over which
to sum where now we will have 57. To accommodate this, we
will replace 2,365 in Eq. 31 with 8,866 and 444 by 1,661. Note
that these have the common ratio of 3.75. The profile struc-
ture is given in Figure 6a. In Figure 6b we plot the fluxes vs.
Figure 5b. Local difference ⌬ p between computational
profiles and straight line profiles.
the component number. We do this problem primarily to show
in Figure 6c that what appear to be straight lines are begin-
ning to look more and more like straight lines when the
straight line comparison is made; however, comparison of
Figures 5b and 6c shows that the convergence may be slow.
In Figure 6d is the plot of flux vs. straight line slope and it
appears that k is close to 2r5 so
2
N Ž p . sy D sy cD w x Ž p,l . y x Ž p,0 . x ;
5 l dz 5
so that the coefficient of diffusion coefficient for this case is
2r5D where D was the normalizing diffusivity in ␤i j . We will
show in the Theoretical section that this is a natural conse-
quence of the continuous distribution case. Figure 6d also
indicates that the convergence to straight line profiles may be

Figure 5a. Fifteen-component case with essentially Figure 5c. Fluxes for Example 5: formula diffusivities
straight line profiles. (䢇) and randomly chosen diffusivities (U ).

826 April 2003 Vol. 49, No. 4 AIChE Journal

Figure 6a. Case with 57 components and three different
sets of profiles.
Figure 6c. Local difference between calculated profiles
and linear profiles.
slow. Figure 6a is a little confusing since the profiles for 12 F The difference is less, but changing slowly.
pF16 do not emanate from the corners and in fact there is a
profile among 26 - i- 40 for which it will be almost flat.
angles. We will use the same distribution, but we must mod-
Example 7 ify its application somewhat. At ␰ s 0, we will write

This example is presented to illustrate the case where con-

Ž pi y2.Ž pi y16. 2 ; 6 F pi F16,
tinuous distribution is augmented by three components which
have a finite contribution. With our lumping procedure, this
is no more difficult than the others. This time, for the lump-
y Ž pi ,0 . s
½ 0; 16 F pi F 20,

ing, the continuous distribution will contain 57 species and, Ž pi y12.Ž 30y pi . ; 12 F pi F 20,
with the finite three, we have
y Ž pi ,l . s
½ 0; 6 F pi F12.

iy1 57 57
pi s6q , is1,2, . . . ,57, We set Sum l s Ý y Ž pi ,l .; Sum0 s Ý y Ž pi ,0. and let
4 is 1 is 1
p58 s 3, p59 s 4, p60 s 5.
x Ž pi ,0 . s 0.8 y Ž pi ,0 . rSum0 , is1,2, . . . ,57,
In this case rather than using integers we will again be using
x Ž pi ,l . s 0.8 y Ž pi ,l . rSum l , is1,2, . . . ,57,
jumps of one-quarter unit from ps6 to ps 20, hence, ap-
proximating the continuous distribution by a finer set of rect-

Figure 6d. Dimensionless slope vs. dimensionless flux

Figure 6b. Fluxes for 57-component case. showing tendency toward linearity.

AIChE Journal April 2003 Vol. 49, No. 4 827

and with

x Ž p58 ,0 . s 0.04; x Ž p58 ,l . s 0.12,

x Ž p59 ,0 . s 0.06; x Ž p59 ,l . s 0.02,
x Ž p60 ,0 . s 0.10; x Ž p60 ,l . s 0.06.

Also

8.09
␤i j s ; .25F ␤i j F1.
pi7r12 p7jr12

The results of the calculations are presented in Figure 7a,

and one immediately notices that the profiles are again
straight lines or approximately so. If one makes the same cal-
culations, but this time using diffusivities taken from a ran-
dom number generator in the interval 0.15- ␤i j -1.25, one
obtains what appear to be the same profiles. The fluxes vary
obviously, depending upon the field of random numbers se- Figure 7a. Concentration vs. mol fraction with 57 com-
lected. The dots are the diffusivities chosen from the Fuller ponents in the distributed section and three
formula, while the pluses and asterisks are the two results finite compositions.
from different domains of random numbers. The comparison
of the three is in Figure 7b.
Calculations were also made in order to determine the ef-
fect of the continuous domain on the solution profiles for the which can be written as
discrete domain, if any. For a very small continuous domain,
the solution for the discrete domain is similar in shape to
Figure 7a with slightly curved profiles as before. Changing
⭸x
the diffusion coefficients, such as ␤i j , 58F i- j F60, from
these values determined by the above formula to strongly dif- ⭸ x
2
pg x ⭸x pg ⭸␰
Hp dqy Hp dq
ferent coefficients Žsuch as, the ones used in Example 2. pro- pg Z ⭸␰ 2 l
␤ ⭸␰ l
␤
Hp dqs
duces a strong effect on the solution profiles for the discrete
l
␤ ⭸x
domain with little or no effect on the solution profiles for the ⭸x pg x pg ⭸␰
continuous domain. Hp dqy x Hp dq
⭸␰ l
␤ l
␤

Theoretical Justification
We have seen in the numerical solutions for the continu-
ous distribution cases that the concentration profiles are
straight lines or at least very close to straight lines. This seems
like something that should be evident from the equations.
Suppose we write Eq. 29 in the form

⭸ x Ž p, ␰ . pg ZŽ q.
q x Ž p, ␰ .
ZŽ p. s y
ž ⭸␰
Hp
l
␤ Ž p,q .
dq
/
x Ž q, ␰ .
y1
pg

žHpl ␤ Ž p,q .
dq
/ .

The lefthand side is not dependent upon ␰ , so the derivative

of the righthand side with respect to ␰ must be zero giving

⭸x
⭸ 2x pg x ⭸x pg ⭸␰ ⭸x pg x pg Z
0 sy Hp dqq H dqq Hp dq Hp dq
⭸␰ 2
l
␤ ⭸␰ p l
␤ ⭸␰ l
␤ l
␤
⭸x Figure 7b. Dimensionless flux vs. component number
with formula based diffusivities and vs. two
pg ⭸␰ pg Z
yx Hp dq Hp dq sets of randomly chosen ones with total of
l
␤ l
␤ 60 components.

828 April 2003 Vol. 49, No. 4 AIChE Journal

The left hand side is independent of ␰ so that differentiation where D is the normalizing diffusivity; for this to be valid, k
with respect to ␰ , and, after some manipulation, using must be a constant since otherwise the integral sum of the
fluxes will not be zero.
⭸x We see now that the continuous distribution case for the
pg x ⭸A pg ⭸␰ one-dimension diffusion will be
As Hp dq; s Hp dq,
␤ ⭸␰ ␤
l l
⭸ x Ž p, ␰ . pg x Ž q, ␰ . pg ZŽ q.
y sZ Hp dqy x Ž p, ␰ . Hp dq
we obtain ⭸␰ l
␤ Ž p,q . l
␤ Ž p,q .

⭸x ⭸A ⭸A ⭸ 2x ⭸ 3x ⭸ 2x ⭸A ⭸ x ⭸ 2A which holds for 0 - ␰ -1; and x Ž p,0. s f 0 Ž p . and x Ž p,1. s

ž A
⭸␰
yx
⭸␰ / ⭸␰ ⭸␰ 2
qA
⭸␰ 3
y
⭸␰ 2 ⭸␰
y
⭸␰ ⭸␰ 2
f 1Ž p . where f 0 and f 1 are prescribed functions of p where
each is always non-negative and

⭸ 2x ⭸ 2A ⭸ 2x ⭸ x ⭸A pg pg
y A yx A y s0 f 0 Ž q . dqs1, and f 1Ž q . dqs1
⭸␰ 2
⭸␰ 2
⭸␰ 2
⭸␰ ⭸␰ Hp Hp
l l

which will certainly be equal to zero if x is a linear function where f i Ž q . dq stands for the mol fraction of component q.
of ␰ . So, we choose as that linear function We know also that

x Ž p, ␰ . s w x Ž p,1 . y x Ž p,0 . x ␰ q x Ž p,0 . . pg

Hp Z Ž q . dqs 0
l
Since now the solution profiles are established, one must
check on the fluxes. Substitution of the linear solution in the
which follows from the basic assumption of component bal-
Stefan-Maxwell equations gives
ancing. We also know that ␤ Ž p,q . ) 0. The sort of problem
in which we are interested is in which the number of
y w x Ž p,1 . y x Ž p,0 . x molecules per unit of volume is always fixed and constant,
w x Ž q,1 . y x Ž q,0 . x ␰ q x Ž q,0 . and that the reservoirs at ␰ s 0 and ␰ s1 are of fixed com-
Pg
s ZŽ p. Hp dq position.
l
␤ Because of the fact that the profiles of concentration are
straight lines, the problem may be considered to fall into two
pg ZŽ q.
y w Ž x Ž p,1 . y x Ž p,0 . . ␰ q x Ž p,0 . x Hp dq. cases. If f 0 and f 1 have no components in common, then all
1
␤ of the profiles pass through either ␰ s 0 or ␰ s1 and have
their termini on ␰ s1 or ␰ s 0, respectively. On the other
Again, the lefthand side is independent of ␰ , so differentia- hand, if f 0 and f 1 have components in common, the corre-
tion gives sponding profile will run from the mol fractions on ␰ s 0 to
the mol fraction on ␰ s1. The profile will have slope zero if
pg w x Ž q,1 . y x Ž q,0 . x the corresponding mol fractions are also equal as it should
ZŽ p. Hp dq be.
l
␤

pg ZŽ q. Conclusions
y w x Ž p,1 . y x Ž p,0 . x Hp dqs 0
1
␤ Some of the following problems have been faced in this
article successfully and others somewhat less so:
Thus Ž1. The ternary case for the Stefan-Maxwell equations has
resulted in a successful inversion and has produced a simpler
Z Ž p . sy k w x Ž p,1 . y x Ž p,0 . x presentation than might have been expected because of Eq.
6.
Ž2. The quaternary case resulted in the inversion in a
and because the profiles are straight lines, we must have
quasi-symmetric form and a simpler presentation was not very
simple.
dx Ž p, ␰ . Ž3. A numerical scheme was presented which seems to be
Z Ž p . sy k , k ) 0.
d␰ very robust for an arbitrary number of components with rea-
sonable convergence speed and was used on all examples
Now, since Z and the derivative are both dimensionless, k presented.
must be dimensionless and so, reverting to dimensions again Ž4. The case with a continuous distribution of components
was solved, and the character of the solution profiles satisfac-
kcD dx dx torily explained. With several components, the solution pro-
Nsy sy kcD , files are very nearly straight lines. The lumping scheme will
l d␰ dz always give slightly curved solutions.

AIChE Journal April 2003 Vol. 49, No. 4 829

Acknowledgment
The work leading to this article was not supported by any alpha-
betical agencies, either public or private, except the University of
Houston whom we applaud. We thank Jiwen He for valuable consul-
tations.
Literature Cited
Burden, R. L., and J. D. Faires, Numerical Analysis, BrooksrCole,
Pacific Drove, CA Ž2001..
Fuller, E. N., P. D. Schettler, and J. C. Giddings, ‘‘A New Method
for Prediction of Binary Gas-Phase Diffusion Coefficients,’’ Ind.
Eng. Chem., 58, 19 Ž1966..
Hsu, H. W., and R. B. Bird, ‘‘Multicomponent Diffusion Problems,’’
AIChE J., 6, 516 Ž1960..
Krishna, R., and G. L. Standard, ‘‘A Multicomponent Film Model
Incorporating an Exact Matrix Method for Solution to the
Maxwell-Stefan Equations,’’ AIChE J., 22, 383 Ž1976..
830 April 2003 Vol. 49, No. 4
Stewart, W. E., and R. Prober, ‘‘Matrix Calculation of Multicompo-
nent Mass Transfer in Isothermal Systems,’’ Ind. Eng. Chem. Fun-
dam., 3, 224 Ž1964..
Taylor, R., and R. Krishna, Multicomponent Mass Transfer, Wiley,
New York Ž1993..
Taylor, R., ‘‘More on Exact Solutions of the Maxwell-Stefan Equa-
tions for the Multicomponent Film Model,’’ Chem. Eng. Commun.,
14, 361 Ž1982.
Taylor, R., and D. R. Webb, ‘‘Film Models for Multicomponent Mass
Transfer Computational Models,’’ Comput. Chem. Eng., 5, 61
Ž1981..
Toor, H. L.,‘‘Diffusion in Three Component Gas Mixtures,’’ AIChE
J., 3, 198 Ž1957..
Toor, H. L., ‘‘Solution of the Linearized Equations of Multicompo-
nent Mass Transfer,’’ AIChE J., 10, 448 Ž1964..
Manuscript recei®ed July 12, 2002, and re®ision recei®ed Oct. 10, 2002.
AIChE Journal