V8n3a1 PDF

Revista Mexicana de Ingeniería Química
Vol. 8, No. 3 (2009) 213-243
DERIVATION AND APPLICATION OF THE STEFAN-MAXWELL EQUATIONS
DESARROLLO Y APLICACIÓN DE LAS ECUACIONES DE STEFAN-MAXWELL
Stephen Whitaker*
Department of Chemical Engineering & Materials Science

University of California at Davis
Received 5 of June 2009; Accepted 9 of November 2009
Abstract
The Stefan-Maxwell equations represent a special form of the species momentum equations that are used to
determine species velocities. These species velocities appear in the species continuity equations that are used to
predict species concentrations. These concentrations are required, in conjunction with concepts from
thermodynamics and chemical kinetics, to calculate rates of adsorption/desorption, rates of interfacial mass
transfer, and rates of chemical reaction. These processes are central issues in the discipline of chemical engineering.
In this paper we first outline a derivation of the species momentum equations and indicate how they simplify
to the Stefan-Maxwell equations. We then examine three important forms of the species continuity equation in
terms of three different diffusive fluxes that are obtained from the Stefan-Maxwell equations. Next we examine the
structure of the species continuity equations for binary systems and then we examine some special forms associated
with N-component systems. Finally the general N-component system is analyzed using the mixed-mode diffusive
flux and matrix methods.
Keywords: continuum mechanics, kinetic theory, multicomponent diffusion.
Resumen
Las ecuaciones de Stefan-Maxwell representan una forma especial de las ecuaciones de cantidad de movimiento de
especies que son usadas para determinar las velocidades de especies. Estas velocidades de especies aparecen en las
ecuaciones de continuidad de especies que son usadas para predecir las concentraciones de especies. Estas
concentraciones son requeridas, en conjunción con los conceptos de termodinámica y cinética química, para calcular las
velocidades de adsorción/desorción, las velocidades de transferencia de masa interfacial, y las velocidades de reacción
química. Estos procesos son elementos centrales en la disciplina de la ingeniería química.
En este artículo presentamos primeramente un desarrollo de las ecuaciones de cantidad de movimiento de
especies e indicamos como se simplifican a las ecuaciones de Stefan-Maxwell. Posteriormente examinamos tres formas
importantes de la ecuación de continuidad de especies en términos de tres diferentes fluxes difusivos que se obtienen de
las ecuaciones de Stefan-Maxwell. Más adelante examinamos la estructura de las ecuaciones de continuidad de especies
para sistema binarios y examinamos algunas formas especiales asociados con sistemas de N-componentes. Finalmente se
analiza el sistema general de N-componentes usando métodos matriciales y de flux difusivo de modo mixto.
Palabras clave: mecánica del continuo, teoría cinética, difusión multicomponente.
* Corresponding author. E-mail: whitaker@mcn.org 213
Publicado por la Academia Mexicana de Investigación y Docencia en Ingeniería Química A.C.

S. Whitaker / Revista Mexicana de Ingeniería Química Vol. 8, No. 3 (2009) 213-243
Contents
1. Introduction 2
1.1 Conservation of mass 3
1.2 Laws of mechanics 5
2. Mass continuity equation 15
3. Molar continuity equation 16
4. Mixed-mode continuity equation 19
5. Binary systems 20
5.1 Mass diffusive flux 20
5.2 Molar diffusive flux 23
5.3 Mixed-mode diffusive flux 25
6. Special forms for N-component systems 26
6.1 Dilute solution diffusion equation 26
6.2 Dilute solution convective-diffusion equation using J*A 28
6.3 Dilute solution convective-diffusion equation using J A 30
6.4 Diffusion through stagnant species 31
7. General form for N-component systems: Constant total molar concentration 33
8. General form for N-component systems: Constant total mass density 37
9. Conclusions 41
Nomenclature 41
Acknowledgment 41
References 44
Appendix A: Chemical reaction and linear momentum 46
Appendix B: Thermodynamic pressure 50
Appendix C: Algebraic relations 55
Appendix D: Assumptions, Restrictions and Constraints 58
Appendix E: Restrictions for constant total concentration and density 62
214 www.amidiq.org
1. Introduction
Our derivation of multi-component transport

equations is based on the concept of a species body.
In Part I of Fig. 1 we have illustrated a system
containing both species A and species B and these
are illustrated as discrete particles. We have also
illustrated a region from which we have cut out both
a species A body and a species B body. In Part II of
Fig. 1 we have indicated that the species A body will
be treated as a continuum while the discrete
character of species B is retained for contrast. As
time evolves the two species bodies separate because
their velocities are different. This separation is
illustrated in Part III of Fig. 1 where we have also
indicated that the species B body will be treated as a
continuum. The continuum velocities of species A
and species B are designated as vA and vB. In
general, the continuum hypothesis should be
satisfactory when the distance between molecules is
very small compared to a characteristic length for
the system.
1.1 Conservation of mass
In terms of the species A body illustrated in Fig. 1, Fig.1. Motion of species A and species B bodies
we state the two axioms for the mass of multi-
component systems as In order to extract a governing differential equation
from Eq. 1, we make use of the general transport
Axiom I:
equation (Whitaker, 1981, Sec. 3.4, with w = v A )
d
dt V A∫( t )
ρ A dV = ∫ rA dV , A = 1, 2, ...., N (1) d ∂ρ A
dt V A∫( t )
VA (t ) ρ A dV = ∫ dV
V A ( t ) ∂t
A= N
(5)
Axiom II: ∑ rA = 0 (2) + ∫
AA ( t )
ρ A v A ⋅ n dA , A = 1, 2, ...., N
A =1
Here ρ A represents the mass density of species A and the divergence theorem (Stein and Barcellos,
and rA represents the net mass rate of production per 1992, Sec. 17.2)
unit volume of species A owing to chemical reaction.
In Eqs. (1) and (2) we have used a mixed-mode ∫ ρ A v A ⋅ n dA = ∫ ∇ ⋅ ( ρ A v A ) dV ,
nomenclature making use of both letters and AA ( t ) VA (t ) (6)
numbers to identify individual species. For example, A = 1, 2, ...., N
Axiom II could be expressed in terms of alphabetic
subscripts as in order to express Eq. 1 in the form
⎡ ∂ρ A ⎤
Axiom II: rA + rB + rC + rD + ..... + rN = 0 (3) ∫ ⎢ ∂t + ∇ ⋅ ( ρ A v A ) − rA ⎥ dV = 0 ,
VA (t ) ⎣ ⎦ (7)
or we could use numerical subscripts to represent A = 1, 2, ...., N
this axiom as
Since V A (t ) illustrated in Fig. 1 is arbitrary, and
Axiom II: r1 + r2 + r3 + r4 + ..... + rN = 0 (4)
since it is plausible to assume that the integrand in
This latter result can obviously be compacted to Eq. (7) is continuous, the integrand in Eq. (7) must
produce Eq. 2; however, the use of alphabetic be zero and the governing differential equation
subscripts to represent molecular species is prevalent associated with Eq. 1 is given by
in the chemical engineering literature. Because of ∂ ρA
this we will use alphabetic subscripts to identify + ∇ ⋅ ( ρ A v A ) = rA , A = 1, 2, ...., N (8)
distinct molecular species, and we will use the ∂t
nomenclature contained in Eq. 2 to represent the
various sums that appear in this paper.
www.amidiq.org 215
If we sum Eq. (8) over all species and impose the d

dt VA∫( t )
Axiom II we obtain ρ A v A dV = ∫ ρ A b A dV
VA (t )
∂ρ B= N
∂t
+ ∇ ⋅ (ρ v )=0 (9) Axiom I: + ∫
AA ( t )
t A( n ) dA + ∫ ∑P
V A ( t ) B =1
AB dV (15)
in which the total density and the total mass flux are + ∫
VA (t )
rA v∗A dV , A = 1, 2 , ..., N
determined by
A= N A= N With an appropriate interpretation of the
ρ= ∑ρ
A =1
A , ρv= ∑ρ
A =1
A vA (10) nomenclature, one finds that this result is identical to
the second of Eqs. 5.10 of Truesdell (1969, page 85)
The mass average velocity, v, can be expressed in provided that one interprets Truesdell’s growth of
linear momentum as the last two terms in Eq. (15). In
terms of the mass fraction, ωA, and the species
terms of the forces acting on species A, we note that
velocity, vA, according to
ρ A b A represents the body force, t A(n ) represents the
A= N
surface force, and PAB represents the diffusive force
v= ∑ω A v A , ωA = ρ A ρ , A = 1, 2,..., N (11)
A =1 exerted by species B on species A. This diffusive
force is constrained by
The typical treatment of Eq. (8) involves the solution
of N − 1 species continuity equations along with a PAA = 0 , A = 1, 2,3,..., N (16)
solution of Eq. (9). This suggests a decomposition of
the species velocity into the mass average velocity, The last term in Eq. (15) represents the increase or
v, and the mass diffusion velocity, uA decrease of species A momentum resulting from the
increase or decrease of species A caused by chemical
vA = v + uA , A = 1, 2, ..., N (12) reaction, and this term is discussed in Appendix A.
The angular momentum principle for the
so that the species continuity equations take the form species A body is given by
d
dt V A∫( t )
∂ρA r × ρ A v A dV = ∫ r × ρ Ab A dV
+ ∇ ⋅ ( ρ A v) = − ∇ ⋅ ( ρ Au A ) + rA ,
∂t

N VA (t )
N convective chemical
diffusive (13) B=N
∫ ∑ r×P
reaction
∫
accumulation transport transport
Axiom II: + r × t A( n ) dA + dV (17)
A = 1, 2, ...., N − 1 AA ( t ) V A ( t ) B =1
AB
Here we note that only N − 1 of the diffusive + ∫

VA (t )
r × rA v∗A dV , A = 1, 2, ..., N
transport terms are independent since Eqs. (10) and
(12) require the constraint
in which r represents the position vector relative to a
A= N fixed point in an inertial frame. Truesdell (1969,
∑ρ
A =1
A uA = 0 (14) page 84) presents a more general version of Axiom II
in which a growth of rotational momentum is
In order to solve Eqs. (9) and (13) we need included, and Aris (1962, Sec. 5.13) considers an
governing differential equations for the mass analogous effect for polar fluids. The analysis of Eq.
diffusion velocity, uA, and the mass average velocity, (17) is rather long; however, the final result is simply
v. These are determined by the axioms for the the symmetry of the species stress tensor as indicated
mechanics of multi-component systems. by
TA = TA , A = 1, 2, ..., N
T
(18)
1.2 Laws of mechanics
The constraint on PAB is given by Truesdell (1962,
Our approach to the laws of mechanics for multi- Eq. 22) as
component systems follows the work of Euler and
A= N B = N
Cauchy (Truesdell, 1968), the seminal works of
Chapman & Cowling (1939) and Hirschfelder,
Axiom III: ∑ ∑P
A =1 B =1
AB =0 (19)
Curtiss & Bird (1954), along with the recent work of
Curtiss & Bird (1996, 1999). In terms of the species and a little thought will indicate that this is satisfied
body illustrated in Fig. 1 the linear momentum by
principle for species A is given by
PAB = − PBA (20)
216 www.amidiq.org
One can think of this as the continuum version of This result is identical to Eq. 4.20 of Bearman and
Newton’s third law of action and reaction Kirkwood (1958) for the case in which rA = 0
(Whitaker, 2009a). provided that one takes into account the different
Hirschfelder et al. (1954, page 497) point out nomenclature indicated by (with the subscript
that “even in a collision which produces a chemical α = A)
reaction, mass, momentum and energy are Bearman and Kirkwood:
conserved” and the continuum version of this idea
for linear momentum gives rise to the constraint: ρ Ab A = c A X A , ∇ ⋅ ( TA − ρ Au Au A ) = − ∇ ⋅ σ A ,
A= N B= N (27)
Axiom IV: ∑ rA v∗A = 0 (21) ∑P
B =1
AB = c A FA(1)∗
A =1
This result, along with Eq. (19), is contained in the Bearman and Kirkwood refer to σ A as the partial
second of Eqs. 5.12 of Truesdell (1969). stress tensor and note that it consists of a “molecular
force contribution” represented by − TA and a
Returning to the linear momentum principle,
we note that the analysis associated with Cauchy’s “kinetic contribution” represented by ρ Au Au A .
fundamental theorem (Truesdell, 1968) can be
applied to Eq. (15) in order to express the species Equation (23) can be represented in more
stress vector in terms of the species stress tensor compact form using the species continuity equation.
according to We begin by multiplying Eq. (8) by the species
velocity to obtain
t A(n ) = n ⋅ TA (22)
⎡ ∂ρ ⎤
v A ⎢ A + ∇ ⋅ ( ρ A v A ) ⎥ = rA v A , A = 1, 2, ......, N (28)
This representation can be used in Eq. (15), along ⎣ ∂ t ⎦
with the divergence theorem and the general
transport theorem, to extract the governing Subtraction of this equation from Eq. (23) leads to
differential equation for the linear momentum of
species A given by ⎛ ∂ vA ⎞
ρA ⎜ + v A ⋅ ∇v A ⎟ = ρ Ab A + ∇ ⋅ TA
⎝ ∂ t ⎠
∂ (29)
( ρAvA ) + ∇ ⋅ ( ρ A v A v A ) = ρ Ab A + ∇ ⋅ TA B= N
∂
t
local

convective

N
body
N
surface
force force
+ ∑ PAB + rA ( v∗A − v A ) ,
B =1
A = 1, 2, ..., N
acceleration
acceleration
B=N
(23)
Bird (1995) has pointed out that Chapman and
+ ∑P AB + rA v∗A
N
, A = 1, 2, ..., N Cowling (1939) first obtained this result1 for dilute

B =1

source of momentum
gases by means of kinetic theory provided that
diffusive owing to reaction
force
rA = 0 . From the continuum point of view, Eq. (29)
Equation (23) is identical to Eq. A2 of Curtiss and is given by Truesdell and Toupin (1960, Eq. 215.2),
Bird (1996) for the case in which rA = 0 provided Truesdell (1962, Eq. 22), and Curtiss and Bird (1996,
Eqs. 7b and A7) all with rA = 0 . The correspondence
that one takes into account the different
nomenclature indicated by with Truesdell (1962) is based on the nomenclature
ρ Ab A = G A , ∇ ⋅ TA = − ∇ ⋅ σ A , ρ Ab A = ρ Af A , ∇ ⋅ TA = div t A ,
Truesdell: B= N (30)
∑P
Curtiss & Bird: B= N (24) = ρ pˆ A
∑P
B =1
AB = FA
B =1
AB
One can make use of the identity In its present form, Eq. (29) represents a governing
equation for the species velocity, vA, and we want to
ρ A v A v A = ρ A ( v A v + vv A − vv ) + ρ Au Au A (25) use this result to derive a governing equation for the
mass diffusion velocity, uA. To carry out this
in order to express Eq. (23) in the from derivation, we need the total momentum equation
that is developed in the following paragraphs.
∂
( ρ A v A ) + ∇ ⋅ ⎣⎡ ρ A ( v A v + v v A − v v )⎦⎤
∂t
= ρ Ab A + ∇ ⋅ ( TA − ρ Au Au A ) (26)
B= N
+ ∑P
B =1
AB + rA v∗A , A = 1, 2, ..., N
1
See species momentum equation following Eq. 6 on page
135.
www.amidiq.org 217
1.2.1 Total momentum equation ⎛ ∂v ⎞ ⎡ A= N ⎤

ρ⎜ + v ⋅∇v ⎟ = ρ b + ∇ ⋅ ⎢ ∑ ( TA − ρ Au Au A ) ⎥ (38)
The traditional analysis of momentum transport in ⎝ ∂t ⎠ ⎣ A=1 ⎦
multi-component systems makes use of the sum of
Eqs. (23) over all N species to obtain the total In order to use this result to predict the mass average
momentum equation that is used to determine the velocity, we need a constitutive equation for the sum
mass average velocity, v. The remaining N − 1 of the species stress tensors. This problem is
independent species momentum equations can then considered in the following paragraphs.
used to determine the individual species velocities, 1.2.2 Governing equation for the mass diffusion
v A , v B , … v N −1 . We begin by taking into account velocity
Axioms III and IV so that the sum of Eq. (23) leads
to Our objective here is to develop the governing
differential equation for the mass diffusion velocity,
∂ A= N A= N uA. We begin by multiplying Eq. (38) by the mass
∂t
∑ρ
A =1
A v A + ∇ ⋅ ∑ ρAvAvA
A =1
fraction ωA
A= N A= N
(31)
= ∑ ρ Ab A + ∇ ⋅ ∑ TA ρA ⎜
⎛ ∂v ⎞
+ v ⋅∇v ⎟ = ρ Ab
A =1 A =1
⎝ ∂ t ⎠
(39)
⎡ A= N ⎤
+ ω A∇ ⋅ ⎢ ∑ ( TA − ρ Au Au A ) ⎥
The first and third terms in this result can be
simplified by the definitions ⎣ A =1 ⎦
A= N A= N
∑ρ
A =1
A vA = ρv , ∑ρ
A =1
A b A = ρb (32) and subtracting this result from Eq. (29) to obtain the
desired governing differential equation given by
and Eqs. (10) and (12) can be used to obtain ⎛ ∂u A ⎞

ρA ⎜ + v A ⋅ ∇u A + u A ⋅ ∇v ⎟ = ρ A (b A − b)
A= N A= N ⎝ ∂ t ⎠
∑ρ A v A v A = ρ vv + ∑ρ A v Au A (33)
⎡ A= N ⎤
A =1 A =1
+∇ ⋅ TA − ω A ∇ ⋅ ⎢ ∑ ( TA − ρ Au Au A ) ⎥ (40)
⎣ A=1 ⎦
Application of Eq. (14) allows us to simplify the B=N
convective momentum transport to the form + ∑P AB + rA ( v∗A − v A ) , A = 1, 2,..., N − 1
B =1
A= N A= N
∑ρ
A =1
A v A v A = ρ vv + ∑ρ
A =1
A u Au A (34) Here it is important to note that this result is based
only on the two axioms for mass given by Eqs. 1 and
and substitution of Eqs. (32) and (34) in Eq. (31) 2, and the four axioms for the mechanics of multi-
provides component systems given by Eqs. (15), (17), (19)
and (21). In addition, we have made use of classical
∂ ⎡ A= N ⎤ continuum mechanics to obtain the result given by
( ρ v ) + ∇ ⋅ ( ρ vv ) = ρ b + ∇ ⋅ ⎢ ∑ ( TA − ρ Au Au A )⎥ (35) Eq. (22).
∂t ⎣ A=1 ⎦
At this point we need to be specific about the
Concerning the last term in this result, we note that
species stress tensor, TA , and to guide our thinking
Truesdell and Toupin (1960, Sec. 215) refer to
ρ Au Au A as the “apparent stresses arising from and constrain the subsequent development, we
propose that:
diffusion” and we note that this term also appears in
the analysis of Curtiss and Bird (1996, Eq. A7). In The analysis is restricted to mixtures
that case one needs to make use of the second of Eqs. that behave as Newtonian fluids
(24) along with (Serrin, 1959, Sec. 59; Aris, 1962,
A= N A= N
Sec. 5.21).
π= ∑π
A =1
A = − ∑ ( TA − ρ Au Au A )
A =1
(36) Given this restriction for the mixture, we
follow Slattery (1999, Sec. 5.3) and write
to complete the correspondence. At this point we can A= N
use Eq. (9) to obtain T= ∑T
A =1
A = − pI + τ (41a)
⎡∂ρ ⎤
v⎢ + ∇ ⋅ (ρ v) ⎥ =0 (37) in which p is the thermodynamic pressure and τ is
⎣ ∂t ⎦
the extra stress tensor given by (Serrin, 1959, Eq.
and this allows us to express Eq. (35) in the form 61.1; Slattery, 1999, Eq. 5.3.4-3; Bird et al., 2002,
page 843)
218 www.amidiq.org
τ = μ ( ∇v + ∇vT ) + λ (∇ ⋅ v) I (41b) In Appendix A we show that difference between v∗A

and v A should be on the order of the diffusion
Given these results, Eq. (38) provides the Navier- velocity
Stokes equations containing an additional term
associated with the sum of the diffusive stresses. (v ∗
A − v A ) = O(u A ) (47)
Here we need to point out that Eqs. (41) can
be obtained by following a classic continuum Arguments are given elsewhere (Whitaker, 1986,
mechanics analysis; or this result can be obtained 2009b) indicating that several of the terms in Eq.
from kinetic theory (Hirschfelder, Curtiss & Bird, (46) are generally negligible. This leads to the
1954, Eqs. 7.2-45 and 7.6-29). The advantage of this simplifications given by
latter approach is that a method of calculating the
∂u A
coefficients μ and λ is created within the framework ρA << ∇p A (48a)
of the theory. The disadvantage is that the ∂t
calculations associated with the determination of μ
for a dense gas or a liquid may be much more ρ A ( v A ⋅ ∇u A + u A ⋅ ∇v ) << ∇p A (48b)
difficult than the associated experiment.
A= N
Given that the behavior of the mixtures under ω A∇ ⋅ ∑ ρ Au Au A << ∇p A (48c)

A =1
consideration is described by Eqs. (41), we propose
that the species stress tensor can be represented by
(ω A∇ ⋅ τ − ∇ ⋅ τ A ) << ∇p A (48d)
Proposal: TA = − p AI + τ A , A = 1, 2,..., N (42)
rA ( v∗A − v A ) << ∇p A (48e)
in which pA is the partial pressure defined by
(Truesdell, 1969, page 97) The first of these indicates that the governing
equation for u A is quasi-steady; the second indicates
pA = ρ 2
A ( ∂ψ A ∂ ρ A )T , ρ , ρB C ,.......
(43)
that diffusive inertial effects are negligible, the third
indicates that the diffusive stresses are negligible, the
Here ψ A is the Helmholtz free energy of species A fourth indicates that viscous effects are negligible,
per unit mass of species A. In general it is more and the final inequality indicates that the effects of
convenient to work with the internal energy and homogeneous chemical reactions are negligible.
define the partial pressure by (Whitaker, 1989,
Chapter 10) When the restrictions given by Eqs. (48) are
imposed, the governing equation for the mass
p A = ρ A2 ( ∂ eA ∂ ρ A ) s , ρ (44) diffusion velocity takes the form
B , ρC ,.......
∇p A − ω A ∇p − ρ A (b A − b)
in which eA is the internal energy of species A per
B=N (49)
unit mass of species A. A detailed discussion of the
partial pressure and the total pressure is given in
= ∑P
B =1
AB , A = 1, 2,..., N − 1
Appendix B. At this point we define the total
pressure and the total viscous stress tensor by Truesdell (1962, Eq. 7) represents the left hand side
A= N A= N
of this result by p d A and cites Hirschfelder, Curtiss
p= ∑p
A =1
A , τ= ∑τ
A =1
A (45) & Bird (1954) as the source. Curtiss & Bird (1999,
Eq. 7.6) represent the left hand side of Eq. (49) by
cRT d A and refer to it as the generalized driving
and we use these definitions along with Eq. (42) in
order to express Eq. (40) as force for diffusion. At this point we make use of the
identity
⎛ ∂u A ⎞
ρA ⎜ + v A ⋅ ∇u A + u A ⋅ ∇ v ⎟ 1 p
⎝ ∂ t ⎠ ∇ ( pA p ) = ∇p A − A2 ∇p (50)
p p
⎡ A= N ⎤
− ω A ∇ ⋅ ⎢ ∑ ρ Au Au A ⎥ = − (ω A ∇ ⋅ τ − ∇ ⋅ τ A ) in order to express Eq. (49) in the form
⎣ A=1 ⎦ (46)
− ∇p A + ω A∇p + ρ A (b A − b) ∇ ( p A p ) + p −1 ⎡⎣( p A p ) − ω A ⎤⎦ ∇p
B= N
+ ∑P + rA ( v∗A − v A ) B=N (51)
B =1
AB − p −1 ρ A (b A − b) = p −1 ∑ PAB , A = 1, 2,..., N − 1
B =1
www.amidiq.org 219
In order to see how this result is related to the work ∇ ( p A p ) + p −1 ⎣⎡( p A p ) − ω A ⎦⎤ ∇p

of Hirschfelder, Curtiss & Bird (1954), we make use
pressure diffusion
of their Eqs. 7.4-48 and 7.3-27 (in terms of the B=N
x A xB
nomenclature used in this work) to obtain − p −1 ρ A (b A − b) =

∑
B =1 DAB
( vB − v A ) (55)
forced diffusion
Hirschfelder et al.
B= N
x A xB ⎛ DBT DT ⎞
∇ x A + p −1 [ x A − ω A ] ∇ p + ∑ ⎜
B =1 DAB ⎝ ρ B
− A ⎟∇ ln T , A = 1, 2,..., N − 1
ρA ⎠
B= N
x A xB
− p −1 ρ A (b A − b) = ∑
B =1 DAB
( vB − v A ) (52) thermal diffusion
Here we have explicitly identified the terms

B= N
x A xB ⎛ DBT DAT ⎞
+ ∑
B =1
⎜
DAB ⎝ ρ B
− ⎟∇ ln T , A = 1, 2,..., N − 1
ρA ⎠
associated with pressure diffusion, forced diffusion,
and thermal diffusion. This form of the species
momentum equation is restricted by the following:
The right hand side of this result is approximate in
that (1) it is based on dilute gas kinetic theory, and I. The basic assumptions associated with
(2) the binary diffusivities, DAB , have been used in continuum mechanics.
place of the coefficient of diffusion, DAB (see II. The constitutive equation given by Eq. (42)
Hirschfelder, Curtiss & Bird, 1954, page 485). The
III. The simplifications indicated by Eqs. (48).
left hand side of Eq. (52) is identical to the left hand
side of Eq. (51) provided that p A is replaced by IV. The form of the terms that appear on the
x A , p , and this is consistent with the idea that Eq. right hand side of Eq. (55).
(52) was developed for ideal gases. In terms of the One should remember that Eq. (55) is the governing
work of Chapman and Cowling (1970), we note that equation for the diffusion velocity, and this becomes
their Eqs. 18.2,6 and 18.3,13 lead to Eq. (52) with more apparent if we replace v B − v A with u B − u A .
the last term in Eq. (52) expressed as
In general, thermal diffusion creates very
B= N
x A xB ⎛ DBT DAT ⎞
kTA∇ ln T = ∑B =1
⎜
DAB ⎝ ρ B
− ⎟ ∇ ln T
ρA ⎠
(53) small fluxes that are difficult to measure (Whitaker
and Pigford, 1958) and in this study we will neglect
this term to obtain
When dealing with ideal gases, one can proceed with
Eq. (52); however, for more general cases that are ∇ ( p A p ) + p −1 ⎡⎣( p A p ) − ω A ⎤⎦ ∇p
consistent with Eq. (42), one should make use of Eq.
pressure diffusion
(51) and this means dealing with the force, PAB . B=N
x A xB
− p −1 ρ A (b A − b) =

∑
B =1 DAB
( v B − v A ), (56)
1.2.3 Non-ideal mixtures forced diffusion
The simplest approach for non-ideal mixtures is to A = 1, 2,..., N − 1

use the form associated with dilute gas kinetic theory
Chapman & Cowling (1970, page 257) discuss the
in order to represent the right hand side of Eq. (51)
impact of pressure diffusion on the distribution of
as
chemical species in the atmosphere, and both Deen
Proposal: (1998, page 452) and Bird et al. (2002, page 772)
⎛ DBT DAT ⎞ provide an example of this effect in terms of a
x A xB x x
p −1PAB = ( vB − v A ) + A B ⎜ − ⎟ ∇ ln T (54) separation process using an ultracentrifuge. The
DAB DAB ⎝ ρB ρA ⎠ process of forced diffusion of electrically charged
particles is analyzed by Chapman & Cowling (1970,
Here the diffusion coefficients are to be determined Chap. 19) among others.
experimentally with the idea that this form for PAB is
an acceptable approximation, and that Eq. (20) Estimates (Whitaker, 2009b, Sec. 5.6) of the
would be utilized as a solution to Axiom III. terms on the left hand side of Eq. (56) indicate that
Truesdell (1962, Sec. 6) refers to this approximation these terms are generally quite small leading to the
as the special case of binary drags. However, relatively simple relation given by
multicomponent diffusion in liquids is more complex B= N
x A xB
than suggested by Eq. (54), and Rutten (1992), 0 = −∇ ( p A p ) + ∑
B =1 DAB
( vB − v A ) ,
(57)
among many others, has documented these
complexities for ternary systems. Putting aside the A = 1, 2,..., N − 1
seminal problem associated with PAB , we make use
of Eq. (54) in Eq. (51) to obtain Here one should remember that the first term in this
result is based on the use of Eq. (42) and that the
220 www.amidiq.org
second term represents a less than robust model for The mass diffusive flux, ρ Au A , is often represented
non-ideal mixtures in which the binary diffusivities, as (Bird et al, 2002, page 537)
DAB , must be determined experimentally.
jA = ρ Au A (63)
1.2.4 Ideal mixtures
so that Eq. (62) takes the form
At this point we are ready to make the final
simplification given by Species Mass:
pA = xA p , ideal mixture (58) ∂ ρA

+ ∇ ⋅ ( ρ A v ) = −∇ ⋅ jA + rA , A = 1, 2, ..., N − 1 (64)
∂t
in order to obtain the classic Stefan-Maxwell
equations that will be examined in the remainder of Here we note that the mass diffusive fluxes are
this paper constrained by
Species Momentum: A= N
B= N
x A xB ( v B − v A )
∑j A =0 (65)
∑
A =1
0 = − ∇x A + , A = 1, 2,..., N − 1 (59)
DAB
B =1
and we need to determine N − 1 of these diffusive
fluxes in order to develop a solution for Eq. (64).
To complete our formulation of the mechanical
problem, we recall Eq. (38) in the form of the In many liquid-phase diffusion processes, the
Navier-Stokes equations governing equation for the total density given by Eq.
(61) is replaced by the assumption
Total Momentum:
Assumption: ρ = constant (66)
⎛∂v ⎞
ρ⎜ + v ⋅∇v ⎟ = ρ b − ∇p + μ∇ 2 v (60)
⎝ ∂ t ⎠ and we need only solve the N − 1 species continuity
equations given by Eqs. (64).
in which the diffusive stresses have obviously been
neglected. The determination of v A , v B , …, v N 3. Molar continuity equation
using Eqs. (59) and (60) is a very complex problem Chemical engineers are primarily interested in
and the chemical engineering literature contains chemical reactions, interfacial mass transfer, and
many simplified treatments of this problem. adsorption/desorption phenomena, thus molar
However, the domain of validity of these simplified concentrations and mole fractions are more useful
treatments is not always clear, and in the following than mass densities and mass fractions. Because of
sections we attempt to clarify the basis for some of this, the molar form of the species continuity
the special forms of the Stefan-Maxwell equations. equation is often preferred. This form is obtained
from Eqs. (8) by the use of the relations
2. Mass continuity equation
We begin this study with the total mass continuity ρ A = c A M A , rA = RA M A , A = 1, 2, ..., N (67)
equation [see Eq. (9)]
This leads to the species molar continuity equation
∂ρ given by
Total Mass: + ∇ ⋅ (ρ v) = 0 (61)
∂t ∂ cA
+ ∇ ⋅ (c A v A ) = RA , A = 1, 2, ..., N (68)
along with N − 1 species mass continuity equations ∂t
[see Eqs. (13)]
while the constraint on the mass rate of reaction
Species Mass: given by Eq. 2 provides
A= N
∂ ρA
+ ∇ ⋅ ( ρ A v) = − ∇ ⋅ ( ρ Au A ) + rA , ∑R A MA = 0 (69)
∂t (62) A =1
A = 1, 2, ..., N − 1 The total molar continuity equation is analogous to

Eq. (61) and it is developed by constructing the sum
These equations can (in principle) be used to of Eqs. (68) over all species to obtain
determine all the species mass densities, ρ A , ρ B ,…,
ρ N in the same way that the momentum equations ∂c B= N
Total Molar: + ∇ ⋅ (c v∗ ) = ∑ RB (70)
represented by Eqs. (59) and (60) can be used to ∂t B =1
determine all the species velocities, v A , v B , …, v N .
www.amidiq.org 221
Here the total molar concentration and the molar B=N B=N B=N
average velocity are defined by ∑ω

B =1
B vB = ∑ω
B =1
B v∗ + ∑ ωB u∗B
B =1
(78)
A= N A= N
c= ∑c
A =1
A , c v∗ = ∑c
A =1
A vA (71) On the basis of the second of Eqs. (10) this takes the
form
The development in Sec. 2 indicates that Eq. (70) B=N
should be solved along with the N − 1 species v = v∗ + ∑ ωB u∗B (79)

B =1
continuity equations given by
and we are now confronted with the mixed-mode
Species Molar:
term ωB u∗B that involves a mass fraction and a molar
∂ cA diffusion velocity. We would like to express ωB u∗B in
+ ∇ ⋅ ( c A v A ) = RA , A = 1, 2, ..., N − 1 (72)
∂t terms of molar diffusive fluxes, and to do so we
manipulate this term as follows
This allows for the determination of all the species
molar concentrations, c A , cB ,…, cN . ρ B u∗B
ωB u∗B =
ρ A + ρ B + ρC + ..... + ρ N
The form of Eqs. (70) through (72) suggests
(but does not require) a decomposition of the species M B cB u∗B
= (80)
velocity given by M A c A + M B cB + ....... + M N cN
v A = v∗ + u*A , A = 1, 2, ..., N M B J ∗B
(73) =
c ( x A M A + xB M B + ......xN M N )
in which u*A is the molar diffusion velocity. A little
thought will indicate that the molar diffusion If we define the mean molecular mass as
velocities are constrained by
M = x A M A + xB M B + ... + xN M N (81)
A= N
∑c
A =1
A u*A = 0 (74) we can express Eq. (80) in compact form according
to
When Eq. (73) is used in Eq. (72) the transport of M B J ∗B
species A can be represented in terms of a convective ωB u∗B = (82)
cM
part, c A v∗ , and a diffusive part, c Au*A , leading to
At this point we return to Eq. (79) to develop the
∂ cA
+ ∇ ⋅ (c A v∗ ) = −∇ ⋅ (c Au*A ) + RA , following relation between the molar average
∂t (75) velocity and the mass average velocity:
A = 1, 2, ...., N − 1
1 B= N M B ∗
v∗ = v − ∑ JB (83)
The molar diffusive flux, c Au∗A , is often identified as c B =1 M
(Bird et al, 2002, page 537) Substitution of this expression for the molar average
velocity into Eq. (77) allows us to express that form
J ∗A = c Au∗A (76)
of the species continuity equation as
so that Eq. (75) takes the form Species Molar:
∂ cA ∂ cA ⎛ B= N
⎞
+ ∇ ⋅ (c A v∗ ) = −∇ ⋅ J ∗A + RA , M
+ ∇ ⋅ (c A v) = −∇ ⋅ ⎜ J ∗A − x A ∑ B J ∗B ⎟ + RA ,
∂t (77) ∂t ⎝ B =1 M ⎠ (84)
A = 1, 2, ...., N − 1 A = 1, 2, ...., N − 1
This result is similar in form to Eq. (64) for the in which the molar diffusive fluxes are constrained
species mass density; however, there is no governing by
equation for the molar average velocity, v∗ , whereas
A= N
∑J
the mass average velocity in Eq. (64) can be *
A =0 (85)
determined by the application of Eq. (60). In order to A =1
eliminate the molar average velocity from Eq. (77)
we return to Eq. (73), multiply by ω A , and sum over Here we can see that this convection-diffusion
all species to obtain problem is inherently nonlinear in terms of the
diffusive flux; however, if the mole fraction of
222 www.amidiq.org
species A is sufficiently small it is possible that the ∂ cA

term involving the sum of the diffusive fluxes in Eq. + ∇ ⋅ (c A v) = −∇ ⋅ (c A u A ) + RA ,
∂t (89)
(84) can be neglected. By “sufficiently small” we
mean that the following inequality A = 1, 2, ...., N − 1
B= N Here the diffusive flux is represented in terms of a

MB ∗
xA ∑ J B << J ∗A (86) molar concentration and a mass diffusion velocity.
B =1 M
This mixed-mode diffusive flux is often referred to
is satisfied and Eq. (84) becomes linear in the molar as a hybrid flux and identified as (Bird et al, 2002,
page 537)
diffusive flux, J ∗A .
J A = cA u A (90)
To complete our formulation of the molar
forms of the species continuity equation, we make Use of this representation in Eq. (89) leads to
use of Eq. (83) in Eq. (70) to obtain
Species Molar:
Total Molar:
∂ cA
∂c B=N
M B=N + ∇ ⋅ (c A v) = −∇ ⋅ J A + RA ,
+ ∇ ⋅ (c v) = ∇ ⋅ ∑ B J ∗B + ∑ RB (87) ∂t (91)
∂t B =1 M B =1
A = 1, 2, ..., N − 1
This total molar transport equation should be
compared with Eq. (61) in order to appreciate the The constraint on this diffusive flux is more complex
complexity associated with the molar form of the than that for either the mass diffusive flux or the
species transport equations. In many gas phase mass molar diffusive flux and is given by
transfer processes, Eq. (87) can be replaced by the A= N
assumption ∑M
A =1
A JA = 0 (92)
Assumption: c = constant (88)
This hybrid diffusive flux, JA, lacks popularity;
and we need only solve the N − 1 species however, the transport equation given by Eq. (91)
continuity equations given by Eqs. (84). has the advantage that it is linear in the diffusive
flux. In terms of the mixed-mode diffusive flux, the
4. Mixed-mode continuity equation total molar continuity equation takes the form
The motivation for a mixed-mode or hybrid species Total Molar:
continuity equation is based on the applications that
are dominant in the area of chemical engineering, ∂c B= N B= N
+ ∇ ⋅ (cv) = −∇ ⋅ ∑ J B + ∑ RB (93)
and on the mechanical problem under consideration. ∂t B =1 B =1
To be explicit, we note two facts:
and we are still confronted with a complex form of
(1) Chemical reactions and interfacial
the total molar transport equation. This complexity
mass transfer are usually represented
often serves to generate the assumption that the total
in terms of the molar concentration,
molar concentration is constant as indicated by
cA, or the mole fraction, xA, thus we
are motivated to use the molar form of Assumption: c = constant (94)
the continuity equation given by Eq.
(68) as opposed to the mass form Often gas phase diffusion problems lead to the use of
given by Eq. (62). a molar form of the species continuity equation
(2) The species continuity equation because Eq. (94) provides a reasonable
involves velocities that must be simplification. On the other hand, liquid phase
determined by the laws of mechanics, diffusion problems suggest the use of the mass form
thus we are motivated to use the mass of the species continuity equation because Eq. (66)
decomposition of the species velocity provides a reasonable simplification. The author is
given by Eq. (12) as opposed to the unaware of any solution to a diffusion problem that
molar decomposition given by Eq. does not make use of either Eq. (66) or Eq. (94), and
(73). removing these assumptions remains as a significant
challenge.
In order to obtain a mixed-mode or hybrid continuity
equation, we begin with the species mass continuity 5. Binary systems
equation given by Eq. (62) and divide by the
molecular mass of species A to obtain Binary systems are often used to introduce the
phenomena of diffusion, and we will follow that
approach in order to explore the nature of the mass,
www.amidiq.org 223
molar, and mixed-mode forms of the species ωA = ρ A ρ , xA = cA c ,

continuity equation. (103)
M = x A M A + xB M B + ..... + xN M N
5.1 Mass diffusive flux
in which MA represents the molecular mass of
For a binary system, Eq. (59) reduces to species A. In addition to these results, we make use
of
x A xB ( v B − v A )
0 = −∇x A + (95)
DAB cA = ρ A M A , c = ρ M , A = 1, 2,...N (104)
and we think of this as the governing differential to obtain the following relations between the mole
equation for vA. The value of vB is available from a fractions and the mass fractions
solution for v and vA which can be used in the second
cA cA M ( ρ A M A ) M M
of Eqs. (10) to obtain xA = = = = ωA ,
c ρ ρ MA (105)
1
vB = ( v − ωA v A ) (96) A = 1, 2,...N
ωB
At this point we direct our attention to binary
For a binary system, the two mass continuity systems and make use of the following relations
equations are given by
∇ x A = − ∇x B , ω B = 1 − ω A ,
∂ ρA
+ ∇ ⋅ ( ρ A v ) = −∇ ⋅ ( ρ Au A ) + rA (97) 1 ω ω (106)
∂t = A + B
M MA MB
∂ρ
+ ∇ ⋅ ( ρ v) = 0 (98) along with several algebraic steps (see Appendix C)
∂t to arrive at
and we need to determine uA and v in order to solve M2

these equations. Given the form of Eq. (97) it will be ∇x A = ∇ω A (107)
M AM B
convenient to express Eq. (95) in terms of mass
diffusion velocities, and the use of Eq. (12) leads to Substitution of this expression for the gradient of the
mole fraction of species A into Eq. (102) leads to
x A xB (u B − u A )
0 = −∇x A + (99)
DAB M2 1 x A xB
0=− ∇ω A − ( ρ Au A ) (108)
M AM B ρ DAB ω AωB
For a binary system, Eq. (14) provides
ω Au A + ωB u B = 0 (100) From Eqs. (105) we see that
x A xB M2
and this can be used in Eq. (99) to obtain = (109)
ω AωB M AM B
1 x A xB
0 = −∇x A − (ω Au A ) (101)
DAB ω AωB and Eq. (108) simplifies to the classic form of Fick’s
Law given by
Multiplying and dividing the second term by the total Fick’s Law: j A = ρ Au A = − ρ DAB ∇ω A (110)
density allows us to express this result as
1 x A xB Returning to Eq. (97), we make use of this form of

0 = −∇x A − (ρ u ) (102) Fick’s Law to obtain the following governing
ρ DAB ω AωB A A equation for the species density, ρ A
Here we have a mixed-mode representation in which
∂ ρA
the mass diffusive flux, ρ Au A , is expressed in terms + ∇ ⋅ ( ρ A v ) = ∇ ⋅ ⎡⎣ ρ DAB ∇ ( ρ A ρ ) ⎤⎦ + rA (111)
∂t
of the gradient of the mole fraction, ∇x A , along with
the mixed-mode term, x A xB ω AωB . For liquid systems this result can often be simplified
on the basis of the assumption
Before attacking the binary result given by
Eq. (102) it is convenient to list some results for N- Assumption: ρ = constant (112)
component systems. We begin with the definitions
for the mass fraction, ω A , the mole fraction, x A , and which leads to
the mean molecular mass, M . These are given by
224 www.amidiq.org
∂ ρA and employ the form of Eq. (76) for both species to

+ ∇ ⋅ ( ρ A v) obtain
∂t
(113)
⎧ ρ = constant x A J *B − xB J *A
= ∇ ⋅ ( DAB ∇ρ A ) + rA , ⎨ 0 = −∇x A + (119)
⎩ binary system c DAB
Here we have an attractive, linear transport equation Application of the binary version of Eq. (85)
for the species density, ρ A .
J *A + J *B = 0 (120)
When confronted with chemical reactions and
interfacial transport, we generally prefer to work allows us to express Eq. (119) in the classic form of
with the molar form of the species continuity Fick’s Law given by
equation. This form can be extracted from Eq. (113)
by the use of Fick’s Law: J ∗A = c A u*A = −c DAB ∇x A (121)
ρ A = c A M A , rA = RA M A (114) This is the molar analogy of Eq. (110), and

substitution of this result into Eq. (116) leads to the
which leads to molar analogy of Eq. (111).
∂ cA ∂c A
+ ∇ ⋅ (c A v ) + ∇ ⋅ (c A v ) = ∇ ⋅ ⎡⎣( M B M ) cDAB ∇ ( c A c ) ⎤⎦ + RA (122)
∂t ∂t
(115)
⎧ ρ = constant
= ∇ ⋅ ( DAB ∇cA ) + RA , ⎨ If we ignore variations in the total molar
⎩ binary system concentration on the basis of the assumption
This is an attractive form to use with liquids where Assumption: c = constant (123)
the assumption of a constant density is likely to be a
valid approximation. When the total density is not we see that Eq. (122) takes the form
constant, one must solve Eq. (111) simultaneously
∂ cA
with Eq. (98). + ∇ ⋅ (c A v ) = ∇ ⋅ ⎡⎣( M B M ) DAB ∇c A ⎤⎦
∂t
(124)
⎧ c = constant
+ RA , ⎨
5.2 Molar diffusive flux ⎩ binary system
Because of the prevalence of molar concentrations
and mole fractions in chemical engineering analysis, in which the presence of M leads to the non-
the species molar continuity equation is generally linearity associated with
preferred. This form can be extracted from Eq. (84) −1
according to M B M = ⎡⎣ xA ( M A M B − 1) + 1⎤⎦ (125)
Species Molar: In order to obtain the so-called dilute solution form

of Eq. (124), we impose
∂ cA ⎛ B=2
M ⎞
+ ∇ ⋅ (c A v ) = −∇ ⋅ ⎜ J ∗A − x A ∑ B J ∗B ⎟ + RA (116)
∂t ⎝ B =1 M ⎠ Restriction: x A ( M A M B − 1) << 1 (126)
while the total molar continuity equation given and Eq. (124) simplifies to the classic convective-
earlier by Eq. (87) takes the form diffusion equation given by
∂c A
+ ∇ ⋅ (c A v ) = ∇ ⋅ ( DAB ∇c A )
Total Molar:
∂t
∂c B=2
M
+ ∇ ⋅ (c v ) = ∇ ⋅ ∑ B J ∗B + ( RA + RB ) (117) ⎧ c = constant (127)
∂t B =1 M
⎪
+ RA , ⎨ x A ( M A M B − 1) << 1
Ignoring for the moment the difficulties associated ⎪
with the total molar continuity equation, we direct ⎩ binary system
our attention to the molar diffusive flux represented Here it is very important to note that this result is
by J*A . We begin by using Eq. (73) to express the identical to Eq. (115). However, Eq. (127) is not
single Stefan-Maxwell equation as based on the constraint that the density is constant.
Instead, Eq. (127) is based on the assumption of
x A xB (u*B − u*A ) “constant total molar concentration” indicated by Eq.
0 = −∇x A + (118)
DAB (123), and the assumption of a “dilute solution”
www.amidiq.org 225
indicated by Eq. (126). For binary systems we have Substitution of this result into Eq. (131) provides the
(see Eqs. (103) and (104)) following governing equation for the species A molar
concentration
ρ = c ( x A M A + xB M B ) (128) ∂ cA
+ ∇ ⋅ (c A v ) = ∇ ⋅ ⎣⎡( M B M ) c DAB ∇ ( c A c ) ⎦⎤ + RA (136)
∂t
which can be arranged in the form
ρ = c { x A ⎣⎡( M A M B ) − 1⎦⎤ + 1 } M B
This result is identical to Eq. (122) indicating that
(129) both the molar representation given by Eq. (116) and
the mixed-mode representation given by Eq. (131)
When the two restrictions associated with Eq. (127) lead to the same result for a binary system.
are imposed, the total density is essentially constant
and Eq. (127) is consistent with Eq. (115). It is of some interest to note that the mixed-
mode diffusive flux can be expressed as
Returning to Eq. (127), we note that the
maximum value of the mole fraction for species A 1
will usually be known a priori, and this allows us to cA u A = ρA uA (137)
MA
express the constraint associated with Eq. (127) as
and on the basis of Eq. (110) this takes the form
Constraint: ( x A ) max ( M A M B − 1) << 1 (130)
1
One should remember that there is a restriction cA u A = − ρ DAB ∇ω A (138)
MA
associated with every assumption and when one
imposes the restriction one always assumes that Use of this result in Eq. (131) yields what appears to
small causes give rise to small effects (Birkhoff, be an unattractive form given by
1960). In addition, one should remember that behind
every restriction there is a constraint (see Appendix ∂c A ⎡ 1 ⎤
D); however, constraints can often be very difficult + ∇ ⋅ (c A v ) = ∇ ⋅ ⎢ ρ DAB ∇ω A ⎥ + RA (139)
to develop. ∂t ⎣MA ⎦
5.3 Mixed-mode diffusive flux However, if we impose the condition
In this case we return to the mixed-mode species Assumption: ρ = constant (140)

continuity equation [see Eq. (89)]
and make use of the first of Eqs. (114) we find
∂c A
+ ∇ ⋅ (c A v ) = −∇ ⋅ (c A u A ) + RA (131) ∂c A
∂t + ∇ ⋅ (c A v ) = ∇ ⋅ ( DAB ∇c A )
∂t
(141)
and direct our attention to the single Stefan-Maxwell ⎧ ρ = constant
equation given by Eq. (101) and repeated here as + RA , ⎨
⎩ binary system
1 x A xB
0 = −∇x A − (ω Au A ) (132) which was given earlier by Eq. (115).
DAB ω AωB
6. Special forms for N-component systems
It is convenient to rearrange this result in the form
Given the complexity of the binary forms described
1 xB in the previous sections, we should expect additional
0 = −∇x A − (c A u A ) (133)
c DAB ωB complexities for N-component systems. This
naturally leads to the search for simplifications, and
in order to obtain the mixed-mode diffusive flux we will examine some of these simplifications in this
given by section.
ωB 6.1 Dilute solution diffusion

cA u A = − c DAB ∇x A (134)
xB There are mass transfer processes in which all the
molar fluxes are the same order of magnitude and the
At this point we can use Eq. (105) to obtain the dominant diffusing species is dilute. In this special
mixed-mode form of Fick’s Law given by case, it is convenient to represent the Stefan-
Maxwell equations in terms of the molar flux defined
Fick’s Law:
by
J A = c A u A = − ( M B M ) c DAB ∇xA (135) N A = cA v A , A = 1, 2,..., N (142)
226 www.amidiq.org
which allows us to express Eqs. (59) as N A = −c DAm ∇x A , A = 1, 2, ..., G < N (149)

At this point we recognize that Eqs. (72) can be
B=N
x A N B − xB N A
0 = −∇x A + ∑
B =1 c DAB
, A = 1, 2, ..., N − 1 (143) expressed in terms of N A to obtain
B≠ A
∂ cA
+ ∇ ⋅ N A = RA , A = 1, 2, ...., N − 1 (150)
At this point we separate the second term to obtain ∂t
B=N B=N
NB xB
0 = − c ∇x A + x A ∑D
B =1
− NA ∑D
B =1
, and that Eq. (149) can be used to obtain a dilute
solution diffusion equation given by
B≠ A
AB
B≠ A
AB (144)
A = 1, 2, ..., N − 1 ∂c A
= ∇ ⋅ ( c DAm ∇x A ) + RA , A = 1, 2, ...., G < N (151)
∂t
and we define the mixture diffusivity by
B=N
We are still confronted with the complexity of the
1 x
= ∑ B , A = 1, 2, ..., N − 1 (145) transport equation for the total molar concentration
DAm B = 1 D AB given by Eq. (87), and this difficulty is classically
B≠ A
avoided by assuming that the total molar
so that the Stefan-Maxwell equations can be concentration is a constant in order to obtain
expressed as ∂c A
= ∇ ⋅ ( DAm ∇c A ) + RA ,
B=N DAm ∂t
0 = −c DAm ∇x A + x A ∑D NB − N A ,
⎧ G<N
(152)
B =1
B≠ A
AB (146) A = 1, 2, ..., G , ⎨
⎩ other conditions
A = 1, 2, ..., N − 1
in which the other conditions associated with this
For some processes, such as diffusion in porous result are given by
media (Whitaker, 1999) in which the flux of all the
species is driven by heterogeneous reaction or by Assumption: c = constant (153a)
adsorption/desorption, we can impose the
simplification Restriction: N B = O ( N A ) , B = 1, 2, ..., N (153b)
B=N DAm Constraint: ( x A ) max << 1 , A = 1, 2, ..., G < N (153c)
xA ∑D N B << N A ,
B =1
B≠ A
AB (147)
The constraint identified by Eq. (153c) is generally
A = 1, 2, ..., G , G < N available in terms of the problem statement, and
when this constraint is satisfied it is probable that the
when the following two conditions are satisfied: assumption given by Eq. (153a) and the restriction
given by Eq. (153b) are also valid.
Constraint: ( x A ) max << 1 , A = 1, 2, ..., G < N (148a)
Restriction: N B = O ( N A ) ,
6.2 Dilute solution convective-diffusion equation
B = 1, 2, ..., N (148b)
using J*A
The first of Eqs. (148) is identified as a constraint In order to develop the convective-diffusion version
since the maximum values of the mole fractions are of Eqs. (152), we begin with the generally valid form
generally known a priori, while the second given by Eqs. (84) and repeated here as
inequality is identified as a restriction since it is not
expressed in terms of quantities that are known a ∂c A ⎛ B= N
M ⎞
priori. Equation 148b should be interpreted to mean + ∇ ⋅ (c A v) = −∇ ⋅ ⎜ J ∗A − x A ∑ B J ∗B ⎟
∂t ⎝ B =1 M ⎠ (154)
that N B is not significantly larger than N A and if
+ RA , A = 1, 2, ...., N
species B is stagnant, N B would be zero.
In Eqs. (147) and (148a) we have indicated The Stefan-Maxwell equations can be expressed as
that our N-component system contains G
B= N
x A J ∗B − xB J ∗A
0 = −∇x A + ∑
components that are dilute. For example, if we may
, A = 1, 2, ..., N − 1 (155)
have a five-component mixture in which three B =1 cDAB
components have mole fractions that are small B≠ A
compared to one, we have G = 3 and Eqs. (147) and

and the summation can be separated leading to
(148a) applies to these three components. Use of Eq.
(147) allows us to express the dilute forms of Eqs.
(146) as
www.amidiq.org 227
B= N
J ∗B B=N ∂c A
+ ∇ ⋅ (c A v ) = ∇ ⋅ ( c DAm∇x A ) + RA ,
x
0 = −c∇x A + xA ∑ − J ∗A ∑ B ,
B =1 DAB B =1 DAB (156) ∂t (163)
B≠ A B≠ A
A = 1, 2, ..., G < N
A = 1, 2, ..., N − 1
In addition to the inequalities given by Eqs. (160)
The definition of the mixture diffusivity given by Eq. and (162), we assume that the total molar
(145) can be used to express this result in the form concentration is constant in order to obtain the
classic linear convective-diffusion equation for
B= N DAm
0 = −c DAm ∇x A + x A ∑ J ∗B − J ∗A , species A.
B =1 DAB (157)
B≠ A ∂c A
+ ∇ ⋅ (c A v) = ∇ ⋅ ( DAm∇c A )
A = 1, 2, ..., N − 1 ∂t
⎧ G<N (164)
In making judgments about this result, we need to ⎪
remember that the diffusive fluxes are constrained by + RA , A = 1, 2, ..., G , ⎨ c = constant
⎪ x << 1
B=N
⎩ A
∑J
B =1
*
B =0 (158)
This special form of the species continuity equation
is ubiquitous in the chemical engineering literature;
indicating that the diffusive fluxes tend to be the however, the simplifications associated with this
same order of magnitude. This means that the result are generally not made clear. In addition to the
following inequality dominant restrictions listed in Eq. (164), one should
keep in mind the restriction given by Eq. (162) that
Restriction: would appear to be automatically satisfied by Eqs.
B= N D
xA ∑ J ∗B << J ∗A , A = 1, 2, ..., G < N (158) and (160) unless there is a serious disparity in
Am
(159)
B =1 DAB the molecular masses.
B≠ A
6.3 Dilute solution convective-diffusion equation

has considerable appeal when the mole fraction of using J A
species A is small compared to one as indicated by
In this case we begin with Eq. (91)
Restriction: x A << 1 , A = 1, 2, ..., G < N (160)
∂c A
Use of the inequality given by Eq. (159) in the + ∇ ⋅ (c A v ) = −∇ ⋅ J A + RA , A = 1, 2, ...., N − 1 (165)
Stefan-Maxwell equations given by ∂t
Eqs. (157) leads to the multi-component form of
and note that the Stefan-Maxwell equations can be
Fick’s Law
expressed as
“Fick’s Law” B=N
x A J B − xB J A
∗ ⎧ G<N
0 = −∇x A + ∑ c DAB
, A = 1, 2, ..., N − 1 (166)
J = −c DAm ∇x A , A = 1, 2, ..., G , ⎨ (161) B =1
B≠ A
⎩ x A << 1
A
Separating the terms in the sum leads to

which is analogous to the result for binary systems
B=N B=N
given by Eq. (121). JB xB
0 = − c ∇x A + x A ∑D
B =1
− JA ∑D
B =1
,
We now turn our attention to the species B≠ A
AB
B≠ A
AB (167)
continuity equation given by Eq. (154). Use of the
dilute solution condition indicated by Eq. (160) and A = 1, 2, ..., N − 1
the constraint on the diffusive fluxes given by Eq.
and use of the definition of the mixture diffusivity
(158) leads to the restriction
given by Eq. (145) provides
B= N
MB ∗
Restriction: xA ∑ J B << J ∗A (162) B=N DAm
B =1 M 0 = −c DAm ∇x A + x A ∑D JB − J A ,
B =1
B≠ A
AB (168)
Use of this inequality along with the multi-
component form of Fick’s Law given by A = 1, 2, ..., N − 1
Eq. (161) in Eq. (154) leads to the following form of
In making judgments about this result we need to
the convective-diffusion equation
remember that the diffusive fluxes are constrained by
228 www.amidiq.org
A= N Use of the definition of the mixture diffusivity given

∑M
A =1
A JA = 0 (169) by Eq. (145) immediately leads to
N A = −c DAm ∇x A (176)
thus if the mole fraction of species A is small
compared to one, we can make use of the restriction
Note that this result is not restricted to a dilute
given by
solution; however, we have imposed the condition on
Restriction: the velocities given by
B=N D
xA ∑ J B << J A , A = 1, 2, ..., G < N v B = v C = ... = v N = 0
Am
(170) Assumption: (177)
B =1 D AB
B≠ A
This assumption could be replaced with the
restriction
Under these circumstances, the Stefan-Maxwell
equation for species A takes the form Restriction: v B << v A , v C << v A , ..., v N << v A (178)
“Fick’s Law”:
in which the use of the absolute values of the
J A = −cDAm ∇x A , A = 1, 2, ..., G < N (171) velocities is understood. Here one should remember
that we are repeatedly relying on Birkhoff’s (1960)
Use of this result in Eq. (165) leads to the following plausible intuitive hypothesis that small causes give
form of the convective-diffusion equation rise to small effects. Use of Eq. (176) in Eq. (150)
leads to
∂c A
+ ∇ ⋅ (c A v) = ∇ ⋅ ( cDAm∇x A ) + RA (172) ∂c A
∂t = ∇ ⋅ ( cDAm ∇x A ) + RA (179)
∂t
This result, based on the single restriction given by
Eq. (170), is identical to that given earlier by Eq. and we can assume that the total molar concentration
(163). To complete the analysis of the mixed-mode is constant to obtain
diffusive flux, we assume that the total molar
concentration is constant so that Eq. (172) takes the ∂c A ⎧ c = constant
= ∇ ⋅ ( DAm ∇c A ) + RA , ⎨ (180)
form ∂t ⎩ other conditions
∂c A
+ ∇ ⋅ (c A v) = ∇ ⋅ ( DAm∇c A ) + RA , where the other conditions are those indicated by
∂t Eqs. (178). This result is identical to Eq. (152)
⎧ G<N (173) except for the fact that there is only a single
⎪ component that could satisfy this equation. As a
A = 1, 2, ..., G , ⎨ c = constant
⎪ x << 1 reminder of the difference between Eq. (180) and Eq.
⎩ A (152) we summarize the conditions upon which it is
based
Certainly the route to Eq. (173) is simpler than that
followed in the development of Eq. (164); however, Restriction: v B << v A , v C << v A , ..., v N << v A (181)
the preferred approach might still be considered to be
a matter of choice. Restriction: x A∇c << ∇c A (182)
6.4 Diffusion through stagnant species Comparing these two restrictions with Eqs. (153)
indicates that Eqs. (152) and (180) describe rather
The case of binary transport of species A through a
different physical phenomena even though the two
stagnant species B has been treated in terms of the
equations are identical. In reality, it seems unlikely
classic Stefan diffusion tube (Whitaker, 2009b, Sec.
that a process restricted by Eq. (181) could involve
2.7). Moving beyond the binary system, we consider
significant homogeneous reaction, thus a more
the case in which species A is diffusing and all other
realistic version of Eq. (180) would require that we
species are stagnant. Under these circumstances, the
set RA equal to zero. Nevertheless, the fact that Eq.
Stefan-Maxwell equation for species A reduces to
(152) and Eq. (180) are identical in form suggests
B=N
x A xB v A that we must be very careful to understand the
0 = −∇x A − ∑
B =1 DAB
(174) precise meaning of the special forms of Eq. (68).
B≠ A
7. General solution for N-component systems:
and this can be arranged in the form Constant total molar concentration
B=N From the analysis in previous sections, it seems clear

xB
0 = −c∇x A − ∑
B =1 DAB
NA (175) that the most efficient route to the determination of
the molar concentration is via the mixed-mode
B≠ A
www.amidiq.org 229
continuity equation described in Sec. 4. This is using v ∗ and J ∗A with A = 1, 2,..., N − 1 is given by
especially true for the case in which we develop an Bird et al. (2002, Sec. 22.9). In addition, Quintard et
exact solution of the Stefan-Maxwell equations. In al. (2006) have studied the formulation and the
this section we consider the case of constant total numerical solution for this problem using both the
molar concentration and in the next section we
molar forms, v ∗ and J ∗A , and the mass forms, v and
examine the case of constant total mass density. The
completely general case for which neither ρ nor c is jA .
constant remains as a challenge.
We begin our analysis of the diffusive flux
In this treatment we make use of Eq. (91) repeated with the Stefan-Maxwell equations given by Eq.
here as (166), and we make use of the mixture diffusivity
defined by Eq. (145) to obtain
∂c A
+ ∇ ⋅ (c A v) = −∇ ⋅ J A + RA , B=N
DAm
∂t (183) J A = −c DAm ∇x A + x A ∑ JB ,
DAB
A = 1, 2, ..., N − 1 B =1
B≠ A
(187)
A = 1, 2, ...., N − 1
along with the constraint on the mixed-mode
diffusive flux given by
We want to use Eq. (184) to eliminate J N and it will
A= N be convenient to express that constraint on the
∑M
A =1
A JA = 0 (184) mixed-mode diffusive fluxes in the alternate form
given by
For N-component systems, it is convenient to work A= N
in terms of matrices, thus we define the following ∑ J (M A A MN ) = 0 (188)
column matrices that will be used in subsequent A =1
paragraphs.
At this point we extract J N from the sum in Eq.
⎡ cA ⎤ ⎡ ∇c A ⎤ (187) in order to obtain
⎢ c ⎥ ⎢ ∇c ⎥
⎢ B ⎥ ⎢ B ⎥ B = N −1
DAm D
⎢ c ⎥ ⎢ ∇cC ⎥ J A = −c DAm ∇x A + x A ∑ J B + x A Am J N ,
[c ] = ⎢ ...C ⎥ , [∇c ] = ⎢ ... ⎥ , B =1 DAB DAN (189)
⎢ ⎥ ⎢ ⎥ B≠ A
⎢ ... ⎥ ⎢ ... ⎥ A = 1, 2, ..., N − 1

⎢ ⎥ ⎢ ⎥
⎣⎢ cN −1 ⎦⎥ ⎣⎢∇cN −1 ⎦⎥ and from Eq. (188) we have the following
⎡ ∇x A ⎤ ⎡ JA ⎤ representation for J N
⎢ ∇x ⎥ ⎢ J ⎥
⎢ B ⎥ ⎢ B ⎥ B = N −1
⎢ ∇xC ⎥
[∇x ] = ⎢ ... ⎥ ,
⎢ J ⎥
[ J ] = ⎢ ...C ⎥ ,
JN = − ∑
B =1
JB (M B M N ) (190)
⎢ ⎥ ⎢ ⎥
⎢ ... ⎥ ⎢ ... ⎥ In order to use this result with Eq. (189), we need to
⎢ ⎥ ⎢ ⎥ condition the sum with the constraint indicated by
⎣⎢∇xN −1 ⎦⎥ ⎣⎢ J N −1 ⎦⎥ B ≠ A and this leads to
⎡ RA ⎤
⎢ R ⎥ B = N −1
⎢ B ⎥ JN = − ∑ JB (M B M N ) − J A (M A M N ) (191)
⎢ R ⎥ B =1
[ R ] = ⎢ ...C ⎥ B≠ A
⎢ ⎥ Use of this result in Eq. (189) provides the following

⎢ ... ⎥
⎢ ⎥ form of the Stefan-Maxell equations
⎢⎣ RN −1 ⎥⎦ (185)
⎛ M D ⎞ B = N −1
⎛M D D ⎞
Use of the first, fourth and fifth of these matrices J A ⎜ 1 + x A A Am ⎟ + x A ∑ ⎜ B Am − Am ⎟ J B
⎝ M D
N AN ⎠ B =1 ⎝ M D
N AN DAB ⎠ (192)
allows us to express Eq. (183) as B≠ A
= −c DAm ∇x A , A = 1, 2, ...., N − 1
∂[c]
+ ∇ ⋅ ([c] v ) = −∇ ⋅ [ J ] + [ R ] (186)
∂t This can be expressed in compact form according to
and our single objective at this point is to develop a

useful representation for [ J ] . A similar approach
230 www.amidiq.org
⎡ JA ⎤ ⎡ D Am ∇x A ⎤ so that the matrix representation for the mixed-mode

⎢J ⎥ ⎢ D ∇x ⎥ diffusive flux becomes
⎢ B ⎥ ⎢ Bm B ⎥
⎢ JC ⎥ ⎢ D Cm ∇xC ⎥
[ H ] ⎢ ... ⎥ = − c ⎢ ... ⎥ ⎡ JA ⎤
⎢ ⎥ ⎢ ⎥
(193)
⎢J ⎥ ⎡ DAm ....
0 0 0 ⎤ ⎡ ∇x A ⎤
⎢ B ⎥ ⎢ 0 D 0 .... 0 ⎥⎥ ⎢ ∇xB ⎥
⎢ ... ⎥ ⎢ ... ⎥ ⎢ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ JC ⎥ Bm
⎥ = −c [ H ] ⎢ 0
−1
⎢⎣ J N −1 ⎥⎦ ⎢⎣ D N −1m ∇xN −1 ⎥⎦ ⎢ 0 D .... Cm 0 ⎥ ⎢ ∇xC ⎥
⎢ ... ⎥ ⎢ ⎥⎢ ⎥
⎢ ... ⎥ ⎢ . . . .... . ⎥⎢ .. ⎥
in which [ H ] is an ( N − 1) × ( N − 1) square matrix ⎢ ⎥ ⎢
⎣ 0 0 0 .... DN −1m ⎦⎥ ⎢⎣∇xN −1 ⎥⎦
⎢⎣ J N −1 ⎥⎦
⎡ H AA H AB . . . H AN −1 ⎤ (198)
⎢ H H BB . . . H BN −1 ⎥⎥
⎢ BA The diffusivity matrix is now defined by
⎢ H . . . . . ⎥
[ H ] = ⎢ CA .⎥ (194) ⎡ DAm ....
0 0 0 ⎤
⎢ . . . . . . ⎥ ⎢ 0
⎢ . ⎥ ⎢ D 0 ....
Bm 0 ⎥⎥
. . . . .
⎢ ⎥ [ D] = [ H ] ⎢ 0
−1
0 D .... 0 ⎥ (199)
⎣⎢ H N −1 A H N −1N −1 ⎦⎥ Cm
. . . . ⎢ ⎥
⎢ . . . .... . ⎥
having the elements defined by ⎢ 0 0 0 .... DN −1m ⎥⎦
⎣
M A DAm
H AA = 1 + x A , and this allows us to express Eq. (198) as
M N DAN (195a)
A = 1, 2,..., N − 1 ⎡ JA ⎤ ⎡ ∇x A ⎤
⎢J ⎥ ⎢ ∇x ⎥
⎢ B ⎥ ⎢ B ⎥
⎛M D D ⎞ ⎢ JC ⎥ ⎢ ∇xC ⎥
H AB = x A ⎜ B Am − Am ⎟ , ⎢ ⎥ = − c [ D] ⎢ ⎥ (200)
⎝ M N DAN DAB ⎠ (195b)
⎢ ... ⎥ ⎢ ... ⎥
A, B = 1, 2,.., N − 1 , A ≠ B ⎢ ... ⎥ ⎢ ... ⎥
⎢ ⎥ ⎢ ⎥
We assume that the inverse of [ H ] exists in order to ⎣⎢ J N −1 ⎦⎥ ⎣⎢∇xN −1 ⎦⎥
express the column matrix of the mixed-mode with the compact form given by
diffusive flux vectors in the form
[J] = − c [ D ] [ ∇x ] (201)
⎡ JA ⎤ ⎡ DAm ∇x A ⎤
⎢J ⎥ ⎢ D ∇x ⎥ This represents the N-component analog of Fick’s
⎢ B ⎥ ⎢ Bm B ⎥
⎢ JC ⎥ ⎢ D ∇ x ⎥ Law given by Eq. (135) that we recall here as
⎥ = − c[H ] ⎢
−1 Cm C
⎢ ⎥ (196)
⎢ ... ⎥ ⎢ ... ⎥ Fick’s Law: J A = −c ⎣⎡( M B M ) DAB ⎦⎤ ∇x A (202)
⎢ ... ⎥ ⎢ ... ⎥
⎢ ⎥ ⎢ ⎥
⎣⎢ J N −1 ⎦⎥ ⎢⎣ DN −1m ∇xN −1 ⎥⎦ Use of Eq. (201) in Eq. (186) leads to
∂[c]
The column matrix on the right hand side of this + ∇ ⋅ ([c]v ) = ∇ ⋅ ( c [ D ][∇x ]) + [ R ] (203)
result can be expressed as ∂t
⎡ DAm ∇x A ⎤ Once again we may be faced with the difficult task

⎢ D ∇x ⎥ of determining the total molar concentration on the
⎢ Bm B ⎥ basis of Eq. (93), and to avoid this problem we
⎢ DCm ∇xC ⎥ restrict Eq. (203) to the case of constant total molar
⎢ ⎥
⎢ ... ⎥
concentration. This leads to
⎢ DN −1m ∇xN −1 ⎥
⎣ ⎦ ∂[c]
(197) + ∇ ⋅ ([ c ] v ) = ∇ ⋅ ([ D ] [ ∇ c ] ) + [ R ]
⎡DAm 0 0 .... 0 ⎤ ⎡ ∇x A ⎤ ∂t
⎢ 0 D .... 0 ⎥⎥ ⎢ ∇xB ⎥ c = constant
(204)
⎢ Bm 0
⎢ ⎥ ⎧
⎨
=⎢ 0 0 DCm .... 0 ⎥ ⎢ ∇xC ⎥ ⎩ N − component system
⎢ ⎥⎢ ⎥
⎢ . . . .... . ⎥⎢ .. ⎥ Here it is important to remember that [ D] depends
⎢ 0 0 0 .... DN −1m ⎥⎦ ⎢⎣∇xN −1 ⎥⎦
⎣ explicitly on the mole fractions, as indicated by the
www.amidiq.org 231
definitions given in Eq. 195 and implicitly as M ⎡ B= N

M ⎤
indicated by the definition of the mixture diffusivity ∇x A = ⎢ ∇ω A − ω A ∑ ∇ω B ⎥ ,
given by Eq. (145). This means that a trial-and-error
MA ⎣ B =1 M B ⎦ (210)
numerical solution will be necessary in which the A = 1, 2,...N
assumed values used for the mole fractions are
upgraded after each iteration. The solution for [c ] At this point we can make use of the fact that the
sum of the mass fractions is equal to one so that the
will provide values of c A , cB , ...., cN −1 and the
gradients are related by
concentration cN can be determined by the first of
∇ωN = − ( ∇ω A + ∇ωB + ∇ωC + ..... + ∇ω N −1 ) (211)
Eqs. (71). Similarly, the solution for [ R] will
provide values of RA , RB , ...., RN −1 and the reaction This allows us to eliminate ∇ω N from Eq. (210) and
rate RN can be determined by Eq. (69). In the case express that result in the form
of complex kinetics, the column matrix of reaction
M ⎡ B = N −1
⎛ M M ⎞ ⎤
⎢∇ω A − ω A ∑ ⎜
rates will need to be expressed as
∇x A = − ⎟ ∇ωB ⎥ ,
MA ⎣⎢ B =1 ⎝ M M N ⎠ ⎦⎥ (212)
[ R] = [F (cA , cA , ..., cN )] (205)
B
A = 1, 2,...N − 1
and the trial-and-error procedure will be more
complex. Here we need to condition the sum with the
constraint indicated by B ≠ A and this leads to
8. General solution for N-component systems: M ⎧⎪ ⎡ ⎛ M M ⎞⎤
Constant total mass density ∇x A = ⎨ ⎢1 + ω A ⎜ − ⎟ ⎥ ∇ω A
M A ⎩⎪ ⎣⎢ ⎝ N
M M A ⎠⎦⎥
In addition to the N-component form of the species (213)
⎫
continuity equation based on the assumption of a B = N −1
⎛ M M ⎞ ⎪
constant total molar concentration, it would be +ω A ∑ ⎜ − ⎟ ∇ωB ⎬ , A = 1, 2,...N − 1
B =1 ⎝ M N MB ⎠ ⎪
useful to develop the analogous form for constant B≠ A ⎭
total density. Our starting point for this analysis is
Eq. (203) and the analysis requires that we express which can be expressed as a matrix equation given
[∇x ] in terms of the gradient of the mass fractions, by
∇ω A , ∇ωB , etc. We begin the analysis with Eq. ⎡ ∇x A ⎤ ⎡ WAA WAB WAC . . WAN −1 ⎤ ⎡ ∇ω A ⎤
(105) repeated here as ⎢ ∇x ⎥ ⎢ W WBB . . . WBN −1 ⎥⎥ ⎢⎢ ∇ωB ⎥⎥
⎢ B ⎥ ⎢ BA
M ⎢ ∇xC ⎥ ⎢ WCA . . . . . ⎥ ⎢ ∇ωC ⎥
⎢ ⎥=⎢
xA =
MA
ωA , A = 1, 2,...N (206)
⎢ . ⎥ ⎢ . . . . . . ⎥⎥ ⎢⎢ . ⎥⎥
⎢ . ⎥ ⎢ . . . . . . ⎥⎢ . ⎥
in which the mean molecular mass can be expressed ⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢⎣∇xN −1 ⎥⎦ ⎢⎣WN −1 A WN −1B . . . WN −1N −1 ⎥⎦ ⎢⎣∇ωN −1 ⎥⎦
as in terms of the mass fractions in order to obtain
(see Eq. C11 in the Appendix C) (214)
1 ω ω ω ω Here the elements of this ( N − 1) × ( N − 1) square

= A + B + C + ... + N (207) matrix are defined as
M M A M B MC MN
M M ⎛ M M ⎞
We can use Eq. (206) to express the gradient of the WAA = + ωA ⎜ − ⎟,
mole fraction as MA MA ⎝ MN MA ⎠ (215a)
A = 1, 2,..., N − 1
∇M M
∇x A = ωA + ∇ω A , A = 1, 2,...N (208)
MA MA M ⎛ M M ⎞
WAB = ω A ⎜ − ⎟,
while the gradient of the mean molecular mass is MA ⎝ MN MA ⎠ (215b)
given by A, B = 1, 2,..., N − 1 , A≠ B
B= N
∇ω B
∇M = − M 2 ∑
At this point we recall Eq. (200) and make use of Eq.
(209)
B =1 MB (214) to obtain
Use of Eq. (209) in Eq. (208) leads to
232 www.amidiq.org
⎡ JA ⎤ ⎡ ∇ω A ⎤ ⎡ JA ⎤ ⎡M A 0 0 . . 0 ⎤ ⎡ ∇c A ⎤
⎢J ⎥ ⎢ ∇ω ⎥ ⎢J ⎥ ⎢ 0 MB 0 . . 0 ⎥⎥ ⎢⎢ ∇cB ⎥⎥
⎢ B ⎥ ⎢ B ⎥ ⎢ B ⎥ ⎢
⎢ JC ⎥ ⎢ ∇ωC ⎥ ⎢ JC ⎥ ⎢ 0 0 MC . . . ⎥ ⎢ ∇cC ⎥
⎥ = − M [ D] [W ] ⎢
−1
⎢ ⎥ = −c [ D ][W ] ⎢ ⎥ ⎢
. ⎥⎥ ⎢⎢ . ⎥⎥
(216)
⎢ ... ⎥ ⎢ . ⎥ ⎢ ... ⎥ ⎢ . . . . .
⎢ ... ⎥ ⎢ . ⎥ ⎢ ... ⎥ ⎢ . . . . . . ⎥⎢ . ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢⎣ J N −1 ⎥⎦ ⎢⎣∇ωN −1 ⎥⎦ ⎢⎣ J N −1 ⎥⎦ ⎢⎣ 0 0 . . . M N −1 ⎥⎦ ⎢⎣∇cN −1 ⎥⎦
(221)
in which the square matrix [W ] is defined explicitly
in which it is understood that the total mass density
by
is assumed to be constant. We can represent this
⎡ WAA WAB WAC . . WAN −1 ⎤ result in compact form
⎢W WBN −1 ⎥⎥
⎢ BA WBB . . . [ J ] = − [D][∇c ] (222)
⎢ WCA . . . . . ⎥
[W ] = ⎢ . . . . . . ⎥⎥
(217) in which the new diffusivity matrix is given by
⎢
⎢ . . . . . . ⎥
⎢ ⎥ ⎡M A M 0 0 . . 0 ⎤
⎣⎢WN −1 A WN −1B . . . WN −1N −1 ⎦⎥ ⎢ ⎥
⎢ 0 MB M 0 . . 0 ⎥
⎢ 0 0 MC M . . . ⎥
Use of the third of Eqs. (104) leads to the total mass [D] = − [ D][W ] ⎢ ⎥
density as a multiplier and Eq. (216) takes the form ⎢ . . . . . . ⎥
⎢ . . . . . . ⎥
⎡ JA ⎤ ⎡ ∇ω A ⎤ ⎢ ⎥
⎢J ⎥ ⎢ ∇ω ⎥ ⎣⎢ 0 0 . . . M N −1 M ⎦⎥
⎢ B ⎥ ⎢ B ⎥
(223)
⎢ JC ⎥ ⎢ ∇ ω ⎥
⎢ ⎥ = −ρ M
−1
[ D] [W ] ⎢ C
⎥ (218)
. Use of Eq. (222) in Eq. (186) yields
⎢ ... ⎥ ⎢ ⎥
⎢ ... ⎥ ⎢ . ⎥
∂[c]
⎢ ⎥ ⎢ ⎥ + ∇ ⋅ ([c]v ) = ∇ ⋅ ([D][∇c]) + [ R ]
⎣⎢ J N −1 ⎦⎥ ⎣⎢∇ω N −1 ⎦⎥ ∂t
(224)
We are now in a position to impose the condition ⎧ ρ = constant
⎨
that the total mass density is a constant in order to ⎩ N − component system
express the mixed-mode fluxes in the form
In the trial-and-error solution of this transport
⎡ JA ⎤ ⎡ ∇ρ A ⎤ equation, values of the mole fractions will be
⎢J ⎥ ⎢ ∇ρ ⎥ required as in the solution of Eq. (204); however, in
⎢ B ⎥ ⎢ B ⎥
⎢ JC ⎥ ⎢ ⎥ this case it is the total mass density, ρ, that is a

∇ ρ
⎢ ⎥=−M
−1
[ D] [W ] ⎢ C
⎥, specified constant and not the total molar
⎢ ... ⎥ ⎢ . ⎥ (219) concentration, c. This requires that we first determine
⎢ ... ⎥ ⎢ . ⎥ ρ N and then cN according to
⎢ ⎥ ⎢ ⎥
⎣⎢ J N −1 ⎦⎥ ⎣⎢∇ρ N −1 ⎦⎥ A = N −1
ρ = constant ρN = ρ − ∑
A =1
c A M A , cN = ρ N M N (225)
At this point we make use of the first of Eqs. (104) to

The mole fractions required for the evaluation of
express the column matrix of the gradients of the
species densities as [ D ] would then be determined by
⎡ ∇ρ A ⎤ ⎡ M A 0 0 . . 0 ⎤ ⎡ ∇c A ⎤ B= N
⎢ ∇ρ ⎥ ⎢ 0 MB 0 . . 0 ⎥⎥ ⎢⎢ ∇cB ⎥⎥
xA = cA ∑c B (226)
⎢ B ⎥ ⎢ B =1
⎢ ∇ρ C ⎥ ⎢ 0 0 MC . . . ⎥ ⎢ ∇cC ⎥
⎢ ⎥=⎢
. ⎥⎥ ⎢⎢ . ⎥⎥
(220) while the mass fractions required for the evaluation
⎢ . ⎥ ⎢ . . . . . of [W ] would be calculated according to
⎢ . ⎥ ⎢ . . . . . . ⎥⎢ . ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥ B=N
⎣⎢∇ρ N −1 ⎦⎥ ⎣⎢ 0 0 . . . M N −1 ⎦⎥ ⎣⎢∇cN −1 ⎦⎥
ωA = cA M A ∑c MB (227)
B
B =1
Substitution of this result into Eq. (219) leads
to The result for constant total density given by Eq.
(224), along with that for constant total molar
www.amidiq.org 233
concentration given by Eq. (204), should prove to be NA c A v A , molar flux of species A, mole/m2s
useful for a wide range of mass transfer problems, n unit normal vector
provided that the Stefan-Maxwell equations are an PAB force per unit volume exerted by species
acceptable representation for the diffusive fluxes. A
discussion of the conditions for which the total molar B on species A, N/m3
A= N
concentration and total mass density may be treated
as constants is given in Appendix E.
p ∑p
A =1
A , total pressure, N/m2
pA partial pressure of species A, N/m2

9. Conclusions
rA net mass rate of production of species A
In this study we have examined the derivation of the owing to homogeneous reactions, kg/m3s
Stefan-Maxwell equations and we have explored the RA net molar rate of production of species A
structure of these equations in terms of the mass
diffusive flux, the molar diffusive flux, and the owing to homogeneous reactions,
mixed-mode diffusive flux. Several classic special moles/m3s
cases have been examined and the assumptions, R gas constant, J/mol K
restrictions and constraints have been identified t time, s
whenever possible. A general method of solution of t A(n ) stress vector for species A, N/m2
the Stefan-Maxwell equations has been presented in TA stress tensor for species A, N/m2
terms of the mixed-mode diffusive flux. U total internal energy in a volume V, J
Nomenclature uA v A − v , mass diffusion velocity, m/s
u∗A v A − v∗ , molar diffusion velocity, m/s
AA (t ) surface area of a species A material
vA velocity of species A, m/s
volume, m2
A= N
bA body force per unit mass exerted on v ∑ω A v A , mass average velocity, m/s
species A, N/kg A =1
b body force per unit mass exerted on the A= N
mixture, N/kg v∗ ∑x
A =1
A v A , molar average velocity, m/s
cA molar concentration of species A,
v∗A velocity associated with the net rate of
moles/m3 production of species A momentum
c total molar concentration, moles/m3 owing to chemical reaction, m/s
dA driving force for diffusion of species A in V A (t ) volume of a species A body, m3
an ideal solution, m−1 xA c A / c , mole fraction of species A
DAB DBA , binary diffusion coefficient for
species A and B, m2/s Greek Letters
DAm mixture diffusivity for species A, m2/s ρA mass density of species A, kg/m3
[ D] diffusivity matrix used with constant total ρ total mass density, kg/m3
molar concentration, m2/s μ viscosity, N/m2s
[D] diffusivity matrix used with constant total τ viscous stress tensor, N/m2
mass density, m2/s τA viscous stress tensor for species A, N/m2
DAT thermal diffusion coefficient for species
A, kg m 2s ωA ρ A / ρ , mass fraction of species A
G number of molecular species that are Acknowledgment
dilute
g gravitational body force per unit mass, This paper grew out of a presentation at the Second
N/kg International Seminar on Trends in Chemical
jA ρ Au A , mass diffusive flux of species A, Engineering, the XXI Century, Mexico City, January
kg/ m2s 28 – 29, 2008. The encouragement of students from
J ∗A c A u∗A , molar diffusive flux of species A, Puebla to prepare a more complete discussion of the
Stefan-Maxwell equations is greatly appreciated. In
moles/ m2s addition, the thoughtful comments of Francois
JA c Au A , mixed-mode diffusive flux of Mathieu-Potvin helped to clarify some of the issues
species A, moles/ m2s treated in this work. Finally, the comments of
MA molecular mass of species A, g/mole Professor R.B. Bird have clarified my understanding
M mean molecular mass of a mixture, of the complex process of multicomponent mass
g/mole transfer.
N total number of molecular species
234 www.amidiq.org
References Slattery, J.C. (1999). Advanced Transport

Phenomena, Cambridge University Press,
Aris, R. (1962). Vectors, Tensors, and the Basic Cambridge.
Equations of Fluid Mechanics, Prentice-Hall, Stein, S.K. and Barcellos, A. (1992). Calculus and
Englewood Cliffs, New Jersey. Analytic Geometry, McGraw-Hill, Inc., New
Bearman, R.J. and Kirkwood, J.G. (1958). Statistical York.
mechanics of transport Processes. XI Truesdell, C. and Toupin, R. (1960). The Classical
Equations of transport in Multicomponent Field Theories, in Handbuch der Physik, Vol.
systems. Journal of Chemical Physics 28, III, Part 1, edited by S. Flugge, Springer
136-145. Verlag, New York.
Bird, R.B., Stewart, W.E. and Lightfoot, E.N. Truesdell, C. (1962). Mechanical basis of diffusion,
(1960). Transport Phenomena, First Edition. Journal of Chemical Physics 37, 2336-2344
John Wiley and Sons, Inc., New York. Truesdell, C. (1968). Essays in the History of
Bird, R.B. (1995) personal communication. Mechanics, Springer-Verlag, New York.
Bird, R.B., Stewart, W.E. and Lightfoot, E.N. Truesdell, C. (1969). Rational Thermodynamics,
(2002). Transport Phenomena, Second McGraw-Hill Book Company, New York.
Edition. John Wiley and Sons, Inc., New Truesdell, C. (1971). The Tragicomedy of Classical
York. Thermodynamics, Springer-Verlag, New
Bird, R.B. (2009). Notes for the 2nd edition of York.
Transport Phenomena, Whitaker, S. and Pigford, R.L. (1958). Thermal
http://www.engr.wisc.edu/che/faculty/bird_by diffusion in liquids. Measurements and a
ron.html. molecular model. Industrial and Engineering
Birkhoff, G. (1960). Hydrodynamics, A Study in Chemistry 50, 1026-1032.
Logic, Fact, and Similitude, Princeton Whitaker, S. (1981). Introduction to Fluid
University Press, Princeton, New Jersey. Mechanics, R.E. Krieger Pub. Co., Malabar,
Chapman, S. and Cowling, T.G. (1939). The Florida.
Mathematical Theory of Nonuniform Gases, Whitaker, S. (1986). Transport processes with
First Edition, Cambridge University Press. heterogeneous reaction, pages 1 to 94 in
Chapman, S. and Cowling, T.G. (1970). The Concepts and Design of Chemical Reactors,
Mathematical Theory of Nonuniform Gases, edited by S. Whitaker and A.E. Cassano,
Third Edition, Cambridge University Press. Gordon and Breach Publishers, New York.
Curtiss, C.F. and Bird, R.B. (1996). Multicomponent Whitaker, S. (1988). Levels of simplification: The
diffusion in polymeric liquids. Proceedings of use of assumptions, restrictions, and
the National Academy of Sciences USA 93, constraints in engineering analysis. Chemical
7440-7445. Engineering Education 22, 104-108.
Curtiss, C.F. and Bird, R.B. (1999). Multicomponent Whitaker, S. (1989). Heat transfer in catalytic packed
diffusion. Industrial and Engineering bed reactors, in Handbook of Heat and Mass
Chemistry Research 38, 2515-2522. Transfer, Vol. 3, Chapter 10, Catalysis,
Deen, W. M. (1998). Analysis of Transport Kinetics & Reactor Engineering, edited by
Phenomena. Oxford University Press, New N.P. Cheremisinoff, Gulf Publishers,
York. Matawan, New Jersey.
Gibbs, J.W. (1928). The Collected Works of J. Whitaker, S. (1999). The Method of Volume
Willard Gibbs, Volume I: Thermodynamics, Averaging, Kluwer Academic Publishers,
Longmans, Green and Co., New York. Dordrecht.
Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B. Whitaker, S. (2009a). Newton’s laws, Euler’s laws,
(1954). Molecular Theory of Gases and and the speed of light. Chemical Engineering
Liquids, John Wiley & Sons, Inc., New York. Education, Spring.
Quintard, M., Bletzacker, L., Chenu, D. and Whitaker, S. (2009b). Chemical engineering
Whitaker, S. (2006). Nonlinear, multi- education: Making connections at interfaces,
component mass transfer in porous media, Revista Mexicana de Ingeniería Química 8, 1-
Chemical Engineering Science 61, 2643- 32.
2669.
Rutten, Ph.W.M. (1992). Diffusion in Liquids (PhD
thesis), Delft University Press, The Appendix A: Chemical Reaction and Linear
Netherlands. Momentum
Serrin, J. (1959). Mathematical Principles of The rate of change of linear momentum of species A
Classical Fluid Mechanics, in Handbuch der
owing to chemical reaction, rA v∗A , can be caused
Physik, Vol. VIII, Part 1, edited by S. Flugge
and C. Truesdell, Springer Verlag, New York. either by the increase of species A (production) or by
the decrease of species A (consumption). If species A
is consumed by chemical reaction, it seems plausible
www.amidiq.org 235
Fig. A1. Reaction of species B to form species A Fig. A2. Reaction of B and C to produce A and D
that the rate of change of linear momentum is given

⎧ rate of change of ⎫
by rA v A . Here we need to note that the molecular ⎪ ⎪
⎨linear momentum ⎬ = rA v A = rA v A + rA ( v B − v A ) (A6)
∗
velocity (Hirschfelder et al., 1954, page 453) of ⎪ of species A ⎪
species A is much larger than the continuum velocity, ⎩ ⎭
v A ; however, the average velocity associated with
The species velocities can be expressed in terms of
the consumption of species A should be adequately
the mass average velocity and the diffusion velocity
represented by v A . If species A is produced by a to obtain
chemical reaction, the rate of change of linear
momentum depends on the velocities of the species vA = v + uA , vB = v + uB (A7)
that react to form species A.
and these results can be used in Eq. (A6) so that the
The simple reaction illustrated in Fig. A1 can rate of change of momentum of species A takes the
be described as 2B → A , and we assume that the form
loss of momentum by species B is equal to the gain
of momentum of species A. We express this idea as ⎧ rate of change of ⎫
⎪ ⎪
[see Eq. (21)] ⎨linear momentum ⎬ = rA v A + rA ( u B − u A ) (A8)
⎪ of species A ⎪
r v∗B + rA v∗A = 0 (A1) ⎩ ⎭
B

loss gain
This leads to the estimate
and note that conservation of mass [see Eq. 2]
requires ⎧ rate of change of ⎫
⎪ ⎪
⎨linear momentum ⎬ = rA v A + O ( rAu A ) (A9)
rB + rA = 0 (A2) ⎪ of species A ⎪
N N ⎩ ⎭
loss gain
suggested by Whitaker (1986, Eqs. 1-19 and 1-55).

On the basis of the argument given above, we
assume that If we consider the slightly more complex
reaction illustrated in Fig. A2, the concepts
v∗B = v B (A3) illustrated in Eqs. (A2) and (A4) take the form
and Eq. (A1) takes the form r + rC + rA + rD = 0 (A10)

N N
B
∗ gain gain
r v B + rA v = 0
loss
(A4)
B

A
loss gain
rB v B + rC v C + rA v∗A + rD v∗D = 0 (A11)

loss gain gain

When Eq. (A2) is used with this result we find that
v∗A is given by In this case constructing a value for v ∗A is not as
v∗A = v B (A5) simple as the result illustrated by Eq. (A5). In terms
of molar rates of reaction, we have
and the rate of change of linear momentum of
rB = M B RB , rB = M B RB ,
species A can be expressed as (A12)
rB = M B RB , rB = M B RB
in which M B represents the molecular mass of

species B and RB represents the molar rate of
236 www.amidiq.org
reaction for species B. In terms of molecular masses in which eA is the internal energy of species A per
and molar rates of reaction, we can express Eq. unit mass of species A, and ρ A is the mass density of
(A11) as
species A. We defined the total pressure in terms of
M B RB v B + M C RC v C + M A RA v∗A the partial pressures according to

A= N
loss gain
(A13) p= ∑p A (B2)
+ M D RD v∗D = 0 A =1

However, the total pressure, p, can also be expressed
gain
as
and Eq.(A10) can be replaced by p = ρ 2 ( ∂e ∂ρ )s , ρ , ρ ,..., ρ (B3)
B C N
RB = RC , RC = − RA , RA = RD in which e is the total internal energy defined by

A= N
Use of this constraint on the molar rates of reaction

e= ∑ω
A =1
A eA (B4)
in Eq. (A13) leads to In this appendix we wish to show that there is no
∗ ∗ conflict between Eqs. (B1), (B2) and (B3), and this
M B v B + M C vC = M A v + M D v
A D (A14) requires that we demonstrate the following:
A= N
⎛ ∂e ⎞ ⎛ ∂e ⎞
We now express the species velocities in terms of the
∑ ρ A2 ⎜ A ⎟
∂ ρ
= ρ2 ⎜ ⎟
∂ρ ⎠ s , ρB , ρC ,..., ρ N
(B5)
mass average velocity and the diffusion velocities in A =1 ⎝ A ⎠ s , ρ , ρ ,..., ρ
B C N
⎝
order to obtain
In order to illustrate how the thermodynamic
vA = v + uA , vB = v + uB , definition of the partial pressure is related to the
(A15) thermodynamic definition of the total pressure, we
vC = v + uC need the following theorem:
A= N
∂Λ ∂Λ
Use of these relations in Eq. (A14) provides Theorem: ∑
A =1
ρA A = ρ
∂η ∂η
(B6)
M B u B + M C u C + ( M B + M C ) v = M A ( v∗A − v A ) Here Λ A is a partial mass quantity such as the

(A16)
+ M D ( v∗D − v D ) + M A u A + M D u D + ( M A + M D ) v species internal energy represented in Eq. (B1),
while Λ is a total mass quantity defined by
A= N
that can be simplified to Λ= ∑ω
A =1
A ΛA (B7)
M A ( v∗A − v A ) + M D ( v∗D − v D ) In Eq. (B6) we have used η to represent some

(A17)
= ( M B u B + M C uC ) − ( M A u A + M D u D ) thermodynamic state variable such as the
temperature, the total mass density, etc.
Provided that the molecular masses are all the same We begin this proof with some variable Ω
order of magnitude, this result suggests that the that can be represented as
difference, v∗A − v A , is on the order of the diffusion ρΩ = ρ A Ω A + ρ B Ω B + ... + ρ N Ω N (B8)
velocities. Given the general constraint on the or in a manner identical to Eq. (B7)
diffusion velocities [see Eq. (14)], the result given by Ω = ω A Ω A + ωB Ω B + ... + ω N Ω N (B9)
Eq. (A17) suggests that
Here the mass fractions are defined by the second of
v∗A = v A + O ( u A ) (A18) Eqs. (11) and they are constrained by
which is equivalent to Eq. (A9). ω A + ωB + ... + ω N = 1 (B10)
The two cases represented in Figs. A1 and A2 Because of this constraint all the mass fractions are
are especially simple; however, most chemical not independent and the functional representation for
reactions are likely to be binary in nature, thus Eq. Ω is given by
(A18) represents a plausible estimate of the velocity
Ω = Ω ( ρ , T , ω A , ωB , ...., ω N −1 ) (B11)
v∗A .
If we differentiate Ω with respect to ω A we can
Appendix B: Thermodynamic pressure hold all the mass fractions constant except one. For
convenience we choose this one to be ω N and write
The decomposition given by Eq. (42) indicates that (Slattery, 1999, page 447)
TA is represented in terms of the partial pressure, ( ∂ Ω ∂ ω A )ρ , T , ω ( B ≠ A, N ) = Ω A − Ω N (B12)
B
p A , and the viscous stress tensor, τ A . The partial This allows us to express Ω A as
pressure of species A can be defined by (Whitaker,
1989, Chapter 10) Ω A = ( ∂ Ω ∂ ω A )ρ , T , ω + ΩN (B13)
B ( B ≠ A, N )
p A = ρ A2 ( ∂ eA ∂ ρ A )s , ρ , ρ ,....... (B1) and Eq. (B8) can be used to obtain

B C
www.amidiq.org 237
A= N A= N
⎛ ∂Ω ⎞ eA = eA ( s, ρ A , ρ B , ...., ρ N ) (B26b)
ρΩ = ∑ρ ΩA = ∑ρ ⎜ ⎟ + ρ ΩN
⎝ ∂ ω A ⎠ ρ , T , ωB ( B ≠ A , N )
A A
A =1 A =1 In addition, the density of species A, for example,
can be expressed as
(B14)
Subsequently we will use this result in the form ρ A = ρ − ( ρ B + ρC + .... + ρ N ) (B27)
A= N
⎛ ∂Ω ⎞ or in the functional form given by
∑ ρA ⎜ ⎟ = ρΩ − ρ Ω N (B15) ρ A = ρ A ( ρ , ρ B , .... , ρ N ) (B28)
A =1 ⎝ ∂ ω A ⎠ ρ , T , ω ( B ≠ A, N )
On the basis of this representation for ρ A we can
B
At this point we consider the special case in which

express Eq. (B26b) as a composite function given by
∂Λ ∂Λ A
Ω= , ΩA = , A = 1, 2,..., N (B16) eA = eA ⎡⎣ s, ρ A ( ρ , ρ B , ... , ρ N ) , ρ B , ...., ρ N ⎤⎦ (B29)
∂η ∂η
Use of this result in Eq. (B15) yields Directing our attention to the derivative on the left
A= N ∂ ( ∂Λ ∂η ) ∂Λ ∂Λ N
hand side of Eq. (B24) we note that it can be
∑A =1
ρA
∂ω
=ρ
∂ η
−ρ
∂η
(B17) expressed as (Stein and Barcellos, 1992, page 149)
A
ρ , T , ωB ( B ≠ A , N )
Here we write Eq. (B13) for the variables Λ A and Λ ⎛ ∂eA ⎞ ⎛ ∂e ⎞ ⎛ ∂ρ A ⎞

⎜ ⎟ =⎜ A ⎟ ⎜ ⎟
to obtain ⎝ ∂ρ ⎠ s , ρ B , ρC , .... ρ N ⎝ ∂ρ A ⎠ s , ρ B , ρC , .... ρ N ⎝ ∂ρ ⎠ ρB , ρC , .... ρ N
Λ A = ( ∂ Λ ∂ ω A ) ρ , T , ω ( B ≠ A, N ) + Λ N (B18) (B30)
B
Use of this result in the left hand side of the theorem Since the mass density for species A can be
we wish to prove leads to expressed as
⎡ ∂ ⎛ ∂Λ ⎞ ρ A = ωA ρ (B31)
A= N
∂Λ A= N
∂Λ N ⎤
∑ ρA A = ∑ ρA ⎢ ⎜
∂η
⎟ +
⎢⎣ ∂η ⎝ ∂ ω A ⎠ ρ , T , ωB ( B ≠ A, N ) ∂η ⎥⎦
⎥ we have
A =1 A =1
( ∂ρ A ∂ρ ) ρ , ρ , .... ρ = ω A
B C N
(B32)
(B19)
and Eq. (B30) takes the form
Changing the order of differentiation in the first term
and carrying out the summation with the second term ( ∂eA ∂ρ )s , ρ , ρ , .... ρ = ω A ( ∂eA ∂ρ A )s , ρ
B C N B, ρC , .... ρ N
(B33)
provides Use of this relation in Eq. (B25) leads to
A= N
∂Λ A= N ∂ ( ∂Λ ∂η ) ∂Λ N A= N
⎛ ∂e ⎞ ⎛ ∂e ⎞
∑ ρA A = ∑ ρA
η ω
+ρ
∂η
∑ ρ A2 ⎜ A ⎟
ρ
= ρ2 ⎜ ⎟ (B34)
A =1 ∂ A =1 ∂ A
ρ , T , ωB ( B ≠ A , N )
A =1 ⎝ ∂ ⎠ s , ρB , ρC , .... ρ N ⎝ ∂ρ ⎠ s , ρ B , ρC ,..., ρ N
(B20) and on the basis of the definition of the partial
Substitution of this result in Eq. (B17) provides the pressure, this takes the form
desired proof given by A= N
⎛ ∂e ⎞
A= N
∂Λ ∂Λ ∑ pA = ρ 2 ⎜ ⎟ (B35)
Theorem: ∑ ρA A = ρ (B21) A =1 ⎝ ∂ρ ⎠ s , ρB , ρC ,..., ρ N
A =1 ∂η ∂η We now define the total pressure according to [see
At this point we want to verify the relations Eq. (B2)]
contained in Eq. (B5), and we begin with the A= N
following representation of the partial pressure p= ∑p A (B36)
p A = ρ A2 ( ∂ eA ∂ ρ A )s , ρ , ρ ,.......
A =1
(B22)
B C which leads to
which can be summed over all species to obtain ⎛ ∂e ⎞
A= N A= N p = ρ2 ⎜ ⎟ (B37)
∑p
A =1
A = ∑ ρ (∂ e
A =1
2
A A ∂ ρ A )s, ρ
B , ρC ,..., ρ N
(B23) ⎝ ∂ρ ⎠ s , ρ B , ρC ,..., ρ N
At this point we have proved Eq. (B5).
Our objective now is to represent the right hand side
To complete this discussion we need to
of this result in terms of the total thermal energy. We indicate how this representation of the total pressure
begin with Eq. (B21)in the form is related to the classic description for equilibrium
A= N
⎛ ∂e ⎞ ⎛ ∂e ⎞
∑ ρA ⎜ A ⎟ = ρ⎜ ⎟ (B24) systems. If we represent the volume per unit mass as
A =1 ⎝ ∂ρ ⎠ s , ρB , ρC ,..., ρ N ⎝ ∂ρ ⎠ s , ρ B , ρC ,..., ρ N v =1 ρ (B38)
and multiply by the total density to obtain we see that Eq. (B37) leads to the following
A= N
⎛ ∂e ⎞ ⎛ ∂e ⎞ expression for the total pressure
∑ ρρ A ⎜ A ⎟
ρ
= ρ2 ⎜ ⎟ (B25) ⎛ ∂e ⎞
A =1 ⎝ ∂ ⎠ s , ρ B , ρC ,..., ρ N ⎝ ∂ρ ⎠ s , ρB , ρC ,..., ρ N p = −⎜ ⎟ (B39)
⎝ ∂v ⎠ s , ρB , ρC ,..., ρ N
The functional dependence of eA can be represented
In terms of thermo-statics (Truesdell, 1971), we
in terms of the mass fractions or the species densities consider a system at equilibrium having a mass m
as indicated by with the volume and internal energy given by
eA = eA ( ρ , s, ω A , ωB , ...., ωN −1 ) (B26a) V =mv, U =me (B40)
238 www.amidiq.org
Under these circumstances the equilibrium pressure M2

takes the classic form (Gibbs, 1928, page 33) given x A xB = ω AωB , A, B = 1, 2,..., N (C14)
M AM B
by
p = − ( ∂U ∂V )S
In order to develop a relation between the gradient of
(B41)
the mole fraction, ∇x A , and the gradient of the mass
Appendix C. Useful algebraic relations fraction, ∇ω A , for a binary system we begin with
Eq. (C13) and take the gradient to obtain
We begin by noting that the total mass density and ∇M M
total molar concentration for a N-component system ∇x A = ωA + ∇ω A , A = 1, 2,..., N (C15)
MA MA
are given by
In terms of binary systems, the gradient of the mean
ρ = ρ A + ρ B + ρC + ... + ρ N (C1a)
molecular mass takes the form
c = c A + cB + cC + ... + cN (C1b) ∇ M = ( ∇ x A ) M A + ( ∇ xB ) M B
The mass fractions and mole fractions take the form (C16)
= ( M A − M B ) ∇x A
ρ c
ωA = A , xA = A , and use of this result in the binary form of Eq. (C15)
ρ c (C2) provides
A = 1, 2,..., N ω M
and the constraints on these quantities are given by ∇x A = A ( M A − M B ) ∇x A + ∇ω A ,
MA MA (C17)
B=N B= N
∑ω
B =1
A =1, ∑x
B =1
A =1 (C3) A = 1, 2
Collecting terms leads to
The mean molecular mass is defined by
⎡ ω ⎤ M
M = x A M A + xB M B + xC M C + ... + xN M N (C4) ∇x A ⎢1 − A ( M A − M B ) ⎥ = ∇ω A ,
and multiplication by the total molar concentration ⎣ MA ⎦ MA (C18)
gives A = 1, 2
c M = cA M A + cB M B + cC M C + ... + cN M N (C5) which can be simplified to the form
The species molar concentration and the species ∇x A ( ω B M A + ω A M B ) = M ∇ω A ,
(C19)
mass density are related by A = 1, 2
ρ A = cA M A , cA = ρ A M A (C6) At this point we use Eq. (C13) to obtain
and the use of the first of these in Eq. (C5) provides ⎡x M M x M M ⎤
c M = ρ A + ρ B + ρC + ... + ρ N (C7) ∇x A ⎢ B B A + A A B ⎥ = M ∇ω A ,
⎣ M M ⎦ (C20)
Use of Eq. (C1a) allows us to express this result as
A = 1, 2
cM =ρ (C8)
which can be simplified to (Bird, 2009)
We can use Eq. (C1b) and the second of Eqs. (C6) to
M2
obtain ∇x A = ∇ω A , A = 1, 2 (C21)
ρA ρB ρC ρN M AM B
c= + + (C9) + ... + This result is Eq. (107) in the section on binary
M A M B MC MN
systems.
Dividing both sides by the total mass density
provides the following result Appendix D: Assumptions, restrictions and
ωA ωB ωC ωN constraints
c ρ= + + + ... + (C10)
M A M B MC MN
and on the basis of Eq. (C8) we have Throughout this paper we have imposed various
1 ω ω ω ω assumptions associated with the analysis. The most
= A + B + C + ... + N (C11) frequent of these concerned the total mass density
M M A M B MC MN and the total molar concentration, and an example
For a N-component system, the mole fraction of concerning the total mass density is given in Eq.
species A is given by (63). In engineering analysis there is a logical
c c M (ρ MA )M M sequence of events that begins with a simplifying
xA = A = A = A = ωA , assumption, or an idea, and leads to a theory with an
c ρ ρ MA (C12)
identifiable domain of validity. In this section we
A = 1, 2,..., N wish to illustrate this sequence of events with an
and in compact form we express this result as example from fluid mechanics where the path from
M an assumption to a constraint is well known
xA = ωA , A = 1, 2,..., N (C13) (Whitaker, 1988)
MA
A large class of fluid mechanical problems
For use in the Stefan-Maxwell equations we need the can be described by the continuity equation for
product form of this result that is given by incompressible flow
www.amidiq.org 239
∇⋅v = 0 (D1) Here λ is a unit tangent vector to a streamline and v

and the Navier-Stokes equations is the magnitude defined by
⎛ ∂v ⎞ v= v ⋅ v (D7)
ρ⎜ + v ⋅ ∇v ⎟ = − ∇p + ρ g + μ∇ 2 v (D2)
⎝ ∂t ⎠ The representation given by Eq. (D6) allows us to
The development of these two equations requires express the inertial term as
assumptions that may be supported by restrictions or ρ v ⋅ ∇v = vλ ⋅ ∇v (D8)
reinforced by constraints; however, we will simply in which λ ⋅ ∇ is known as the directional derivative
accept Eq. (D1) and Eq. (D2) without inquiring into (Stein and Barcellos, 1992, Sec. 14.7). The
their limitations. directional derivative can be expressed as
As an illustration of the development of d
assumptions, restrictions and constraints, we λ ⋅∇ = (D9)
ds
consider Eq. (D2) and assume that the convective
where s represents the arc length measured along a
inertial effects are negligible in order to obtain
streamline. Use of Eq. (D9) in Eq. (D8) provides the
∂v
ρ = − ∇p + ρ g + μ∇ 2 v (D3) following exact representation of the inertial term
∂t dv
This linear form can be easily solved for a wide ρ v ⋅ ∇v = ρ v (D10)
ds
variety of initial and boundary conditions, whereas
the general form given by Eq. (D2) represents a While this form is not often used in the development
difficult problem. It is important to clearly identify of solutions of the Navier-Stokes equations, it is
the assumption that leads one from Eq. (D2) to Eq. extremely useful in the development of an estimate
(D3) , and one way to express this is of the magnitude of the inertial term. To do so, we
Assumption: ρ v ⋅ ∇v = 0 (D4) need only think about how the velocity vector
changes as we proceed along a streamline and this
Equation (D4) indicates exactly what is being done
suggests that we define an inertial length, Lρ , by the
in a mathematical sense, but it is not necessarily a
precise description of the physics of any particular estimate
dv
fluid mechanical process. Strictly speaking, Eq. (D4) = O ( v Lρ ) (D11)
can only be true when the velocity vector is a ds
constant and this is not likely to occur in any real One should think of the inertial length as being the
flow. distance, along a streamline, over which significant
From a physical point of view, the changes in the velocity take place. A little thought
simplification of Eq. (D2) to Eq. (D3) is based on the will indicate that the estimate of Lρ requires an
idea that the convective inertial term, ρ v ⋅ ∇v , is
intuitive knowledge of the flow field, and this
negligible. This immediately raises the question:
intuitive knowledge is based primarily on a
“Negligible relative to what?” and one answer is that
knowledge of the no-slip condition. Use of Eq. (D11)
the convective inertial term is negligible relative to
allows us to estimate the inertial term as
the viscous term. This represents the second level in
dv
our process of simplification, and in this case we ρ v ⋅∇v = ρ v = O ( ρ v 2 Lρ ) (D12)
express our simplification as a restriction. ds
Restriction: ρ v ⋅∇v << μ ∇ 2 v (D5) and we need only develop an estimate of the viscous
In writing inequalities of this type, it is understood term in Eq. (D5) to complete our analysis.
that the comparison is being made between the We begin developing an estimate of the
absolute values of the vectors under consideration. If magnitude of the viscous term by expanding the
we apply the idea represented by Eq. (D5) to the Laplacian in rectangular, Cartesian coordinates to
Navier-Stokes equations, we again obtain Eq. (D3) obtain
provided we are willing to assume that small causes ∂2 v ∂2 v ∂2 v
∇ v= 2 + 2 + 2
2
(D13)
give rise to small effects (Birkhoff, 1960). Equation ∂x ∂y ∂z
(D5) has a definite advantage over Eq. (D4) since it In terms of order of magnitude estimates, we express
indicates what is required in a physical sense; this result as
however, neither Eq. (D4) nor Eq. (D5) indicate
when Eq. (D3) will be valid in terms of parameters ⎛ Δv ⎞ ⎛ Δv y ⎞ ⎛ Δv ⎞
∇ v = O⎜ 2 x ⎟ + O⎜ 2 ⎟ + O⎜ 2 z ⎟
2
(D14)
that are known a priori. In order to determine ⎜ L ⎟ ⎜ ⎟ ⎜ ⎟
⎝ x ⎠ ⎝ Ly ⎠ ⎝ Lz ⎠
precisely under what circumstances Eq. (D3) will be
valid, one must be able to estimate the magnitude of Here Δv x represents the change of v that takes place
the terms in Eq. (D2). over the distance Lx , and the meaning of Δv y and
We begin our analysis of the inertial term by
expressing the velocity in terms of a unit vector and Δv z is analogous for the y and z-directions. We now
the magnitude according to represent the largest of the three terms on the right
v = vλ (D6)
240 www.amidiq.org
hand side of Eq. as Δv L μ so that our estimate of concentration are small enough so that they can be
2
neglected. The general application of the Stefan-

the viscous term becomes
Maxwell equations requires that we replace c∇x A
∇ v = O ( Δv Lμ )
2 2
(D15)
with ∇c A and this leads to the restriction given by
We refer to Lμ as the viscous length and note that in Restriction: x A∇c << c∇x A (E2)
general it is quite different than the inertial length. If small causes give rise to small effects (Birkhoff,
Once again we note that knowledge of the no-slip 1960), the condition represented by Eq. (E2) will
condition is crucial for the determination of a reliable lead to the multi-component transport equation given
estimate of the viscous length. For a large class of by Eq. (204) and the binary form given by Eq. (124).
problems, Δv in Eq. (D15) is on the order of the In order to identify the conditions under
velocity itself because of the no-slip condition, i.e., which Eq. (E2) is valid, we need to express this
Δv ≈ v , because of the no-slip condition (D16) inequality in terms of parameters that are known a
and this allows us to estimate the viscous term in Eq. priori and to achieve this we follow the approach
(D5) as outlined in Appendix D. While that approach has led
μ∇ v = O ( μ v L2μ ) to an established success, the problem under
2
(D17)
consideration in this appendix is more difficult and
Use of this result, along with the estimate of the further study is in order.
inertial term, in the restriction given by Eq. (D5) As an example we consider the process
leads to the inequality illustrated in Fig. E1 in which the γ-phase represents
ρ v2 μv a flowing fluid. The κ−γ interface might be an
<< 2 (D18)
Lρ Lμ interface at which adsorption or desorption occurs, or
It is traditional to define the Reynolds number in an interface at which a catalytic reaction occurs, or
terms of the viscous length an interface at which mass transfer between the κ-
ρ vLμ phase and the γ-phase occurs. This could occur
Re = (D19) because the κ-phase is a porous catalyst phase or
μ because of a difference in the chemical potential of
and this allows us to express Eq. (D18) as a species A between the κ-phase and the γ-phase. The
constraint that takes the form direction of mean flow is indicated by the unit vector
Constraint: Re ( Lμ Lρ ) << 1 (D20) λ , and the direction orthogonal to the mean flow is
We refer to this as a constraint with the thought that indicated by the unit vector n .
the Reynolds number, the viscous length, and the Directing our attention to Eq. (E2) we express
inertial length will all be known, at least in an that result in the form
approximate sense. This means that the domain of Restriction: c −1 ∇c << x A−1 ∇x A (E3)
validity of Eq. (D3) is established by Eq. (D20), and And note that we need to consider the gradients in
this is something that is not done by either the the direction of flow (the λ -direction indicated in
assumption given by Eq. (D4) or the restriction Fig. 1E) and the direction orthogonal to the direction
given by Eq. (D5). of flow (the n -direction indicated in Fig. 1E). We
represent the two inequalities associated with Eq.
Appendix E: Constraints for constant total molar (E3) as
concentration
Restriction: c −1 λ ⋅∇c << x A−1 λ ⋅∇x A (E4a)
Throughout this paper we have imposed the Restriction: c −1 n ⋅∇c << x A−1 n ⋅∇x A (E4b)
condition of constant total molar concentration and and note that the second of these is likely to be the
constant total mass density in order to obtain most important restriction. We limit our analysis to
transport equations that could be used to determine ideal gases so that the equation of state is given by
the species molar concentration or the species mass pV = nRT , ideal gas (E5)
density. In general the assumption of constant total and the total molar concentration can be expressed as
molar concentration is associated with gas-phase
c = p RT (E6)
diffusion processes, and the assumption of constant
total mass density is associated with liquid-phase For an ideal gas, the left hand side of Eq. (E3) takes
diffusion processes. In this appendix we will treat the form
only the first of these two cases with the thought that c −1∇c = p −1 ∇p − T −1 ∇T (E7)
the second case can be explored on the basis of our and this yields two restrictions associated with Eq.
analysis of the first. (E3) that are given by
The assumption that the total molar Restrictions:
concentration is constant can be expressed as p −1 ∇p << x A−1 ∇x A , T −1 ∇T << x A−1 ∇x A (E8)
Assumption: c = constant (E1)
In this appendix we will consider only the first of
however, nothing is constant and what is meant by these restrictions and that requires an analysis of the
Eq. (E1) is that the variations of the total molar
www.amidiq.org 241
O ( ρ v t ∗ ) + O ( ρ v 2 Lρ )
1 1
p −1 λ ⋅ ∇p =
p p
(E15)
+ O ( ρ g ⋅ λ ) + O ( μ v Lμ )
1 1 2
p p
Here we have assumed that the no-slip condition is
valid at the γ−κ interface, thus Δv can be replaced
with v .
The magnitude of the pressure can be
estimated as
p = O(ρ C2 ) (E16)
where C is the speed of sound (Whitaker, 1981, Sec.
10.3). Use of this result in Eq. (E15) leads to
⎛ M ⎞ ⎛M2 ⎞
p −1 λ ⋅ ∇p = O ⎜ ∗ ⎟ + O ⎜
⎝Ct ⎠ ⎜ L ⎟⎟
⎝ ρ ⎠
Fig. E1. Mass transfer in a two-phase system (E17)
⎛M 1 ⎞
2 ⎛M2 1 ⎞
+O ⎜ + O⎜
Navier-Stokes equations. The second restriction ⎜ Fr L ⎟⎟ ⎜ Re L ⎟⎟
⎝ μ ⎠ ⎝ μ ⎠
requires an analysis of the thermal energy equation, in which the following dimensionless quantities have
and in many cases it would be appropriate to include been used:
the rate of chemical reaction and the heat of reaction.
M = Mach number = v C
We begin our analysis of the first of Eqs. (E8)
by considering the direction of the mean flow Fr = Froude number = v 2 g ⋅ λ Lμ (E18)
illustrated in Fig. E1. This leads to a restriction given Re = Reynolds number = ρ vLμ μ
by
Directing our attention to the right hand side of Eq.
Restriction: p −1 λ ⋅ ∇p << x A−1 λ ⋅ ∇x A (E9)
(E9) we estimate the gradient as
λ ⋅ ∇x A = O ( Δx A Lc )
and we can make use of Eq. (D2) to obtain
(E19)
1 ⎡ ∂ (v ⋅ λ ) ⎤
p −1 λ ⋅∇p = O ⎢ ρ
∂ t ⎥⎦
in which Lc represents the convective length scale
p ⎣
for the transport of species A. Use of this result along
1 1 with Eq. (E17) in Eq. (E9) leads to the constraint
+ O ⎡⎣ v ⋅ ∇( v ⋅ λ ) ⎤⎦ + O ⎡⎣ ρ g ⋅ λ ⎤⎦ (E10)
p p given by
1 ⎛ M ⎞ ⎛M2 ⎞ ⎛M2 1 ⎞
+ O ⎡⎣ μ∇ 2 ( v ⋅ λ ) ⎤⎦ O⎜ ∗ ⎟ + O⎜ ⎟ + O⎜
p ⎝Ct ⎠ ⎜ L ⎟ ⎜ Fr L ⎟⎟
⎝ ρ ⎠ ⎝ μ ⎠
The magnitude of the velocity in the direction of the (E20)
⎛M 1 ⎞
2
⎛ Δx A 1 ⎞
+O ⎜ << O ⎜
⎜ Re L ⎟⎟ ⎟
mean flow is given by
v = v ⋅λ (E11) ⎝ μ ⎠ ⎝ x A Lc ⎠
and this allows us to express Eq. (E10) in the form Since Δx A x A will be less than or on the order of
1 ⎛ ∂ v⎞ 1 one, a conservative representation of this constraint
p −1 λ ⋅ ∇p = O ⎜ ρ ⎟ + O ( ρ v ⋅ ∇v ) is given by
p ⎝ ∂t ⎠ p
(E12) ⎛ M Lc ⎞ ⎛ M 2 Lc ⎞ ⎛ M 2 Lc ⎞
+ O⎜ + O⎜
+ O ( ρ g ⋅ λ ) + O ( μ∇ v ) ⎟ ⎜ Fr L ⎟⎟
1 1 O⎜ ∗ ⎟
2
⎝ Ct ⎠ ⎜ L ⎟
p p ⎝ ρ ⎠ ⎝ μ ⎠
(E21)
Following the development given in Appendix D, we ⎛ M Lc ⎞
2
+O ⎜ << 1
estimate the inertial and viscous terms according to ⎜ Re L ⎟⎟
⎝ μ ⎠
ρ v ⋅∇v = O ( ρ v 2 Lρ ) , μ ∇ v = O ( μ Δv Lμ ) (E13)
2 2
For Mach numbers small compared to one, it will be

and we estimate the local acceleration as difficult to violate this constraint; however, one must
∂v ⎛ ρv ⎞ remember that this constraint is based on Eq. (E9)
ρ = O⎜ ∗ ⎟ (E14)
∂t ⎝ t ⎠ and we also need to consider the restriction given by
Restriction: p −1 n ⋅ ∇p << x A−1 n ⋅ ∇x A ((E22))
in which t ∗ is a characteristic process time. Use of
Eqs. (E13) and (E14) in Eq. (E12) leads to the To explore this restriction, we express Eq. (E10) in
following estimate for the pressure gradient in the the form
direction of the mean flow:
242 www.amidiq.org
1 ⎡ ∂ (v ⋅n ) ⎤ n ⋅ ∇x A = O ( Δx A LD ) ((E29))
p −1 n ⋅ ∇p = O ⎢ρ
p ⎣ ∂ t ⎥⎦ in which LD represents the diffusive length scale for
1 the transport of species A. Use of this result along
+ O ⎡ ρ v ⋅ ∇( v ⋅ n ) ⎤⎦ (E23)
p ⎣ with Eq. (E27) in Eq. (E22) leads to the constraint
⎛ M ⎞ ⎛M2 1 ⎞
1 1
+ O [ ρ g ⋅ n ] + O ⎡⎣ μ∇ 2 ( v ⋅ n ) ⎤⎦ O ⎜ ∗ ⎟ + O ⎡⎣ M 2 κ ⎤⎦ + O ⎜
⎝Ct ⎠ ⎜ Fr L ⎟⎟
p p ⎝ μ ⎠
(E30)
For the special case in which n is replaced by the ⎛M 1 ⎞
2
⎛ Δx A 1 ⎞
+ ⎜ << O ⎜
unit normal vector to a streamline, the inertial term ⎜ Re L ⎟⎟ ⎟
⎝ μ ⎠ ⎝ x A LD ⎠
takes the form (Whitaker, Sec.7.4, 1968)
Since Δx A x A will be less than or on the order of
ρ v ⋅∇( v ⋅ n ) = ρ v ⋅ ∇v ⋅ n = ρ v 2κ (E24)
one, a conservative representation of this constraint
in which κ is the curvature (Stein & Barcellos, Sec.
is given by
13.4, 1992). Here it is important to keep in mind that
n is a constant unit vector as indicated in Fig. E1 ⎛ M LD ⎞ ⎛ M 2 LD ⎞
+ O ⎡⎣ M 2 LD κ ⎤⎦ + O ⎜
⎜ Fr L ⎟⎟
O⎜ ∗ ⎟
while the unit normal to a streamline will be a ⎝ Ct ⎠ ⎝ μ ⎠
function of position. Equation (E24) helps us to (E31)
⎛ M LD ⎞
2
+ ⎜ << 1
⎜ Re L ⎟⎟
estimate the inertial term in Eq. (E23) as
1 1 ⎝ μ ⎠
O ⎡⎣ ρ v ⋅ ∇( v ⋅ n ) ⎤⎦ = ρ v 2 κ (E25)
p p Once again, it will be difficult to violate this
in which κ represents some appropriate mean constraint whenever the Mach number is small
curvature associated with the system illustrated in compared to one.
Fig. E1. About the other terms in Eq. (E23) we can At this point we have developed a constraint
only say that they will be smaller than the analogous associated with the restriction given by the first of
terms in Eq. (E10). This means that we can over Eqs. (E8); however, the second restriction given by
estimate p −1 n ⋅ ∇p as Restriction: T −1 ∇T << x A−1∇x A (E32)
still needs to be explored. This will require an
1 ⎡ ∂(v ⋅ λ ) ⎤ 1
p −1 n ⋅ ∇p = O ⎢ρ ⎥ + O ⎡⎣ ρ v κ ⎤⎦
2 analysis of the thermal energy equation and for most
p ⎣ ∂t ⎦ p realistic systems the thermal energy equation will be
1 1 coupled to a mass transfer and reaction process. In
+ O [ ρ g ⋅ n ] + O ⎡⎣ μ∇ 2 ( v ⋅ λ ) ⎤⎦ addition, the constraints associated with non-ideal
p p
gases need to be developed along with the
(E26) constraints associated with the assumption for liquid
and follow our earlier development to obtain phase mass transfer processes given by
⎛ M ⎞ Assumption: ρ = constant (E33)
p −1 n ⋅ ∇p = O ⎜ ∗ ⎟ + O ⎡⎣ M 2 κ ⎤⎦
⎝Ct ⎠ The constraints associated with this assumption will
(E27) be much more difficult to obtain than those given by
⎛M2 1 ⎞ ⎛M2 1 ⎞
+ O⎜ +
⎜ Fr L ⎟⎟ ⎜⎜ Re L ⎟⎟
Eqs. (E21) and (E31); however, the developments
⎝ μ ⎠ ⎝ μ ⎠ presented in this appendix should provide guidance
In this case the Froude number is defined by for the attack on Eq. (E33).
Fr = Froude number = v 2 g ⋅ n Lμ (E28)
Directing our attention to the right hand side of Eq.
(E22) we estimate the gradient as
www.amidiq.org 243

V8n3a1 PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

V8n3a1 PDF

Transféré par

Droits d'auteur :

Formats disponibles

Revista Mexicana de Ingeniería Química

Vol. 8, No. 3 (2009) 213-243

DERIVATION AND APPLICATION OF THE STEFAN-MAXWELL EQUATIONS

DESARROLLO Y APLICACIÓN DE LAS ECUACIONES DE STEFAN-MAXWELL

Department of Chemical Engineering & Materials Science

Received 5 of June 2009; Accepted 9 of November 2009

Keywords: continuum mechanics, kinetic theory, multicomponent diffusion.

Palabras clave: mecánica del continuo, teoría cinética, difusión multicomponente.

* Corresponding author. E-mail: whitaker@mcn.org 213

Publicado por la Academia Mexicana de Investigación y Docencia en Ingeniería Química A.C.

1.1 Conservation of mass 3

1.2 Laws of mechanics 5

2. Mass continuity equation 15

3. Molar continuity equation 16

4. Mixed-mode continuity equation 19

5.1 Mass diffusive flux 20

5.2 Molar diffusive flux 23

5.3 Mixed-mode diffusive flux 25

6. Special forms for N-component systems 26

6.1 Dilute solution diffusion equation 26

6.2 Dilute solution convective-diffusion equation using J*A 28

6.3 Dilute solution convective-diffusion equation using J A 30

6.4 Diffusion through stagnant species 31

7. General form for N-component systems: Constant total molar concentration 33

8. General form for N-component systems: Constant total mass density 37

Appendix A: Chemical reaction and linear momentum 46

Appendix B: Thermodynamic pressure 50

Appendix C: Algebraic relations 55

Appendix D: Assumptions, Restrictions and Constraints 58

Appendix E: Restrictions for constant total concentration and density 62

Our derivation of multi-component transport

1.1 Conservation of mass

If we sum Eq. (8) over all species and impose the d

Here we note that only N − 1 of the diffusive + ∫

1.2.1 Total momentum equation ⎛ ∂v ⎞ ⎡ A= N ⎤

and Eqs. (10) and (12) can be used to obtain ⎛ ∂u A ⎞

τ = μ ( ∇v + ∇vT ) + λ (∇ ⋅ v) I (41b) In Appendix A we show that difference between v∗A

Given that the behavior of the mixtures under ω A∇ ⋅ ∑ ρ Au Au A << ∇p A (48c)

In order to see how this result is related to the work ∇ ( p A p ) + p −1 ⎣⎡( p A p ) − ω A ⎦⎤ ∇p

Here we have explicitly identified the terms

The simplest approach for non-ideal mixtures is to A = 1, 2,..., N − 1

pA = xA p , ideal mixture (58) ∂ ρA

A = 1, 2, ..., N − 1 The total molar continuity equation is analogous to

determine all the species velocities, v A , v B , …, v N .

average velocity are defined by ∑ω

should be solved along with the N − 1 species v = v∗ + ∑ ωB u∗B (79)

species A is sufficiently small it is possible that the ∂ cA

B= N Here the diffusive flux is represented in terms of a

molar, and mixed-mode forms of the species ωA = ρ A ρ , xA = cA c ,

and we need to determine uA and v in order to solve M2

ω Au A + ωB u B = 0 (100) From Eqs. (105) we see that

1 x A xB Returning to Eq. (97), we make use of this form of

the mean molecular mass, M . These are given by

∂ ρA and employ the form of Eq. (76) for both species to

ρ A = c A M A , rA = RA M A (114) This is the molar analogy of Eq. (110), and

Species Molar: In order to obtain the so-called dilute solution form

5.3 Mixed-mode diffusive flux However, if we impose the condition

In this case we return to the mixed-mode species Assumption: ρ = constant (140)

ωB 6.1 Dilute solution diffusion

which allows us to express Eqs. (59) as N A = −c DAm ∇x A , A = 1, 2, ..., G < N (149)

compared to one, we have G = 3 and Eqs. (147) and

6.3 Dilute solution convective-diffusion equation

Separating the terms in the sum leads to