Separation and Purification Technology 202 (2018) 345–356

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Robust thin film composite PDMS/PAN hollow fiber membranes for water T
vapor removal from humid air and gases
⁎
Can Zeng Liang, Tai-Shung Chung
Department of Chemical & Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, Singapore 117585, Singapore

A R T I C LE I N FO A B S T R A C T

Keywords: Water vapor in humid air and industrial gases needs to be removed for various applications such as air con-
Water vapor ditioning, foodstuffs storage, aviation and space flights. Membranes can remove water vapor from gas streams
Composite hollow fiber membrane without involving phase change and regeneration. A robust and productive membrane is vital for water vapor
Air dehumidification removal but it is challenging to produce it. Here we present a defect-free thin film composite (TFC) poly-
PMDS
dimethylsiloxane (PDMS)/polyacrylonitrile (PAN) hollow fiber membrane with an ultrathin PDMS selective
Gas dehumidification
layer of about 260 nm. Different from previous works, the PAN substrate was spun at a high take-up speed of
60 m/min. It has an average of surface pore size of 5.6 nm and superior mechanical properties. After the optimal
PDMS coating, the composite membrane shows a N2 permeance of about 280 GPU, an O2/N2 selectivity of 2.2
and a water vapor permeance ranging from about 800 to 3700 GPU depending on operation conditions. In
addition, the PDMS/PAN composite membrane has been tested for 48-hour outer door dehumidification and
shown very stable and robust performance with a water vapor removal rate of 65% from the humid air. Besides,
the PDMS/PAN composite membrane is not only able to efficiently dry the humid biogas but also upgrade its
methane content simultaneously. Therefore, the newly developed TFC composite membrane has great potential
for water vapor removal from humid air and various gas streams.

1. Introduction nowadays accounts for about 50% of buildings’ energy consumption

and 20–40% of the total national energy use in developed countries
Water vapor prevalently exists in many gas streams such as natural [4,6,10,12,27].
gas, flue gas and air. The removal of water vapor from specific gas During air conditioning, the cooling machine or chiller removes
streams is important for various applications, such as dehydration of both the sensible heat (dry air enthalpy) and the latent heat (water
natural gas, air conditioning, storage of foodstuffs, aviation and space vapor enthalpy) simultaneously. The latent heat accounts for the largest
flights [1–11]. Conventional methods for water vapor removal include fraction of the total heat of the humid air, particularly for the air in hot
condensation and adsorption. In the condensation method, water vapor and humid regions and periods [2,4,6,9,10,27]. Therefore, there is a
is cooled down below the dew point and liquefied. In the adsorption need to remove the water vapor prior to conditioning the air through
process, water vapor is captured and adsorbed by adsorbents. Both membrane-based dehumidifiers. Currently, most of the air/gas dehu-
methods are energy intensive and expensive. The former involves in midification membranes are made from polymers such as cellulosic
cooling and phase change to condense vaporous water to liquid water, polymers, polyamide, polyimide, polyacrylonitrile (PAN), polyethylene
while the later requires regeneration steps of desiccants at high tem- (PE), polypropylene (PP), polysulfone (PSf) and polydimethylsiloxane
peratures [1,9–13]. It may also suffer from the depletions of desiccants. (PDMS) [9,12,14,16,28]. However, polymeric based membranes suffer
Comparing to condensation and adsorption, membrane-based technol- from a tradeoff relationship between permeability and selectivity. A
ogies offer an attractive alternative to remove water vapor because they polymer with a high permeability usually exhibits a low selectivity and
do not need thermal regeneration and phase change [2,5,8–11,14–23]. vice versa [28]. To overcome this tradeoff, inorganic-based membranes
Additionally, membrane-based technologies are compact, energy-effi- have also been developed and studied [17,19,29–34]. Unfortunately, it
cient, easy to scale up, low initial investment, and simple in operation is difficult to fabricate large pieces of inorganic membranes for com-
and maintenance [24–26]. Therefore, membrane-based dehumidifica- mercial gas separation applications because of poor reproducibility,
tion has received increasing attention recently because air conditioning processability, high cost and difficulties in scaling up [35,36]. To date,

⁎
Corresponding author.
E-mail address: chencts@nus.edu.sg (T.-S. Chung).

https://doi.org/10.1016/j.seppur.2018.03.005
Received 2 February 2018; Received in revised form 3 March 2018; Accepted 3 March 2018
Available online 04 April 2018
1383-5866/ © 2018 Elsevier B.V. All rights reserved.
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

commercial gas and vapor separation membranes are dominant by Table 1

polymeric hollow fiber membranes because they have a number of Conditions and parameters for spinning PAN hollow fiber membranes.
advantages such as good processability, self-supporting characteristics, Conditions/parameters
a high ratio of surface area to volume and ease to scale up [36–39].
Due to the low H2O vapor pressure in air, a desirable de-humidifi- Air gap 5 cm
Take up speed 60.0 m/min
cation membrane must have characteristics of high permeance in order
External coagulant Tap water
to achieve a high water removal rate. To improve the permeance, it Coagulant bath temperature 25 ± 2 °C
requires to minimize the dense-selective thickness of membranes to a Dope and bore fluid temperature 25 ± 2 °C
submicron level (e.g., 0.1–1.0 µm) and to reduce the transport re- Bore fluid flow rate 16.0 ml/min
sistance of substrates [36,40–45]. However, it is technically challenging Bore fluid composition H2O/NMP = 10/90 (wt%)
Polymer dope flow rate 30.0 ml/min
to produce such membranes having an ultrathin dense-selective layer
Polymer dope composition PAN/NMP = 20/80 (wt%)
without causing defects, particularly in large-scale membrane produc-
tion [36]. Therefore, in order to make membranes more technically and
economically viable to compete with conventional technologies, the 2.3. Synthesis of crosslinked PDMS
configuration of thin film composite (TFC) hollow fiber membranes
consisting of a porous substrate and a thin dense-selective layer is The preparation of partially crosslinked PDMS is described as fol-
chosen for this study. A commercially available glassy polyacrylonitrile lows: 10 g of the base (a component of the Sylgard®184 silicone elas-
(PAN) and a rubbery polydimethylsiloxane (PDMS) are selected as the tomer kit) and 1.0 g of the cross-linker (another component of the
membrane materials because both PAN and PDMS have good chemical elastomer kit) were dispensed into a plastic beaker (50 ml) and mixed
resistance and processability [46–48]. In addition, PAN and PDMS for 3 min at room temperature (∼25 °C). Then the beaker containing
polymers are not expensive and have been widely used as membrane the homogeneous mixture was immerged into a 75 ± 1 °C water bath
materials for liquid, gas and vapor separations [41,49–51]. Further- and mixed manually until the mixture became very viscous (close to
more, PDMS has been known as the most gas permeable rubbery gelation). Subsequently, the beaker contained the viscous crosslinked
polymers due to its flexible siloxane linkages and an acceptable H2O/N2 PDMS was removed from the water bath, then PDMS was dissolved with
selectivity (e.g., ∼140) [13,52–54]. cyclohexane and a 5 wt% PDMS stock solution was prepared. The stock
To design robust and defect-free TFC hollow fiber membranes for PDMS solution was kept stirring with the aid of a magnetic stirrer
the removal of water vapor from air and gases, this work focuses on the (2Mag-Magnetic eMotion, MIX 15 eco) at room temperature overnight.
following three parts: (1) developing a highly porous and strong hollow Then the PDMS solution was subjected to further studies and char-
fiber membrane substrate in a productive way; (2) fabricating a defect- acterizations.
free TFC membrane with an ultrathin dense-selective layer while
minimizing the polymer intrusion by applying a solution coating
2.4. Fabrication of PDMS/PAN composite membranes
method; and (3) testing the performance of the composite membrane
using synthetic humid air and biogas, and real outdoor air as feeds
To fabricate a PDMS/PAN composite membrane, the PAN substrate
under various conditions such as RH and flow rate of the feed gas,
was dipped into the PDMS solution for a few seconds (e.g. 5 s), then
applied vacuum pressure and flow rate of sweep gas. This study may
taken out and dried in air at room temperature for at least 2 days. The
offer insightful knowledge for producing practical membranes for ap-
dry composite membrane was subjected to gas permeation tests and
plications such as air and gas dehumidification.
other characterizations. The details of the membrane module fabrica-
tion and the dip coating methods could be found elsewhere [56].
2. Experiments and methods
2.5. Pure gas permeation tests for PAN substrates and the composite
2.1. Materials membranes

The Sylgard®184 silicone elastomer kit was bought from Dow Pure gas permeation tests were carried out at room temperature
Corning (Singapore) Pte Ltd. Polyacrylonitrile (PAN) was provided by (25 ± 2 °C) using a permeation cell system as described in the previous
Chung Yuan Christian University, Taiwan, with the help of Prof. Chein- work [56]. A transmembrane pressure of 5.0 ± 0.1 psi (1 bar = 14.7
Chieh Hu. Cyclohexane (analytical reagent grade, ≥99.99%) was pur- psi) was used to measure the gas permeance of the hollow fiber sub-
chased from Fisher Chemical. N-methyl-2-pyrrolidone (NMP, ≥99.5%), strates. A transmembrane pressure of 2 bar (29.4 ± 0.4 psi) was ap-
polyethylene glycol (PEG) and polyethylene oxide (PEO) were bought plied to measure the gas permeance of the coated/composite mem-
from Merck. Deionized (DI) water was generated by a water pur- branes. The effective membrane length was 15 cm. The gas permeate
ification system (Millipore, Elix). Gases including oxygen (O2, flow rate was determined using a universal gas flowmeter (Agilent,
≥99.9995%), nitrogen (N2, ≥99.9995%), helium (He, ≥99.9995%), ADM1000) which has a measurement range of 0.5–1000 ml/min. For
CH4/CO2 mixed gas (CH4/CO2 = 50/50 vol%) and pure air used in this each experimental condition, at least three membrane modules were
study were supplied by Singapore Oxygen Air Liquide Pte. Ltd. produced for gas permeation tests. Unless stated otherwise, the average
(SOXAL). Pure air mainly consisted of N2 and O2 (i.e., O2 = 21 ± 1%, results were reported in this work.
H2O < 3 ppm, CnHm < 5 ppm). The pure gas permeance ( J ) can be calculated according to the
following equation:
2.2. Preparation of hollow fiber substrates P Q Q
J= = =
L AΔP nπDLm ΔP (1)
The PAN hollow fiber substrates were fabricated via a dry-wet
spinning process. The spinning conditions were tabulated in Table 1. where Q is the gas permeate flow rate (cm3/min), n is the number of
The as-spun hollow fiber substrates were soaked in fresh tap water for fiber(s) in each module, D is the outer diameter of hollow fiber mem-
three days to remove the residual solvent, freeze-dried in a freeze dryer branes (cm), Lm is the effective length of hollow fibers on the module
(CHRIST, Beta 2–8 LDplus) for 48 h, and then stored for further in- (cm) and ΔP is the pressure difference across the membrane (cmHg).
vestigations. The details about hollow fiber membrane spinning op- The unit of permeance can be expressed as GPU (1
erations could be found elsewhere [55,56]. GPU = 1 × 10−6 cm3 (STP)/cm2 s cmHg).

346
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

P H
P
Dry air
MFC Mixed tank

P
Humid air Sweep
P MFC

Demister
MFC
Gas cylinder

Dehumidified
air

Dry air Vacuum Flow

Pump meter
Fig. 1. Schematic of a laboratory-scale dehumidification testing system for synthetic humid air. The feed gas flows along the shell side of the composite hollow fiber
and the permeate gas exits from the lumen side. Dry pure air, or dehumidified air, or pure helium is the sweep gas on the permeate side.

The gas selectivity (αi / j ) is defined as: Humid

outdoor air
Ji Pump
αi / j =
Jj (2)

where Ji and Jj are the permeances of gases i and j , respectively.

H
2.6. Water vapor removal from air and gas by the composite membrane Dry air

2.6.1. Water vapor removal from synthetic humid air

The dehumidification tests were performed for the optimal com-
posite membranes. A membrane module containing 15 composite
Vacuum Flow
membranes were tested under counter-current conditions using an ex- Pump meter
perimental system as presented in Fig. 1. Each fiber had an effective
length of about 21 cm. The synthetic humid air was produced by con- Fig. 2. Schematic of a laboratory-scale dehumidification test system for outdoor
trolling the flow rates of mass flow controllers (MFC) (Bronkhorst®; real humid air. Humid air is blown into the membrane module on the shell side
of the composite hollow fibers and the permeate stream is removed from the
model: F-201CV-1K0-AGD-22-V; range: 0–1000 sccm; accuracy: ±
lumen side by a vacuum pump.
(0.5% Rd + 0.1% FS)). The humid air stream was created by passing
the dry air through a water tank inbuilt with an air stone. The water
droplets or mists in the humid air stream were removed by a demister. The driving force was created by the vacuum pump that took the
The dry air and demisted humid air were mixed in the mixed tank, then permeate stream away. The pressures of feed and retentate were the
the final synthetic humid air was fed into the membrane module for atmospheric pressure. The experimental data within the error of ± 10%
testing. Different values of relative humidity of the synthetic humid air were considered valid and used for analyses. Unless state otherwise, the
were obtained by tuning the flow rates of dry and humid air streams. average results were presented.
The relative humidity (RH) and temperature were determined using a
hygrometer (Rotronic, hydropalm; operating range: −10 to 60 °C,
0–100% RH). The pressures of feed air and retentate streams were zero 2.6.3. Water vapor removal from the synthetic humid biogas
or at atmospheric pressure (14.7 psi). The membrane dehumidification Similar to the generation of the synthetic humid air, the synthetic
was run in three modes: (1) vacuum mode, in which a driving force of humid biogas was made by bubbling a stream of a CH4/CO2 mixed gas
water vapor partial pressure difference was created by a vacuum pump (CH4/CO2 = 50/50 vol%) through a water tank as shown in Fig. 1. The
and the permeate water vapor was carried away by the vacuum pump; water mists in the humid biogas stream were removed by the demister.
(2) sweeping mode, in which a sweep gas stream created the driving The dehumidification of the humid biogas was run in the sweeping
force and took the permeate water vapor away; (3) vacuum-sweep mode, in which the pure helium was used as the sweeping gas. The
mode, in which a fraction of the retentate (i.e., the de-humidified air) pressures of feed, retentate and permeate streams were at atmospheric
was split and returned back as the sweep gas, the driving force was pressure. The flow rates of the feed and the sweep gas were controlled
created by the vacuum pump and returned air. The flow rate of the and determined by a MFC, the flow rates of the retentate and permeate
retentate was measured using a gas flowmeter. The experimental data were measured using a gas flowmeter. The CH4 and CO2 content of the
within the error of ± 5% were considered valid and used for analyses. retentate were tested using a gas chromatography-GC (Agilent, 7890A).
Unless state otherwise, the average results were reported. Unless state otherwise, the average results of at least three measure-
ments were reported.
2.6.2. Water vapor removal from real outdoor air
The membrane air dehumidification was conducted by applying the
2.6.4. Calculation of water permeance
real outdoor air. The experimental system (as shown in Fig. 2) was
The permeance of water vapor was calculated using the trans-
located outdoor, and the air dehumidification tests were carried out
membrane partial pressure difference by the following equations:
outdoor. The humid air was fed into the membrane module by a pump.

347
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Jw =
Qw of the small surface pores of the substrates (e.g. ∼5 nm) and the short
A (Pfw−Ppw ) (3) coating duration (e.g. ∼5 s). Therefore, based on the mass balance and
the PDMS density, the thickness of the PDMS coating layer can be
where Jw is the permeance of water vapor, Q w is the flux of the removed
calculated by measuring the weights of hollow fiber membranes before
water vapor, A is the effective membrane area, Pfw is the average partial
and after the coating, as described in our previous publication [56].
pressure of water vapor in the retentate/feed side, Ppw is the average
partial pressure of water vapor in the permeate side. The water vapor
2.7.4. FESEM, SEM-EDX, mechanical properties and contact angle
content (g water vapor/kg dry air) were calculated by using the mea-
The morphologies of hollow fiber substrates and PDMS/PAN com-
sured values of temperature, RH and psychometric models [57,58].
posite membranes were observed using field emission scanning electron
Since the operation of the synthetic biogas dehumidification was at a
microscopy (FESEM, JEOL, JSM-6700LV). The PDMS coating layers
low pressure (e.g. the atmospheric pressure), the biogas stream was
were studied and analyzed using scanning electron microscopy (SEM,
treated as an air stream for simplicity and convenience. Thus, the water
JEOL JSM-5600LV) equipped with an Oxford INCA energy dispersion X-
vapor content of the biogas could be determined by using a hygrometer
ray system (SEM-EDX). To prepare FESEM/SEM-EDX membrane sam-
which measured the relative humidity of the gas streams. The partial
ples, membranes were immersed in liquid nitrogen and fractured. Then,
pressure and the water vapor volume in the gas stream can be calcu-
the samples were attached to stubs and coated with platinum (Pt) by a
lated based on the water vapor content and the given gas pressure. The
sputter coater (JEOL JFC-1300). The mechanical properties of hollow
stage cut (a ratio of the permeate flow rate to the feed flow rate)
fiber membranes were determined by using a tensile meter (Instron,
was < 0.1.
5542) equipped with a Bluehill software. The contact angle of the
crosslinked PDMS flat-sheet membranes was measured by a contact
2.7. Characterizations angle goniometer (Rame Hart, USA) at room temperature (e.g. ∼25 °C).
DI water was used as the drop-forming liquid. At least ten random spots
2.7.1. Pore size and pore size distribution of the PAN substrate of the membrane were tested and the average result was presented in
Based on the solute transport theory [59–61], the pore size and pore the work. The average contact angle of the PDMS was about 103° (see
size distribution of the PAN substrates were measured using neutral supporting information, Fig. S1).
solutes such as PEG and PEO of Mw = 400, 2 k, 6 k, 12 k, 35 k, 100 k
and 200 k g/mol. These PEG and PEO were dissolved in DI water. The
3. Results and discussion
tests were performed using an in-house ultrafiltration (UF) setup under
a transmembrane pressure of 1 bar. Two membrane modules, each
3.1. PAN hollow fiber membrane substrate
consisting of three as-spun PAN substrates, were constructed and used
for the UF tests. The solute and permeate concentrations were de-
3.1.1. Effects of spinning conditions on mechanical properties
termined using a total organic carbon analyzer (Shimadzu, TOC ASI-
In order to obtain a robust PAN hollow fiber membrane substrate
5000A).
with a high gas permeance and also lower its production cost, the newly
The solute rejection (%) to PEG or PEO can be calculated by the
developed hollow fiber membrane was spun at a high spinning speed of
following equation:
60 m/min. Table 2 compares the effects of spinning conditions on
Cp physical and mechanical properties of PAN substrates between this and
Rs = (1− ) × 100
Cf (4) previous studies using the same spinneret [56,63]. A comparison be-
tween P20-15S and P20-15L indicates that the outer diameter (OD) and
where Cp and Cf are the solute concentrations (ppm) of the permeate the wall thickness of the hollow fibers increase by about 20% and 18%,
and feed solutions, respectively. Once the solute rejections were de- respectively, by doubling the flow rates of bore fluid and polymer dope
termined, the pore size and pore size distribution of the PAN substrates simultaneously. It also results in an increment of about 30% in the
can be calculated as elaborated in previous literatures [56,62]. maximum load (or force). When the take-up speed rises from 15 to
DI water was used to determine the pure water permeability (PWP, 60 m/min by approximately tripling the flow rates of bore fluid and
L/m2 h bar) according to the following equation: polymer dope at the same time (P20-60L vs. P20-15L), the maximum
Q load, tensile stress and modulus are further improved. Although the
PWP =
AΔP (5)
Table 2
where Q is the water permeation rate (liter/hour = L/h), A is the ef- Effects of spinning conditions on physical and mechanical properties of PAN
fective filtration area (m2) and ΔP is the transmembrane pressure (bar). hollow fiber membrane substrates.
Code name of membrane P20-15S P20-15L P20-60L
2.7.2. Density of PDMS dense film
The density of the crosslinked PDMS dense film membrane was Spinning PAN concentration (%) 20.0 20.0 20.0
measured using a balance (Mettler Toledo, XS205, accuracy: 0.0001 g) conditions Take up speed 15.0 15.0 60.0
Bore fluid flow rate (ml/ 2.5 5.0 16.0
and a density measurement kit according to the Archimedes’ principle.
min)
The membrane density (ρP ) could be calculated by the following Polymer dope flow rate 5.0 10.0 30.0
equation: (ml/min)
Draw ratio 4.20 2.76 2.92
MA
ρP = ρ Fiber Dimension Outer diameter (µm) 720 894 887
MA−ML 0 (6)
Inner diameter (µm) 414 520 536
where MA is the membrane weight in air, ML is the membrane weight in Wall thickness (µm) 153 186 175
the auxiliary liquid and ρ0 is the density of the auxiliary liquid. Ethanol Mechanical Maximum load (N) 1.38 2.02 2.28
was used as the auxiliary liquid. The density of the crosslinked PDMS properties Maximum tensile stress 5.06 4.88 5.85
(Mpa)
was found to be 1.02 g/cm3.
Maximum tensile strain 0.30 0.22 0.22
(mm/mm)
2.7.3. Thickness of the coating layer Young's modulus (Mpa) 232.57 235.59 248.71
Comparing to the PDMS amount coated on the PAN substrate, the Reference [56] [63] This work
PDMS amount intruding into the PAN substrate was negligible because

348
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Fig. 3. Morphology and structure of the as-spun hollow fiber substrate. (A) Cross-section. (B) Enlarged cross-section. (C) Outer skin layer. (D) Outer surface. (E)
Outer-inner interface. (F) Inner surface.

dimensions of the P20-15L and P20-60L membranes are similar, the time of about 5 s. Fig. S3 portrays the dip coating process.
latter one is about 10% stronger than the former in terms of the max-
imum load. This may result from many factors such as enhanced
3.2.1. Effects of the PDMS concentration
polymeric chain elongation and orientation under the high take-up
In order to obtain the optimal coating conditions, PDMS solutions
speed. In addition, the high dope flow rate may alter the mutual dif-
with different concentrations are deployed to fabricate PDMS/PAN
fusion process between the solvent (NMP) and non-solvent (water)
composites. Fig. 5 shows the effects of PDMS concentration on gas se-
during the phase inversion, thereby forming a denser outer surface and
paration performance. O2 and N2 are chosen to benchmark the per-
stronger structure [40,45,64–66].
formance because they are permanent gases and major components in
air. Fig. 5A illustrates that the O2 and N2 permeances decrease with an
3.1.2. Morphologies and gas transport properties of the PAN substrate increase in PDMS concentration because a higher PDMS concentration
Fig. 3 shows the morphology and structure of the as-spun hollow results in a thicker PDMS coating layer, as demonstrated in Fig. 6. In
fiber substrate. The PAN substrate has characteristics of a highly porous contrast, Fig. 5B displays that the corresponding O2/N2 selectivity in-
structure full of macrovoids. Its inner surface is fully porous with a creases initially and then levels off at around 2.2. This up and plateau
honeycomb shape morphology, while its outer skin is smooth and trend is ascribed to the fact that the initial low PDMS concentration of
highly porous with many small surface pores. The inner-surface mor- 0.25 wt% results in a defective PDMS/PAN composite membrane, while
phology is favourable for gas/vapor transport with minimal resistance, a high PDMS concentration of ≥0.5 wt% leads to a defect-free com-
while the outer-surface morphology is preferable for the solution posite membrane with a gas selectivity equal to the intrinsic selectivity
coating. Fig. 4 presents the pore size distribution, PWP and mechanical of the dense PDMS films (Table 3). Interestingly, the gas permeance
properties of the PAN substrate. The average surface pore size is 5.6 nm, becomes almost constant when the PDMS concentration increases from
which is about 70% smaller than the previous PAN membrane (P20- 3.0 to 4.0 wt%. This phenomenon is probably owed to the effect of
15L) of 9.6 nm even though they both have similar fiber dimensions solution intrusion. A lower PDMS concentration of 3.0 wt% may lead to
[63]. Fig. S2 illustrates the gas transport properties of the PAN sub- a thinner PDMS coating layer (Fig. 6) but result in more intrusion be-
strate. Even though it has a smaller surface pore size than the P20-15L cause of its lower bulk viscosity. As a result, it has an apparent dense
sample, its gas (N2) permeance is as high as around 52,000 GPU which layer thickness of PDMS almost equal to the one coated with a higher
is slightly less than the latter of ∼56,000 GPU. This suggests that the as- PDMS concentration of 4.0 wt%. Although a higher PDMS concentra-
spun PAN substrate has a small pore size but a high surface porosity, tion results in a thicker PDMS coating layer, it has less polymer intru-
which are evidenced by its morphology as shown in Fig. 3D. Fig. S2B sion due to its high bulk viscosity.
shows that the PAN substrate has an O2/N2 selectivity of 0.93. This
value is very close to the theoretical O2/N2 selectivity of Knudsen dif-
3.2.2. Morphology and architecture of the optimal PDMS/PAN composite
fusion [16,41–45,67], where the O2/N2 selectivity is equal to the ratio
membrane
of the inverse square root of their molecular weights.
The PDMS/PAN composite membrane coated with 0.5 wt% PDMS is
the optimal composite membrane, which is defect-free and offers a high
3.2. PDMS/PAN composite hollow fiber membranes gas permeance of ∼280 GPU for N2. As shown in Fig. 6, it has an ul-
trathin PDSM coating layer of about 260 nm confirmed from the FESEM
The PDMS/PAN TFC hollow fiber membranes can be produced by image in Fig. 7. Fig. 8 illustrates the architecture and element analyses
directly dipping the PAN substrate into the PDMS solution for a short of the optimal PDMS composite membrane. The sharp peak of the SEM-

349
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Fig. 4. The pore size distribution, PWP and mechanical properties of the PAN substrate.

EDX silicon element spectrum depicted in Fig. 8B and C indicates that a

PDMS layer has been coated on the outer surface of the hollow fiber
substrate because the PAN substrate contains no silicon element.

3.3. Water vapor removal from air by the optimal composite membrane

A membrane module consisting of 15 PDMS/PAN TFC hollow fiber

is fabricated by applying the optimal coating conditions (i.e., 0.5 wt%
crosslinked PDMS and a coating duration of ∼5 s) to remove water
vapor from the synthetic humid air, real outdoor air and the synthetic
humid biogas. Each fiber has a length of ∼20 cm and the total mem-
brane area is about 90 cm2. Fig. S4 describes the process and me-
chanism of water vapor removal from air by the composite hollow fiber
membrane. The feed air and retentate are on the shell side of the
module while the permeate exits from the lumen side. The vacuum
and/or sweep gas provide the driving force and take the permeate
Fig. 6. The thickness of the PDMS dense layer as a function of PDMS coating
(e.g.·H2O vapor) away. The separation of water vapor from air/gas is
concentration. (The PDMS thickness was determined by the principle of mass
based on the solution-diffusion mechanism. The dehumidification can
balance).
be operated in different modes, whose effects are discussed as follows.

Fig. 5. The gas permeation of the PDMS/PAN composite membranes coated with different PDMS concentration. (A) Effects of PDMS concentrations on gas per-
meance (B) (A) Effects of PDMS concentrations on O2//N2 selectivity.

350
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Table 3 pressure. This is because a low flow rate offers a longer contact time
Pure gas permeability of PDMS dense film membrane.a between the water vapor and membrane. Consequently, more water
Permeability (Barrer) Ideal selectivity vapor can permeate to the lumen side of the membrane, resulting in a
drier retentate. Although the permeance of water vapor increases with
N2 287 an increase in vacuum pressure, Fig. 10B shows that its value remains
O2 614 O2/N2 2.1
almost similar regardless of the feed flow rates because of the solution-
CH4 1200 H2O/N2 140
CO2 3195 H2O/CH4 33
diffusion mechanism. In other words, the driving force for water vapor
H2O 40,000 H2O/CO2 13 removal is mainly determined by the differences in RH and partial
water vapor pressure across the membrane. Since gas transport in the
a
The permeabilities of N2, O2 and CO2 are cited from Ref. [56]; CH4 is cited shell and lumen sides are much faster than inside the membrane, the
from Ref. [8]; H2O is cited from Ref. [53]. 1 Barrer = 1 × 10−10 cm3 (STP) cm/ solution-diffusion mechanism across the membrane becomes the rate
cm2 s cmHg. determining step. Consequently, the feed flow rate in the shell side of
modules plays a less important role than others.
3.4. Water vapor removal from synthetic humid air by the optimal
composite membrane 3.4.2. Effects of sweep flow rate and relative humidity of the feed
In order to study the effects of sweep flow rate, RH and feed flow
3.4.1. Effects of vacuum, relative humidity and feed flow rate rate on dehumidification performance, the dehumidification system
To study the effects of vacuum pressure, relative humidity (RH) and was run in a sweeping mode where the water vapor partial pressure
feed air flow rate on dehumidification performance, the dehumidifi- difference is created by a dry sweeping gas (i.e., pure air) that carries
cation system was firstly run in a vacuum mode, where a vacuum pump the permeate water vapor out of the module. Fig. 11A and B show the
creates the partial pressure difference of water vapor across the mem- effects of sweep gas flow rate and feed RH on water vapor removal,
brane. Fig. 9 displays the effects of vacuum pressure and air RH on respectively. The RH of the retentate declines with an increase in sweep
water vapor removal. At a given feed flow rate (e.g., 1000 ml/min), gas flow rate because a higher flow rate of sweep gas can dilute the
Fig. 9A shows that the RH of the retentate decreases with an increase in permeate water vapor more effectively and create a larger partial
vacuum pressure. In addition, the higher RH in the feed air, the higher pressure difference of water vapor across the membrane. In contrast,
RH drop in the retentate. This is because the combination of a higher the corresponding permeance of water vapor initially increases with the
vacuum pressure and a higher RH provide a much larger driving force, augment in sweep gas flow rate, then becomes almost constant when
which enables more water vapor diffusing through the membrane. the flow rate is above 600 ml/min. The reason for this up-then-plateau
Fig. 9B displays that the water vapor permeance rises with an increment trend arises from the fact that the dilution effect of the sweep gas be-
in vacuum pressure or vapor pressure difference but the drier feed (i.e., comes negligible when the sweep flow rate reaches a certain level. A
lower HR) tends to have a lower water vapor permeance. At the max- similar phenomenon has been observed by others [8]. These trends and
imum vacuum pressure (e.g., 1000 mbar), the maximum and minimal results suggest that there is an optimal sweep gas flow rate when the
permeances of water vapor are about 1700 and 800 GPU for feed air membrane dehumidification system is operated in a sweeping mode.
containing RH of 90 and 10%, respectively. Interestingly, these values Although the maximum water vapor permeance varies from about
are still higher than the pure O2 permeance of about 620 GPU and the 3100 to 3700 GPU for the feed RH range of 20–90%, a higher feed RH
pure N2 permeance of about 280 GPU of the same membrane, as shown results in a lower water vapor permeance. This surprising phenomenon
in Fig. 5. is probably due to the effects of internal concentration polarization
Table 4 compares the basic physicochemical properties of gases and (ICP) on water vapor transport [8,53,70]. In other words, when the feed
vapor (e.g. O2, N2 and water vapor) [8,68,69]. Water vapor has a much air has a high RH, the driving force for water vapor transport across the
higher critical temperature (Tc) and boiling point (Tb) than the two membrane becomes large. However, this may result in a high water
permanent gasses (N2 and O2), but the former has a much smaller ki- vapor concentration inside the inner layer of the membrane because the
netic diameter and critical volume than the latter. As a result, water PAN substrate has a cross-section full of long finger-like macrovoids
vapor has a higher condensability (i.e., solubility) and a greater diffu- (Fig. 3B). The accumulated and stagnant water vapor reduces the ef-
sivity than the two gases [24]. Since permeability is a product of so- fective driving force. Therefore, a higher feed RH results in a severer
lubility and diffusivity, water vapor has a greater permeance across the ICP and a lower water vapor permeance.
membrane than the two permanent gases.
Fig. 10 exhibits the effects of vacuum pressure and feed flow rate on 3.4.3. Combined effects of vacuum and sweep
water vapor removal. When the RH of the feed is given (e.g. 80%), the Based on the above results, the sweeping method offers better air
RH of retentate drops with an increase in vacuum pressure, and the feed dehumidification efficiency than the vacuum one. However, the sweep
with a lower flow rate shows more water removal at a given vacuum gas is not always available and may come with a price. Therefore, a

Fig. 7. The cross-section morphology of the optimal PDMS/PAN composite hollow fiber membrane. (A) The cross-section image of the PAN hollow fiber substrate,
(B) The cross-section image of the optimal composite.

351
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Fig. 8. The architecture and element analyses of the optimal PDMS/PAN composite hollow fiber membrane. (A) The chemical composition in each layer. (B) The
cross-section image with SEM-EDX mapping. (C) The SEM-EDX element spectrum of silicon (Si).

Fig. 9. Effects of vacuum pressure and air relative humidity on water vapor removal. The feed flow rate is 1000 ml/min.

combined vacuum and sweep operation mode is studied, as described in becomes the sweep gas). Clearly, the efficiency of water vapor removal
Fig. 12A. A fraction of the retentate or the dehumidified air is diverted increases as the applied vacuum and sweep gas flow rate increase.
and returned to the lumen side of the membrane by tuning the valve. Without sweeping, only a small percentage (0.3–5%) of humidity is
The returned gas stream can flow and sweep with the aid of a vacuum removed by the membrane under low vacuum pressures (e.g., 200 and
pump so that the permeate water vapor is removed by the sweeping gas 400 mbar). When approximately 30% of the dehumidified feed air is
and vacuum pump. Fig. 12B and C show the combined effects of va- returned as the sweep gas, the RH of the retentate drops from 80% to
cuum and sweep. The RH of the retentate decreases with an increase in about 70% under a vacuum pressure of 200 mbar and further to 60%
vacuum pressure and sweep gas flow rate (i.e., the fraction of the feed under a vacuum pressure of 400 mbar, as illustrated in Fig. 12B and C.

352
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Table 4 3.6. Water vapor removal from the humid biogas by the optimal composite
Physical properties of the permanent gases and water vapor.a membrane
Gas/vapor Condensability Size
The dehumidification of biogas is necessary in the industry in order
Critical Boiling point Critical Kinetic to avoid the occurrence of water condensation, which may accelerate
temperature (K) temperature (K) volume diameter (Å) the corrosion of pipelines and equipment [71]. The synthetic humid
(cm3/mole)
biogas was generated via a humid gas generation system (Fig. 1) and
N2 126 77 90.1 3.64 the dehumidification was operated under the sweeping mode using
O2 154 90 73.4 3.46 helium as sweeping gas. Fig. 14 shows the effects of sweep gas flow rate
CH4 190 112 98.6 3.80 on biogas dehumidification. As presented in Fig. 14A, the RH of the
CO2 304 195b 94.1 3.30
biogas retentate decreases from about 88% to 40% as the sweep gas
H2O 647 373 56.0 2.50
flow rate increases from 0 to 1000 ml/min because a higher sweep gas
a
Cited from references: critical temperature, Lennard-Jones temperature flow rate provides a larger driving force. Thus, more water vapor
and critical volume [68]; and kinetic diameter [8,69]. permeates to the lumen side of the membrane. Interestingly, the re-
b
Cited from Ref. [69]. tentate has a higher CH4 content and a lower CO2 content with an in-
crease in sweep gas flow rate because CO2 has a permeability about
Although more water vapor is removed by the combination of vacuum three times larger than CH4 (Table 3). Therefore, the membrane de-
and sweep methods, it comes with penalties such as an increase in the humidification process not only reduces the water moisture in biogas
working load of the vacuum pump and a reduction of the product air. but also upgrade its methane content. The upgraded biogas is com-
mercially favourable because the biogas with a higher methane content
has a higher calorific value [71]. Fig. 14B presents the corresponding
3.5. Water vapor removal from the real outdoor air by the optimal water vapor permeance. It increases with an increment in sweep gas
composite membrane flow rate, then reaches a plateau. This trend is similar to the afore-
mentioned air dehumidification (Fig. 11B) and other reported ob-
The stability and robustness of the newly developed TFC hollow servations [8]. The maximum water vapor permeance in the biogas
fiber membranes were put to test using real outdoor air. The whole dehumidification, is about 3200 GPU.
membrane dehumidification system (Fig. 2) was deployed outside the
lab (outdoor) and run under a vacuum mode for air dehumidification.
The dehumidification period ranged from 10 am to 10 pm. The humid 4. Conclusions
air was fed into the membrane module by a pump, but the feed and
retentate pressures were kept at the atmospheric pressure. Fig. 13 A thin film composite (TFC) PDMS/PAN hollow fiber membrane
shows the air dehumidification results. The outer door’s RH and tem- with an ultrathin PDMS selective layer of about 260 nm has been fab-
perature fluctuate considerably, while the RH of the retentate (dehu- ricated in this work. Different from previous works, the PAN substrate
midified air) exhibits relatively stable. As shown in Fig. 13 A, the was spun at 60 m/min. It has an average of surface pore size of 5.6 nm.
membrane system can reduce the RH of the humid air from about 80% After dip coating the PAN substrate in a 0.5 wt% PDMS solution for a
to 28%. In other words, about 65% water vapor in the humid air can be short period of time for about 5 s, the resultant composite membrane
removed. Fig. 13B displays the water vapor permeance as a function of has a N2 permeance of about 280 GPU and an O2/N2 selectivity of 2.2,
time. The permeance exhibits a stable trend, and the average per- indicating that the membrane is defect-free. The newly developed
meance is approximately 1600 GPU. This stable 48-hour performance composite membrane is efficient for water vapor removal from humid
suggests that the PDMS/PAN composite hollow fiber membrane is ro- air. It has a water vapor permeance ranging from about 800 to 3700
bust and applicable for real air dehumidification applications. GPU depending on operation conditions such as vacuum pressure, flow
rate and relative humidity of the feed and sweeping flow rate. The
membrane system has been tested for 48-hour outdoor air

Fig. 10. Effects of vacuum pressure and feed flow rate on water vapor removal. The relative humidity of the feed is 80%.

353
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Fig. 11. Effects of sweep gas flow rate and feed’s relative humidity on water vapor removal. The flow rate of the feed is 1000 ml/min, the relative humidity of the
sweep gas is considered to be 0%.

Humid air
Humid air

A
Fraction of dehumidified air
was returned for sweeping
Dehumidified
air

100 100
B C
Relative humidity of retentate (%)

Relative humidity of retentate (%)

Applied vacuum: Applied vacuum:

90 90
200 mbar 200 mbar
400 mbar 400 mbar
80 80

70 70

60 60
RH of feed: 80% RH of feed: 80%
Flow rate of feed: 1000 ml/min Flow rate of feed: 600 ml/min
50 50
0 10 20 30 0 10 20 30
Fraction of the feed as sweep gas (%) Fraction of the feed as sweep gas (%)
Fig. 12. Combined effects of vacuum and sweep. (A) The sketch of the vacuum-sweep mode for membrane air dehumidification. (B and C) The effects of vacuum
pressure and fraction of the feed as sweep gas under different feed flow rates.

dehumidification and shown very stable and robust performance with a may offer a viable approach to advance the membrane-based technol-
water vapor removal rate of 65% from the humid air. In addition, the ogies for water vapor removal from humid air and various gas streams.
newly developed composite membrane is also capable of simulta-
neously drying and upgrading the humid biogas. Therefore, this work

354
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

Fig. 13. The performance of the long-term

membrane air dehumidification. (A) The
change of relative humidity of the outdoor
air and dehumidified air, outdoor tempera-
ture versus dehumidification time. (B) The
change of water vapor permeance versus
dehumidification time. (The feed flow rate is
around 500 ml/min and the dehumidifica-
tion system is run under a vacuum mode).

Fig. 14. Effects of sweep gas flow rate on the dehumidification of the synthetic humid biogas. The flow rate of the feed is 1000 ml/min, the sweep gas is helium.

Acknowledgement Appendix A. Supplementary material

This research grant is supported by the Singapore National Research Supplementary data associated with this article can be found, in the
Foundation with the project entitled, “Membranes to remove air pol- online version, at https://doi.org/10.1016/j.seppur.2018.03.005.
lutants and excess humidity for better indoor air quality and energy
saving” (grant number: R-279-000-453-279). Special thanks to Dr. Wai References
Fen Yong, Mr. Bai wang Zhao, and Ms. Liang-Yi Wang for their supports
and help. [1] P. Scovazzoa, J. Burgosa, A. Hoehnb, P. Todd, Hydrophilic membrane-based hu-
midity control, J. Membr. Sci. 149 (1998) 69–81.

355
C.Z. Liang, T.-S. Chung Separation and Purification Technology 202 (2018) 345–356

[2] H.T. El-Dessouky, H.M. Ettouney, W. Bouhamra, A novel air conditioning system Netherlands, 1996.
membrane air drying and evaporative cooling, Chem. Eng. Res. Des. 78 (2000) [38] W.J. Koros, R. Mahajan, Pushing the limits on possibilities for large scale gas se-
999–1009. paration: which strategies? J. Membr. Sci. 175 (2000) 181–196.
[3] P. Mazzei, F. Minichiello, D. Palma, HVAC dehumidification systems for thermal [39] R.W. Baker, Future directions of membrane gas separation technology, Ind. Eng.
comfort: a critical review, Appl. Therm. Eng. 25 (2005) 677–707. Chem. Res. 41 (2002) 1393–1411.
[4] K.J. Chua, S.K. Chou, New hybrid drying technologies, in: D.-W. Sun (Ed.), [40] D.T. Clausi, W.J. Koros, Formation of defect-free polyimide hollow fiber membranes
Emerging Technologies for Food Processing, 2005, pp. 535–551. for gas separations, J. Membr. Sci. 167 (2000) 79–89.
[5] F. Krull, C. Fritzmann, T. Melin, Liquid membranes for gas/vapor separations, J. [41] J.M.S. Henis, M.K. Tripodi, Composite hollow fiber membranes for gas separation:
Membr. Sci. 325 (2008) 509–519. the resistance model approach, J. Membr. Sci. 8 (1981) 233–246.
[6] L. Pérez-Lombard, J. Ortiz, C. Pout, A review on buildings energy consumption [42] I. Pinnau, J.G. Wijmans, I. Blume, T. Kuroda, K.-V. Peinemann, Gas permeation
information, Energy Build. 40 (2008) 394–398. through composite membranes, J. Membr. Sci. 31 (1988) 81–88.
[7] G. Ge, F. Xiao, X. Xu, Model-based optimal control of a dedicated outdoor air- [43] T.-S. Chung, A review of microporous composite polymeric membrane technology
chilled ceiling system using liquid desiccant and membrane-based total heat re- for air-separation, Polym. Polym. Compos. 4 (1996) 269–283.
covery, Appl. Energy 88 (2011) 4180–4190. [44] A.F. Ismail, K.C. Khulbe, T. Matsuura, Gas Separation Membranes-Polymeric &
[8] H. Lin, S.M. Thompson, A. Serbanescu-Martin, J.G. Wijmans, K.D. Amo, Inorganic: Polymeric and Inorganic, Springer International Publishing, Switzerland,
K.A. Lokhandwala, T.C. Merkel, Dehydration of natural gas using membranes. Part 2015.
I: Composite membranes, J. Membr. Sci. 413–414 (2012) 70–81. [45] N. Peng, N. Widjojo, P. Sukitpaneenit, M.M. Teoh, G.G. Lipscomb, T.-S. Chung, J.-
[9] S. Roy, C.M. Hussain, S. Mitra, Poly(acrylamide-co-acrylic acid) hydrophilization of Y. Lai, Evolution of polymeric hollow fibers as sustainable technologies: past, pre-
porous polypropylene membrane for dehumidification, Sep. Sci. Technol. 107 sent, and future, Prog. Polym. Sci. 37 (2012) 1401–1424.
(2013) 54–60. [46] Y. Shi, C. Burns, X. Feng, Poly(dimethyl siloxane) thin film composite membranes
[10] K. Zaw, M.R. Safizadeh, J. Luther, K.C. Ng, Analysis of a membrane based air-de- for propylene separation from nitrogen, J. Membr. Sci. 282 (2006) 115–123.
humidification unit for air conditioning in tropical climates, Appl. Therm. Eng. 59 [47] E. Favre, Carbon dioxide recovery from post-combustion processes: can gas per-
(2013) 370–379. meation membranes compete with absorption? J. Membr. Sci. 294 (2007) 50–59.
[11] K.L. Chong, N. Peng, H. Yin, G.G. Lipscomb, T.-S. Chung, Food sustainability by [48] M. Ai, S. Shishatskiy, J. Wind, X. Zhang, C.T. Nottbohm, N. Mellech, A. Winter,
designing and modelling a membrane controlled atmosphere storage system, J. H. Vieker, J. Qiu, K.J. Dietz, A. Golzhauser, A. Beyer, Carbon nanomembranes
Food Eng. 114 (2013) 361–374. (CNMs) supported by polymer: mechanics and gas permeation, Adv. Mater. 26
[12] F.W. Yu, K.T. Chan, Experimental determination of the energy efficiency of an air- (2014) 3421–3426.
cooled chiller under part load conditions, Energy 30 (2005) 1747–1758. [49] N. Stafie, D.F. Stamatialis, M. Wessling, Effect of PDMS cross-linking degree on the
[13] B. Bolto, M. Hoang, Z. Xie, A review of water recovery by vapour permeation permeation performance of PAN/PDMS composite nanofiltration membranes, Sep.
through membranes, Water Res. 46 (2012) 259–266. Purif. Technol. 45 (2005) 220–231.
[14] K.L. Wang, S.H. McCrayb, D.D. Newboldb, E.L. Cussler, Hollow fiber air drying, J. [50] I.C. Kim, H.Gu. Yun, K.H. Le, Preparation of asymmetric polyacrylonitrile mem-
Membr. Sci. 72 (1992) 231–244. brane with small pore size by phase inversion and post-treatment process, J.
[15] L.Z. Zhang, J.L. Niu, Energy requirements for conditioning fresh air and the long- Membr. Sci. 199 (2002) 75–84.
term savings with a membrane-based energy recovery ventilator in Hong Kong, [51] B. Jung, J.K. Yoon, B. Kim, H.-W. Rhee, Effect of molecular weight of polymeric
Energy 26 (2001) 119–135. additives on formation, permeation properties and hypochlorite treatment of
[16] K. Kneifel, S. Nowak, W. Albrecht, R. Hilke, R. Just, K. Peinemann, Hollow fiber asymmetric polyacrylonitrile membranes, J. Membr. Sci. 243 (2004) 45–57.
membrane contactor for air humidity control: modules and membranes, J. Membr. [52] H. Lin, B.D. Freeman, Materials selection guidelines for membranes that remove
Sci. 276 (2006) 241–251. CO2 from gas mixtures, J. Mol. Struct. 739 (2005) 57–74.
[17] L.-Z. Zhang, Y.-Y. Wang, C.-L. Wang, H. Xiang, Synthesis and characterization of a [53] S. Metz, W. Vandeven, J. Potreck, M. Mulder, M. Wessling, Transport of water vapor
PVA/LiCl blend membrane for air dehumidification, J. Membr. Sci. 308 (2008) and inert gas mixtures through highly selective and highly permeable polymer
198–206. membranes, J. Membr. Sci. 251 (2005) 29–41.
[18] S.-M. Huang, L.-Z. Zhang, Researches and trends in membrane-based liquid de- [54] P. Bernardo, E. Drioli, G. Golemme, Membrane gas separation: a review/state of the
siccant air dehumidification, Renew. Sustain. Energy Rev. 28 (2013) 425–440. art, Ind. Eng. Chem. Res. 48 (2009) 4638–4663.
[19] R. Xing, Y. Rao, W. TeGrotenhuis, N. Canfield, F. Zheng, D.W. Winiarski, W. Liu, [55] S. Bonyadi, T.S. Chung, W.B. Krantz, Investigation of corrugation phenomenon in
Advanced thin zeolite/metal flat sheet membrane for energy efficient air dehumi- the inner contour of hollow fibers during the non-solvent induced phase-separation
dification and conditioning, Chem. Eng. Sci. 104 (2013) 596–609. process, J. Membr. Sci. 299 (2007) 200–210.
[20] T.D. Bui, F. Chen, A. Nida, K.J. Chua, K.C. Ng, Experimental and modeling analysis [56] C.Z. Liang, W.F. Yong, T.-S. Chung, High-performance composite hollow fiber
of membrane-based air dehumidification, Sep. Sci. Technol. 144 (2015) 114–122. membrane for flue gas and air separations, J. Membr. Sci. 541 (2017) 367–377.
[21] B.W. Zhao, N. Peng, C.Z. Liang, W.F. Yong, T.S. Chung, Hollow fiber membrane [57] I. Urieli, Engineering thermodynamics – a graphical approach, Chapter 10: air-
dehumidification device for air conditioning system, Membranes 5 (2015) 722–738. water vapor mixtures, < https://www.ohio.edu/mechanical/thermo/ > .
[22] B. Zhao, W.F. Yong, T.-S. Chung, Haze particles removal and thermally induced [58] ASHRAE Handbook, 1997 Fundamentals, Chapter 6 Psychrometrics (1997).
membrane dehumidification system, Sep. Sci. Technol. 185 (2017) 24–32. [59] S. Singh, K.C. Khulbe, T. Matsuura, P. Ramamurthy, Membrane characterization by
[23] J.S. Cha, R. Li, K.K. Sirkar, Removal of water vapor and VOCs from nitrogen in a solute transport and atomic force microscopy, J. Membr. Sci. 142 (1998) 111–127.
hydrophilic hollow fiber gel membrane permeator, J. Membr. Sci. 119 (1996) [60] P. Aimar, M. Meireles, V. Sanchez, A contribution to the translation of retention
139–153. curves into pore size distributions for sieving membranes, J. Membr. Sci. 54 (1990)
[24] R.W. Baker, Gas separation, in: Membrane Technology and Applications, John 321–338.
Wiley and Sons Ltd, United Kingdom, 2012, pp. 325–375. [61] B. Bruggen, C. Vandecasteele, Modelling of the retention of uncharged molecules
[25] M. Carta, R. Malpass-Evans, M. Croad, Y. Rogan, J.C. Jansen, P. Bernardo, with nanofiltration, Water Res. 36 (2002) 1360–1368.
F. Bazzarelli, N.B. McKeown, An efficient polymer molecular sieve for membrane [62] K.Y. Wang, T. Matsuura, T.S. Chung, W.F. Guo, The effects of flow angle and shear
gas separations, Science 339 (2013) 303–307. rate within the spinneret on the separation performance of poly(ethersulfone) (PES)
[26] L.-Z. Zhang, Progress on heat and moisture recovery with membranes: from fun- ultrafiltration hollow fiber membranes, J. Membr. Sci. 240 (2004) 67–79.
damentals to engineering applications, Energy Convers. Manage. 63 (2012) [63] C.Z. Liang, J.T. Liu, T.-S. Chung, High-performance multiple-layer PIM composite
173–195. hollow fiber membrane for gas separation, J. Membr. Sci. (2018) (submitted for
[27] P. Scovazzo, A.J. Scovazzo, Isothermal dehumidification or gas drying using va- publication).
cuum sweep dehumidification, Appl. Therm. Eng. 50 (2013) 225–233. [64] M.R. Kosuri, W.J. Koros, Defect-free asymmetric hollow fiber membranes from
[28] J.R. Du, L. Liu, A. Chakma, X. Feng, Using poly(N,N-dimethylaminoethyl metha- Torlon®, a polyamide-imide polymer, for high-pressure CO2 separations, J. Membr.
crylate)/polyacrylonitrile composite membranes for gas dehydration and humidi- Sci. 320 (2008) 65–72.
fication, Chem. Eng. Sci. 65 (2010) 4672–4681. [65] N. Peng, T.S. Chung, K.Y. Wang, Macrovoid evolution and critical factors to form
[29] L.M. Robeson, The upper bound revisited, J. Membr. Sci. 320 (2008) 390–400. macrovoid-free hollow fiber membranes, J. Membr. Sci. 318 (2008) 363–372.
[30] M. Asaeda, L.D. Du, K. Ikeda, Experimental studies of dehumidification of air by an [66] S.P. Sun, T.A. Hatton, T.S. Chung, Hyperbranched polyethyleneimine induced cross-
improved ceramic membrane, J. Chem. Eng. Jpn. 19 (1986) 238–240. linking of polyamide-imide nanofiltration hollow fiber membranes for effective
[31] S. Sircar, Moisture removal from a wet gas, in: U.S. Patent (Ed.), 1993. removal of ciprofloxacin, Environ. Sci. Technol. 45 (2011) 4003–4009.
[32] A. Ito, Dehumidification of air by a hygroscopic liquid membrane supported on [67] A. Marković, D. Stoltenberg, D. Enke, E.U. Schlünder, A. Seidel-Morgenstern, Gas
surface of a hydrophobic microporous membrane, J. Membr. Sci. 175 (2000) 35–42. permeation through porous glass membranes Part I. mesoporous glasses-effect of
[33] Hyunho Kim, Sungwoo Yang, Sameer R. Rao, Shankar Narayanan, Eugene pore diameter and surface properties, J. Membr. Sci. 336 (2009) 17–31.
A. Kapustin, Hiroyasu Furukawa, Ari S. Umans, Omar M. Yaghi, E.N. Wang, Water [68] B.E. Poling, J.M. Prausnitz, J.P. O’Connell, The Properties of Gases and Liquids, fifth
harvesting from air with metal-organic frameworks powered by natural sunlight, ed., McGraw-Hill, New York, 2001.
Science 356 (2017) 430–434. [69] K. Tanaka, A. Taguchi, J. Hao, H. Kita, K. Okamoto, Permeation and separation
[34] L.Z. Zhang, Fabrication of a lithium chloride solution based composite supported properties of polyimide membranes to olefins and paraffins, J. Membr. Sci. 121
liquid membrane and its moisture permeation analysis, J. Membr. Sci. 276 (2006) (1996) 197–207.
91–100. [70] Y. Wu, X. Peng, J. Liu, Q. Kong, B. Shi, M. Tong, Study on the integrated membrane
[35] Y. Yampolskii, B. Freeman, Membrane Gas Separation, John Wiley & Sons Ltd., processes of dehumidification of compressed air and vapor permeation processes, J.
United Kingdom, 2010. Membr. Sci. 196 (2002) 179–183.
[36] R.W. Baker, B.T. Low, Gas separation membrane materials: a perspective, [71] E. Ryckebosch, M. Drouillon, H. Vervaeren, Techniques for transformation of biogas
Macromolecules 47 (2014) 6999–7013. to biomethane, Biomass Bioenergy 35 (2011) 1633–1645.
[37] M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publishers,

356