You are on page 1of 15

ChBE 6260 Topic 2

Diffusion Coefficients of Gases in Different Environments


〈𝑅2 〉 1
𝐷𝑥 = = 𝑓𝐿2 is a starting point
2𝑡 6

Let’s establish some rules of thumb for diffusivities, which are often the parameters you need to
estimate
“Bulk” [Gas A diffusing in Gas B] Transition- Gas A diffusing in other gases and
𝑐𝑚2 frequently striking walls between gas collisions
𝐷𝐴𝐵 ~0.1 − 0.9 𝑎𝑡 1 𝑎𝑡𝑚 25°𝐶 𝑐𝑚2
𝑠 −2
𝐷𝐴 ~10 − 0.1
𝑠
25°𝐶, 1 𝑎𝑡𝑚

“Knudsen” Gas A undergoes free flight with Surface- Gas A moves along a solid surface as an
other gas between its collisions with pore walls adsorbed species in small 2D hops
𝑐𝑚2
𝐷𝐴,𝐾 ~10−5 − 10−2 𝑐𝑚2
𝑠 𝐷𝐴,𝑠 = 10 −6
− 10 −3
25 °𝐶 𝑃 ≪ 1 𝑎𝑡𝑚 𝑠

Liquid and Solid- Gas undergoes small hops from one sorbed cage to another when the medium
opens a gap next to the sorbed gas
𝑐𝑚2
𝐷𝐴 = 10−15 − 10−5
𝑠

Page 1 of 15
ChBE 6260 Topic 2

Starting with gases allows us to develop estimates of diffusion coefficients from kinetic theory.
We will first illustrate and rationalize the relationship between diffusivity, molecular velocity,
and free path of the gas, i.e.,
̅𝜆
𝐷𝛼𝑈
̅ = 𝑀𝑒𝑎𝑛 𝑚𝑜𝑙𝑒𝑐𝑢𝑙𝑎𝑟 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝜆 = 𝑚𝑒𝑎𝑛 𝑓𝑟𝑒𝑒 𝑝𝑎𝑡ℎ
𝑈
We will also show how “random walk” in the presence of thermodynamic driving forces
produces net motion in a certain direction.
Start with the following idealization:

➢ These planes all contain molecules


➢ Molecules all move at arithmetic average
kinetic speed:
̅ = √ 8𝑘𝑇
➢ 𝐶1 (−𝜆) > 𝐶1 (𝜆) 𝑈
𝜋𝑀𝑤
➢ Boltzmann (from Maxwellian
Distribution)
➢ All molecules move in one of six
coordinate directions with equal
likelihood; “off perpendicular” angles are
neglected

𝜆 is the mean free path


➢ Distance a molecule travels before making a collision
Since 𝜆 is the distance our molecules move at 𝑈̅ the molecules in plane “A” that have a +𝑧̂
1
component of velocity ( of total ) will reach plane “0” right as they have a collision and change
6
velocity to some uncorrelated direction. Therefore, the number of molecules arriving at plane “0”
𝜆
at ∆𝑡 = ̅
from plane “A” with +z velocity is:
𝑈

1
𝑁𝑧+ = 𝑈̅𝐶 (−𝜆)
6 1
𝜆
Similar arguments indicate that the number of molecules arriving at “0” at ∆𝑡 = ̅
from plane
𝑈
“B” with −𝑧̂ velocity vector is:
1
𝑁𝑧− = 𝑈̅𝐶 (𝜆)
6 1
Finally, the net flux crossing “0” in the +𝑧̂ direction at 𝑡 = ∆𝑡 (or any time at steady state)
1
𝐽𝑧 = 𝑁𝑧+ − 𝑁𝑧− = ̅[𝐶1 (−𝜆) − 𝐶1 (𝜆)]
𝑈
6

Page 2 of 15
ChBE 6260 Topic 2

̅ ∆𝐶1
𝑈
= [ ∆𝑧]
6 ∆𝑧
̅ 𝜕𝐶1
𝑈
= [ 2𝜆]
6 𝜕𝑧
𝜆𝑈̅ 𝜕𝐶1 𝜆𝑈̅
𝐽𝑧 = − 𝑎𝑛𝑑 𝐷 =
3 𝜕𝑧 3
Let’s use this as a starting point to describe diffusion from a microscopic level
𝒄𝒎𝟐
Bulk Gas Diffusion (𝟎. 𝟏 − 𝟎. 𝟗 𝒂𝒕 𝟏 𝒂𝒕𝒎 𝒂𝒏𝒅 𝟐𝟓°𝑪)
𝒔

➢ Gas diffusing in other gases 1


Recall: 𝐷 = 𝑓𝐿2
6
➢ The jump frequency is
determined by the collision 𝑓: The jump frequency in any of the
frequency between gas 6 coordinate directions
molecules. Jump length is the
𝐿: the length of an average
mean free path
differential step "𝑑𝐴 " or the collision
diameter. “Can be thought of as a
Leonard Jones or some other
diameter”

𝑘𝑇
𝐿=𝜆=
√2𝜋𝑑𝐴2 𝑃
̅
𝜆𝑈
Combining jump length and frequency gives 𝐷 = ̅ = √ 8𝑘𝑇 )
(Recall that 𝑈
3 𝜋𝑀𝑤

Stick all these together and you get:


1 3
2 𝑘3 2 𝑇2
𝐷𝐴𝐴 = [ 3 ] [ ] 𝑤𝑖𝑡ℎ ± 5% 𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦
3 𝜋 𝑀𝑤𝐴 𝑃𝑑𝐴2

𝐷𝐴𝐴 = 𝐴 𝑑𝑖𝑓𝑓𝑢𝑠𝑖𝑛𝑔 𝑖𝑛 𝐴
𝐷𝐴𝐵 = 𝐴 𝑑𝑖𝑓𝑓𝑢𝑠𝑖𝑛𝑔 𝑖𝑛 𝐵

1 3 1
2 𝑘3 2 𝑇2 1 1 2
𝐷𝐴𝐵 = [ 3] 1 [ + ]
3 𝜋 2𝑀𝑤𝐴 2𝑀𝑤𝐵
𝑑𝐴 + 𝑑𝐵 2
[𝑃 ( 2 ) ]

Page 3 of 15
ChBE 6260 Topic 2

3
1.49∗10−3 𝑇 2
Here are some simplifications, 𝐷𝐴𝐴 = 1
2 𝑃𝑑 2
𝑀𝑤𝐴 𝐴

1
3 1 1 2
4.2 ∗ 10−3 𝑇 2 [
𝑀𝑤𝐴 𝑀𝑤𝐵 ]
+
𝐷𝐴𝐵 =
[𝑑𝐴 + 𝑑𝐵 ]2 𝑃
𝑐𝑚2
𝐷[=] , 𝑃 [=] 𝑎𝑡𝑚, 𝑑𝐴 [=] Å,
𝑠
𝑔
𝑇 [=] °𝐾, 𝑎𝑛𝑑 𝑀𝑤[=]
𝑚𝑜𝑙
These are accurate at low pressures (< 10 atm)
Another approach: Hirshfelden-Curtis-Bird (Gas Mutual Diffusion Coefficients)
3 1
𝑐𝑚2 1.85 ∗ 10−3 𝑇 2 1 1 2
[=] 𝐷𝐴𝐵 = 2 ( + )
𝑠 𝑃𝜎𝐴𝐵 Ω𝐷 𝑀𝑤𝐴 𝑀𝑤𝐵
𝑔
𝑃 [=] 𝑎𝑡𝑚 𝜎𝐴𝐵 [=] Å 𝑇 [=] °𝐾 𝑀𝑤[=] Ω = 𝐶𝑜𝑙𝑙𝑖𝑠𝑖𝑜𝑛 𝐼𝑛𝑡𝑒𝑔𝑟𝑎𝑙
𝑚𝑜𝑙 𝐷
Ω𝑑 𝑖𝑠 𝑏𝑎𝑠𝑒𝑑 𝑜𝑛 𝜖 𝑎𝑛𝑑 𝑘𝑇
Direct Table entries in Table 2.3
𝜎𝐴𝐵 12 𝜎𝐴𝐵 6
𝜙𝐴𝐵 (𝑟) = 4𝜖𝐴𝐵 [( ) −( ) ]
𝑟 𝑟
𝜎𝐴𝐵 = 𝑐𝑜𝑙𝑙𝑖𝑠𝑖𝑜𝑛 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟

𝜎𝐴 + 𝜎𝐵
𝑟 = 𝑀𝑜𝑙𝑒𝑐𝑢𝑙𝑎𝑟 𝑠𝑝𝑖𝑛 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝜎𝐴𝐵 =
2

𝜖𝐴𝐵 ~√𝜖𝐴 𝜖𝐵

1
𝑚3
For estimates 𝜎(Å)~8.33𝑉̂𝐶3 𝑉̂𝐶 [=]
𝑚𝑜𝑙
1
𝑜𝑟, 𝜎(Å)~11.8𝑉̂𝑏3 (𝑉̂𝑏 𝑎𝑛𝑑 𝑇𝑏 , 𝑠𝑎𝑚𝑒 𝑢𝑛𝑖𝑡𝑠)
𝜖 𝜖
𝑎𝑛𝑑, ~ − .75𝑇𝐶 𝑜𝑟 ~1.21𝑇𝑏
𝑘 𝑘
This is a good approach for estimating 𝐷𝐴𝐵 , but the range of values at a fixed temperature is so
𝑐𝑚2
small (e.g., 0.1 − 0.9 ) that good accuracy is to be expected.
𝑠

Page 4 of 15
ChBE 6260 Topic 2

Gas diffusion may also occur inside systems with boundaries. The gas molecules will experience
a collision with the boundary and will be redirected at a random angle.
These types of collisions have significant effect on the gas diffusivity
(Note: Molecular diffusion can also be considered bulk diffusion)

Knudsen Diffusion
Gas undergoes free flight with no intermolecular
collisions between collisions with walls

−5 −2
𝑐𝑚2
10 − 10 𝑎𝑡 25°𝐶 𝑃 ≪ 1 𝑎𝑡𝑚
𝑠
30.78𝑇
𝜆=
𝑃𝑑 2
𝜆
≫1
2𝑟𝑝

𝜆 𝑖𝑠 𝑢𝑠𝑢𝑎𝑙𝑙𝑦 ~500 − 1000 Å 𝑎𝑡 1 𝑎𝑡𝑚 𝑎𝑛𝑑 25°𝐶

Knudsen processes can control for 100-1000 Å pores


The “jump frequency” is determined by the collision frequency of the free-flying molecules and
the jump length is approximately equal to 2𝑟𝑝 , so:
1
1 2 𝑇 2
𝐷𝐴,𝐾 = 𝑈̅𝐿 = 𝑈̅𝑟𝑝 = 9.7 ∗ 10−5 [ ] 𝑟
3 3 𝑀𝑤 𝑝
𝑐𝑚2
𝐷𝐴,𝐾 [=] 𝑎𝑛𝑑 𝑟𝑝 [=] Å
𝑠
Such pores are often in an interconnected network so the “effective” diffusivity can be written
as:
𝜖
𝐷𝐴,𝐾,𝑒 = 𝐷𝐴,𝐾 𝑤ℎ𝑒𝑟𝑒 𝜖 = 𝑝𝑜𝑟𝑜𝑠𝑖𝑡𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑜𝑙𝑖𝑑 𝑎𝑛𝑑 𝜏 𝑖𝑠 𝑡ℎ𝑒 tortuosity
𝜏
𝑟𝑝 can be estimated using various porosimetry methods (e.g. H2, N2, Hg, TEM, etc.)

For random, disordered porous 𝑙 = 𝑠𝑎𝑚𝑝𝑙𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠


solids, a useful rule of thumb is:
1
𝜖~
𝜏
𝐿 𝑇 = 𝐴𝑐𝑡𝑢𝑎𝑙 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑝𝑜𝑟𝑒 𝑙𝑒𝑛𝑔𝑡ℎ
Page 5 of 15 𝐿𝑇
𝜏= →6>𝜏>1
𝑙
ChBE 6260 Topic 2

In a pinch, you can use the following equation


2𝜖
𝑟𝑝 = 𝑤ℎ𝑒𝑟𝑒 𝜌𝑔 = 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑜𝑓 𝑛𝑜𝑛𝑝𝑜𝑟𝑜𝑢𝑠 𝑠𝑜𝑙𝑖𝑑
𝜌𝑔 𝐴̂𝑠

𝑚2
𝑎𝑛𝑑 𝐴̂𝑠 𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝑠𝑢𝑟𝑓𝑎𝑐𝑒 𝑎𝑟𝑒𝑎 𝑜𝑓 𝑝𝑜𝑟𝑜𝑢𝑠 𝑠𝑜𝑙𝑖𝑑 𝑖𝑛 , 𝑑𝑒𝑡𝑒𝑟𝑚𝑖𝑛𝑒𝑑 𝑏𝑦 𝐵𝐸𝑇
𝑔
Transition Gases- Gas diffuses in bulk, but also frequently hits walls
As a rule: If,
𝜆 > 10𝑟𝑝 → 𝐾𝑛𝑢𝑑𝑠𝑒𝑛

𝜆 < 0.1𝑟𝑝 → 𝐵𝑢𝑙𝑘 𝑜𝑟 𝑀𝑜𝑙𝑒𝑐𝑢𝑙𝑎𝑟


𝜆
𝐼𝑓 ~1, both mechanisms must be considered
2𝑟𝑝

𝑐𝑚2
𝐷𝐴 ~ 10−2 − 10−1 is typical. The presence of
𝑠
additional collisions with the wall (in addition to
those in the gas phase) reduces the effective “L” in
̅
𝜆𝑈
the expression: 𝐷 = , so even if 𝑈 ̅ is not affected,
3
𝐷𝐴,𝑒 , the effective diffusion coefficient of A in the
pore is reduced below its bulk value. The added
hindrance is generally handled by the following
equation:
1 1 1
= +
𝐷𝐴,𝑒 𝐷𝐴,𝐾,𝑒 𝐷𝐴,𝐵,𝑒

Surface Diffusion: A complex mechanism that is often superimposed on bulk and Knudsen
processes
Surface transport is often important when a high 𝑇𝑐 gas (e.g. Kr) is present.

Page 6 of 15
ChBE 6260 Topic 2

Liquid Systems: Molecular size – not molecular weight – is the dominant factor, and the
penetrant molecule waits in its “cage” for a transient gap to open next to it, thereby allowing a
jump. Similar case for polymers!

The Wilke-Chang correlation (for simple liquids) is qualitatively built on this physical picture,
and is still about the best we can do:
1
°
[𝜉𝐵 𝑀𝐵 ]2 𝑇 𝑐𝑚2
𝐷𝐴𝐵 = 1.17 ∗ 10−9 [ ]
𝑉̂𝐴0.6 𝜇𝐵 𝑠

𝑘𝑔 𝑚3
𝑀𝐵 [=] 𝑉 [=] 𝑇[=] °𝐾 𝜇𝐵 [=]𝑐𝑃
𝑘𝑔𝑚𝑜𝑙 𝐴 𝑘𝑔𝑚𝑜𝑙
𝜉𝑏 (𝑊𝑖𝑙𝑘𝑒 − 𝐶ℎ𝑎𝑛𝑔) = 1 𝑓𝑜𝑟 𝑛𝑜𝑛𝑝𝑜𝑙𝑎𝑟, 1.5 𝑓𝑜𝑟 𝑒𝑡ℎ𝑎𝑛𝑜𝑙, 1.9 𝑓𝑜𝑟 𝑚𝑒𝑡ℎ𝑎𝑛𝑜𝑙, 2.6 𝑓𝑜𝑟 𝑤𝑎𝑡𝑒𝑟
As 𝜇𝐵 decreases, the likelihood of opening a gap increases, and as 𝑉̂𝐴 decreases, the likelihood
that the gap will be big enough to allow “A” to jump through increases.
Hydrodynamic Interpretation of D:
This approach will help us develop diffusion relationships outside of the infinitely dilute regime.
Various potentials are useful for driving movement of different components with respect to the
center of mass velocity (chemical, electrical, thermal, magnetic, etc.).
Formally the negative gradient of a potential corresponds to a force. If we imagine a sphere
“creeping” through a medium without bulk flow, we can imagine a situation where the chemical
potential force is offset by the drag force (or when sphere drift velocity reaches a constant value).

(𝐹𝑖𝑧 )𝑐 = (𝐹𝑖𝑧 )𝑑
𝜕𝜇 1
(𝐹𝑖𝑧 )𝑐 = − ∗
𝜕𝑧 𝑁𝐴𝜋
𝑁𝐴𝜋

The hydrodynamic drag force is given by Stokes Law:

Page 7 of 15
ChBE 6260 Topic 2

(𝐹𝑖𝑧 )𝑑 = 3𝜋𝜇𝑈𝑖𝑧 𝑑 𝑤ℎ𝑒𝑟𝑒:


𝜇 = 𝑣𝑖𝑠𝑐𝑜𝑠𝑖𝑡𝑦, 𝑈𝑖𝑧 = 𝑑𝑟𝑖𝑓𝑡 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑜𝑓 𝑖 𝑖𝑛 𝑡ℎ𝑒 𝑧 𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑜𝑛, 𝑎𝑛𝑑 𝑑 = ℎ𝑦𝑑𝑟𝑜𝑑𝑦𝑛𝑎𝑚𝑖𝑐 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟

For small, molecular sized penetrants, the concept of a surrounding fluid flowing past the sphere
with its continuum viscosity is replaced with a “frictional coefficient”, 𝜁𝑖𝑗 , characteristic of the ij
mixture. Thus,

(𝐹𝑖𝑧 )𝑑 = 𝑈𝑖𝑧 𝜁𝑖𝑗

We can relate the chemical potential of the component in the solution to its concentration
𝜕𝜇𝑖 𝑅𝑇 𝐶𝑖 𝜕 ln 𝛾𝑖 𝜕𝐶𝑖
= (1 + )
𝜕𝑧 𝐶𝑖 𝜕𝐶𝑖 𝜕𝑧
𝐶𝑖 𝜕 ln 𝛾𝑖
= 0 𝑓𝑜𝑟 𝑖𝑑𝑒𝑎𝑙 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠
𝜕𝐶𝑖
𝑅𝑇 𝜕𝐶𝑖 1
Thus, 𝑈𝑖𝑧 𝜁𝑖𝑗 = − ∗
𝐶𝑖 𝜕𝑧 𝑁𝐴𝜋

𝑅𝑇 𝜕𝐶𝑖 𝑅𝑇 𝑘𝑇
𝐶𝑖 𝑈𝑖𝑧 = − = 𝐽𝐴∗ → 𝐹𝑜𝑟 𝑖𝑑𝑒𝑎𝑙 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛𝑠: 𝐷 = , 𝑆𝑡𝑜𝑘𝑒𝑠 𝑙𝑎𝑤 𝑙𝑖𝑚𝑖𝑡: 𝐷 =
𝑁𝐴𝜋 𝜁𝑖𝑗 𝜕𝑧 𝑁𝐴𝜋 𝜁𝑖𝑗 3𝜋𝜇𝑑

𝑅𝑇 𝐶𝑖 𝜕 ln 𝛾𝑖 (𝐶𝑖 )
𝐹𝑜𝑟 𝑛𝑜𝑛 𝑖𝑑𝑒𝑎𝑙 𝑐𝑎𝑠𝑒𝑠: 𝐷 = [1 + ]
𝑁𝐴𝜋 𝜁𝑖𝑗 𝜕𝐶𝑖
1 𝐶𝑖 𝜕 ln 𝛾𝑖 (𝐶𝑖 )
Define a “mobility” factor, Μ𝑖 = → 𝐷𝑖 = Μ𝑖 𝑅𝑇 [1 + ]
𝜁𝑖𝑗 𝑁𝐴𝜋 𝜕𝐶𝑖

This can be written in terms of mole fractions as well:


𝑥𝑖 𝜕 ln 𝛾𝑖 (𝑥𝑖 )
𝐷𝑖 = Μ𝑖 𝑅𝑇 [1 + ] **
𝜕𝑥𝑖

The “frictional factor”, 𝜁𝑖𝑗 , should be equivalent for I acting on j and j acting on I so that
𝜁𝑖𝑗 = 𝜁𝑗𝑖 at some local composition.

However, one would expect that there should be some variation between (𝜁𝑖𝑗 )𝑥 →0 & (𝜁𝑖𝑗 )𝑥 →1 .
𝑖 𝑖
A common way to account for these composition-dependent friction factors is to use a Vignes-
type relation.
° 𝑥𝐵 ° 𝑥𝐴
𝑅𝑇 (𝐷𝐴𝐵 𝜇𝐵 ) (𝐷𝐵𝐴 𝜇𝐴 )
= (𝐷𝐴𝐵 )𝑖𝑑𝑒𝑎𝑙 = where ° shows infinite dilution
𝜁𝑖𝑗 NAπ 𝜇𝑚𝑖𝑥

° ° 𝑥𝐵 𝑥𝐴
Usually 𝜇𝑚𝑖𝑥 is not available so: 𝐷𝐴𝐵,𝑖𝑑𝑒𝑎𝑙 ≈ (𝐷𝐴𝐵 ) (𝐷𝐵𝐴 ) at some 𝑥𝐴
°
You can use Wilke-Chang to estimate 𝐷𝐴𝐵 , or you can use the Sitaraman equation;

Page 8 of 15
ChBE 6260 Topic 2

1 0.93
0.5 ̂ 3
°
𝑀𝑤𝐵 ∆𝐻𝐵 𝑇 𝑐𝑚2
𝐷𝐴𝐵 = 1.679 ∗ 10−9 ( 1 ) [ ]
𝑠
𝜇𝐵 𝑉̂𝐴 ∆𝐻
2 ̂𝐴
0.3

𝑚3 𝐽
𝜇 [=] 𝑐𝑃 𝑇[=] °𝐾 𝑉̂𝐴 [=] ̂𝑖 [=]
𝐻 𝑎𝑡 𝑇𝑏 ,
𝑘𝑚𝑜𝑙 𝑘𝑔
°
Obtain 𝐷𝐵𝐴 simply by rotating subscripts
𝑐𝑚2
➢ For most low Mw mixtures of A & B (i.e. MW<300 Da) 𝐷𝐴𝐵 ~0.5 ∗ 10−5 − 5 ∗ 10−5
𝑠

We can finally use this information in parallel with ** to get a concentration-dependent


diffusivity:
𝑥𝐴 𝜕 ln 𝛾𝐴 (𝑥𝐴 )
𝐷𝐴𝐵 = (𝐷𝐴𝐵 )𝑖𝑑𝑒𝑎𝑙 [1 + ] ***
𝜕𝑥𝐴

Hines & Maddox gives an example of this effect for


ethanol/water:

°
𝑐𝑚2 ° 𝑐𝑚2
𝐷𝐸𝑊 = 1.28 ∗ 10−5 𝐷𝑊𝐸 = 1.13 ∗ 10−5
𝑠 𝑠
𝛾𝐸 𝑥𝐸 1 + 𝜕 ln 𝛾𝐸 /𝜕 ln 𝑥𝐸 𝑐𝑚2
𝐷𝐸𝑊 ( )
𝑠
1.95 0 - 1.28 ∗ 10−5
1.45 0.1 0.754 0.95 ∗ 10−5
1.27 0.3 0.751 0.93 ∗ 10−5
1.10 0.5 0.756 0.91 ∗ 10−5
1.00 1.0 - 1.13 ∗ 10−5

°
In cases where 𝛾𝐴 < 1 we see the opposite case with respect to 𝑥𝐴 , so that (𝐷𝐴𝐵 )𝑥𝐴 =0.5 > 𝐷𝐴𝐵

Except for very careful work, the concentration dependence of liquid mixtures can usually be
disregarded. In rare cases when comparing A-B & B-C, and 𝛾𝐴 < 1 in one system and > 1 in the
° °
other, ±2𝑥 (±100%) may result if both have similar 𝐷𝐴𝐵 = 𝐷𝐵𝐴 (pretty rare!)

Page 9 of 15
ChBE 6260 Topic 2

Diffusion of gases in polymers and other solids:


➢ No reliable ‘a priori’ predictions exist for 𝑓, 𝐿, 𝑜𝑟 𝐷
➢ However a generalized correlation does exist for amorphous polymers

Rubbery

Glassy

Page 10 of 15
ChBE 6260 Topic 2

Single File Diffusion


➢ If the pore size continues to decrease to the point that guest molecules cannot pass each
other, we arrive in a regime known as “single file diffusion”.
➢ In this regime, diffusive jumps can only occur if there is a “site” (or space) open for the
diffusing molecule.

Relevant to:
Ion channels in biological membranes
Catalytic reactions in zeolites
Drug or small molecule delivery

Set τ to be the mean time between subsequent jump attempts. Jumps occur equally probably in
either direction.

1−𝜃 2 𝑡 𝑓𝑖𝑙𝑙𝑒𝑑 𝑠𝑖𝑡𝑒𝑠


〈𝑥 2 〉 = 𝜆2 ( ) √ √ 𝑤ℎ𝑒𝑟𝑒 𝜃 =
𝜃 𝜋 𝜏 𝑇𝑜𝑡𝑎𝑙 𝑠𝑖𝑡𝑒𝑠

A “ mobility” factor in single file diffusion is defined as:


1−𝜃 1
𝐹 = 𝜆2 ( )
𝜃 √2𝜋𝜏
𝜋𝐹 2
𝑆𝑜 〈𝑥 2 〉 = 2𝐹 √𝑡 𝑎𝑛𝑑 𝐷𝑆𝐹 =
𝑙2

Page 11 of 15
ChBE 6260 Topic 2

Activated Zone Diffusion:


Similar to diffusion in transient “holes” in liquid systems, diffusion in solids is an activated
process
𝐸𝐷
𝐷 = 𝐷𝑜 exp [− ]
𝑅𝑇
In many solid systems, the “guest” molecule is trapped in a “cage” and must jump through an
“activated state” into another cage to make a successful diffusive step.

In these systems, most guest


molecules can freely rotate and
translate in their sorbed environments

For instance, O2 and N2 have 3 degrees of translational motion, 2 degrees of rotational motion,
and 1 degree of vibrational motion in the “normal” state. This yields the “partition function”, 𝐹.
𝐹 = 𝐹𝑡𝑟𝑎𝑛𝑠 𝐹𝑟𝑜𝑡 𝐹𝑣𝑖𝑏
However, in the activated state, the partition function does not contain a translational component
in the direction of diffusion. Moreover, some degrees of freedom are lost in the activated state
due to constrictions from the solid.
This loss of conformational states is quantified by the ratio of the partition functions in the
activated state,
𝐹 ∗ < 𝐹, 𝐹 ∗ = 𝑎𝑐𝑡𝑖𝑣𝑎𝑡𝑒𝑑 𝑠𝑡𝑎𝑡𝑒
And the “activation entropy” of diffusion is proportional to that ratio
𝑆𝐷∗ 𝐹∗
exp [ ]𝛼
𝑅 𝐹

Maxwell-Stefan Descriptions of Mixture Transport:


➢ Originally developed to describe diffusion in a homogeneous gas or liquid phase
➢ Imagines the diffusion coefficient as inverse drag coefficients representing the
interchange of momentum between different types of molecules
➢ The general form of the equation is:
𝐶𝑖 𝜕𝜇𝑖 𝐶𝑗 𝐽𝑖 − 𝐶𝑖 𝐽𝑗
=∑
𝑅𝑇 𝜕𝑧 𝐶Ð𝑖𝑗
𝑗≠1

Page 12 of 15
ChBE 6260 Topic 2

Important: These “MS” diffusivities do not correspond to Fickian diffusivities


Fortunately, a Fickian diffusivity (e.g., "𝐷12 ") can be converted into a Maxwell-Stefan diffusivity
(e.g. “Ð12 ”)
𝑥1 𝜕 ln 𝛾1
𝐷12 = Ð12 (1 + )
𝜕𝑥1
Ð12 is ostensibly only weakly dependent on solution composition and is identical to the Fickian
diffusivity at infinite dilution.
One of the clear take aways from the Maxwell-Stefan approach is that the appropriate driving
force for diffusion driven flux is the chemical potential (or fugacity) gradient (probably first
recognized by Einstein).
Experimental evidence for this is found in liquid diffusion systems near the critical solution
temperature.

Example on
toluene/n-
hexane/n-
tetradecene

Frame of Reference Effects:


Our previous discussions have framed the diffusional flux in terms of a chemical potential or
concentration gradient. Both are acceptable (concentration gradients need to be used carefully),
however we should now be more precise about mass or molar “flux of A” and mass or molar
“diffusional flux of A”.
Consider a stream of uniform concentrations containing A and B passing through the imaginary
plane at “1” in the pipe below:

With no chemical potential gradient at the plane “1”, no net movement occurs of A relative to B.
in other words, we expect no “diffusional flux of A” relative to some frame of reference moving
with the bulk fluid even though there is a net total flux of A through the plane “1”.

Page 13 of 15
ChBE 6260 Topic 2

For this case where no gradient acts at “1”, we don’t need to be precise about the reference frame
velocity, since no movement occurs relative to either a mass average or a molar average velocity.
When spreading does occur due to a chemical potential gradient, the flux will be different for a
mass average velocity frame of reference and a molar average velocity frame of reference (unless
𝑀𝑤𝐴 = 𝑀𝑤𝐵 )
I use BSL’s terminology, which is more commonly used than Hines & Maddox
[𝑇𝑜𝑡𝑎𝑙 𝑀𝑜𝑙𝑎𝑟 𝐹𝑙𝑢𝑥 𝑜𝑓 𝐴 𝑡ℎ𝑟𝑜𝑢𝑔ℎ "1"]
= [convection of A through plane “1”]
+ [Diffusion of A relative molar average frame of reference ]
𝑁𝐴 = 𝑥𝐴 (𝑁𝐴 + 𝑁𝐵 ) − 𝐶𝐷𝐴𝐵 ∇𝑥𝐴
𝑎𝑛𝑑 𝐽𝐴∗ = −𝐶𝐷𝐴𝐵 ∇𝑥𝐴
Simple isobaric, isothermal diffusive interchange between one mole each of two gases:
(A=He and B=SF6):
𝑃
Clearly; 𝐶 = = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 but,
𝑅𝑇
𝑃𝐴 𝑃𝐵
𝐶𝐴 = & 𝐶𝐵 = are not
𝑅𝑇 𝑅𝑇
constant

Since this system stays at constant pressure,


𝑁𝐴 = −𝑁𝐵 , thus, 𝑁𝐴 = 𝑥𝐴 (𝑁𝐴 + 𝑁𝐵 ) − 𝐶𝐷𝐴𝐵 𝑥𝐴
𝑁𝐴 = 𝐽𝐴∗ = −𝐶𝐷𝐴𝐵 ∇xA
Thus we have a simple result if we chose a molar average velocity reference frame
[𝑇𝑜𝑡𝑎𝑙 𝑀𝑎𝑠𝑠 𝐹𝑙𝑢𝑥 𝑜𝑓 𝐴 𝑡ℎ𝑟𝑜𝑢𝑔ℎ "1"]
= [Mass convection of A through plane “1”]
+ [Diffusion of A relative mass average frame of reference ]
𝑛𝐴 = 𝜔𝐴 (𝑛𝐴 + 𝑛𝐵 ) − 𝜌𝐷𝐴𝐵 ∇𝜔𝐴 𝑎𝑛𝑑 𝑗𝐴 = −𝜌𝐷𝐴𝐵 ∇𝜔𝐴
So,
𝑁𝐴 ∗ 𝑀𝑤𝐴 = 𝜔𝐴 (𝑁𝐴 𝑀𝑤𝐴 + 𝑁𝐵 𝑀𝑤𝐵 ) − 𝑗𝐴 𝑎𝑛𝑑 𝑁𝐴 = −𝑁𝐵
So,
𝑀𝑤𝐵
𝑗𝐴 = 𝑀𝑤𝐴 ∗ 𝑁𝐴 [1 − 𝜔𝐴 (1 − )] 𝑎𝑛𝑑 𝑁𝐴 = 𝐽𝐴∗
𝑀𝑤𝐴
Thus,

Page 14 of 15
ChBE 6260 Topic 2

𝑀𝑤𝐵
𝑗𝐴 = 𝑀𝑤𝐴 𝐽𝐴∗ [1 − 𝜔𝐴 (1 − )]
𝑀𝑤𝐴
And finally
𝑀𝑤𝐵
𝑗𝐴 = 𝑀𝑤𝐴 [1 − 𝜔𝐴 (1 − )] 𝐷𝐴𝐵 ∇𝐶𝐴
𝑀𝑤𝐴
This is a messier form that effectively has a concentration dependent diffusivity, since 𝜔𝐴 is not
constant unless 𝑀𝑤𝐴 = 𝑀𝑤𝐵 .
𝑀𝑤𝐵
𝐷 = 𝐷𝐴𝐵 [1 − 𝜔𝐴 (1 − )]
𝑀𝑤𝐴
Thus, for gases it is most convenient to use 𝑁𝐴 𝑎𝑛𝑑 𝐽𝐴∗ .
On the other hand, for liquid and solid diffusion problems, it is often more convenient to use
𝑛𝑎 𝑎𝑛𝑑 𝑗𝐴 . First, even though 𝑁𝐴 ≠ 𝑁𝐵 in general in liquids it may turn out that 𝑛𝐴 ~ − 𝑛𝐵 or
𝑛𝐵 ~0. Also, in solids of complex composition and molecular weight (e.g. polymers) it can be
very difficult to define 𝑥𝐴 , whereas 𝜔𝐴 is easily known.
All of the various forms are acceptable in principle – but it pays to think about your choice
before starting a problem to simplify the math and physical interpretation of results.
Table of Flux Expressions
Type Basis Flux Gradient Expression
Total Mass 𝑛𝐴 = 𝜌𝐴 𝑣𝐴 ∇𝜔𝐴 𝑛𝐴 = 𝜔𝐴 (𝑛𝐴 + 𝑛𝐵 ) − 𝜌𝐷𝐴𝐵 ∇𝜔𝐴

Molar 𝑁𝐴 = 𝐶𝐴 𝑣𝐴 ∇𝑥𝐴 𝑁𝐴 = 𝑥𝐴 (𝑁𝐴 + 𝑁𝐵 ) − 𝐶𝐷𝐴𝐵 ∇𝑥𝐴

Diffusive Mass 𝑗𝐴 = 𝜌𝐴 (𝑣𝐴 − 𝑣) ∇𝜔𝐴 𝑗𝐴 = −𝜌𝐷𝐴𝐵 ∇𝜔𝐴

Molar 𝐽𝐴∗ = 𝐶𝐴 (𝑣𝐴 − 𝑣 ∗ ) ∇𝑥𝐴 𝐽𝐴∗ = −𝐶𝐷𝐴𝐵 ∇𝑥𝐴

𝑣𝐴 = 𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑜𝑓 𝐴 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑡𝑜 𝑠𝑡𝑎𝑡𝑖𝑜𝑛𝑎𝑟𝑦 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒𝑠


𝑣 = 𝑚𝑎𝑠𝑠 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 = 𝜔𝐴 𝑣𝐴 + 𝜔𝐵 𝑣𝐵
𝑣 ∗ = 𝑚𝑜𝑙𝑎𝑟 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑣𝑒𝑙𝑜𝑐𝑖𝑡𝑦 = 𝑥𝐴 𝑣𝐴 + 𝑥𝐵 𝑣𝐵

Page 15 of 15

You might also like