joum uo

MEMBRANE
SCIENCE
ELSEVIER Journal of Membrane Science 135 (1997) 147-159

Some observations about the application of Fick's first law

for membrane separation of multicomponent mixtures
H. Denny Kamaruddin, William J. Koros*
Department of Chemical Engineering, The University of Texas at Austin, Austin, TX 78712, USA

Received 10 October 1996; received in revised form 8 April 1997; accepted 10 April 1997

Abstract

According to Fick's first law of diffusion, permeation flux of a permeant through a pore-free dense polymeric membrane
with respect to a fixed frame of reference is equal to the sum of the bulk and diffusional flux. Under ideal conditions and low
downstream (permeate) pressure, membrane permeability, PA, of component A is equal to the product of the mobility
coefficient, DA, and the solubility coefficient, SA. The definition of membrane permeability as stated above can be derived
using Fick's first law of diffusion when the frame of reference (bulk flux) term is negligible. When analyzing the transport of
permeant through pore-free dense polymeric membrane, the bulk flux contribution is usually assumed to be negligible when
the sorption levels of the components are small. However, this assumption can be erroneous in the case of multicomponent
mixtures when the permeation flux of one of the permeant is much higher than the others. Two examples will be discussed to
demonstrate that neglecting the bulk flux contribution might lead researchers to incorrect conclusions on the material science
aspects of the membrane separation. The first example is C02/CI-I4 gas separation using a dense pore-free glassy polymeric
membrane and the second example is the removal of traces of phenol from water using a polyether-block-polyamidepolymer.

Keywords: CO2/CH 4 gas separation; Pervaporation of VOCs; Membrane transport; Fick's first law of diffusion; Solution-
diffusion

1. Introduction the diffusional velocity. In this case, flux of the

permeating component is caused by the diffusional
It is well known that the transport process through a velocity relative to the bulk velocity. In the case of
pore-free polymeric medium can be described by fixed frame of reference transport, the membrane can
F i c k ' s first law of diffusion. W h e n describing the be used as the frame of reference because the mem-
transport process, it is necessary to specify the frame brane is stationary at steady state. For this static
of reference of the transport process [ 1]. There are two reference frame, the penetrant transport process is a
c o m m o n frames of reference: fixed and moving. In the combination of diffusional and bulk (convective) flux.
moving frame of reference, the mass average bulk When describing membrane-related transport pro-
velocity is typically used as the frame of reference for cesses, it is more convenient to use the fixed frame
of reference because the experimental permeation rate
is measured with respect to a fixed frame of reference,
*Corresponding Author. that is, the membrane.

0376-7388/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved.

PII S0376-7388(97)00142-7
148 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159

When describing binary mixture permeation, the ~ x

(Op2 ~ 1
system comprises a ternary mixture of the polymer ~, = 1- O)AI - LOBI
p Dpm dO)pt dx
and two permeants. Eqs. (1) and (2) shown below are Permeate

transport equations for the two permeants and Eq. (3) (n,+%+%)~p
np = 0
1,
is the transport equation for the polymer.
rTA = {nA + riB) 0)A - pDkmd(~A/dx
p DAm dfOA/dx II=
&CA
na = (ha +nB d- np)cda - - pDAm dx (1) O)AI
nB =(=?A + ne) ~ - pD~d(O~/dx

dUB
nB: (HA ~- nB@ rip)biB - - pOBm ~ (2)
~t

n p = (na + n B + np)Wp -- p D pm d~vp

~ (3) ( 0 ~ = o~e~ ~ 0

( low downstream pressure )

A and B refer to the permeating components and p

9 f
refers to the polymer. Dim is the effective binary
diffusivity for diffusion of i in the ternary mixture Fig. 1. Illustration of the transport process of binary mixture
[ 1], that is, polymer, component A and B. ni is the mass membrane permeation.
flux (g/cm 2 s) of permeant i with respect to a fixed
frame of reference and ~vi is the concentration of RA = n bulk + n diff (5)
permeant i in the membrane (g/g). Eqs. (1)-(3) can
be extended to describe multicomponent mixtures of n~ ulk ---- (ha +nB + rip)We (5a)
n. For the sake of simplicity, only binary mixture
ndiff - d~A
permeation will be discussed in this paper. =--D[-)Am ~ (5b)
Eqs. (1)-(3) are written in mass flux unit (g/cm 2 s);
they can also be written in terms of gas molar flux unit Paul and Ebra-Lima [2-7] recognized the impor-
(cm 3 (STP)/cm 2 s). Eq. (4) is the transport equation tance of the bulk flux contribution in single component
for component A in term of gas molar flux. permeation in a highly swollen membrane. In pure
component permeation, the bulk flux is a function of
the sorption level of the penetrant; however, in the case
N ~ VACA _Damd_~A (4)
NA = (NA + NB + pl 22400 of binary mixtures, bulk flux is a function of both the
sorption level and flux of both mobile components.
When describing permeation through a pore-free poly- When the mass flux of one component is much higher
meric medium, it is more convenient to use equations than the other, omitting the frame of reference term
with mass flux unit instead of those using gas molar can be erroneous when describing the transport of the
flux unit because the molar volume of the sorbed slower permeating component. Therefore, it is neces-
penetrants, l?i, might not be readily available, espe- sary to distinguish between the permeability calcu-
cially in the case of gaseous permeants. lated based on the total mass flux and permeability
The polymer mass flux, np, is equal to zero at steady calculated based on the diffusion flux. The later part of
state since the membrane is stationary. Therefore, the this paper will explore the difference, which is not
diffusional mass flux of the polymer is equal in value usually recognized, between the diffusion-based and
and opposite in the direction to the flow with respect to the observed membrane permeability.
its bulk (convective) mass flux. The total mass flux of
A, nA, is comprised of the mass flux resulting from the
bulk motion of the permeants and mass flux resulting 2. Diffusion-based permeabilities of binary
from diffusion relative to the bulk flux, and is given by mixtures
Eq. (5). The same definition also applies to compo-
nent B and p. The schematic of the transport process The bulk flux contributions in Eqs. (1) and (2) are
for the three species in the thickness direction of a often neglected based on the assumption that the
membrane of thickness { is illustrated in Fig. 1. amount of sorption of the penetrants, WA and ~B,
 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159 149

are negligible. In some cases this assumption is rea- The following assumptions were used in Eq. (9):
sonable, such as the sorption of simple gases, that is, 1. The density of the membrane is assumed constant
H2, He, 02 and N2. This assumption has convention- throughout the membrane thickness.
ally been casually extended to most gases and solvents 2. Diffusion and solubility coefficients of the per-
so that bulk flux contributions are almost always meating species are constant.
assumed to be negligible. Transport equations dis- The density of the membrane varies in the thick-
cussed below are obtained using this conventional ness direction due to different levels of sorption,
assumption that bulk flux is minimal with respect to but an average density can often be used to a
the diffusional flux. reasonable approximation. The average density can
be estimated as the mean of the feed and permeate
n bulk = (nA q- riB)idA ,~ 0 (6) side density. A more complex expression of density
can also be used, but it will only complicate the
Therefore transport equation of component A becomes picture without adding insight on the problem
being considered here. Since the feed and permeate
do3A
nA ~, ndAiff = --pDArn ~ (7) side density do not vary too much in the case
of low to moderately swelling systems, using the
The total flux of A, ha, in Eq. (7) is the mass flux simple average density value will not cause much
resulting from diffusion transport. The above equation error. On the other hand, the assumption of a
can also be written in terms of gas molar flux, NA constant diffusion coefficient can be erroneous
(cm 3 (STP)/cm 2 s), which is obtained from Eq. (4) because diffusion coefficients can be a strong func-
with the assumption of minimal bulk flux. tion of the local permeant concentrations [8-10].
Therefore, the diffusion coefficient in the equation
Ndiff d CA is actually the average diffusion coefficient of the
"~ --DAm --~ (8)
permeant in the membrane of thickness (. Keeping
this in mind and using the assumptions stated above
the diffusion coefficients of Eqs. (7) and (8) have the and integrating Eq. (9) with the following boundary
same value and units. In the case of minimal bulk flux, conditions:
equations with molar gas flux unit are commonly used
to describe transport through a pore-free polymeric X ~ 0; PA = Pal
medium because molar volumes of penetrants and x ~- ~; PA = PA2
polymer are not required (Eq. (4)). The concentration
of A inside the membrane, C A (cm 3 (STP)/cm3), is a Eq. (10) is obtained
function of the upstream partial pressure of A. The DArnSA - NAdiffd
solubility (sorption) coefficient of the membrane can (10)
PAl -- PA2
be obtained from the sorption isotherm. The solubility
coefficient value is the slope of the sorption isotherm, where the average effective diffusion coefficient of A,
SA = CA/PA. In the case of an ideal mixture, the Dim, is defined as follows
partial pressure can be used to determine the solubility •CA1 D(c)dc
coefficient but fugacity has to be utilized when the gas bArn = dCa2 (11)
behaves non-ideally. When the feed solution is a liquid CAI -- CA2
mixture, the partial pressure or fugacity term does not Permeability is defined as the thickness and pressure
refer to the static pressure. Instead, the partial pressure normalized flux; therefore, the product of the solubi-
or fugacity term refers to the equilibrium vapor pres- lity and diffusion coefficient is the membrane perme-
sure of the liquid feed mixture. Assuming ideality and ability. Since the definition of membrane permeability
substituting the definition of the solubility coefficient, as stated in Eq. (10) was derived under the limiting
SA, into Eq. (8) condition of zero bulk mass flux, it is therefore more
NAdiff = - - D A m S A dpa appropriate to refer to the membrane permeability in
-~ (9)
Eqs. (10) and (12) as a 'diffusion-based' membrane
150 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159

permeability, or PA*: ary conditions:

Ndiffd X ~ 0~ ~a)A = 03A1 ~ ~-OB = 0.)BI

PA* -- -- DAmSA (12) X = ~; 03A : 03A2 ~ 03B : ~0B2

PAl --PA2
when permeate (down stream) pressure approaches to
This will differentiate the diffusion-based permeabil- z e r o , O)A2 and wn2 -~ 0. In the calculation, the bulk flux
ity from the experimentally observed membrane per- contribution will be fully accounted for. The assump-
meability, PA, which is based on both the bulk and tions of constant diffusion coefficients and density will
diffusional flux. Therefore, in the limit when bulk flux be used as described previously.
is zero, the diffusion-based membrane permeability is
equal to the product of a kinetic factor, the average nAE =
pDamln Fl-~2(l+l/r)]
Ll-wA'(l+l/r)J
p[~amln 1--wAl(l+l/r)
mobility (diffusion) coefficient (Dim) and a thermo- (1 + l/r) (1 + l/r)
dynamic factor, the average solubility coefficient
(14)
(Sire). The term of solution-diffusion transport
mechanism comes from the definition of diffusion- pDBmln [1-we2(l+r)l p/)Bmln [ ~ ]
nJ = L1-~.,(l+r)],~
based permeability, that is, Eq. (12). In the case of an
(1 + r) (1 + r)
ideal mixture and permeate pressure approaching
zero, the ideal membrane selectivity of A relative to 05)
nA
B is the ratio of the diffusion-based permeabilities where r = - - ; (nA > riB) (16)
nB
PA* OamSA
(13) The fraction of bulk flux contribution, 1-I~ulk and
OgA/B -- PB* DBm SB
i--[bulk
B , of each component permeating through a mem-
DAm brane of thickness : is the ratio of bulk mass flux to the
/)Bin = mobility selectivity total mass flux as shown in Eqs. (17) and (18). Since
concentration gradients exist in the direction of mem-
SA -- solubility selectivity brane thickness, average concentrations, WAvg and
SB avg
WB , should be used in estimating the approximate
Ideal membrane selectivity can also be taken to be fraction of bulk flux contributions.
equal to the product of the mobility and solubility
selectivities. The mobility selectivity is a measure of (17)
the relative ease with which the components perform
diffusive jumps while solubility selectivity is a mea- [q-bulk __ (hA ~ avg
q-nB)WB -- (1 q- r)WBvg
1 IB (18)
sure of the relative condensibility of the components nB
and affinity of the polymer to the components. Using In earlier paragraphs, it was stated that when sorption
the mobility and solubility aspects of the components levels are small, neglecting the frame of reference
to explain permeation properties of membrane sys- (bulk flux) seems to be a reasonable estimation.
tems has been instrumental in bringing about a However, Eq. (18) shows that if the mass flux of
better understanding of transport in polymeric component A is much higher than B ( ' r ' is high), then
materials. It is therefore imperative to use the correct the fraction of bulk flux for B can still be significant
permeability equation to calculate the diffusion even though ~vg value is small. Looking back at
coefficient. Eq. (2), when wB is small, the bulk flux of B might
be small compared to the total mass flux of A and B,
(ha -~- riB). However, the bulk flux of B might still be
3. Permeabilities with respect to fixed frame o f significant relative to the diffusional flux of B. There-
reference fore, the bulk flux of B cannot be neglected.
On the other hand, Eq. (17) shows the impact of
The mass flux of A and B can be obtained by 'r' on component A to be the opposite of component
integrating Eqs. (1) and (2) with the following bound- B. The value of I-I~uL~ approaches WAvg when 'r' is
 H.D. Kamaruddin,W.J.Koros/Journalof MembraneScience135 (1997)147-159 151

large. This picture is actually quite logical when Hbulk ( ~) OdA1+ ~A2 ~ (1 + !) 03AI
a ~ 1+ 2 ~- (23)
one component has a much higher flux than the
other component, the slower component is 'swept'
bulk ~B1 71-~')B2 ~ (1 + ~ 03B1 (24)
along by the faster component. Obviously, the flux HB ~ (1 + r) 2 r)
of the slower component is too small to influence
~')A2 = C'0B2 ~ 0
the bulk flux of the faster component. In extreme
cases where the level of sorption of B and the value Clearly, the error of Eqs. (23) and (24) can be
of 'r' are high, the bulk flux can actually dominate significant if the fraction of bulk flux is significant.
the total flux of B. While these issues are clear, we The ri~ ulk obtained from Eqs. (23) and (24) is 3.5%
do not believe that they have been pointed out lower than that of Eqs. (21) and (22) when
previously, and they can be rather important in H bulk = 0.1 and 20% lower when I-Ibulk = 0.5. There-
practical situations. fore, the H bulk calculated using Eqs. (23) and (24)
Average concentrations for each component, WAvg should only be used as an approximation.
avg
and oJB , can be obtained by first determining the With the value of the fraction of the bulk flux in
concentration profiles inside the membrane, WA(X) and hand, the diffusion-based permeability of the per-
0JB (x), and averaged as follows: meant can then be calculated based on diffusional
mass flux which represents the true diffusional
feo OJA(X)dX transport.
Hoax ( ) a)AI(I+I/F)PA
1 WA1 PA* = l - - I ~ ulk PA=ln[1/(l_~Al(l+l/r))]
(19)
-- (1 + 1/r) I n [ l / ( 1 -- WAI(1 + I/r))] (25)
and ( ) wm(1 + F)P,
PB. = 1 - I I bulk P n = l n [ 1 / ( l _ o . ) m ( l + r ) ) ]
1 O.)Bl
(26)
f0Cdx (l+r) ln[1/(1-Wel(1 +r))]
PA and PB are the experimentally observed membrane
(20)
permeability of component A and B. The experimen-
Detailed manipulation of Eqs. (19) and (20) is dis- tally observed membrane permeability for component
cussed in the Appendix A. i, Pi, can be obtained by normalizing the observed
Expressions for bulk flux contributions are obtained diffusional mass flux, n. with the membrane thickness
by substituting Eqs. (19) and (20) to Eqs. (17) and and driving force (partial pressure or fugacity differ-
(18) ence across the membrane), Api.

bulk 1 WAI(1 + 1/r) (21) 22400 nif

HA Pi -- (27)
= ln[1/(1 - OJAI(i + l/r))] M,Api
If the observed permeability (Eq. (27)), is calculated
I-I bulk 1 ~m(1 + r)
8 = ln[1/(1 - wal(1 + r))] (22) based on the total (bulk and diffusion) mass flux, and is
used instead of the diffusion-based permeability
When the fraction of the bulk flux with respect to (Eqs. (25) and (26)) to calculate the average diffusion
the total mass flux is minimal, the concentration coefficients, the value obtained will overestimate the
profile inside the membrane approaches linearity; actual value. The extent of the error depends on the
therefore, the average concentration inside the mem- fraction of bulk flux contribution, the higher the bulk
avg
brane, 0:~vg and we , can be estimated to be the simple flux contribution the higher the inaccuracy.
average of the feed and permeate side concentrations. The diffusion-based permeability is the appropriate
Substituting these values to Eqs. (17) and (18) gives a property of the material since it is dependent only on
simpler method of estimating the bulk flux contribu- the material property, that is, diffusion and solubility.
]~bulk
tion, ~~i • The importance of the distinction between the calcu-
152 H.D. Kamaruddin, W.J.Koros/Journal of Membrane Science 135 (1997) 147-159

lated permeability based on the diffusional flux and to as Langmuir's population while those occupying
overall flux distinction will be illustrated with two the dense matrix are referred to as Henry's population.
membrane separation examples. The first example is The dual-mode equation for single component
the gas separation of C O 2 / C H 4 using 6FDA-TADPO sorption is shown in Eq. (28); the equations for
polypyrrolone membrane; the second is the removal of binary mixture components are shown in Eqs. (29)
phenol from water (pervaporation) using a polyether- and (30).
block-polyamide (Pebax ®) membrane.
C--HAtbAPA~ MA (28)
033 = kDAPA~ 1 + bAPAJ 22400 p
4. CO2/CH 4 separation using 6FDA-TADPO Ctta'bAPA_ .~ MA (29)
polypyrrolones ~3A = kDAFAq 1 + bAPA + bAPBJ 2 2 4 0 0 p

CI4,'bBp8 ~ MB
The data used in this example are cited from Ref. cob = kosps q 1 + bAPA q- bAps J 22400 p (30)
[11] and the mixed-gas data are summarized in
Table 1. 6FDA-TADPO polypyrrolone is a rigid koi is the 'Henry's law constant' which characterizes
semi-ladder polymer. Its open structure and chain the sorption in the dense region of the polymer matrix,
rigidity give it desirable membrane properties, that Pi is the partial pressure of component, b, and is a
is, high permeability and selectivity. Its unique proper- measure of the affinity of the penetrant to the Lang-
ties enable it to resist CO2 plasticization better than muir packing 'defect' sites and Ct4' is the Langmuir
most other polymers. Due to the unique properties of capacity constant. The density term, p, refers to the
6FDA-TADPO polypyrrolone, it is an ideal example density of the system, the polymer and permeants.
to demonstrate the importance of the bulk flux in However, in the case of non-swelling (low level of
membrane separation of binary mixtures. sorption) the density of the system can be approxi-
The sorption isotherms of both gases can be mod- mated to be the polymer density without causing
eled with a high degree of accuracy using the dual- major error. When describing mixtures which exhibit
mode model. The dual-mode sorption model assumes, non-ideal behavior, fugacity instead of partial pressure
for convenience, that glassy polymers have two dis- should be used. Since CO2/CI-I4 gas mixtures are non-
tinct types of sorption environments, free volume ideal mixtures, therefore fugacity has to be utilized
(defects) and dense matrix [12-15]. The free volume when describing the permeants' transport through the
(defects) region exists because of the inability of the membrane. The fugacity of pure and mixed CO2/CH4
polymer chains to pack perfectly at temperatures can be calculated using the virial equation of state
lower than their glass transition temperature (Tg). In [19]. Fig. 2 shows the pure and mixed (50/50
effect, glassy polymers have not reached equilibrium CO2/CH4) CO2 and CH4 sorption isotherms as a
packing conditions; however, the state of the polymer function of fugacity. The percentage decrease of sorp-
is metastable because the relaxation time of the poly- tion isotherm of mixed CH4 is much larger than CO2
mer chains is extremely long. The population of the when compared to their pure component sorption
components sorbed in the free-volume sites is referred isotherms. This is because in the case of mixed gas

Table 1
Permeation properties of the separation of 50/50 CO2/CH4 using 6FDA-TADPO polypyrrolone membrane [11]
~.t~,
°bile ~ C ~ ile I"1"bulk l-[bulk
Fugacity Pco2 Pcr~ co2 nco2: x 109 r 11co2 llco,
(psia) (Barrer) (Barrer) (g/g) (g/g) (g c m / c m 2 s)
139.3 30.0 0.573 0.0110 0.0008 2.18 143,4 0.0056 0.0583
409.1 22.8 0.508 0.0273 0.0024 5.13 121.9 0,0138 0.1565
532.2 20.9 0.506 0.0354 0.0032 6.31 112.6 0.0180 0.1955
682.8 19.9 0.530 0.0459 0.0042 8.02 107.8 0.0234 0.2413
767.8 18.7 0.528 0.0522 0.0049 8.65 96.6 0.0266 0.2638
Barrer: 10 -1° cm3(STP) c m / c m 2 s cmHg.
 H.D. Kamaruddin, W.J.Koros/Journal of Membrane Science 135 (1997) 147-159 153

0.14 . . . . i . . . . i . . . . i . . . . i . . . . i . . . . i . . . .
Table 2
: 6FDA-TADPO polypyrrolone
50•50 CO2/CH4, 35"C Fugacity-based dual-mode and partial immobilization parameters
0.12
of 6FDA-TADPO polypyrrolone for CO2 and CH4 at 35°C
A . . . . pure gas sorption .- ..
g 0.10 mixed gas sorption ." " Carbon dioxide CO2 Cth
B

0.08 kD (cm3(STP)/cm3 atm) 1.526 0.327

CH' (cm3(STP)/cm3) 34.084 22.838
g 0.06
== b (atm q) 1.023 0.160
Do (cm2/s) 1.196e-7 1.12e-8
o
0.04
F (Dt4/DD) 0.084 0.026
0.02
............ Methane
. . . . . . . . . ,','70; 7 . . . . . . . . . . . . . . . . . . . . . .
0.00
0 1 O0 200 300 400 SO0 600 700
P* : kDcu4ODcH4(1 + FcH4KCH4
gugacity (psia)
CH4 1 ÷ bco~fco2 + bCH4fCH4.]
(32)
Fig. 2. Pure and mixed (50/50 CO2/CH4 mol%)CO 2 and CH4 The values of the dual-mode and partial immobiliza-
sorption isotherms.
tion parameters are summarized in Table 2.
The mobile concentration of CO2 and CH4 is shown
sorption, CH 4 has to compete with C O 2 for available in Eqs. (33) and (34). As discussed in the previous
Langmuir's sites and CO2 has a higher affinity for the paragraph, the partial immobilization theory requires
free volume 'defect' sites than CH4. mobile concentration to be used when describing
The transport of penetrants through rubbery and permeant transport in a glassy polymer. Therefore
glassy polymer is slightly different. As explained in w~°bile and w~B°bile terms have to be used in transport
the previous paragraph, sorbed populations in glassy equations previously discussed, Eqs. (1)-(26).
polymers are made of Langmuir's and Henry's popu-
lations. Koros et al. [16] discussed the partial immo- mobile kDc%fC02MC02
bilization theory for gas mixtures which suggested Wc°2 - 22400 p
that only a fraction of Langmuir's population is able to FC02 KCO2 .')
perform the diffusive jump. It is assumed that Henry's × 1 -~ 1 + bco~fco2 + bcn4fcn~/ (33)
population is completely mobile, that is, they are able
to fully participate in performing the diffusive jump. mobile kDc.,fcH4MCH~
Therefore, in glassy polymers, the mobile species are ~JCH4 -- 22400 p
made of Henry's population and a fraction of Lang-
muir's population. When discussing transport of per- FcmKcH~ -~ (34)
× 1 -~ 1 + bco~fco2 -4- bcnffcrh//
meants through glassy membrane polymers using
Fick's first law of diffusion equation, ~MA and cob refer
Fig. 3 shows CO2 and CH4 mobile concentrations of
to the concentration of the mobile species. If the entire
the pure and 50/50 mol% CO2/CI-I4. The isotherms
sorbed population is used, it will cause a significant
are calculated using Eqs. (33) and (34). Fig. 3 shows
error in the calculated value. A detailed description of
that the sorption isotherms of the mobile species are
the partial immobilization theory is discussed in Ref.
much lower than the sorption isotherm of the total
[ 16]. The value of the diffusion coefficients of Henry's
sorbed species. The concentration of both mobile CO2
and Langmuir's population can be obtained using
and CH 4 are only slightly lower in mixed-gas than
Eqs. (31) and (32). The constant 'F' in Eqs. (31)
pure gas. This is because most of the mobile species
and (32) is the ratio of diffusion coefficient of Lang-
are due to Henry's population and only a small fraction
muir's to Henry's populations.
of Langmuir's population is mobile. The dual-mode
theory assumes Henry's population to be the same in
* ( Fc°2Kc°2 .'~ pure and mixed-gas case.
Pc% = kDco2DDco2 1 -+ 1 + bco2fco 2 -~- bCHffCH4J Fig. 4 shows the experimental and predicted
(31) C O 2 / C H 4 mixed-gas ideal membrane selectivity.
154 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159

0.10 .... i .... i .... i .... i .... 14 x l 0 "a i i i i i

6FDA-TADPO polypyrrolone COz-total (pure)
35"C Carbon d i o x i d e
12 x l 0 "a -,.-._..-,--.,----,.,--,-.-., .... ;---.
A 0.08
1 0 x l 0 "a

g o.o6 C a r b o n dioxide Methane

8 x l O "a
~ . - A" " "~ CO2"mobile o mixed-gas zx mixed-gas
/ . ." ° (pure and mixed) • pure g a s • pure g a s
" 0.04 / .~-.~" 6 x l O "a
O
.... theoretical . . . . theoretical
4 x l 0 "a

O 0.02 / , a , -''''''~°'" CH4-total (pure)

~, 2 x l O "a Methane
-'CA- . . . . • .... /r . . . . ~r-- -•- - ~-- -~ - -
o.oo, ~. . . .
. . i. . .. .. .... . . . . . . .J. . ..... .. . . i
~-~o-~e'~;.~ed)
. ,4.. i ....
- I I I I 1 I

100 200 300 400 500 0 100 200 300 400 500 600

Fugaclty (psia)
Fugacity (psia)

F i g . 3. P u r e a n d m i x e d (50/50 C02/CH4r e e l % ) mobile CO2 and Fig. 5. Pure, mixed and predicted CO2/CH 4 mobility and
CH 4 sorption isotherms. solubility coefficients of 6FDA-TADPO polypyrrolone.

~>" 80 .... i .... i .... i .... i .... i , . , can be used as an initial estimate. The calculations
1; 6FDA-TADPO polypyrrolone were performed using a spreadsheet program and it is
e 50150 C O 2 / C H ,, 3 5 " C
Ol 70 actually quite straightforward. The calculated values
"".....
of the observed membrane selectivity are in good
~ 60

E .................
agreement with the experimental data. Fig. 4 also
~ so shows that significant errors can arise if diffusion-
"o
based permeabilities (neglecting bulk flux contribu-
E
~ 4o tion) are used to calculate the observed membrane
,o o Experimental
0
--- Neglecting the
selectivity as illustrated by the dotted line. It over-
=" 30 bulk flux contribution predicts the experimentally observed membrane selec-
0 -- Calculated
0 .... i .... i .... i .... i , i i. J .... tivity.
0 20
Fig. 5 shows that the pure, mixed-gas and the
100 200 300 400 500 600
CH 4 Feed Fugaclty (psla)
theoretical value of diffusion (mobility) coefficient
of CO2 and C H 4 are approximately equal when the
Fig. 4. Comparison of the experimental (solid circle) and bulk flux contributions are not assumed to be negli-
calculated (solid line) observed membrane selectivities of 50/50 gible. The diffusion coefficients of the mixed-gas were
CO2/CH4 mol%. The calculated selectivity was based on dual-
mode and partial immobilization parameters obtained from pure
calculated by using Eqs. (14) and (15). Therefore, the
CO2 and CI-I4 sorption and permeation data. The dotted line lower value of the experimentally observed selectivity
represents the predicted selectivity when the bulk flux contribution (Fig. 4) in the mixed-gas case is not due to slight
is neglected. membrane plasticization caused by the presence of
CO2 as originally thought. This phenomenon can be
The solid circles represent the experimental data and fully explained by taking into account the bulk flux
the solid line is the calculated value predicted using contribution. Pure CO2 permeation data [11] also
pure component sorption and permeation data. First, confirmed the absence of plasticization by CO2.
the mobile concentration of CO2 and C H 4 is calculated Fig. 6 shows the comparison of the fraction of bulk
using equations Eqs. (33) and (34). Then the mass mass flux of CO2 in the pure- and mixed-gas case. The
fluxes of CO2 and C H 4 c a n be calculated using curves were calculated using Eq. (21). As discussed
Eqs. (14) and (15). The calculation will require itera- earlier, the bulk mass flux contribution of the compo-
tions because the values of nco2, ncH4 and 'r' are nent with the higher mass flux in a binary mixture
unknown. Therefore, the 'r' value obtained from membrane permeation can be estimated to be approxi-
Eqs. (31) and (32) (partial immobilization theory) mately equal to its average sorption level when the
 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159 155

0.10 . . . . i . . . . i . . . . i . . . . i . . . . i . . . . i , , .

mass flux. This is because the mobility and solubility

6 F D A - T A D P O polypyrrolone
35"C of the penetrants in the membrane can only be
~ 0.08
Carbon dioxide
described using the diffusion-based permeabilities.
8 o mixed-gas In this example, taking into account the frame of
. 0.06 50150 C O 2 / C H 4
- -e- - pure gas . .e reference enables us to come to a different conclusion
m= 0.04
.-
.w
o
. -
about the permeation properties of the membrane.
"6 .• oo° •'°

°l.O -''~
0.02 oo°.r "o"
.m 5. Removal of phenol from water using a
0.00
~'," ,'~. i . . . . i . . . . i . . . . i . . . . i . . . . q .... pervaporation process
100 ZO0 300 400 500 600 700

Carbon dioxide Fugaclty (paid) In the previous example, the bulk flux contribution
Fig. 6. Bulk flux contribution of C O 2 in a 50/50 C O 2 / C H 4 mixed-
has a negative impact on the separation. However, the
gas experiments as a function of its fugacity. bulk flux contribution can be positive or negative for
the separation performed. It depends on the desired
component to be separated, it is positive if the flux of
the component to be separated is enhanced by the bulk
0.30 . . . . ~ . . . . i . . . . , . . . . i . . . . i . . . .

g
°°...--°'" flux more than the other component and vice versa. In
6FDA'TAD PO5.1~olyPyrr°l° n e .o. -
o.2s . o °
the case of the removal of phenol from water, the flux
o- of phenol is enhanced more than the flux of water by
t~ 0.20 sO
the bulk flux contribution. Therefore, the bulk flux
,.r 0.15 •
0
Methane
contribution actually benefits the separation because
- - o - - mmxeo-gas
phenol is the desired component to be separated.
'3 o.~o
• " 50150 CO~ICH 4
The data utilized in this example are summarized in
,-e"" - -= -. pure gas
Table 3 and they are obtained from Ref. [17]. The
0.05 e•
removal of phenol from water was done using a
0.00 ~-.-.-rQ-~.T-t~-,-9-r-,--,-r-?n-~e~. -,l-.'T-,'-.-T-:-l'T-T-7-.' pervaporation process utilizing a polyether-block-
0 tOO 200 300 400 500 600 polyamide (Pebax®(Atochem)) membrane. The
Methane Fugacity ( p s l a ) membrane selectivitely removes phenol from water.
Fig. 7. Bulk flux contribution of CH4 in a 50/50 C O 2 / C H 4 mixed- The overall pervaporation separation factor, /~per~ap,
gas experiments as a function of its fugacity. can be thought of as equal to the product of evapora-
tion separation factor, flevap,and membrane separation
factor, flr,~m [18]. The evaporation separation factor,
value of 'r' is large (Eq. (23)). On the other hand, as flevap, is a measure of the relative volatility of the
shown in Fig. 7, the fraction of the bulk mass flux of components while the membrane separation factor,
CH4 in the mixed-gas case is much higher than the
pure-gas case. The fraction of bulk mass flux of CH4 in
Table 3
the mixed-gas case was calculated using Eq. (22). In Removal of phenol from water using a polyether-block-polyamide
the mixed-gas permeation, CH4 molecules were membrane [16]
'swept' along by the much faster CO2 molecules. ,. ,mobile bulk
Phenol ~phenol nphenol r ]-[phenol
Therefore, a considerable portion of the mass flux
Feed conc. (ppm) (g/g) (g/m2h) (nU~o/nphenol)
is a result of bulk flux which if not accounted for can
31 0.00105 0.4 434.5 0.25
lead researchers to incorrectly conclude that lower 48 0.00200 0.6 290.7 0.33
mixed-gas selectivity is due to the loss of the mobility 85 0.00330 1.1 161.5 0.30
selectivity, that is, as a result of plasticization. 153 0.00600 1.9 93.3 0.32
In order to understand the materials science aspect 324 0.01300 4.1 44.1 0.34
585 0.02300 7.5 24.6 0.34
of the polymer-penetrant interaction, the bulk mass
994 0.03500 12.7 14.9 0.32
flux contribution has to be subtracted from the total
156 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159

0.5
/~mem, is a measure of the relative solubility and ' • I , ' • I • , • I ' ' ' i , ,

polyether-block-polyamides
diffusivity of the permeants in the membrane. At 8 46 lttm, 50"C
low permeate pressure, the membrane separation fac- o.4
tor, /~mem,is approximately equal to the ideal mem- C
o • ___., . . . . . . . ".. . . . . . . . . . . _". . . . . . . . . . . . . . . . . .
0
brane selectivity• 0.3
;-r
Pebax ® membrane is selective toward phenol over
water, that is, phenol has a higher permeability than ~ 0.2
water• In this example, phenol has a lower mass flux o
than water because phenol has a much smaller driving ~ 0.1
force than water due to low phenol concentration in ,,.r
the feed solution• Since water has a higher mass flux 0 I i i i I , , , I , , , I , , ,

than phenol, the mass flux of water has a positive 200 400 600 800 1000
effect on phenol bulk mass flux, much more than the Phenol Feed Concentration (ppm)
effect of phenol mass flux on water• The linearity of
Fig. 9. Bulk flux contribution of phenol as a function of phenol
concentration of phenol as a function of phenol feed
feed concentration.
concentration suggests the sorbed population is mostly
Henry's (Fig. 8). Therefore, it is assumed that all of the
sorbed phenol in the polymer is mobile• The summary improving the removal of phenol from the feed solu-
of the data is shown in Table 3 and Fig. 8. The fraction tion.
of phenol bulk flux contribution is illustrated in Fig• 9.
Fig. 9 shows that the bulk flux contribution of phenol
is approximately 30% of the total mass flux. The 6. Conclusion
scatter of data in Fig. 9 is presumably due to an error
caused by translating the sorption data from the phenol The objective of this paper is to show that the frame
sorption graph in Ref. [17]. The water sorption level of of reference term cannot always be ignored even if the
polyether-block-polyamide membrane is approxi- sorption level is minimal. In the case of membrane-
mately 0.02 gH20/g at phenol feed concentration of based separation of multicomponent mixtures, the
31-994 ppm. Thus, referring to Eq. (17) or Eq. (21) it assumption of negligible bulk flux based on low
is clear that the bulk flux contribution for water is sorption levels is erroneous when the relative mass
minimal• As a result bulk mass flux has benefitted the flux of the other component is high. In the two
separation process by enhancing the mass flux of examples discussed above, the sorption level of the
phenol to a much larger degree than water, thus components is relatively small; however, significant
bulk flux contribution is observed• The above discus-
sion shows that keeping the frame of reference term in
0.040 • ' • I ' ' ' ~ ' ' '

polyether-block-polyam ides the transport equation can make a difference in the

A
0.035 46 Ilm, 50"C fundamental interpretation of experimental data.
0.030 In the case of binary systems in which the mass flux
8
¢=
0.025
of the component to be separated is much higher than
the other component (high 'r') and the slower corn-
0.020 / avg \
0
ponent has a moderate-to-high sorption level (w8 ),
0.015 the separation process can be inefficient even though
=.=
0

0.010 the relative mobility and solubility favor the desired

r,

0.005
component to be separated (component A) tremen-
D•" • , I , , , I , , , i , , , i
dously. The first example showed that even at low
0.000 -

sorption levels of CH4, the separation process

0 200 400 600 800 1000
becomes less efficient due to the enhanced bulk flux
Phenol Feed Concentration (ppm)
of CH4. If the sorption level of mobile CH4 is doubled,
Fig. 8. Phenolsorption level as a functionof feed concentration. the result would have been very detrimental to the
 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159 157

separation process. It is therefore imperative to reduce kD dual-mode Henry's constant (atm -1)
the bulk flux in order to achieve better separation. This M molecular weight (g/mol)
point should be kept in mind when selecting a mem- n mass flux (g/cm 2 s)
brane for a particular separation process. Since bulk n bulk bulk mass flux (g/cm 2 s)
mass flux of the slower component (component B) is n aiff diffusion mass flux (g/cm 2 s)
proportional to the sorption level and (1 + r), there- N gas molar flux (cm3(STP)/cmz s)
fore, minimizing the sorption level of the slower p pressure (cmHg)
component will improve the separation. Reducing P observed membrane permeability (cm 3
'r' is not desirable because the ideal membrane selec- (STP) cm/cm 2 s cmHg)
tivity is also reduced. If the sorption level of the slower P. diffusion-based membrane permeability
component is reduced without the expense of the (cm3(STP) cm/cm 2 s cmHg)
mobility and solubility selectivities then the mem- r ratio of mass flux, nA/n8
brane separation will be improved. The improvement S solubility coefficient (cm3(STP)/cm 3 cmHg)
to the separation is a function of the reduction of t h e ~"i sorbed molar volume (cm3/mol)
sorption level of the slower component, a ideal membrane selectivity
In the derivation of the equations, the diffusion /~pervap overall pervaporation separation factor
(mobility) coefficient of the components is assumed /~evap evaporation separation factor
to be constant. This assumption should be quite a c c u - flmem membrane separation factor
rate in the case of the two examples illustrated in this ~v concentration (g/g)
paper because of the low level of sorption. However, p density (g/cm 3)
this assumption can be erroneous in the case of a 1-[bulk fraction of bulk flux contribution
highly swelling system, that is, high sorption level of
the components where the diffusion coefficient of the 7,1. Subscripts
permeants can be a strong function of the local
permeant concentration. The concentration profiles i component i
of the components in the membrane can deviate from 1 feed side
the predicted value using constant diffusion coeffi- 2 permeate side
cient, Dim. To get an accurate representation of the
permeant transport, a relationship of diffusion coeffi-
cient as a function of the penetrant concentration has Acknowledgements
to be obtained and utilized in the transport equations.
Another way would be to employ a finite element The authors would like to express their gratitude to
method. the Department of Energy (DOE) and Separation
Research Program (SRP) at the University of Texas
at Austin for sponsoring this project.
7. Notations

b dual-mode affinity constant (atm -1) Appendix A

C concentration (cm3(STP)/cm 3)
Cn' dual-mode capacity constant (cm3(STP)/ Average concentrations for each component,
c m 3) avg
Wavg and tv~ , can be obtained by first determin-
Dim diffusion coefficient (cm2/s) ing the concentration profiles inside the membrane,
D~ diffusion coefficient of Langmuir's popula- (.z3A (X) and ~v8 (x), which is obtained by integrat-
tion (cm2/s) ing (1) and (2) with the following boundary condi-
Do diffusion coefficient of Henry's population tions:
(cm2/s)
fugacity (cmHg) X = 0; 0.)A = OJA1 ~ 03B = OJB1
f~
F~ DHi/DDi = X; ~ = ~(X), ~B = ~B(X)
158 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159

nA = (hA Jr- nB Jr np)OJA -- pDAm- dWA-~ 1 WA1 (A.9)

(A.1) waVg -- (1 + l/r) ln[1/(1 -- WAt(1 + i/r))]
b dwa (A.2) Eqs. (A.10) and (A.11) are obtained by evaluating
n B = (nA + nB+ np)Wa -- p Bm - ~
Eq. (A.4) with the following boundary conditions:
1-[1- WAt(1 + 1/r)]exp [ ~ ] X = O; We(X) ~---(a)B1
WA(X)= (1 + 1/r) 1 -[1 -win(1 +r)]
(A.3) ~BI = (1 + r) (A.10)

and x = ~; WB(X)= 0 (down stream pressure is

approaching zero)
1 - [1 - WSl(1 + r)]exp [nB(~r)x]
L pusm .I (A.4) [nB((lp_~B_~
+ r).]
~B(X) = (1 + r) 0 = 1 -[1 - wBI(1 + r)]exp [ (A.11)
The concentration of A and B inside the membrane of
A simpler expression of w~vg is obtained by rearran-
thickness f can be averaged as follows:
ging Eq. (A.4) and substituting Eqs. (A.10) and
1 (A.11)
avg -- f~&A(X)dX
~.z3A ~ -- ( ( 1 - I - I / r ) )
~dBvg __ 1 0)B1
(l+r) In[l/(1 - a:m (1 + r))] (A.12)
× (1--(.(1- ;1/r)p[)Am)~~/r)"/I
~at(1-k-n~l

[" fna((1-t-1/r)"~ 1)}

× kexpk pb-~-m ) - (A.5)
References
and
[1] R.B. Bird, W.E. Stewart and E.N. Lightfoot, Transport
Phenomena, John Wiley and Sons, New York, 1960.
w~vg -
f~wB(x)dx ( 1 ) [2] D.R. Paul and O.M. Ebra-Lima, Pressure-induced diffusion of
organic liquids through highly swollen polymer membranes,
J. Appl. Polym. Sci., 14 (1970) 2201.
x (1-(.(1- r)
~Bl(lnB((l+--r)P[)~m)"~+] [3] D.R. Paul and D.H. Carranza, Pressure-induced diffusion in
swollen butyl rubber membranes, J. Polym. Sci., 41 (1973)
69.
× fk e x p [[nB((1
+ - 1)
PDBmr)J}_ (A.6) [4] D.R. Paul, Relation between hydraulic permeability and
diffusion in homogeneous swollen membranes, J. Polym. Sci.,
11 (1973) 289.
Eqs. (A.7) and (A.8) are obtained by evaluating [5] O.M. Ebra-Lima and D.R. Paul, Hydraulic permeation of
Eq. (A.3) with the following boundary conditions: liquids through swollen polymeric networks. I. Poly(vinyl
alcohol)-water, J. Appl. Polym. Sci., 19 (1975) 19.
X=0; OJA(X) =~A1 [6] D.R. Paul, J.D. Paciotti and O.M. Ebra-Lima, Hydraulic
I - [ 1 -WAI(1 + l/r)] (A.7) permeation of liquids through swollen polymeric networks. II.
03A1 : (1 + l/r) Liquid mixtures, J. Appl. Polym. Sci., 19 (1975) 1837.
[7] D.R. Paul and O.M. Ebra-Lima, Hydraulic Permeation
x = f; WA (X) = 0 (down stream pressure is approa- of liquids through swollen polymeric networks, llI
A generalized correlation, J. Appl. Polym. Sci., 19 (1975)
ching zero) 2759.
[8] J. Comyn, in J. Comyn (Ed.), Polymer Permeability, Elsevier
0=1-- [1--WAl(l+!)]exp[nAf(~lp~A1/r) ] Appl. Sc. Ltd, London, 1985, Ch. 1.
[9] J.L. Duda, J.S. Vrentas, S.T. Ju and H.T. Liu, Prediction of
(A.8) diffusion coefficients for polymer-solvent systems, AIChE J.,
28 (1982) 279.
the expression for Wavg can be simplified by substitut- [10] J.S. Vrentas and J.L. Duda, Molecular diffusion in polymer
ing Eqs. (A.7) and (A.8) to Eq. (A.5). solutions, AIChE J., 25(1) (1979) 1.
 H.D. Kamaruddin, W.J. Koros/Journal of Membrane Science 135 (1997) 147-159 159

[11] D.R.B. Walker, Ph.D. Thesis, Synthesis and characterization [16] W.J. Koros, R.T. Chern, V. Stannett and H.B. Hopfenberg,
of polypyrrolones for gas separation membranes, The A model for permeation of mixed gases and vapors in
University of Texas at Austin, 1994. glassy polymers, J. Polym. Sci. Polym. Phys. Ed., 19 (1981)
[12] W.J. Koros, Ph.D. Thesis, Sorption and transport in glassy 1513.
polymers, The University of Texas at Austin, 1977. [17] K.W. Boddeker and G. Bengston, Pervaporation of low
[13] W.R. Vieth, J.M. Howell and J.H. Hsieh, Dual-sorption volatility aromatics from water, J. Memb. Sci., 53 (1990) 143.
theory, J. Memb. Sci., 1 (1976) 177. [18] J.G. Wijmans and R.W. Baker, A simple predictive treatment
[14] H.B. Hopfenberg and V. Stannet, in R.N. Haward (Ed.), The of the permeation process in pervaporation, J. Memb. Sci., 79
Physics of Glassy State, Appl. Sci. Publ. Ltd., London, 1973, (1993) 101.
Ch. 9. [19] J.M. Prausnitz, R.N. Lichtenthaler and E.G. de Azevedo,
[15] W.J. Koros, Model for sorption of mixed gasses in glassy Molecular Thermodynamics of Fluid-phase Equilibria, 2nd
polymers, J. Polym. Sci. Polym. Phys. Ed., 18 (1980) 981. edn., Prentice-Hall, NJ, 1986.