Thermodynamic Model of Gas Permeability

in Polymer Membranes
DO KYOUNG LEE, YONG WOO KIM, KYUNG JU LEE, BYOUNG RYUL MIN, JONG HAK KIM
Department of Chemical Engineering, Yonsei University, 134 Shinchon-dong, Seodaemun-gu, Seoul 120-749, South Korea

Received 24 August 2006; revised 24 October 2006; accepted 12 November 2006

DOI: 10.1002/polb.21062
Published online in Wiley InterScience (www.interscience.wiley.com).

ABSTRACT: A new molecular thermodynamic model is developed of the gas permeabil-

ity in polymer membranes on the basis of congurational entropy and Flory-Huggins
theory to predict permeability dependence on the concentration of penetrant. Three
kinds of congurational entropy are taken into account by this model; that is, the dis-
orientation entropy of polymer, the mixing entropy, and specic interaction entropy of
polymer/gas. The validity of the mathematical model is examined against experimen-
tal gas permeability for polymer membranes. Agreement between experimental and
predicted permeability is satisfactory. V
C 2007 Wiley Periodicals, Inc. J Polym Sci Part B:

Polym Phys 45: 661665, 2007

Keywords: congurational entropy; Flory-Huggins theory; gas permeation; mem-
brane; thermodynamics

INTRODUCTION polymeric chains, leading to the increase of gas

Gas separation via polymer membranes has
received much attention as an alternative tech-
nology to traditional methods such as cryogenic
distillation, pressure swing adsorption, and
amine-based absorption. During the last two
decades, numerous polymeric materials have
been developed for applications to gas separa-
tion membranes.14 One of the greatest objec-
tives in gas separation membranes is to achieve
high performance membranes with high perme-
ability as well as high selectivity.
A plot of gas permeability versus gas pressure
(or the concentration of gas) is of crucial impor-
tance in the pressure-driven gas separation pro-
cess. It is also helpful in understanding plastici-
zation phenomenon, which is commonly observed
in high performance glassy polymer mem-
branes.5,6 A typical effect of plasticization is the
reduced interaction strength between adjacent
Correspondence to: J. H. Kim (E-mail: jonghak@yonsei.
ac.kr)
Journal of Polymer Science: Part B: Polymer Physics, Vol. 45, 661665 (2007)
V
C 2007 Wiley Periodicals, Inc.
permeability but the decrease of selectivity. From
this background, Freeman group7 presented a
new model describing the concentration and tem-
perature dependences of gas and vapor perme-
ability in rubbery polymer membranes. This
model also gives useful information to estimate
the gas permeability over a wide range of condi-
tions as well as actual operating conditions.
A congurational entropy model for polymer
properties has been originally developed by
Gibbs and Di Marzio8 and Gibbs and co-
workers9,10 and adequately used to interpret the
conductivity behavior of polymer electrolytes.1113
For example, the basis of the Adam-Gibbs
congurational entropy9 model and the Florys
congurational entropy14 in a lattice treatment
has been used to describe the behavior of ionic
conductivity in polymer electrolytes.11,12 Mac-
Farlane et al.13 also investigated the effect of
plasticizer on the conductivity using Adam-
Gibbs model combined with Florys congura-
tional entropy.
In this study, we propose a simple, molecular
thermodynamic model of gas permeability in
661
662 LEE ET AL.


polymer membranes on the basis of the congu-
Hs
rational entropy model810 and Flory-Huggins S S0 exp
5
RT
T0
theory14 to predict the permeation behavior as a
function of gas concentration. The proposed con- From the assumptions by Adam and Gibbs,9
gurational entropy model consists of three con- i.e., (T
T0)  T ln(T/T0), eq 5 can be written as
tributions; i.e., the disorientation entropy of


polymer, the mixing entropy, and specic inter-
Hs
action entropy of polymer/gas. We quantitatively S S0 exp
6
RT lnT=T0
present the gas permeability in polymer mem-
branes dependent on the gas concentration. A relationship between total molar congura-
tional entropy (STotal
c ) and temperature has been
well described as810

ZT
MODEL DEVELOPMENT
Cp
STotal
c T
STotal
c T0 dT
Cp lnT=T0
T
Gas transport through a polymer membrane is T0
considered as an activated process, which is usu- 7
ally described by an Arrhenius type of equa-
tion.15 This equation demonstrates that temper- with DCp the difference in a specic heat
ature may have a large effect on the gas perme- between the equilibrium melt and the glass at
ation rate. The following eq 1 expresses the T0 STotal
c T0 0. Substituting eq 7 into eq 6
temperature dependence of permeability (P) gives
through the membranes.


 B0
Ep S S0 exp
Total 8
P A exp
1 Sc =kB
RT

where A is temperature-independent constant,
Hs
Cp

Ep is the energy of activation, R is the gas con- B0 9
kB RT
stant, and T is temperature.
The permeability depends on both solubility Because permeability is the product of solubil-
and diffusivity, which are the thermodynamic ity and diffusivity, the permeability yields
and transport parameters, respectively. Both pa-


rameters can be also expressed by the Arrhenius B0
P S  D S0 exp
Total
type of equation as follows. Sc =kB



 B0

Hs  D A exp
Total 10
S S0 exp
2 Sc =kB
RT

 where A is assumed the product of S0 and diffu-

Ed sivity (D), which is based on the fact that solu-
D D0 exp
3
RT bility signicantly changes with temperature
compared to diffusivity.
where S0 and D0 are temperature-independent Equation 10 has an analogous form of rela-
constants, DHs is the heat of solution, and Ed is tionship showing ionic conductivity of polymer
the activation energy for diffusion. electrolytes as a function of congurational en-
Combination of eqs 1, 2, and 3 gives eq 4 tropy.11,12 This equation demonstrates that the


 increase of congurational entropy in polymer/

Hs Ed Ep gas system produces the enhancement of the gas
P S0 D0 exp
P0 exp
4
RT RT permeability through the polymer membranes.
Sc for polymer (1)/gas (2) systems consist of
If we assume T is much higher than T0 (the the disorientation entropy of polymer 1 (Sdis
1),
temperature at which the free volume disap- the mixing entropy (Smix
12), and the specic
pears), eq 2 goes to interaction entropy (Sspe
12).
Journal of Polymer Science: Part B: Polymer Physics
DOI 10.1002/polb

Figure 1. The change of congurational entropies of
polymer (1)/gas (2) systems as a function of gas vol-
ume fraction.
STotal
c n1 ; n2 ; T Sdis
1 Smix
12 Sspe
12 11
Since the specic interaction between polymer
and gas reduces the overall entropy, the value of
Sspe
12 is negative. However, other two entropy
terms increase the overall entropy, thereby
being positive values. Each entropy terms are
expressed as follows.11,12,1618

 
z
1
Sdis
1 kB n1 lnr1 r1
1ln 12
e volume fraction of CO2. Experimental data have
Smix
12 kB n1 ln/1 n2 ln/2  13
 
z
1 /1 /
Sspe
12 cspe kB ln ln/1 2 ln/2
e r1 r2
where kB is the Boltzmann constant. /1 r1 n1 =
r1 n1 r2 n2 and /2 r2 n2 =r1 n1 r2 n2 are the
volume fractions of components 1 and 2, respec-
tively. r1 v1/v0 and r2 v2/v0 where v1 and v2
are the molar volumes of components 1 and 2,
respectively, and v0 is the unit lattice volume. z
( 6) is the lattice coordination number. cspe is a
proportionality constant representing the specic
interaction between polymer and gas. From this
model, we can predict a gas permeability through
a polymer membrane with various compositions
of gas.
Journal of Polymer Science: Part B: Polymer Physics
DOI 10.1002/polb
THERMODYNAMIC MODEL OF GAS PERMEABILITY 663
RESULTS AND DISCUSSION
Figure 1 shows the change of congurational en-
tropy of polymer (1)/gas (2) system as a function
of gas concentration. To simplify the system, the
physical parameters for theoretical calculation
are assumed to have the following values; Mw1
1  105, Mw2 44, density1 1.1, density2
0.8, z 6. The disorientation entropy of poly-
mer (Sdis
1) linearly decreased with the increase
of gas concentration, probably because of the
reduced number of available polymer chains.
The mixing entropy (Smix
12) exhibited a maxi-
mum at the volume fraction of gas of 0.35. This
biased behavior comes from the large difference
of molar volume between polymer and gas. The
specic interaction entropy (Sspe
12) always
showed a negative entropy value, implying that
the interaction between two components reduced
the overall entropy of the system. Thus the total
congurational entropy had a maximum of en-
tropy at around 0.1 of gas volume fraction,
above which it approached zero with the in-
crease of gas concentration.
The gas permeabilities by the current theoret-
ical prediction model were compared with the
experimental values. Figure 2 presents the per-
meabilities of CO2 through pristine cellulose ac-
etate (CA) and CA/poly(ethylene glycol) (PEG)
blend (60/40 wt %) membranes as a function of

14
Figure 2. Permeabilities of CO2 through pristine
CA and CA/PEG blend (60/40 wt %) membranes as a
function of volume fraction of CO2. Experimental data
were measured by Li et al.19 The solid lines were cal-
culated from eqs 510, where cspe 2.2, A 6.3, B0
0.008, r1 1, r2 1653 for CA and cspe 1.9, A
7.4, B0 0.14 for CA/PEG membranes.
664 LEE ET AL.
Table 1. List of Molecular Weight, Density, and
Molar Volume for a Model System
Molar
Density Volume
System Mw (g/mol) (g/cm3) (cm3/mol)
CA 1.0  105 1.1 90,910
CA/PEG 6.8  104 1.1 61,820
(60/40 wt %)
EC 1.0  105 1.1 90,910
CO2 44 0.6a 73.3
CH4 16 0.43a 37.2
N2 28 0.8a 35.0
a
The value at liquid state.
been measured by Li et al.19 The solid lines
were calculated from eqs 1014 by using the
physical properties in Table 1. The best ts to
the curvature of the permeability plots were
obtained with adjustable parameters of cspe
2.2, A 6.3, B0 0.008 for CA and cspe 1.9, A
7.4, B0 0.14 for CA/PEG membranes (Table 2).
The CA/PEG blend membrane exhibited higher
permeabilities to CO2 than pristine CA mem-
brane. Li et al.19 demonstrated that the intro-
duction of PEG to CA decreased the solubility of
CO2 in the membranes but greatly increased the
diffusivity, nally leading to the signicant
enhancement of CO2 permeability through the
membranes. The value of cspe in the current ther-
modynamic model represents the strength of spe-
cic interaction between polymer and gas, which
might be strongly correlated with the gas solubil-
ity in the membranes. The lower value of cspe
( 1.9) for CA/PEG blend membrane presents
the weaker specic interaction between polymer
and gas than that of cspe ( 2.2) for CA mem-
brane, which is coincident with the experimental
results of CO2 solubilities in the membranes.19
The pre-exponential factor A would be directly
proportional to the permeability values. Thus,
the CA/PEG blend membrane with higher perme-
ability of CO2 exhibited larger value of A than
pristine CA membrane. The adjustable parame-
ter of B is expected to be directly related to the
activation energy of permeation. In this system,
however, the CA/PEG blend membrane has
higher activation energy than pristine PEG
membrane, indicating that the values of activa-
tion energy in the current model can not be
related explicitly to the ease of permeation.
Figure 3 presents the permeabilities of CH4
and N2 through EC membranes as a function of
Table 2. List of Adjustable Parameters of Given
Systems for Congurational Entropy
Permeability Model
Polymer Gas cspe A B0
CA CO2 2.2 6.3 0.008
CA/PEG CO2 1.9 7.4 0.14
(60/40 wt %)
EC CH4 2.9 11.3 0.005
EC N2 1.9 6.5 0.002
volume fraction of gas. Experimental data have
been reported by Houde and Stem.20 The solid
lines were calculated from the proposed model
with the adjustable parameters of cspe 2.9, A
11.3, B0 0.005 for CH4 and cspe 1.9, A
6.5, B0 0.002 for N2. The proposed perme-
ability model describes very well the gas perme-
ation behavior for both CH4 and N2, in which
the former plasticizes the polymer membranes
signicantly. It has been well known that a
decreasing permeability with increasing gas con-
centration is adquately characterized by dual-
mode model whereas an increasing permeability
is by plasticization phenomena in the mem-
branes.5 The lower value of cspe ( 2.9) for CH4
presents the stronger specic interaction be-
tween polymer and gas than that for N2 (1.9),
explaining the higher solubility of the former in
the membranes. When the volume fraction of
CH4 reached 0.85, the permeability values started
Figure 3. Permeabilities of CH4 and N2 through EC
membranes as a function of volume fraction of gas.
Experimental data were measured by Houde et al.20
The solid lines were calculated from eqs 510, where
cspe 2.9, A 11.3, B0 0.005 for CH4, and cspe
1.9, A 6.5, B0 0.002 for N2.
Journal of Polymer Science: Part B: Polymer Physics
DOI 10.1002/polb

to increase with the increase of CH4 concentra-
tion, which is general permeation behavior of
penetrant-induced plasticization in glassy poly-
mer membranes. The higher pre-exponential fac-
tor (A) of CH4 would be due to its higher perme-
ability than N2. It was also found that the adjust-
able parameter of B was not greatly different
between two gases.
CONCLUSIONS
A new mathematical model was derived based
on congurational entropy and Flory-Huggins
theory to predict the gas permeability through
polymer membranes. The congurational en-
tropy model consists of the disorientation en-
tropy of polymer 1 (Sdis
1), the mixing entropy
(Smix
12), and the specic interaction entropy
(Sspe
12), adequately describing the permeation
behavior in binary polymer (1)/gas (2) systems.
The value of cspe is related to the specic inter-
action between polymer and gas, explaining the
solubility behavior in the membranes. The value
of A is a pre-exponential factor in the Arrhenius
type of equation, which is directly proportional
to the permeability values. The adjustable pa-
rameter of B is presumably related to the activa-
tion energy of permeation. The advantage of our
current model is that a simple algebraic form
with three model parameters appears to be suit-
able for representing the gas permeability
behavior in the membranes.
This work was supported by a Korea Research Foun-
dation Grant (KRF-2005-005-J01402).
Journal of Polymer Science: Part B: Polymer Physics
DOI 10.1002/polb
THERMODYNAMIC MODEL OF GAS PERMEABILITY 665
REFERENCES AND NOTES
1. Toy, L. G.; Nagai, K.; Freeman, B. D.; Pinnau, I.;
He, Z.; Masuda, T.; Teraguchi, M.; Yampolskii, Y. P.
Macromolecules 2000, 33, 2516.
2. Merkel, T. C.; Bondar, V.; Nagai, K.; Freeman, B. D.;
Yampolskii, Yu, P. Macromolecules 1999, 32, 8427.
3. Dai, Y.; Guiver, M. D.; Robertson, G. P.; Kang, Y. S.
Macromolecules 2005, 38, 9670.
4. Dai, Y.; Guiver, M. D.; Robertson, G. P.; Kang, Y. S.;
Lee, K. J. Macromolecules 2003, 36, 6807.
5. Ismail, A. F.; Lorna, W. Sep Purif Technol 2002,
27, 173.
6. Zhong, Z.; Zheng, S.; Mi, Y. Polymer 1999, 40,
3829.
7. Prabhakar, R. S.; Raharjo, R.; Toy, L. G.; Lin, H.;
Freeman, B. D. Ind Eng Chem Res 2005, 44, 1547.
8. Gibbs, J. H.; Di Marzio, E. A. J Chem Phys 1958, 28,
373.
9. Adam, G.; Gibbs, J. H. J Chem Phys 1965, 43, 139.
10. DiMarzio, E. A.; Gibbs, J. H. J Polym Sci Part A:
Gen Pap 1963, 1, 1417.
11. Kim, J. Y.; Bae, Y. C. J Appl Polym Sci 1999, 73,
1891.
12. Kim, J. Y.; Bae, Y. C. Fluid Phase Equilib 1999, 163,
291.
13. MacFarlane, D. R.; Sun, J.; Fasoulopoulos, P.;
Hey, J. Electrochim Acta 1995, 40, 2131.
14. Flory, P. J. Principles of Polymer Chemistry; Cor-
nell University Press: Ithaca, NY, 1953.
15. Ghosal, K.; Freeman, B. D. Polym Adv Technol 1994,
5, 673.
16. Kim, J. Y.; Hong, S. U.; Won, J.; Kang, Y. S. Mac-
romolecules 2000, 33, 3161.
17. Kim, J. H.; Min, B. R.; Won, J.; Kang, Y. S. J Phys
Chem B 2003, 107, 5901.
18. Kim, J. H.; Min, B. R.; Kang, Y. S. Macromole-
cules 2006, 39, 1297.
19. Li, J.; Wang, S.; Nagai, K.; Nakagawa, T.; Mau,
A. W.-H. J Membr Sci 1998, 138, 143.
20. Houde, A. Y.; Stem, S. A. J Membr Sci 1997, 127,
171.