The Maxwell-Stefan Equations

ChEn 6603

Wednesday, February 1, 12

Outline
Diffusion in ideal, binary systems
Particle dynamics Maxwell-Stefan equations Ficks Law

Diffusion in ideal multicomponent systems

Example: Stefan tube Matrix form of the Maxwell-Stefan equations Ficks Law for multicomponent systems Reference velocities again

Wednesday, February 1, 12

Particle Dynamics
Conservation of momentum:
m1 (u1 uf 1 ) + m2 (u2 uf 2 ) = 0
For molecules, inelastic collisions are known by another name ... what is it?

T&K 2.1.1-2.1.2

u1

m2

Conservation of kinetic energy (elastic collision):

m1 (u2 1 u2 f 1) + m2 (u2 2 u2 f 2) =0

m1

u2

Solve for nal particle velocities:

uf 1 uf 2 = = u1 (m1 m2 ) + 2m2 u2 , m1 + m2 u2 (m2 m1 ) + 2m1 u1 m1 + m2

Momentum exchanged in a collision:

m1 (u1 uf 1 ) = =

m1 m1 u1 (u1 (m1 m1 + m2 2m1 m2 (u1 u2 ) . m1 + m2

m2 ) + 2m2 u2 ) ,

Sum of forces acting on particles of type 1 per unit volume

Rate of change of momentum of particles of type 1 per unit volume

Momentum exchanged per collision between 1 and 2

Rate of 1-2 collisions per unit volume

u1

u2

x1 x2
3

Wednesday, February 1, 12

u1

m2

Sum of forces acting on particles of type 1 per unit volume

m1

u2

Momentum exchanged per collision between 1 and 2

Rate of 1-2 collisions per unit volume

Assume:
System pressure is constant Collisions are purely elastic (kinetic energy is conserved in collisions) No shear stress (negligible velocity gradients)

dS

V(t)

x1 p a dS =
S(t )

Z Z

V (t )

r(px1 ) dV pr x 1 d V

V (t )

Why the (-) sign? f12: drag coefcient for drag that particle 1 feels as a result of interactions with particles of type 2 What is the binary diffusivity a function of? What about x2?

So our force (momentum) balance becomes:

px1

x1 x2 (u1

u2 ), u2 )
p = f12

= f12 x1 x2 (u1

Dene a binary diffusion coefcient as D12

x1 =
diffusion driving force for species 1
Wednesday, February 1, 12

x1 x2 (u1 u2 ) Maxwell-Stefan Equations for a binary, ideal mixture. D12

drag force on species 1 (resisting diffusion)

Ficks Law - Binary Ideal System

x1 = x1 x2 (u1 u2 ) D12
Maxwell-Stefan equations for a binary, ideal system at constant pressure.

rx1

= =

x2 N1 x1 N2 ct D12 x2 J1 x1 J2 ct D12

can you show this?

rx1 =

J1 ct D12
Ficks law for a binary, ideal system at constant pressure

J1 =

ct D12 rx1

Wednesday, February 1, 12

Re-Cap
12 can be interpreted as an inverse drag coefcient. 12 = 21 (symmetric due to momentum conservation) 12 depends on the characteristics of species 1 and 2 (molecule shapes, etc.), but not on their relative compositions. 12 may depend on temperature and pressure. We call 12 the Maxwell-Stefan diffusivity or Binary diffusivity. There are no 1-1 interactions here - 11 is not dened.

Wednesday, February 1, 12

Multicomponent Systems
Binary system: x1 =
x1 x2 (u1 u2 ) D12
u1 m2 u3

T&K 2.1.3-2.1.4

Ternary system: must consider 1-2, 1-3, and 2-3 interactions. x1 x2 (u1 u2 ) x1 x3 (u1 u3 ) x1 = D12 D13
x2 = x1 x2 (u2 u1 ) D12 x2 x3 (u2 u3 ) D23

m1

u2

m3

Multicomponent system: must consider i-j interactions.

rx i =
n X xi xj ( u i
j =1 j 6=i

uj )

in general...

ij

di =

Recall:
di

Ni = xi cui
n X xj Ni j 6=i n X j 6=i

Ji = Ni
xi Nj ,

xi cu

n X xi xj ( u i
j =1 j 6=i

uj )

What about i=j?

ij

di =

xi

= =

cDij

Assumptions:
Constant Pressure Ideal mixture (elastic collisions)
Conservation of translational energy. Where else could the energy go?
7

(so far)

xj Ji xi Jj cDij

Wednesday, February 1, 12

Example: Stefan Tube

Air

T&K Example 2.1.1

z=

Liquid Mixture

Species balance equations (no reaction): i i = = ni , t t cxi ci = = Ni t t At steady state (1D),

ni = i =
i

z=0

Given: ij, xi(z=0), xi(z=), =0.238 m, T=328.5 K Find xi(z)

Acetone (1), Methanol (2), Air (3) x1(z=0)=0.319, x2(z=0)=0.528

Ni

Convectiondiffusion balance...

From the Maxwell-Stefan equations:

dxi = dz

n X xj Ni j 6=i

xi Nj

ct ij

D12 D13 D23

= = =

8.48 13.72 19.91

mm2 /s mm2 /s mm2 /s

Wednesday, February 1, 12

A semi-analytic solution
Maxwell-Stefan Equations
1 d xi d

dxi = dz
n X xi Nj
j =1 j 6=i

n X xj Ni j 6=i

xi Nj

ct ij

Normalized coordinate:

z ,

d d d 1 d = = dz d dz d

x j Ni

ct ij

We need to eliminate xn from the equation so that we have unknowns x1 ... xn-1.
n 1 X j 6=i

n X Nj = xi ct ij j 6=i

x n Ni ct in

x j Ni ct ij

Eliminate xn by x = 1 n substituting:

n 1 X j =1

xj = 1

xi

n 1 X j 6=i

xj

1 dxi d

= dxi = d

n 1 n 1 n X X X xj Ni Nj Ni @ rearrange a bit, collecting A 1 xi xi xj , terms on xi, xj. ct ij ct in ct ij j 6=1 j 6=i j 6=i 0 1 n n 1 X X Nj A Ni Ni Ni Ni xi @ + + xj , move l over. ct in ct ij ct in ct in ct ij j 6=i j 6=i 0 1 n 1 n X X Nj A Ni Ni Ni @ Ni + xi + xj ct in / ct ij / ct in / ct ij / ct in / j 6=i | j 6=i {z } | {z } | {z } i ij
ii

almost there...

Wednesday, February 1, 12

dxi @ Ni Nj A Ni Ni Ni = + xi + xj d ct in / ct ij / ct in / ct ij / ct in / j 6=i j 6=i | {z } | {z } | {z } i ij

ii

n X

n 1 X

d (x) = [ ] ( x) + ( ) d

ii

= = =

A system of linear ODEs with constant coefcients (ct, Nj are constant)

ij

Ni Nk + , ct in / ct ik / k6=i 1 1 Ni , ct in / ct ij / Ni ct in /

n X

Analytic solution (assuming Ni are all constant)

see T&K 8.3 and Appendix B

(x) = exp [[ ] ] (x0 ) + [exp [[ ] ]]

Matrix exponential! exp[] [exp(ij)]!
In Matlab, use expm

[I ]

Note: if we had not eliminated the nth equation, we could not form the inverses required here.

ih i

()

Algorithm:
1. Guess Ni 2. Calculate [], () 3. Calculate (x) at =1 (z=l)
Wednesday, February 1, 12

4. If (xl) matches the known boundary condition, we are done. Otherwise return to step 1.

Note: we could also solve the equations numerically in step 3 and eliminate step 2 (work straight from the original Maxwell-Stefan equations)

Later in the course, we will show another way of getting Ni.

10

Matrix Form of Maxwell-Stefan Equations

di = =
n

T&K 2.1.5

n X xj Ni j 6=i n X j 6=i

xi Nj

cDij

Eliminate Jn from the set of n equations n-1 equations.

n 1 n 1

Jn =
j =1

Jj =
n X xj Ji j 6=i

Ji
j =1 j =i

Jj

xj Ji xi Jj cDij

di ct di

= =

xi Jj

di = 0
i=1

Easily shown for the case we have addressed thus far, di = xi.

ct Dij

,
Split the summation into individual terms.
Recall that we dont have a ii term!

Only n-1 of these equations are independent.

For a binary system, we have:
x1 x2 = = x2 N1 x1 N2 , ct D12 x1 N2 x2 N1 ct D21

n n X X xj Jj Ji + xi , Dij Dij j =1 j =1
j 6=i j 6=i

n n 1 X X xj Jj Jn Ji + xi + xi , Isolate the nth Dij Dij Din diffusive ux. j =1 j =1

j 6=i j 6=i

eliminated the nth diffusive ux

n n 1 X X xj Jj + xi Ji Dij Dij j =1 j =1

Show that these sum to zero.

Gather Ji and Jj terms Dene diagonal and off-diagonal matrix entries.

B xi Ji @ Din Bii Ji

j 6=i

j 6=i

n 1 X j 6=i

n n 1 X X xj C 1 + + x A i D Dij ij j =1 j =1
j 6=i j 6=i

X C xi B Jj A , @Ji + Din j =1
j 6=i

n 1

1 Din

Jj

Bij Jj
11

Wednesday, February 1, 12

ct di

xi Ji + Din Bii Ji

n
j =1 j =i

n 1 j =i

xj + xi Dij

n 1
j =1 j =i

1 Dij

1 Din

Jj ,

n-1 dimensional form of the Maxwell-Stefan equations

Bij Jj

n-1 dimensional matrix form:

ct (d) =

[B ](J)

Bii Bij

= =

xi Din xi

n xj + , Dij j =i

1 Dij
1 1

1 Din

ct

dn

d1 d2 . . .
1

B1,1 B2,1 . . . Bn
1 ,1

B1,2 B2,2 . . . Bn
1 ,2

.. .

B1,n B2,n . . . Bn

1,n 1

Jn

J1 J2 . . .
1

Note: we can write this in n-dimensional form, but then [B]-1 cannot be formed.

Wednesday, February 1, 12

12

Ficks Law
Maxwell-Stefan Equations (matrix form) Ficks Law (matrix form)

T&K 3.2

ct (d) =

[B ](J)

(J)

= =

ct [B ] (d) ct [D](x)
1

so far, di = xi.

Some Observations:
For an ideal gas mixture, the ij are largely independent of composition (but are functions of T and p), while the Dij are complicated functions of composition. The Fickian diffusion coefcients (Dij) may be negative, while ij 0. The binary diffusivity matrix is symmetric (ij = ji) but the Fickian diffusivity matrix is not symmetric (Dij Dji). Note that ii never enter in to any expression, and have no physical meaning. However, the Fickian Dii enter directly into the expression for the uxes, and represent the proportionality constant between the driving force and the diffusion ux for the ith component. ij are independent of reference frame. Dij is for a molar-averaged velocity reference frame.
Wednesday, February 1, 12 13

Binary/Ternary Comparison
n n

T&K 3.2

xi = 1
i=1 i=1

xi = 0
Ternary Diffusion

Binary Diffusion

x3 = 1

x1

x2

x1
z=0

x2 = 1

x1

x1

x2 = 1
z=0 z=L

x1

x3

z=L

x2 =
Wednesday, February 1, 12

x1

x1 =

x2

x3
14

Diffusion Regimes
(J ) = ct [D](rx)
aw

Binary Diffusion J1
's L Fic k
diffusion barrier

Ternary Diffusion J1
reverse diffusion

's k ic

w a L

D12

osmotic diffusion

x1

"normal" diffusion

x1

J1 =

ct Dx1

J1 J2

= =

ct D11 x1 ct D21 x1

ct D12 x2 , ct D22 x2 .
15

Wednesday, February 1, 12

Multicomponent Effects
(J) =
J1 J2 . . .
1

ct [B ]
ct

(d)
D1,2 D2,2 . . . .. .
1,2

Ternary Diffusion J1
reverse diffusion

k' c i F

aw L s

Jn

D1,1 D2,1 . . . Dn
1 ,1

D1,n D2,n . . . Dn

1 1

Dn

1,n 1

d1 d2 . . . dn

diffusion barrier

osmotic diffusion

"normal" diffusion

x1

For multicomponent effects to be important, Dijxj must be signicant compared to Diixi.

Dij xj Dii xi |Dij /Dii | xj = 0

O(1) O(1)
16

Wednesday, February 1, 12

Ficks Law & Reference Velocities

How do we write Ficks law in other reference frames?
(J) =
(j) =
(JV ) =
t [D

T&K 3.2.4, 1.2.1

c[D](x)
[DV ](c)

Molar diffusive ux relative to a molar-averaged velocity. Mass diffusive ux relative to a mass-averaged velocity. Molar diffusive ux relative to a volume-averaged velocity.

]( )

Option 1: Option 2:

Start with GMS equations and write them for the desired diffusive ux and driving force. Then invert to nd the appropriate denition for [D]. Given [D], dene an appropriate transformation to obtain [D] or [DV].
uo Bik ou Bik

[D ] = [B uo ] 1 [ ][x] 1 [D][x][ ] 1 [B uo ] = [B ou ][ ][x] 1 [D][x][ ] 1 [B ou ] 1

= =

ik

xk k

ik

i 1

xn n n xk xn k

[D ] = [B
V

Vu

][D][B

Vu

Vu Bik uV Bik

= =

ik

= [B V u ][D][B uV ]
Wednesday, February 1, 12

ik

xi V k Vn Vt Vk xi 1 Vn
17

T&K Example 3.2.1

Given [DV] for the system acetone (1), benzene (2), and methanol (3), calculate [D].
(J) = c[D](x)
1

T&K 3.2.3-3.2.4

(j) =
x1 0.350 0.766

t [D ]( )
x2 0.302 0.114 0.790 0.500 0.150 0.548 0.795 0.132 0.298
V D11

1 = 74.1 106 V V 2 = 89.4 106 V

(J ) =
V D12

Diffusivities in units of 10 m /s

3 = V

m3 mol m3 mol , m3 40.7 106 mol -9 2 V D21

[DV ](c)
V D22

[DV ] = [B V u ][D][B V u ] = [B
1 V 2 V 3 V = 74.1 = 89.4 = 40.7
Vu

][D][B
6

uV

3.819 4.440 4.472 4.434 3.192 3.513 3.502 3.115 3.050

0.420 0.721 0.962 1.866 0.277 0.665 1.204 0.138 0.150

-0.561 -0.834 -0.480 -0.816 -0.191 -0.602 -1.130 -0.227 -0.269

2.133 2.680 2.569 1.668 2.368 1.948 1.124 2.235 2.250

18

10 10 10

m3 mol m3 mol m3 mol

0.533 0.400 0.299 0.206 0.102 0.120 0.150

Wednesday, February 1, 12