Journal of Membrane Science 595 (2020) 117470

Contents lists available at ScienceDirect

Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci

Non-ideal modelling of polymeric hollow-fibre membrane systems: Pre- T

combustion CO2 capture case study
Ehsan Soroodan Miandoab, Sandra E. Kentish, Colin A. Scholes∗
Peter Cook Centre for Carbon Capture and Storage Research, Department of Chemical Engineering, The University of Melbourne, Australia

ARTICLE INFO ABSTRACT

Keywords: Some membrane gas separation applications operate at high temperature and pressure. However, the majority of
Membrane membrane gas separation models employ simplifying assumptions which are not realistic under these conditions.
Modelling In this study, a rigorous model is developed for polymeric hollow-fibre membrane modules incorporating non-
Pre-combustion CO2 capture isothermal separation (the Joule-Thomson effect), real gas behavior and concentration polarization. The model
Joule-thomson effect
also accounts for temperature-dependent permeability, friction-based pressure loss on both feed and permeate
Real gas behavior
sides and variable physical and transport properties. The rigorous model is applied for pre-combustion CO2
capture, i.e. CO2/H2 separation, and compared with a simplistic model for various polymeric membranes
through changing temperature-independent activation energy of permeation and pre-exponential factor. Two
types of H2- and CO2-selective membranes are then chosen for further analysis. As feed conditions change, the
deviation between the rigorous and simplistic models ranges approximately from 2 to 12% for stage-cut and
2–20% for the permeate composition. The difference is mostly because of real gas behavior at low stage-cuts,
while the Joule-Thomson effect adds to this behavior at high stage-cuts ( 40%) resulting in the greater devia-
tion. The influence of concentration polarization, however, is found negligible even at high stage-cuts.

1. Introduction
Polymeric membrane gas separation is an attractive technology for
many industrial applications; because the process is conceptually
straightforward and energy-efficient, using a semi-permeable polymeric
membrane that enables some gases to pass through easily, while other
gases experience a barrier [1,2]. This provides greater versatility in
separation applications compared to other technologies. Membrane gas
separation is commercialised in natural gas sweetening and air en-
richment [3,4], and research efforts are focused on applying mem-
branes to other gas separation applications. One area in which mem-
brane gas separation has strong potential is CO2 capture; a strategy
proposed to reduce CO2 emissions from large point sources. Pre-com-
bustion CO2 capture in particular holds promise. In this case, an in-
tegrated gasification combined cycle (IGCC) process generates syngas
(mainly CO and H2) through partial oxidation, which is further con-
verted through the water-gas shift reaction into CO2 and H2. The CO2 is
separated from H2 to yield a low-carbon high heating value fuel in
combined cycle plants for power generation [5].
The driving force for membrane gas separation is the chemical po-
tential difference across the polymeric membrane [6]. However, in the
majority of the membrane gas separation literature, difference between
the partial pressure of components (i.e. ideal gas behavior) in upstream
and downstream is considered the driving force for permeation [7].
Given that chemical potential is a complex function of temperature and
fugacity, a rigorous approach should be employed to better describe
permeation through membranes. For most industrial applications, the
systems under consideration are complex, given the high-pressure feed
streams, mixture of gases and vapours, as well as, significant pressure
drop through the membrane. Robust membrane models are therefore
needed to characterise membrane processes under high pressure and
temperature, to better assist in the development of membrane gas se-
paration technology and ensure accurate process simulations are un-
dertaken. Currently, membrane models are severely limited for in-
dustry, as the majority of commercial process simulators do not include
membrane process unit operations in their software suites, or the
membrane simulation is relatively basic. This is also the case in the
academic literature, where membrane models are often simplistic
[8–10]. For these models, the permeance of components is considered
constant, whereas it is inherently dependent on temperature, pressure
and concentration of individual components [11]. The conventional
models often use simple correlations for required physical and transport
properties and neglect the variations of the properties as a result of
change in concentration, temperature and pressure [12].

∗
Corresponding author.
E-mail address: cascho@unimelb.edu.au (C.A. Scholes).

https://doi.org/10.1016/j.memsci.2019.117470
Received 3 May 2019; Received in revised form 12 September 2019; Accepted 13 September 2019
Available online 13 September 2019
0376-7388/ © 2019 Elsevier B.V. All rights reserved.
E. Soroodan Miandoab, et al.
Pre-combustion CO2 capture represents a good model system to
investigate more complex behavior in membrane processes through
simulations, since in this CO2 capture scenario, high temperature and
pressure gas is fed to the CO2/H2 separation unit. Xu et al. [13] con-
ducted a parametric analysis of pre-combustion CO2 capture using
single and double stage membrane separation but the model assumed
isothermal operation and ignored both concentration polarization and
real gas behavior. Similar models which neglect some or all of these
effects have been also developed by Choi et al. [14] and Giordano et al.
[15]. Franz and Scherer developed a more detailed Aspen model to
evaluate the separation and energy efficiency loss of an IGCC power
plant but the model still assumed constant permeance of components
[16].
A number of robust mathematical models capable of simulating
membrane operations have been presented in the literature. Scholz
et al. [17] provided a detailed membrane gas separation model that
included effects such as real gas behavior, non-isothermal operation
(the Joule-Thomson effect) and concentration polarization. They
showed these non-ideal effects adversely influence membrane gas se-
paration. However, physical and transport properties were taken as
constant at feed and permeate conditions. In reality, these properties
can change, as temperature, pressure and concentration of the gas vary
along the membrane module.
Here, a rigorous membrane model incorporating non-isothermal
operation, real gas behavior, concentration polarization, pressure loss
along the module, temperature-dependent permeance and variable
physical and transport properties is presented and applied to a mem-
brane process undertaking pre-combustion CO2 capture. The rigorous
model is validated against experimental data presented in the literature
and compared with a simplistic model to demonstrate the deviation in
performance due to individual effects of real gas behavior, non-iso-
thermal operation and concentration polarization.
2. Model development
2.1. Methodology
The mathematical model is based on an asymmetric hollow-fibre
membrane module, i.e. a dense polymeric membrane layer on a porous
support. To build the rigorous model, as illustrated in Fig. 1, the
membrane module is discretized into N cells for each of which the mass,
energy and momentum balance equations are simultaneously solved
with the membrane transport equation.
The model can simulate co-current and counter-current operations,
shell-side or lumen-side feed and constant or variable gas perme-
abilities, depending upon the nature of membrane and process being
studied. The model also enables a sweep gas stream to be added, if it is
assumed impermeable to the membrane. This last assumption can be
justified if the feed gas is at high pressure and the sweep gas at ambient
pressure, as the partial pressure of the sweep gas that will permeate is a
trivial fraction of the feed.
Fig. 1. Membrane module consisted of N cells (counter-current operation is
shown here).
2
Journal of Membrane Science 595 (2020) 117470
2.2. Mathematical model
Assumptions governing the proposed model are as follow:
1. The membrane module operates at steady-state.
2. Laminar plug flow occurs on both the shell and lumen sides of
hollow fibres. Depending upon the feed flow rate being studied, the
Reynolds numbers for the current simulation are less than 300 on
both the retentate and permeate sides. By this simplification, the
momentum balance is reduced to the Hagen-Poiseuille equation.
This equation is then used to evaluate the pressure drop in each cell.
Consequently, the pressure profile of the retentate and permeate
sides can be determined along the membrane module.
3. For the current simulation, the permeability of components is as-
sumed independent of pressure and concentration. Permeability and
selectivity are then assumed temperature-dependent. However, at a
later stage the proposed model can be easily modified to account for
pressure and concentration dependent permeabilities.
4. The concentration of components changes from the bulk flow to the
membrane surface and inside the porous support due to con-
centration polarization.
5. The thickness of the dense layer, as well as, the inner and outer
diameter of fibres remain uniform under high-pressure operation.
In this work, the rigorous model is compared with a simplistic
model which ignores real gas behavior, concentration polarization and
pressure drop due to friction and assumes constant permeability, iso-
thermal operation and hence, constant physical and transport proper-
ties along the membrane module.
The degree of freedom (DOF) for the rigorous model in terms of the
number of variables that must be fixed is given by:
DOF = 6 + 2 × (number of components ) (1)
Once the model is provided with the DOF number of fixed variables, a
simulation can be successfully solved. For example, in case of a binary
separation, it is required to fix 10 stream variables (i.e. flow rate, tem-
perature, pressure and composition) in advance to perform the simulation.
Fig. 2 represents a single cell of the membrane module operating at
the counter-current mode where subscripts “R” and “P” represent in-
formation on the retentate and permeate sides, respectively. Super-
scripts “in” and “out” delineate inlet and outlet variables of the single
cell, respectively. In the current simulation, variables F, T, P, x and y
stand for flow rate, temperature, pressure, mole fraction on the re-
tentate side and mole fraction on the permeate side, respectively.
The overall mole balance for the single cell is described as:
FRin + FPin = FRout + FPout (2)
Likewise, the mole balance of an individual component i is given by:
FRin xiin + FPin yiin out out out out
, P = FR x i, R + FP yi, P (3)
,R
Fig. 2. A single cell of the membrane module.
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

for which: where variables and parameters with superscript “S” are evaluated in
C the porous support of the hollow-fibre membranes, Ji and Jt are trans-
xiout
,R = 1
membrane flux of component i and total transmembrane flux of all
i=1 (4) components, respectively, ρt is total molar density of components, k is
C
local mass transfer coefficient, D is diffusivity and ε is the porosity of
yiout the porous support. To obtain the mole fraction of components at the
,P = 1
i=1 (5) membrane surface and in the porous support, mass transfer coefficients
should be calculated by applying appropriate empirical correlations for
where i = 1, 2, 3, …, C represents the index of components and C is the
Sherwood numbers. For laminar flow regime with the Reynolds number
number of streams components. To accomplish a mole balance, an
(Re) less than 2100, the Sherwood numbers (Sh) of the retentate and
appropriate equation is needed to describe the transmembrane flow
permeate streams are evaluated by [21]:
rate. With a conservative approach, the mass transport across the
polymeric membranes is obtained by:
1
ki, R dhyd dhyd 3
ShR = = 1.62 ReR ScR
i Di, R l (15)
FPout yiout FPin yiin
,P = A ( M x M PRout M M out
i, P yi PP )
,P M c i, R i (6)
1
ki, P di d 3
where: ShP = = 1.62 ReP ScP i
Di, P l (16)
d o Nf l
Ac = where di is fibre internal diameter. The hydraulic diameter of the
N (7)
membrane module (dhyd) is given by [19,22,23]:
Here, parameters and variables taking superscript “M” refer to the
dense layer, is permeability, δ is thickness, Ac is the active membrane 1
dhyd = do
(17)
area in a single cell, do is fibre external diameter, Nf is the number of
hollow-fibre membranes, l is active fibre length, N is the number of cells for which the packing fraction of the module, Ψ, is determined by
and, is the fugacity coefficient (the fugacity coefficients and mole [19,22]:
fractions on the right-hand side of Equation (6) are evaluated at the 2
surfaces of the dense layer facing the retentate and permeate sides). For do
= Nf
high-pressure applications, such as pre-combustion CO2 capture, de- dm (18)
viation from the ideal gas behavior may lead to inaccuracy in the si-
where dm is the module diameter. To be consistent with the variables
mulation outcomes. Equation (6) considers the fugacity difference of
used for the present simulation, the Reynolds numbers of the retentate
components in the retentate and permeate sides as the driving force of
and permeate streams are estimated by [23]:
permeation [18]. The fugacity coefficients in Equation (6) were calcu-
lated using the composition of components at the membrane surface, as 4FRout MWR
ReR =
well as, temperature and pressure of the retentate and permeate flow in Nf do µR (19)
each cell. The Soave-Redlich-Kwong (SRK) equation of state (EOS) was
used to calculate the fugacity coefficients [19]. 4FPout MWP
ReP =
Because of concentration polarization, the permeation of the more Nf di µP (20)
permeable component is retarded, while it is accelerated for the less
in which MW and μ delineate total molecular weight of components and
permeable component [20]. If the feed stream enters the shell side and
viscosity, respectively. Additionally, Schmidt numbers (Sc) of the re-
the dense membrane layer is located at the external surfaces of hollow-
tentate and permeate sides are calculated by [24]:
fibres, the governing equations considering concentration polarization
are provided by [21]: µR
ScR =
t , R Di, R (21)
x iout
,R xi Jt
= exp µP
xiM xi t , R k i, R (8) ScP =
t , P Di, P (22)
yiM xi Jt S
= exp Friction forces inside fibres and on shell, create pressure drop (ΔP)
yiS xi t
S
DiS (9) leading to the decrease in the transmembrane driving force of per-
meation along the membrane module. Assuming laminar plug flow for
yiS xi Jt both sides, the pressure drop of the retentate and permeate sides is
= exp
yiout
,P xi t , P k i, P (10) given by [8]:
128µR l
where: PR = FRout
2 2
t , R dhyd (dm Nf do2) N (23)
i M M out M M out
J M ( i, R x i PR i, P yi PP )
xi = i = C 128µP l
Jt i
( M x M PRout
i = 1 M i, R i
M M out
i, P yi PP ) (11) PP = 4
FPout
t , P di Nf N (24)
C
For the polymeric membranes, if permeation is temperature-de-
xiM = 1
i (12) pendent, the permeability of components is usually described by an
Arrhenius type equation [25]:
C
yiM = 1 Ep, i
(13) i = i,0 exp
i RTRout (25)
C
where i,0 and Ep,i are temperature-independent pre-exponential factor
xiS = 1
i
and activation energy of permeation for component i, respectively, and
(14) R is the ideal gas law constant (8.314 J/mol.K). To provide an insight

3
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

into the composition of the retentate and permeate streams, membrane where h and λ are local convective heat transfer coefficient and thermal
selectivity should be taken into consideration [2]. For a binary mixture conductivity, respectively. Here, feed is assumed on the shell side. Assuming
of components i and j, temperature-dependent selectivity of i with re- laminar flow within the shell and in the lumen, convective heat transfer
spect to j is defined as [6,26]: coefficients of the retentate and permeate sides are determined by [19]:

= i
=
i,0 exp ( Ep, i
RTRout )= 0
exp
E
hP = 3.66 P
di (34)
exp ( )
i/j i/j
j
Ep, j RTRout
j,0 RTRout (26) M)
1
R ReR PrR (d o + 2 3
hR = 3.66 + 1.077
in which: do + 2 M 2dhyd
(35)
0
0
i/j = i
0 where the Prandtl number (Pr) of the retentate side is defined as:
j (27)
cR µR
PrR =
E = Ep, i Ep, j (28) R (36)

Selectivity is used as a criterion for determining the purity of where c delineates specific heat capacity. The effective conductivity of the
components in the permeate and retentate streams. Applying Equation porous support (λS) is estimated by the conductivities of the polymeric
(6) for i and j, and dividing the transmembrane flow rate of i by j results material forming the support (λSP) and permeate gas molecules inside the
in: pores of the support (λP) [19]:

FPout yiout
,P FPin yiin
,P
S = SP + (1 ) P (37)
FPout (1 yiout
,P ) FPin (1 yiin
,P ) The thermal conductivity of many polymers are reported in the
M M out M M out literature [28–30]. In the current simulation, the polymer thermal
0 E i, R x i PR i, P yi PP
= i/ j exp R M conductivity was fixed at 0.2 W/mK (λM,λSP) [19].
RTout j, R (1 x iM ) PRout M
j, P (1 yiM ) PPout (29) The rigorous model is coded in Aspen Custom Modeller (V9) which
In the model, appropriate values of ΔE and α are required as input. 0 provides access to the component database and thermodynamic
If the information of the previous cell (indicated by superscript “in”) packages of Aspen Properties. The present model can easily be exported
and the mole fraction of components at the membrane surface are to Aspen Plus to enable simulations of more complex processes.
known, Equation (29) is solved to give the mole fractions of compo- Fig. 3 demonstrates the methodology of modelling. Gas components and
nents in the permeate stream. Once the composition of the permeate the thermodynamic package (Equation of State, EOS) are defined in Aspen
stream is obtained, the mole fraction of components in the retentate Properties and imported to the membrane model. The geometric char-
stream can be calculated by solving the mass balance equations in each acteristics of the hollow-fibre membrane module (number of fibres, inner
cell. and outer diameter etc.) and number of cells (which was fixed at 100 in this
Enthalpy exchange along the membrane module and between the work) must also be input into the model. The membrane model is made up
retentate and permeate streams accounts for temperature variations. of two sub-models namely “cell” and “membrane transport” providing it
This is likely to happen when membrane gas separation occurs at high with a solution to the mass, momentum and energy transport equations and
pressure [27]. When this is the case, the permeability of components the permeability of components, respectively. When the simulation has
may be changing as temperature varies within the membrane module. converged, flow rate, composition, temperature and pressure, as well as,
Hence, to obtain the accurate results of simulation, an energy balance is enthalpy and density of the retentate and permeate streams will be de-
considered. The overall energy balance equation for the single cell is termined through solving transport equations, connecting the cells together
expressed as: and linking them to the feed, retentate and permeate streams.
Aspen Custom Modeller utilises an equation-oriented programming
FRin HRin + FPin HPin = FRout HRout + FPout HPout (30) language (i.e. declarative) in which mathematical equations are decom-
where H indicates enthalpy per unit mole. As the high-pressure gas posed to break down the sets of equations into groups. The groups are then
passes across the membrane into a lower pressure environment, isen- solved sequentially by choosing an appropriate solver [31]. For the present
thalpic expansion happens similar to that which occurs across a throt- simulation, Newton's method was employed. The maximum number of
tling valve. This causes a significant change of gas temperature. Such iterations was fixed at 100 with convergence achieved once the following
temperature variations are often referred to as the Joule-Thomson ef- condition is met:
fect [19,27]. As permeation continues, the permeation enthalpy is
< 10 10
transported from the retentate to the permeate side. Due to gas ex-
(38)
pansion from high to low pressure, a temperature difference between
the retentate and permeate streams may be induced, resulting in a net where represents any unknown variables used in the membrane model
rate of heat transfer (q) across the membrane material. These heat ef- and indicates the difference between the values of the unknown vari-
fects are summarized as: able in two successive iterations. Newton's method relies on an accurate
C
initial guess for unknown variables and hence, a detailed initialization
FRin HRin = FRout HRout + HRin i
A ( M x M PRout M M out procedure was applied to ensure the convergence of the simulation.
M c i, R i i, P yi PP ) +q
i=1 (31) The model also allows for easy calculation of physical and transport
properties including density, viscosity, diffusivity, thermal con-
q = UO Ac (TRout TPout ) (32) ductivity, specific heat capacity and molecular weight of the retentate
for which UO is overall heat transfer coefficient. Coker et al. [19] and permeate sides in each cell. The properties were calculated in each
conducted a theoretical analysis and obtained the overall heat transfer cell by calling built-in procedures of Aspen Custom Modeller. These
coefficient by: procedures need temperature, pressure and mole fraction as input
1
parameters. In this simulation, the temperature, pressure and mole
M M M M
UO =
1 do + 2
+
d o do + 2
ln
do d +
+ o M ln
do + 2
+
1 fraction of the retentate and permeate flow exiting the cells (i.e.
hP di 2 S do di do hR TRout , TPout , PRout , PPout , xiout , yiout ) were considered as input to these built-in
(33) procedures to calculate physical and transport properties. Hence, the

4
E. Soroodan Miandoab, et al.
Fig. 3. Methodology of modelling.
properties varied within cells. The enthalpies of the retentate and
permeate flow in each cell were also calculated in the same manner.
3. Results and discussion
3.1. Model validation
The proposed membrane model is validated with experimental data
published by Pan [32], Sidhoum [33] and Feng et al. [34]. The per-
formance of the membrane modules is represented by the dimensionless
parameter stage-cut which is the ratio of feed flow rate to permeate
flow rate. The activation energies of permeation have not been reported
for the systems described by Pan [32] and Sidhoum [33], and hence,
constant permeabilities have been used in these cases. Temperature-
dependent permeabilities are used for the air separation system re-
ported by Feng et al. [34].
The details of the membrane gas separation systems and the full model
fitted to the experimental data are provided in Table S1 and Figs. S1–S4 of
the Supplementary Information. In the current simulation, the number of
cells (N) was fixed at 100, as selecting values greater than 100 did not
affect the simulation results. This is shown in Fig. S2 of the Supplementary
Information along with the model fitted to the experimental data of Sid-
houm. As N is doubled in value, there is no significant difference between
the model predictions with N = 100 and N = 200. The residual error
between the model calculations and the experimental data of Sidhoum and
Feng et al. was less than 1%, while it was less than 3% for the work of Pan.
This relatively larger error is within the experimental error (3.5%) re-
ported by Pan. The membrane model successfully simulated the different
sets of experimental data highlighting the rigorous approach of the model
to be applied for multi-component systems at both high or low feed
pressures, as well as, temperature-dependent permeation case.
3.2. Pre-combustion CO2 capture simulations
Having validated the model, simulations of CO2/H2 separation were
conducted. These simulations assumed a base case of a feed gas of 60%
CO2 and 40% H2 at 100 °C and 60 bara, with a permeate pressure of 1.5
5
Journal of Membrane Science 595 (2020) 117470
bara. Table 1 represents the geometrical characterises of the hollow-
fibre membrane module used for this case study.
3.2.1. Comparison of the rigorous and simplistic models: effect of membrane
selectivity
The parameters influencing selectivity are the temperature-in-
dependent activation energy of permeation and pre-exponential factors.
The case study varied the selectivity parameters to compare the rigorous
and simplistic models at 50% stage-cut in terms of the compositions of the
retentate and permeate streams (Fig. 4). The parameters of the tempera-
ture-dependent selectivity are provided for selected polymers in Table 2.
While constant H2/CO2 selectivity at the feed temperature (100 °C) is
used for the simplistic model, this changes along the membrane module in
the rigorous model due to temperature variations calculated from the
energy balance. Compared to the simplistic model, the rigorous model
predicts a retentate stream that is 2–5% lower in CO2 concentration and a
permeate stream which has 0.5–8% more CO2 at 50% stage-cut (Fig. 4).
Conversely, at 10% stage cut, the difference between the two models is
less than 1% and 5–12% for the composition of the retentate and permeate
streams respectively (see Fig. S5 of the Supplementary Information).
As ΔE increases, the H2/CO2 selectivity decreases (Equation (26))
and the retentate and permeate mole fractions of CO2 obtained by both
models tend to converge. This behavior is independent of the values of
α0 implying that the separation performance of polymers having large
ΔE, i.e. low H2/CO2 selectivity, can reasonably be simulated by the
simplistic model. Conversely, the use of the simplistic model for the
Table 1
Characteristics of the hollow-fibre membrane module [19].
Feed Shell-side
Fibre internal diameter 150 μm
Fibre external diameter 300 μm
Number of fibres 500000
Fibre length 0.8
Thickness of separative layer 0.5 μm
Porosity of the porous support 0.5
Module diameter 0.3048 m
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

Fig. 4. Mole fraction of CO2 in the retentate and permeate streams obtained by the rigorous (solid line) and simplistic (dash line) models as a function of the
difference in CO2 and H2 activation energies of permeation and pre-exponential factor. Feed is 60% CO2 at 60 bara and 100 °C. Stage-cut is fixed at 50%.

membranes with high H2/CO2 selectivity may make errors in simula-
tion. Polymers having a small gap between the activation energies of
the components can be good candidates for CO2/H2 separation. Refer-
ring to Table 2, this suggests that PBI has potential for CO2/H2 se-
paration, as the difference in the activation energies of CO2 and H2 is
least among similar polymers. More CO2 is recovered from the retentate
side and less CO2 is withdrawn from the permeate side as the pre-ex-
ponential factor increases from 5 to 45 (Fig. 4). A high value is thus
desirable, although this parameter has less impact across the range of
values presented in Table 2 than the activation energy.
3.2.2. Comparison of the rigorous and simplistic models: effect of feed
temperature, pressure and composition
H2-selective ( H0 2 / CO2 = 25, ΔE = 2000 J/mol) and CO2-selective
( CO2/ H2 = 0.2, ΔE = 12000 J/mol) membranes were next compared
0
( E = Ep, H2 Ep, CO2 ). The deviation between the rigorous and sim-
plistic models was highlighted for these two membranes and their
ability to separate CO2/H2 for various feed conditions (composition,
temperature and pressure) was investigated.
The exiting permeate flow rate and stage-cut observed within the
membrane module as a function of feed temperature and flowrate are
provided in Fig. 5. As feed flow rate increases, the proportion per-
meating (the stage-cut) decreases. In turn, this means that the retentate
composition remains closer to that of the original feed. This increases
the average driving force for permeation (the difference in composition
between feed and permeate) leading both to a greater permeate flow-
rate (Fig. 5) and a greater permeate purity (Fig. 6).
Increased feed temperature results in an increased permeate flow
rate for both membranes, and this leads to a greater stage-cut at a fixed
feed flow rate. Similarly, the membrane selectivity decreases with
temperature for both membranes, reflecting the positive values of ΔE
used in Equation (29). This results in greater permeate purities at lower
temperature in both cases (Fig. 6). However, achieving higher permeate
purities comes at the expense of lower purity of the less permeable
component in the retentate stream. The deviation between the sim-
plistic and rigorous models developed here, is also shown in Figs. 5 and
6. The deviation ranges from being negligible to 20% depending upon
the process parameters. The differences are generally greater at lower
feed temperatures, as the Joule-Thomson effect escalates when tem-
perature is reduced (see Section 3.3).
The rigorous model predicts permeate flow rates slightly larger than
what is given by the simplistic model for the H2-selective membrane,
while for the CO2-selective membrane the permeate flow rates are
lower. Similarly, the mole fraction of H2 is somewhat higher for the
rigorous model than the simplistic model (Fig. 6) for the H2-selective
membrane, while the reverse is true for most feed flowrates for the CO2-
selective membrane. These changes reflect the fact that H2 has a fu-
gacity coefficient greater than unity, while that for CO2 is less than
unity (see Section 3.3 below).
Figs. 7 and 8 display the difference between the rigorous and sim-
plistic models as a function of feed pressure and composition at 50%
and 10% stage-cuts, respectively. In pre-combustion CO2 capture, the
feed concentration of CO2 can vary from 15 to 60% (dry basis) with
total feed pressure ranging from 20 to 70 bar [40]. Here, the results are

Table 2
Pre-exponential factor ( H2 / CO2 )
0
and activation energy (ΔE) for the H2/CO2 selectivity of selected polymers.

Polymer Ep (kJ/mol) ΔE (kJ/mol) 0

H2
Reference
0 =
H2 / CO2 0
CO 2
H2 CO2 Ep, H2 Ep,CO2

Polydimethylsiloxane (PDMS) 13.4 2.2 11.2 26 [26]

Poly (ether-b-amide) (PEBAX) 12.3 2.7 9.6 26 [35]
Polyethylene oxide (PEO) 76 70 6 60 [36]
Polybenzimidazole (PBI) 10 9.2 0.8 24 [37]
24 21.9 2.1 28 [38]
Polyurethane (PU) 41 33.8 7.2 33 [39]

6
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

Fig. 5. Permeate flow rate and stage-cut obtained by the rigorous (solid line) and simplistic (dash line) models as a function of feed temperature and flow rate. (a) and
(c): H2-selective ( H0 2 / CO2 =25, ΔE = 2000 J/mol), (b) and (d): CO2- selective ( CO
0
2 / H2 = 0.2, ΔE = 12000 J/mol). Feed is 60% CO2 and at 60 bara.

demonstrated for 20, 40 and 60% CO2 (H2 balance) in the feed at
pressures ranging from 30 to 60 bara. The deviation between the two
models increases at elevated feed pressures. This is due to the stronger
impact of real gas behaviour and Joule-Thomson effects at higher
pressures (see Section 3.3). As CO2 concentration increases, the devia-
tion enlarges for the H2-selective membrane, so that it ranges from
0.2% (20% CO2) to 2% (60% CO2) at 50% stage-cut (Fig. 7a). However,
this difference is only 0.3% on average for the CO2-selective membrane
(Fig. 7b). As stage-cut is reduced to 10%, an average deviation of 0.8%
is calculated for the H2-selective membrane, while an average differ-
ence of 1.2% is obtained for the CO2-selective membrane (Fig. 8b). This
reflects the greater magnitude of the Joule-Thomson effect on CO2 re-
lative to H2 at high stage-cuts and the larger departure from ideal gas
behaviour at lower stage-cuts (see Section 3.3 below).

Fig. 6. Mole fraction of components in the permeate stream obtained by the rigorous (solid line) and simplistic models (dash line for 100 °C, dot line for 200 °C) as a
function of feed temperature and flow rate. (a): H2-selective ( H0 2 / CO2 =25, ΔE = 2000 J/mol), (b): CO2- selective ( CO
0
2 / H2 = 0.2, ΔE = 12000 J/mol). Feed is 60%
CO2 and at 60 bara.

7
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

Fig. 7. Mole fraction of components in the permeate stream obtained by the rigorous (solid line) and simplistic (dash line) as a function of feed pressure and
composition. (a): H2-selective ( H0 2 / CO2 =25, ΔE = 2000 J/mol), (b): CO2- selective ( CO
0
2 / H2 = 0.2, ΔE = 12000 J/mol). Feed is at 100 °C. Stage-cut is fixed at 50%.

3.3. Individual influence of real gas behavior, concentration polarization
and non-isothermal operation
Fig. 9 indicates the influence of real gas behavior on the H2-selective
membrane at 50% stage-cut, by plotting the driving force of permeation
and fugacity coefficient along the membrane module. The fugacity
coefficient of CO2 is approximately 0.9 due to high pressure operation
(60 bara) and mildly decreases along the module on the retentate side.
This, in turn, means that the driving force for permeation decreases
compared to the ideal behavior. On the other hand, the fugacity coef-
ficient of H2 is 1.1 on the retentate side and increases along the module
resulting in a greater driving force of permeation in comparison to the
simplistic model. Consequently, the rigorous model calculates a greater
permeate side mole fraction of H2 than the simplistic model (Fig. 6). In
the Supplementary Information, Fig. S6 provides the H2-selective
membrane with the driving force of permeation and fugacity coeffi-
cients at 10% stage-cut where trends similar Fig. 9 are observed. For
both stage-cuts, the fugacity coefficients of CO2 and H2 are 0.9 and 1.09
on average. This results in an approximately constant difference be-
tween the driving force of permeation (3.4 bara for CO2 and 2.1 bara for
H2). Therefore, real gas behavior exists over a wide range of stage-cuts.
For the CO2-selective membrane, the driving force of permeation
and fugacity coefficients of components are presented in Fig. 10 and
Fig. S7 of the Supplementary Information at 50% and 10% stage-cuts,

Fig. 8. Mole fraction of components in the permeate stream obtained by the rigorous (solid line) and simplistic (dash line) as a function of feed pressure and
composition. (a): H2-selective ( H0 2 / CO2 =25, ΔE = 2000 J/mol), (b): CO2- selective ( CO
0
2 / H2 =0.2, ΔE = 12000 J/mol). Feed is at 100 °C. Stage-cut is fixed at 10%.

8
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

Fig. 9. Driving force of permeation for the H2-selective membrane ( H0 2 / CO2 =25, ΔE = 2000 J/mol). Fugacity coefficients of components in the retentate and
permeate streams along the membrane module. Feed is 60% CO2 at 60 bara and 100 °C. Stage-cut is fixed at 50%.

respectively. With 50% stage-cut, the fugacity coefficient of CO2
slightly increases from 0.9 to 0.92 on the retentate side (Fig. 10). This
reduces permeation flux compared to the simplistic model. Conversely,
the driving force of H2 increases with respect to the simplistic model.
This is because H2 shows a small positive departure from the ideal gas
behavior with the fugacity coefficient above unity. The results clearly
demonstrate the existence of real gas behavior at different stage-cuts.
The ratio of the mole fraction on the retentate side (x iout
, R ) to the mole
fraction at the membrane surface ( x iM ) is defined as parameter and
plotted versus the module length (Fig. 11). At the low stage-cut (10%),
the parameter is essentially equal to unity meaning that the gas
composition does not change from the bulk retentate flow to the
membrane surface. This is expected, as concentration polarization is
reduced at lower stage-cuts [41]. The parameter θ decreases and
becomes less than unity with increasing stage-cut, as concentration
polarization increases due to the greater transmembrane flux [41].
Although the trends observed in Fig. 11 confirms the existence of
concentration polarization, the magnitude of is still close to unity for
even the extreme cases. For example, a stage-cut of 0.9 corresponds to θ
of 0.9985 on average for the membrane systems studied. This confirms
that the influence of concentration polarization on the separation per-
formance of the membrane module will be negligible.
The impact of the non-isothermal operation on the H2-selective
membrane is illustrated in Fig. 12. The temperature change along the
module is directly proportional to the feed pressure and stage-cut and
increases at higher stage-cuts and elevated feed pressures. A similar
trend was observed in the literature [19,42,43]. The temperature
change from feed to retentate is positive, meaning that the retentate

Fig. 10. Driving force of permeation for the CO2-selective membrane CO2- selective ( CO
0
2 / H2 = 0.2, ΔE = 12000 J/mol). Fugacity coefficients of components in the
retentate and permeate streams along the membrane module. Feed is 60% CO2 at 60 bara and 100 °C. Stage-cut is fixed at 50%.

9
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

Fig. 11. The ratio of the mole fraction on the retentate side ( x iout
, R ) to the mole fraction at the membrane surface ( x i ) for H2 and CO2. (a) the H2-selective membrane
M

( H0 2 / CO2 =25, ΔE = 2000 J/mol) and (b) the CO2-selective ( CO 0

/
2 2 H = 0.2, ΔE = 12000 J/mol). Feed is 60% CO 2 at 60 bara and 100 °C.

temperature is higher than the feed temperature. Such a positive tem-
perature change from feed to retentate was predicted by Gorissen [27],
for synthesis gas (carbon monoxide (CO) and H2) separation at 100 bara
and 52% stage-cut. A qualitative explanation for the heating of the
retentate stream may be described by using the isenthalpic Joule-
Thomson coefficient given by [44]:
T 1 V brane, expansion driven cooling is seen for this CO2-selective mem-
µJT = = T V
P H cp T (39)
in which V and cp represent the molar volume and specific heat capa-
city, respectively. The Joule-Thomson coefficients of CO2 and H2 are
calculated at different temperatures (Table 3) using molar volume and
specific heat capacity data provided by Aspen Properties. These values
are in a good agreement with the literature [45,46].
For the H2-selective membrane, H2 is the dominant permeating
component. As the Joule-Thomson coefficient of H2 is negative within
the temperature range of the current study, the overall temperature
deviation is positive, leading to the heating of the retentate stream.
The temperature change along the module for the CO2-selective
membrane is provided in Fig. 13. In contrast to the H2-selective mem-
brane. This is because CO2 is the dominant permeating component and
has positive Joule-Thomson coefficients within the temperature range
of the current study (Table 3). Compared to the H2-selective membrane,
the change in temperature is greater for the CO2-selective membrane, as
the Joule-Thomson coefficient of CO2 is much larger than that for H2.

Fig. 12. Temperature change from feed to retentate for the H2-selective membrane ( H2 / CO2 =25,
0
ΔE = 2000 J/mol) as a function of feed pressure. Feed is 60% CO2.

10
E. Soroodan Miandoab, et al.
Table 3
The Joule-Thomson coefficients of CO2 and H2 at 60 bara.
Temperature (°C) CO2 H2
25 1.24 −0.031
50 0.98 −0.034
100 0.57 −0.038
150 0.45 −0.04
200 0.36 −0.042
For the feed pressure and temperature of 80 bar and 100 °C, respec-
tively, there is a substantial temperature change of 39 °C for a 50%
stage-cut (feed flow rate of 3.2 kmol/h) and 58 °C for a 60% stage-cut
(feed flow rate of 2.3 kmol/h). Such a large temperature drop was also
predicted by Coker et al. for CO2 separation from methane (CH4), with a
temperature drop of 40 °C. This temperature drop was calculated at a
53% stage-cut and corresponded to the feed flow rate of 12.8 kmol/h
[19]. The Coker study neglected real gas behavior and assumed con-
stant physical and transport properties despite operating at elevated
feed pressure and high concentration of CO2 (which would normally
cause non-ideality in the system).
The temperature change should affect the selectivity by changing
permeability (see Equation (25)) and hence, the rigorous model should
diverge from the simplistic model due to these effects. Fig. 14 illustrates
the selectivity of the H2- and CO2-selective membranes along the
membrane module as a result of these temperature variations. How-
ever, for the H2-selective membrane, there is a very small change in H2/
CO2 selectivity, although this variation is slightly greater at 100 °C
compared to 200 °C. For the CO2-selective membrane, the CO2/H2 se-
lectivity increases by around 10% and 40% at the feed temperatures of
200 °C and 100 °C, respectively. The ratio of the Joule-Thomson coef-
ficient of CO2 to H2 drops from 15 to 8.5 as temperature rises from
100 °C to 200 °C. This decline results in the smaller temperature change
at 200 °C, as confirmed by Figs. 12 and 13. This can also be inferred
from Fig. 6, where the mole fraction of CO2 is larger at the lower
temperature (100 °C). The average Joule-Thomson coefficient of the
mixed H2/CO2 gas stream will decrease at 200 °C due to the lower CO2
content, leading to a smaller change in temperature at 200 °C compared
to 100 °C. Consequently, this smaller temperature change means that
Fig. 13. Temperature change between feed and retentate for the CO2-selective membrane
60% CO2.
11
Journal of Membrane Science 595 (2020) 117470
the variation in the selectivity is much less significant at 200 °C
(Fig. 14). Non-isothermal operation clearly has more impact on the
CO2-selective membrane at lower temperatures.
For the H2-selective membrane, concentration polarization does not
affect the simulation results (Fig. 15). This conclusion is in agreement
with the study by Mourgues and Sanchez in which they conducted a
theoretical analysis of concentration polarization in membrane gas se-
paration and came to the conclusion that the phenomenon becomes
noticeable when the permeances of components are greater than 1000
GPU and the selectivity of more permeable to less permeable compo-
nent is higher than 100 [41]. For the current study, the selectivity of
both H2- and CO2-selective membranes is below 100 (Fig. 14) and
hence, concentration polarization is negligible. This is also confirmed
by the discussion presented for Fig. 11. As feed flow rate increases, the
rigorous model converges on the model taking only real gas behavior
into account. This means the rigorous model is essentially described by
real gas behavior at lower stage-cuts. For the feed flow rates less than
20 kmol/h, i.e. at stage-cuts higher than 45%, the rigorous model
slightly diverges from the curve of real gas behavior and approaches to
the model for which only non-isothermal operation, i.e. Joule-Thomson
effect, is considered. This is verified by Fig. 12 where the Joule-
Thomson effect is more pronounced at higher stage-cuts.
Fig. 16 summarises how the individual effects of the rigorous model
influence the CO2-selective membrane. The simplistic model converges
on a model that solely takes concentration polarization into con-
sideration. Just like the H2-selective membrane, this implies that con-
centration polarization has almost no impact on the separation per-
formance of the CO2-selective membrane (see Fig. 11). The rigorous
model lies between the curves of real gas behavior and non-isothermal
operation within the flow rates (i.e. stage-cuts) studied in this simula-
tion. Similar to the H2-selective membrane, the rigorous model tends
towards the real gas behavior at higher feed flow rates, i.e. lower stage-
cuts. However, compared to the H2-selective membrane (Fig. 15), these
curves are not completely congruent. Given that there is no significant
Joule-Thomson effect at lower stage cuts (Fig. 13), this divergence
implies that CO2 causes greater deviation from ideal gas behavior
compared to H2. By reducing the feed flow rate, i.e. increasing stage-
cut, the influence of the Joule-Thomson effect is more induced (Fig. 13),
as the rigorous model diverges from real gas behavior and rather
0
CO2 / H2 = 0.2, ΔE = 12000 J/mol) as a function of feed pressure. Feed is
E. Soroodan Miandoab, et al. Journal of Membrane Science 595 (2020) 117470

Fig. 14. Selectivity of (a): H2-selective ( H0 2 / CO2 =25, ΔE = 2000 J/mol), (b): CO2- selective ( 0
CO2 / H2 = 0.2, ΔE = 12000 J/mol). membranes along the membrane
module. Feed is 60% CO2 at 60 bara. Stage-cut is fixed at 50%.

Fig. 15. Influence of individual non-ideal effects on the mole fraction of H2 in the permeate stream as a function of the feed flow rate for the H2-selective membrane
( H0 2 / CO2 =25, ΔE = 2000 J/mol). The dashed curve represents both the simplistic model and one including concentration polarization, as these curves are congruent.
Feed is 60% CO2 at 60 bara and 100 °C.

approaches to the curve of non-isothermal operation. This confirms that
it is the combination of these two effects determining the specifications
of the retentate and permeate streams.
4. Conclusion
A model of single-stage membrane gas separation has been devel-
oped for hollow-fibre modules. The model utilises non-isothermal op-
eration with temperature-dependent permeance, concentration polar-
ization, real gas behavior, frictional pressure loss and variable physical
and transport properties along the membrane module. The model has
been coded in Aspen Custom Modeller enabling end-users to have ac-
cess to the massive component database and thermodynamic properties
of Aspen Properties. It is also capable of being exported to Aspen Plus,
as a process simulator. The model has been validated by several ex-
perimental data. A case study of pre-combustion CO2 capture with H2-
selective and CO2-selective membranes was chosen to demonstrate the
difference in performance between this model and more simplistic ones.
This comparison showed that concentration polarization had the least
impact on the separation performance of both CO2-selective and H2-
selective membranes for this application and at these process flowrates.
Real gas behavior was the most important effect for an H2-selective

12
E. Soroodan Miandoab, et al.
Fig. 16. Influence of individual non-ideal effects on the mole fraction of CO2 in the permeate stream as a function of the feed flow rate for the CO2-selective
membrane ( CO 0
2 / H2 = 0.2, ΔE = 12000 J/mol). The dashed curve represents both the simplistic model and one including concentration polarization, as these curves
are congruent. Feed is 60% CO2 at 60 bara and 100 °C.
membrane, whereas a combination of the Joule-Thomson effect and
real gas behavior affected the separation performance of the CO2-se-
lective membrane. Expansion driven heating and cooling of the re-
tentate stream were observed due to the negative and positive Joule-
Thomson coefficients of H2 and CO2, respectively. Overall, the non-
ideal behavior of membrane-based pre-combustion carbon capture was
primarily because of the non-ideal behavior of both CO2 and H2 and the
large Joule-Thomson coefficient of CO2.
Acknowledgements
Ehsan Soroodan Miandoab acknowledges The University of
Melbourne for the Melbourne Research Scholarship.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.memsci.2019.117470.
References
[1] R.W. Baker, Gas separation, Membrane Technology and Applications, John Wiley &
Sons, Ltd, 2004, pp. 301–353, , https://doi.org/10.1002/0470020393.ch8.
[2] D.F. Sanders, Z.P. Smith, R. Guo, L.M. Robeson, J.E. McGrath, D.R. Paul,
B.D. Freeman, Energy-efficient polymeric gas separation membranes for a sustain-
able future: a review, Polymer 54 (2013) 4729–4761, https://doi.org/10.1016/j.
polymer.2013.05.075.
[3] R.W. Baker, Future directions of membrane gas separation technology, Ind. Eng.
Chem. Res. 41 (2002) 1393–1411, https://doi.org/10.1021/ie0108088.
[4] C.A. Scholes, G.W. Stevens, S.E. Kentish, Membrane gas separation applications in
natural gas processing, Fuel 96 (2012) 15–28, https://doi.org/10.1016/j.fuel.2011.
12.074.
[5] M. Kanniche, R. Gros-Bonnivard, P. Jaud, J. Valle-Marcos, J.-M. Amann,
C. Bouallou, Pre-combustion, post-combustion and oxy-combustion in thermal
power plant for CO2 capture, Appl. Therm. Eng. 30 (2010) 53–62, https://doi.org/
10.1016/j.applthermaleng.2009.05.005.
[6] D.R. Paul, Y.P. Yumpolski, Polymeric gas separation membranes, taylor & francis,
n.d. http://www.taylorfrancis.com/books/e/9781351084338 , Accessed date: 19
January 2019.
[7] K. Ghosal, B.D. Freeman, Gas separation using polymer membranes: an overview,
Polym. Adv. Technol. 5 (1994) 673–697, https://doi.org/10.1002/pat.1994.
220051102.
13
Journal of Membrane Science 595 (2020) 117470
[8] D.T. Coker, B.D. Freeman, G.K. Fleming, Modeling multicomponent gas separation
using hollow-fiber membrane contactors, AIChE J. 44 (1998) 1289–1302, https://
doi.org/10.1002/aic.690440607.
[9] P. Cruz, J.C. Santos, F.D. Magalhães, A. Mendes, Simulation of separation processes
using finite volume method, Comput. Chem. Eng. 30 (2005) 83–98, https://doi.org/
10.1016/j.compchemeng.2005.08.004.
[10] R. Khalilpour, A. Abbas, Z. Lai, I. Pinnau, Modeling and parametric analysis of
hollow fiber membrane system for carbon capture from multicomponent flue gas,
AIChE J. 58 (2012) 1550–1561, https://doi.org/10.1002/aic.12699.
[11] M.J. Thundyil, W.J. Koros, Mathematical modeling of gas separation permeators —
for radial crossflow, countercurrent, and cocurrent hollow fiber membrane mod-
ules, J. Membr. Sci. 125 (1997) 275–291, https://doi.org/10.1016/S0376-7388(96)
00218-9.
[12] P.K. Kundu, A. Chakma, X. Feng, Simulation of binary gas separation with asym-
metric hollow fibre membranes and case studies of air separation, Can. J. Chem.
Eng. 90 (2012) 1253–1268, https://doi.org/10.1002/cjce.20631.
[13] J. Xu, Z. Wang, C. Zhang, S. Zhao, Z. Qiao, P. Li, J. Wang, S. Wang, Parametric
analysis and potential prediction of membrane processes for hydrogen production
and pre-combustion CO2 capture, Chem. Eng. Sci. 135 (2015) 202–216, https://doi.
org/10.1016/j.ces.2015.04.033.
[14] J.H. Choi, M.-J. Park, J. Kim, Y. Ko, S.-H. Lee, I. Baek, Modelling and analysis of
pre-combustion CO2 capture with membranes, Korean J. Chem. Eng. 30 (2013)
1187–1194, https://doi.org/10.1007/s11814-013-0042-7.
[15] L. Giordano, J. Gubis, G. Bierman, F. Kapteijn, Conceptual design of membrane-
based pre-combustion CO2 capture process: role of permeance and selectivity on
performance and costs, J. Membr. Sci. 575 (2019) 229–241, https://doi.org/10.
1016/j.memsci.2018.12.063.
[16] J. Franz, V. Scherer, An evaluation of CO2 and H2 selective polymeric membranes
for CO2 separation in IGCC processes, J. Membr. Sci. 359 (2010) 173–183, https://
doi.org/10.1016/j.memsci.2010.01.047.
[17] M. Scholz, T. Harlacher, T. Melin, M. Wessling, Modeling gas permeation by linking
nonideal effects, Ind. Eng. Chem. Res. 52 (2013) 1079–1088, https://doi.org/10.
1021/ie202689m.
[18] R. Wang, S.L. Liu, T.T. Lin, T.S. Chung, Characterization of hollow fiber membranes
in a permeator using binary gas mixtures, Chem. Eng. Sci. 57 (2002) 967–976,
https://doi.org/10.1016/S0009-2509(01)00435-3.
[19] D.T. Coker, T. Allen, B.D. Freeman, G.K. Fleming, Nonisothermal model for gas
separation hollow-fiber membranes, AIChE J. 45 (1999) 1451–1468, https://doi.
org/10.1002/aic.690450709.
[20] G. He, Y. Mi, P. Lock Yue, G. Chen, Theoretical study on concentration polarization
in gas separation membrane processes, J. Membr. Sci. 153 (1999) 243–258, https://
doi.org/10.1016/S0376-7388(98)00257-9.
[21] T. Melin, R. Rautenbach, Membranverfahren: Grundlagen der Modul- und
Anlagenauslegung, Springer-Verlag, 2007.
[22] J.-M. Zheng, Z.-W. Dai, F.-S. Wong, Z.-K. Xu, Shell side mass transfer in a transverse
flow hollow fiber membrane contactor, J. Membr. Sci. 261 (2005) 114–120,
https://doi.org/10.1016/j.memsci.2005.02.035.
[23] S. Shen, S.E. Kentish, G.W. Stevens, Shell-side mass-transfer performance in hollow-
fiber membrane contactors, Solvent Extr. Ion Exch. 28 (2010) 817–844, https://doi.
E. Soroodan Miandoab, et al.
org/10.1080/07366299.2010.515176.
[24] R.E. Treybal, Mass Transfer Operations, third ed., McGraw-Hill College, n.d.
[25] R. Waack, N.H. Alex, H.L. Frisch, V. Stannett, M. Szwarc, Permeability of polymer
films to gases and vapors, Ind. Eng. Chem. 47 (1955) 2524–2527, https://doi.org/
10.1021/ie50552a045.
[26] T.C. Merkel, R.P. Gupta, B.S. Turk, B.D. Freeman, Mixed-gas permeation of syngas
components in poly(dimethylsiloxane) and poly(1-trimethylsilyl-1-propyne) at
elevated temperatures, J. Membr. Sci. 191 (2001) 85–94, https://doi.org/10.1016/
S0376-7388(01)00452-5.
[27] H. Gorissen, Temperature changes involved in membrane gas separations, Chem.
Eng. Process: Process Intensification 22 (1987) 63–67, https://doi.org/10.1016/
0255-2701(87)80032-6.
[28] Z. Han, A. Fina, Thermal conductivity of carbon nanotubes and their polymer na-
nocomposites: a review, Prog. Polym. Sci. 36 (2011) 914–944, https://doi.org/10.
1016/j.progpolymsci.2010.11.004.
[29] C.L. Choy, Thermal conductivity of polymers, Polymer 18 (1977) 984–1004,
https://doi.org/10.1016/0032-3861(77)90002-7.
[30] D.R. Anderson, Thermal Conductivity of Polymers, (n.d.) 14.
[31] Aspen Custom Modeler, Aspen Modeler Reference Guide, (2004) https://esupport.
aspentech.com/S_Article?id=000001852.
[32] C.Y. Pan, Gas separation by high-flux, asymmetric hollow-fiber membrane, AIChE J.
32 (1986) 2020–2027, https://doi.org/10.1002/aic.690321212.
[33] M. SIDHOUM, Experimental behavior of asymmetric cellulose acetate membranes
and its use in novel separation schemes, AICHE Symp. Ser. 261 (1988) 102–112.
[34] X. Feng, J. Ivory, V.S.V. Rajan, Air separation by integrally asymmetric hollow-fiber
membranes, AIChE J. 45 (1999) 2142–2152, https://doi.org/10.1002/aic.
690451013.
[35] C.A. Scholes, J. Bacus, G.Q. Chen, W.X. Tao, G. Li, A. Qader, G.W. Stevens,
S.E. Kentish, Pilot plant performance of rubbery polymeric membranes for carbon
dioxide separation from syngas, J. Membr. Sci. 389 (2012) 470–477, https://doi.
org/10.1016/j.memsci.2011.11.011.
[36] H. Lin, B.D. Freeman, Gas solubility, diffusivity and permeability in poly(ethylene
14
Journal of Membrane Science 595 (2020) 117470
oxide), J. Membr. Sci. 239 (2004) 105–117, https://doi.org/10.1016/j.memsci.
2003.08.031.
[37] L. Zhu, M.T. Swihart, H. Lin, Tightening polybenzimidazole (PBI) nanostructure via
chemical cross-linking for membrane H 2/CO 2 separation, J. Mater. Chem. 5
(2017) 19914–19923, https://doi.org/10.1039/C7TA03874G.
[38] D.R. Pesiri, B. Jorgensen, R.C. Dye, Thermal optimization of polybenzimidazole
meniscus membranes for the separation of hydrogen, methane, and carbon dioxide,
J. Membr. Sci. 218 (2003) 11–18, https://doi.org/10.1016/S0376-7388(03)
00129-7.
[39] M. Azizi, S.A. Mousavi, CO2/H2 separation using a highly permeable polyurethane
membrane: molecular dynamics simulation, J. Mol. Struct. 1100 (2015) 401–414,
https://doi.org/10.1016/j.molstruc.2015.07.029.
[40] D. Jansen, M. Gazzani, G. Manzolini, E. van Dijk, M. Carbo, Pre-combustion CO2
capture, Int. J. Greenhouse Gas Control 40 (2015) 167–187, https://doi.org/10.
1016/j.ijggc.2015.05.028.
[41] A. Mourgues, J. Sanchez, Theoretical analysis of concentration polarization in
membrane modules for gas separation with feed inside the hollow-fibers, J. Membr.
Sci. 252 (2005) 133–144, https://doi.org/10.1016/j.memsci.2004.11.024.
[42] R. Rautenbach, W. Dahm, Gas permeation — module design and arrangement,
Chem. Eng. Process: Process Intensification 21 (1987) 141–150, https://doi.org/10.
1016/0255-2701(87)87003-4.
[43] F. Ahmad, K.K. Lau, A.M. Shariff, Y. Fong Yeong, Temperature and pressure de-
pendence of membrane permeance and its effect on process economics of hollow
fiber gas separation system, J. Membr. Sci. 430 (2013) 44–55, https://doi.org/10.
1016/j.memsci.2012.11.070.
[44] B.-Z. Maytal, J.M. Pfotenhauer, Miniature Joule-Thomson Cryocooling Principles
and Practice, Springer, 2013.
[45] J.R. Roebuck, T.A. Murrell, E.E. Miller, The joule-thomson effect in carbon dioxide,
J. Am. Chem. Soc. 64 (1942) 400–411, https://doi.org/10.1021/ja01254a048.
[46] P. Linstrom, NIST chemistry WebBook, NIST Standard Ref. Database 69 (1997),
https://doi.org/10.18434/T4D303.