membranes

Article
Fabrication of Defect-Free Cellulose Acetate Hollow
Fibers by Optimization of Spinning Parameters
Xuezhong He
Department of Chemical Engineering, Faculty of Natural Sciences, Norwegian University of Science and
Technology, NO-7491 Trondheim, Norway; xuezhong.he@ntnu.no; Tel.: +47-7359-3942; Fax: +47-7359-4080

Academic Editor: Klaus Rätzke

Received: 15 April 2017; Accepted: 31 May 2017; Published: 5 June 2017

Abstract: Spinning of cellulose acetate (CA) with the additive polyvinylpyrrolidone (PVP) in
N-methyl-2-pyrrolidone (NMP) solvent under different conditions was investigated. The spinning
parameters of air gap, bore fluid composition, flow rate of bore fluid, and quench bath temperature
were optimized based on the orthogonal experiment design (OED) method and multivariate analysis.
FTIR and scanning electron microscopy were used to characterize the membrane structure and
morphology. Based on the conjoint analysis in Statistical Product and Service Solutions (SPSS)
software, the importance of these parameters was identified as: air gap > bore fluid composition >
flow rate of bore fluid > quench bath temperature. The optimal spinning condition with the bore fluid
(water + NMP (85%)), air gap (25 mm), flow rate of bore fluid (40% of dope rate), and temperature of
quench bath (50 ◦ C) was identified to make high PVP content, symmetric cross-section and highly
cross-linked CA hollow fibers. The results can be used to guide the spinning of defect-free CA hollow
fiber membranes with desired structures and properties as carbon membrane precursors.

Keywords: spinning; cellulose acetate; hollow fiber membrane; orthogonal experiment design;
conjoint analysis

1. Introduction
Carbon membranes have been studied in the last few years as a promising candidate for
energy-efficient gas separation technology due to their improved permselectivity, thermal and
mechanical stability, and chemical stability compared to those that are already used [1–4]. Hollow
fiber carbon membranes are usually prepared by the carbonization of hollow fiber precursors. How
to prepare cheaper, defect-free hollow fiber precursors becomes a key issue in the fabrication of
carbon membranes. Many studies have reported that polymer membranes such as polyimide,
polyacrylonitrile (PAN), and cellulose were used as the precursor for carbon membranes [2,5–7].
Cellulose is the most abundant biorenewable material with many important commercial applications.
However, the potential of cellulosic materials has not been fully exploited for use as the precursors for
carbon membranes because cellulose cannot be dissolved in conventional solvents due to strong
inter- and intra-molecular hydrogen bonding [8]. The recently developed Lyocell process uses
N-methylmorpholine-N-oxide (NMMO) to dissolve cellulose directly from biomass, and was reported
to have higher efficiency compared to other processes [9], but this process embodies significant
engineering challenges with regard to solvent stability, safety, and recovery. Ionic liquids are green
solvents that have recently been reported to dissolve the cellulose and spin cellulose fibers [8,10,11].
However, ionic liquids are very expensive, and many of them are still not commercially available.
Fortunately, the regeneration of spun cellulose acetate (CA) hollow fiber membranes provided
a potential solution to make cellulosic hollow fibers. Some studies have reported the spinning of
cellulose acetate fibers [12–18], but mainly used in reverse osmosis (RO) and ultrafiltration (UF), and

Membranes 2017, 7, 27; doi:10.3390/membranes7020027 www.mdpi.com/journal/membranes

Membranes 2017, 7, 27 2 of 9

only a few of them were used as precursors for carbon membranes. Defect-free precursors are crucial
for the preparation of high-performance carbon membranes. Thus, in this work, we will investigate
and optimize the spinning parameters to obtain defect-free CA hollow fiber membranes.
In the present work, the well-known dry-jet wet spinning technology was used to fabricate
thin and defect-free CA hollow fiber membranes. This process consists of the formation of nascent
Membranes 2017, 7, 27
membrane, followed by the interfacial phase separation within the air gap. After that, the nascent
membrane onlywas
a fewimmersed
of them were inused
a non-solvent
as precursors(water)
for carbon quench bathDefect-free
membranes. at a certain temperature
precursors where the
are crucial
for the preparation of high-performance carbon membranes. Thus, in this work,
phase separation occurred throughout the rest of the membranes. Many parameters, such as air gap, we will investigate
bore fluidand optimize the spinning
composition, flow rateparameters
of boretofluid,
obtainanddefect-free CA hollow
temperature offiber membranes.
quench bath, etc. can affect the
In the present work, the well-known dry-jet wet spinning technology was used to fabricate thin
final fiber structure and morphology. Qin [13] and Chung [19] reported that air gap length during
and defect-free CA hollow fiber membranes. This process consists of the formation of nascent
the spinning greatly
membrane, affected
followed by the performance
the interfacial phaseof membranes,
separation withinand an gap.
the air increase
After in airthe
that, gap resulted in a
nascent
significant decrease
membrane wasinimmersed
membrane permeation.
in a non-solvent The quench
(water) orthogonal
bath atexperimental design
a certain temperature (OED)
where the method
phasein
is well used separation occurred throughout
the multi-factor the restfield,
optimization of the and
membranes. Many parameters,
can consider the effectssuchofasall
airinvestigated
gap,
parametersborewhile
fluid composition,
significantly flowreducing
rate of bore fluid,
the and temperature
experimental of quench
runs. bath, etc.
This study can to
aims affect the
introduce the
final fiber structure and morphology. Qin [13] and Chung [19] reported that air gap length during
OED method to optimize the spinning parameters. The defect-free CA hollow fiber membranes were
the spinning greatly affected the performance of membranes, and an increase in air gap resulted in a
spun under the optimal
significant decrease spinning
in membrane conditions.
permeation. The cellulosic-based
The orthogonal membranes
experimental design (OED) canmethod
be regenerated
is
from thewell
spun CAinmembranes
used by deacetylation
the multi-factor treatment,
optimization field, and further
and can consider used as
the effects precursors
of all investigatedfor carbon
membranes parameters while significantly reducing the experimental runs. This study aims to introduce the OED
preparation.
method to optimize the spinning parameters. The defect-free CA hollow fiber membranes were spun
underand
2. Materials the optimal
Methods spinning conditions. The cellulosic-based membranes can be regenerated from the
spun CA membranes by deacetylation treatment, and further used as precursors for carbon
membranes preparation.
2.1. Materials
CA 2.(M
Materials and Methods
W 100,000, average acetyl content: 39.8%) was purchased from the ACROS (Pittsburgh,
PA, USA). Polyvinylpyrrolidone (PVP K30, MW 10,000) was purchased from Sigma (Darmstadt,
2.1. Materials
Germany). The solvent, N-methyl-2-pyrrolidone (NMP > 99.5%) was purchased from Merck
CA (MW 100,000, average acetyl content: 39.8%) was purchased from the ACROS (Pittsburgh,
(Darmstadt, Germany).
PA, USA). Polyvinylpyrrolidone (PVP K30, MW 10,000) was purchased from Sigma (Darmstadt,
Germany). The solvent, N-methyl-2-pyrrolidone (NMP > 99.5%) was purchased from Merck
2.2. Spinning of CA Fibers
(Darmstadt, Germany).
CA hollow fiber membranes were spun using the well-known dry-jet wet spinning process [13,20].
2.2. Spinning of CA Fibers
The dope solution consisted of CA, NMP, and the additive PVP (used to increase the porosity of the
CA hollow
carbon membrane). fiber membranes
A schematic diagramwerefor
spun
theusing the well-known
spinning process is dry-jet
shownwet in spinning
Figure 1.process
The extrusion
[13,20]. The dope solution consisted of CA, NMP, and the additive PVP (used to increase the porosity
rate for dope and bore fluid were controlled by two gear pumps, respectively. A double spinneret
of the carbon membrane). A schematic diagram for the spinning process is shown in Figure 1. The
(ID/OD,extrusion
0.5/0.7 rate
mm) forwas
dopeused in this
and bore fluidstudy with the by
were controlled aimtwoofgear
spinning
pumps, defect-free
respectively. hollow
A doublefibers by
controlling the spinning
spinneret parameters.
(ID/OD, 0.5/0.7 mm) was In usedorder tostudy
in this systematically
with the aim investigate the effects
of spinning defect-free of spinning
hollow
fibers
parameters by reduce
and controlling
thethe spinning parameters.
experimental times, In anorder to systematically
orthogonal investigate
experimental the effects
design (OED) of method
spinning parameters and reduce the experimental times, an orthogonal experimental
together with multivariate analysis was introduced to optimize the spinning conditions. The factors design (OED)
method together with multivariate analysis was introduced to optimize the spinning conditions. The
and levels for the OED are given in Table 1, and Statistical Product and Service Solutions (SPSS)
factors and levels for the OED are given in Table 1, and Statistical Product and Service Solutions (SPSS)
softwaresoftware
was used wasto generate
used the the
to generate experimental
experimentalplan.
plan.

Figure 1. Schematic diagram of membrane spinning process.

Figure 1. Schematic diagram of membrane spinning process.
2
Membranes 2017, 7, 27 3 of 9

Membranes 2017, 7, 27
Table 1. Factors and levels for orthogonal experimental design (OED) of spinning conditions.
Table 1. Factors and levels for orthogonal experimental design (OED) of spinning conditions.
Level Bore Fluid Composition Air Gap Flow Rate of Bore Fluid * Quench Bath Temp.
Level
1 Bore Fluid
HComposition
2O
Air15Gap
mm Flow Rate of 20% Bore Fluid * Quench Bath Temp.
20 ◦ C
21 H2O (85%)
H2 O + NMP 1525mm
mm 20%
40% °C◦ C
2050
32 H2O + NMP (85%) 2535mm
mm 60%
40% 50 °C
3 35same
* 100% means the mm flow rate as the
60%
dope flow rate.
* 100% means the same flow rate as the dope flow rate.
2.3. Measurement and Characterization
2.3. Measurement and Characterization
Fourier transform infrared spectroscopy (FTIR) spectra of the samples were obtained from the
Fourier transform infrared spectroscopy (FTIR) spectra of the samples were obtained from the
Bruker Tensor
Bruker 27 FTIR
Tensor instrument
27 FTIR instrument(Billerica,
(Billerica, MA, USA),which
MA, USA), whichwaswas used
used to to determine
determine the acetyl
the acetyl
content,
content, PVP content, and cross-linking degree between CA and PVP in the spun hollow fibers. Thefibers.
PVP content, and cross-linking degree between CA and PVP in the spun hollow
The morphology
morphology ofofspun
spunCACA hollow
hollow fibers
fibers werewere characterized
characterized by a scanning
by a scanning electron microscope
electron microscope (SEM)
(SEM)(Zeiss
(Zeiss SUPRA
SUPRA 55VP,
55VP, Oberkochen,
Oberkochen, Germany).
Germany).

3. Results
3. Results andand Discussion
Discussion

3.1. Experimental
3.1. Experimental Results
Results
The FTIR
The FTIR spectra
spectra of the
of the pureCA,
pure CA,thethepure
pure PVP,
PVP, the
thephysical
physicalmixture
mixture of CA
of CAandand
PVP,PVP,and and
spunspun
hollow fiber membrane were shown in Figure 2. The characteristic
hollow fiber membrane were shown in Figure 2. The characteristic adsorption peaks of 1030adsorption peaks of 1030 cm−1cm
, −1 ,
1230 cm −1, and 1740 cm −1 are attributed to the ether group (  ), acetyl ester group (  ), and
1230 cm−1 , and 1740 cm−1 are attributed to the ether group C(ν OC
C−O−C ), acetyl ester group (νCH3 −C=O ),
CH  C=O 3

carbonyl group
and carbonyl group (νC=O (  ) of CA, respectively. The peak at 1665 cm −1 is attributed
− 1 is attributed to thegroup
to the carbonyl
C=O ) of CA, respectively. The peak at 1665 cm carbonyl
in the PVP. Moreover, the assumption that an additional hydrogen
group in the PVP. Moreover, the assumption that an additional hydrogen bond bond may form
maybetween the CA the
form between
CA andandtertiary
tertiary amide
amide group
group of of PVP
PVPisispossible
possiblebecause
because thethe
IR IR
spectrum
spectrum displays a strong
displays absorption
a strong absorption
band in the region of 2250–2700 cm −
−1 that is characteristic of hydrogen bond for tertiary amide [21],
1
band in the region of 2250–2700 cm that is characteristic of hydrogen bond for tertiary amide [21],
as illustrated in Figure 3.
as illustrated in Figure 3.

3500 2955 2320 1740 1655 1420 1375 1230 1030

O

0,04 CA,C=O CH3-C O

C-O-C

0,03 Membrane NH +
PVP
C=O
C-CH3
Abs

0,02 -OH -CH

-CH stretch

0,01

0,00
0,6 1:4.5 Physical Mixture
0,4
Abs

0,2

0,0
1,2
0,9
CA 100,000
Abs

0,6
0,3
0,0
1,2
PVP K10
0,9
Abs

0,6
0,3
0,0
3500 3000 2500 2000 1500 1000 500
-1
wavenumber (cm )

Figure 2. Fourier
Figure transform
2. Fourier transforminfrared
infrared spectroscopy (FTIR)spectra
spectroscopy (FTIR) spectraof of cellulose
cellulose acetate
acetate (CA)(CA)
and and
polyvinylpyrrolidone (PVP), 1:4.5 Physical Mixture (CA and PVP power mixed) and Membrane
polyvinylpyrrolidone (PVP), 1:4.5 Physical Mixture (CA and PVP power mixed) and Membrane (spun
(spunCA/PVP
CA/PVP hollow
hollow fibers).
fibers).

3
Membranes 2017, 7, 27 4 of 9
Membranes
Membranes2017,
2017,7,7,2727

Figure
Figure3.
Figure Structure
3.3.Structure for
Structurefor the
forthe formation
theformation of
formationof hydrogen
ofhydrogen bond
hydrogenbond between
bondbetween CA
betweenCA and
CAand PVP.
andPVP.
PVP.

The
TheFTIR
The FTIRspectra
FTIR spectrafor
spectra forspun
for spunmembranes
spun membranesof
membranes ofofOED
OEDare
OED shown
areshown
are shownin ininFigures
Figures
Figures444and
and 5.5.The
and5. The absorption
Theabsorption
absorption
ratios
ratiosof
ofA2320
A2320 cm
cm−1
− /A1030
−11/A1030 cm
cm−1−1and
−and
1 A1665
A1665cm cm /A1030
−1−1
/A1030
− 1 cm
cm−1−1were
ratios of A2320 cm /A1030 cm and A1665 cm /A1030 cm were used to characterize were− 1 used
used to
to characterize
characterize the
thecross-
cross-
the
linking degree
linking degree
cross-linking and the
and the
degree PVP
PVP
and content
thecontent in the membrane,
in the membrane,
PVP content in the membrane,respectively.
respectively. The cross sections
The crossThe
respectively. sections of membrane
crossofsections
membraneof
morphology
morphology were
were characterized
characterized by
by SEM.
SEM. Table
Table 2 2 gives
gives the
the OED
OED
membrane morphology were characterized by SEM. Table 2 gives the OED results for differentresults
results for
for different
different spinning
spinning
conditions.
conditions.
spinning conditions.

Figure FTIR
Figure4.4.FTIR
FTIR spectra
spectra ofofthe
spectra ofthe
thespun
spun
spun hollow
hollow
hollow fiber
fiber fibermembranes
membranes
membranes for
for
for the the
theexperiment
experiment
experiment plan,
plan,which
plan, which which
were were
were
shown
shown
Tableinin
inshown 2.Table
Table2.2.

44
Membranes 2017, 7, 27 5 of 9
Membranes 2017, 7, 27

Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27
Membranes 2017, 7, 27

Figure 5. FTIR spectra of spun membranes for holdout experiments, which were shown in Table 2.
Figure 5. FTIR spectra of spun membranes for holdout experiments, which were shown in Table 2.

Table 2. The
Figure OED
5. FTIR results
spectra for the
of spun optimization
membranes of spinning
for holdout parameters.
experiments, which were shown in Table 2.
Thespectra
Table5.2.FTIR
Figure OEDofresults for the optimization
spun membranes of spinning
for holdout experiments, which parameters.
were shown in Table 2.
Figure 5. Flow
FTIRRate
spectraof of spun membranes for holdout Cross-experiments,PVP which were shown in Table 2.
FigureAir
5. FTIR
Gap spectra ofTable spun membranes Quench for Bath
holdout experiments, which were shown in Table 2.
Membrane
No. Bore Fluid Bore Fluid (%) 2. The OED results for the optimization
Linking of spinning Contentparameters.
Air Gap Table
(mm) 2. The
Figure
Flow 5.OED
Rate
Table
FTIR
*spectra
of
2. The
results
spectra for
of the
Temperature
Quench
OEDmembranes
optimization
spun membranes
(°C)
Bath
results for the
offor
spinning
holdout
Cross-Linking
optimization
Degree
parameters.
experiments,
ofc spinning
(%)
PVP which
Content
parameters.
wereMembrane
shown in Table 2.
Morphology
No. Bore Fluid Figure 5. FTIR
Table 2. The OED of of spun
results for the for holdout
◦ C)holdout
of optimization experiments,
c
of spinning which
parameters. were shown in Table 2.
Figure
(mm) 5. Bore
FTIR spectra
Fluid
Air Gap(%) *Flow
spun Rate
membranes
Temperature for
(Quench experiments,
BathDegree
Cross- which (%) PVPshown
were in Table 2.
Morphology
Membrane
Figure
No. Figure
BoreAir5. FTIR
Fluid spectra
Flow of
Rate spun
of
Bore membranes
Fluid (%) for holdout experiments,
Cross- Linking which
PVP were
Content shown in Table 2.
Figure 5.5. FTIR
Gap spectra of spun membranes for
for holdout experiments,
Cross- which were
PVP shown in Table
Table 2.
Flow Rate Quench
of Bath Membrane
No. Bore Fluid FTIRAir (mm)
spectra
Bore ofTable
Fluid
GapRate
Flow (%) *2.
ofspun The
membranesOED results
Temperature
Quench for(°C)
holdout
Bath the optimization
experiments,
Linking
Cross- Degree
of spinning
cwhich were
Content
PVP parameters.
shown inMembrane
(%) Membrane 2.
No. Bore
Air (mm)
Fluid
Gap Table 2. The
BoreOED results
FluidTemperature
(%)
Quench for
Bath the
(°C)optimization of spinning
Linking parameters.
Content
1 water
No. 15
Bore Fluid Table
40 2.* The
(mm)Fluid
Bore (%) OED results20 for the optimization
Temperature Degree cof spinning
5.19
(°C)
Linking 9.41(%) parameters.
Content
(mm) Table
Table 2. The OED results for
Flow Rate
* Temperature the
of optimization of
(°C) spinning
Degree c parameters.
Cross-(%) PVP
1 water 15 40 *2.
Table 2. The
Air
TheFlowOED
Gap
OEDRateresults
of 20 for
results for the
the optimization
Quench
Degree
optimization5.19 of
c Bathspinning
ofCross- (%) parameters.
spinning 9.41
PVP
parameters. Membrane
No. Bore
AirFluid
Gap Flow Rate ofBore FluidQuench
(%) Bath Cross- Linking
PVP Content
Membrane
1
No. water
Bore Fluid Air15Gap (mm)
Bore 40
Fluid
Rate (%) 20 Bath
Quench 5.19
Temperature Linking
(°C) 9.41
Content
1 No.
water Bore Fluid
15 (mm)
Air Gap 40
Flow
Bore
FlowFluid
of
Rate(%)
of 20*Temperature
Quench Bath(°C)5.19
Cross-
Cross- c 9.41Degree
Linking
PVP
c
Content
PVP (%)Membrane
Membrane
No.
1 Bore Fluid
water (mm)
Air15Gap FlowFluid
Bore Rate
* (%)
40 of Temperature
Quench (°C)
20 Bath Cross-
Degree
Linking
5.19 c PVP
(%)
Content
9.41 Membrane
1 No.
water Bore Fluid
15 Air Gap
(mm) 40 *
Bore Fluid Quench Bath
(%) 20 Temperature (°C)
5.19 Degree
Linking9.41 (%)
Content Membrane
No. Bore Fluid (mm) Bore Fluid
* (%) Temperature (°C) Linkingc
Degree c Content
(%)
2 water 35 20(mm) * 50 3.14
Temperature (°C) Degree9.01 (%)
1 water 15 * 40 20 Degree c 5.19 (%) 9.41
2 water 1 35
12
water
water
water
20
15
35
15
40
20
40
50 20
50
20
3.14 5.19
3.14
5.19
9.019.41
9.01
9.41
2 water
1 water35 15 20 40 50 20 3.14 5.19 9.01 9.41
21 water
water 15
35 40
20 20
50 5.19
3.14 9.41
9.01
2 +
water 1
water water
35 15 20 40 50 20 3.14 5.19 9.01 9.41
3 NMP 15 water
2 + 60
water 35 50 20 3.68 50 9.11 3.14 9.01
water + water
2 + water 35 20 50 3.14 9.01
3
(85%)
NMP3 23 15 NMP+
water
water 15
35
60
60
20 5050
50
50 3.68
3.68
3.14 9.11
9.01
9.11
NMP
water
2 + (85%)15
water 35 60 20 50 3.68 3.14 9.11 9.01
32 NMP
water 15
35 60
20 50
50 3.68
3.14 9.11
9.01
(85%)
3 (85%)
NMP2 water
15 35 60 20 50 50 3.68 3.14 9.11 9.01
(85%) water +
(85%) water
3 ++
water NMP 15 60 50 3.68 9.11
4 water 3 25 NMP+
water 60 15 60 20 50 5.63 3.68 10.08 9.11
34 water
NMP
water + 25
(85%)
15 60
60 20
50 5.63
3.68 9.11
4 water
3 water
(85%)
NMP25 + 15 60 60 20 50 5.63 3.68 9.11
4 water 43 25 (85%)
NMP
water 15
25
60 60
60 20 50
20 5.63 3.68
5.63 9.11
10.08
4 3
water NMP
(85%)
25 15 60 60 20 50 5.63 3.68 9.11
(85%)
(85%)
water + water
4 + water 25 60 20 5.63
4 +
water water 25 60 20 5.63
5 NMP 45 35 NMP+
water
water 40 25
35 40
60 20 20
20 8.61 8.61 8.59
5.63
5 NMP
4 +
water (85%)35
water 25 40 60 20 20 8.61 5.63
(85%)
water + 4
54 water
NMP 25
35 60
40 20
20 5.63
8.61
5 (85%)
NMP water
35 25 40 60 20 20 8.61 5.63
5 NMP 35 (85%) water
40 + 20 8.61 8.59
(85%) water
(85%) 5 +++
water
water NMP 35 40 20 8.61
water + 5 +
water NMP+
water 35 40 20 8.61
56 NMP++
NMP
water
water (85%)
25
35 20
40 20
20 3.03
8.61 8.16
6 NMP 6 5 +25
NMP
water water
(85%)
NMP25
(85%)
+ 20 35 20 40 20 20 20 3.033.03 8.61 8.16
8.16
655 (85%)
NMP
NMP 35
25 40
20 20
20 8.61
3.03 8.16
(85%)
6 (85%)
NMP NMP
(85%)
25 35 20 40 20 20 3.03 8.61 8.16
water + (85%)
(85%) water +
(85%) (85%)+
water
6 NMP 25 6 +
water NMP
20 25 20 20 20
3.03 3.03
8.16 8.16
6 NMP+
water 25
(85%) 20 20 3.03 8.16
(85%) 67 water
NMP +
water 25
25 40
20 50
20 4.07
3.03 9.41
8.16
7 6
water water
(85%)
NMP25 + 25 40 20 50 20 4.07 3.03 9.41 8.16
7 water 766 25 (85%)
NMP
water 40 25
25 20
40 50 20
50 4.07 3.03 9.41
4.07 8.16
9.41
7 water NMP
(85%)
25 25 40 20 50 20 4.07 3.03 9.41 8.16
(85%)
(85%)
7 water 25 40 50 4.07 9.41
7 water 25 40 50 4.07 9.41
7 water 78 25 water
water 40
15
25 20
40 50 20
50 4.07 21.47
4.07 9.41
10.83
9.41
8 water
7 water15 25 20 40 20 50 21.47 4.07 10.83 9.41
7
87 water
water 25 40 50 4.07 9.41
8 water
8 water 15 water
15 20 15
25 20 20
40 20 20 20
50 21.47
21.47
21.47
4.07 10.83
10.83
10.83
9.41

8 water 15 20 20 21.47 10.83

8 water 15 20 20 21.47 10.83
89 water
water 35
15 60
20 20
20 7.34
21.47 7.75
10.83
8 water9 water
8
988 15
water35
water
water
15 60
20
15
35
20
20
60 2020 20
20
20
7.34
21.47 21.47 7.75 10.83
21.47
7.34
10.83
10.83
7.75
9 water water
35 15 60 20 20 21.47 7.75 10.83
9 water 35 60 20 20 7.34
7.34 7.75
9 water 35 60 20 7.34 7.75
9 water 35 60 20 7.34 7.75
10(a)
9 water
water 15
35 60
60 20
20 2.79
7.34 7.72
7.75
10(a) 9
water water15 35 60 60 20 20 2.79 7.34 7.72 7.75
9
10(a) water
water 35
15 60
60 20
20 7.34
2.79 7.75
7.72
9 water
10(a) 9 35
water water
15 35 60
60 60 20
20 20 7.34
2.79 7.34 7.72 7.757.75
10(a) water 15 60 20 2.79 7.72
10(a) water 15 60 5 20 2.79 7.72
10(a)
10(a)
water
water
15
15
60
60 5 20
205
2.79
2.79
7.72
7.72
10(a) water 15 60 5 20 2.79 7.72
10(a) water 15 60 20 2.79 7.72
10(a) water 15 60 20 2.79 7.72
10(a) water 15 60
520 5
2.79 7.72
5
5
55
5
Membranes 2017, 7, 27 6 of 9

Table 2. Cont.

Air Gap Flow Rate of Quench Bath Cross-Linking PVP Content Membrane
No. Bore Fluid Membranes 2017, 7, 27
(mm) Bore Fluid (%) * Temperature (◦ C) Degree c (%) Morphology

water + water +
11(a) NMP 11(a)25 NMP 25
60 60 20 20 7.17 7.17 10.37
10.37
(85%) (85%)

12(b) water 25 40 20
12(b) water 25 water + 40 20
water + 13(b) NMP 25 40 20
13(b) NMP 25 (85%) 40 20
(85%) (a): Holdout; (b): Simulation; c hydrogen bond between CA and PVP, the value is calculated according
to the ratio of absorption from IR spectra (A2320 cm−1/A1030 cm−1); * Percentage of the dope flow rate.
NMP: N-methyl-2-pyrrolidone.
3.2. Conjoint Analysis
3.2. Conjoint Analysis
The conjoint analysis in the SPSS package was used to analyze the results of the orthogonal
The conjoint analysis in the SPSS package was used to analyze the results of the orthogonal
experimental design [22]. The utilities (part-worth) reflect the importance of each factor level. The
experimental design [22]. The utilities (part-worth) reflect the importance of each factor level. The
range (highest minus lowest)
range (highestofminus
the utility values
lowest) of for values
the utility each factor
for eachprovides a measurement
factor provides a measurementof of how
how
important the factorimportant
was to the overall
factor preference
was to overall[23]. Factors
preference [23].with larger
Factors utility
with larger ranges
utility play
ranges playaamore
more
significant
significant role compared to role
thosecompared to those with
with smaller smaller
ranges. ranges.
The The importance
importance scorescore
(IMP) (IMP) of factori i(%)
of factor (%) is
is
calculated as:
calculated as:
Rangei Rangei
I MPi = 100 pIMPi  100 p
(1)
(1)
∑ Range i Range i
i 1
i =1
where pIf= factor
where p = factor number. several number. If several
subjects subjects
are used inare
theused in the analysis,
analysis, the importance
the importance of eachfactor
of each factor is
is
separately calculated for each subject, and then averaged. For the prediction, the probability of each
separately calculated for each subject, and then averaged. For the prediction, the probability
simulation (pi) can be estimated according to the three different methods: (1) The maximum utility
of each
simulation (pi ) can be estimated
model determines according to the
the probability three
as the numberdifferent methods:
of respondents (1) The
predicted maximum
to choose utility
the case divided
model determines the by probability
the total numberas theofnumber
respondents;of respondents
(2) The BTL predicted to choose model
(Bradley–Terry–Luce) the case dividedthe
determines by
the total number of probability
respondents; as the(2)
ratio
The of BTL
one case utility to that for all simulation
(Bradley–Terry–Luce) model cases; and (3) The
determines thelogit model is
probability
similar to BTL, but uses the natural log of the utilities instead of the utilities. The conjoint analysis
as the ratio of one case utility to that for all simulation cases; and (3) The logit model is similar to BTL,
method reported in our previous work [23] was used to estimate the part worth of the contribution
but uses the naturalfrom
log each
of the utilities
factor’s level instead of the
in this work, and utilities.
the response The conjoint
is the analysis
combination method
of three subjectsreported
(the PVP
in our previous work [23] was used to estimate the part worth of the contribution fromwill
content, cross-linking degree, and membrane morphology; these parameters each factor’s
accordingly
influence
level in this work, and the microporosity,
the response mechanicalofstrength,
is the combination and structure
three subjects (the PVP and content,
morphology of carbon
cross-linking
membrane). The importance of each factor was calculated separately for each subject, and then
degree, and membrane morphology; these parameters will accordingly influence the microporosity,
averaged. The correlations of Pearson’s R and Kendall’s τ were 0.964 and 0.957, respectively, which
mechanical strength, and structure
indicated and morphology
a good consistency between theof carbonresults
estimated membrane).
from the The modelimportance of each
and the experimental
factor was calculated separately
data. for each
Table 3 shows subject,
the utilities and then
(part-worth) averaged.
of each Theand
factor level, correlations of Pearson’s
averaged importance R
score of
and Kendall’s τ were each factor.
0.964 andHigher utility
0.957, values indicate
respectively, greaterindicated
which preference in the selection
a good of spinning
consistency condition.
between the
estimated results from the model and the
Table experimental
3. Utilities data.
and averaged Table 3scores
importance shows the utilities
for different factors. (part-worth) of
each factor level, and averaged importance score
Factors and Levels
of each factor. Higher utility values
Utility Averaged Importance Score (%)
indicate greater
preference in the selection of spinning condition. water −0.917
Bore fluid composition 28.731
water + NMP(85%) 0.917
Table 3. Utilities and averaged importance15 mm scores −1.111
for different factors.
Air gap 25 mm 0.889 29.467
35 mm 0.222
Factors and Levels Utility Averaged Importance Score (%)
20% −0.778
Bore fluid composition Flow rate water
of bore fluid 40% −0.917 0.889 27.860
28.731
water + NMP(85%) 60% 0.917 −0.111
15 mm 20 °C−1.111 −0.750
Air gap Quench 25bath
mm Temp. 13.942
50 °C 0.889 0.750 29.467
35 mm (Constant) 0.222 5.556
20% −0.778
Flow rate of bore fluid 40% 0.889 27.860
60% −0.1116
20 ◦ C −0.750
Quench bath Temp. 13.942
50 ◦ C 0.750
(Constant) 5.556
Membranes 2017, 7, 27 7 of 9

Membranes
The 2017,
range7, of
27 the utility values (averaged importance score) for each factor provides a measure of
the importance of each factor to the overall performance. Factors with greater averaged importance
The range of the utility values (averaged importance score) for each factor provides a measure
score play a more significant role than those with smaller values. From Table 3, one could find that the
of the importance of each factor to the overall performance. Factors with greater averaged importance
importance of these four factors was sorted as follows:
score play a more significant role than those with smaller values. From Table 3, one could find that
•the air gap > bore
importance of fluid
thesecomposition > flow
four factors was rate as
sorted of bore fluid > quench bath temperature.
follows:
 aircan
It gapbe > bore
clearlyfluid
seencomposition
that air gap > flow
was theratemost
of bore fluid > quench
important bath of
parameter temperature.
the spinning process,
It can be clearly seen that air gap was the most important parameter
which kept a good consistency with the previous results [13,19]. Therefore, the length of the spinning process,
of air gap needs
which kept a good consistency with the previous results [13,19]. Therefore, the length
to be well controlled during the spinning process to prepare defect-free CA hollow fiber membranes of air gap needs
to be desired
with well controlled
structure during the spinning
and property. processalltothe
Moreover, prepare defect-free
utilities CA hollow
are expressed fiber membranes
in a common unit, and
with desired structure and property. Moreover, all the utilities are expressed
can be summed to give the total utility of any combination. Table 4 shows the comparison betweenin a common unit, and
an
can be summed to give the total utility of any combination. Table 4 shows the comparison
arbitrary selected combination of factor level and the optimal spinning condition. It can be found that between
an arbitrary
utility of the selected
optimal combination
condition (Caseof factor level higher
2) is much and thecompared
optimal spinning
to Case 1.condition. It can be found
that utility of the optimal condition (Case 2) is much higher compared to Case 1.
Table 4. An example for combination of different spinning conditions.
Table 4. An example for combination of different spinning conditions.
Utility
Utility
Case Bore Fluid Flow Rate Total Utility Remarks
Case Bore Fluid Flow Rate ofof Quench
Quench Bath
Bath Total Utility Remarks
Composition AirAir
GapGap
Bore Fluid Temp
Constant
Constant
Composition Bore Fluid Temp
(−0.917) 35 mm (−0.111) ◦ (−0.75)
11 water
water (−0.917) 35 mm (0.222) 60%
(0.222) 60%(−0.111) 2020°C C
(−0.75) 5.556
5.556 44
water + NMP
water + NMP
22 (85%) (0.917)
25 mm
25 mm (0.889) 40%
(0.889) 40% (0.889)
(0.889) °C◦ C
5050 (0.75)
(0.75) 5.556
5.556 9.001 Optimal
Optimal
(85%) (0.917)

3.3. Predictions
The real power of conjoint analysis is to predict the performance (structure and property) of the
hollow fibers
spun hollow fibersthat
thathave
havenotnotbeen
beeninvestigated
investigatedininthe
the experiments—the
experiments—the simulation
simulation cases
cases (see(see
No.No.12
12 and 13 in Table 2). The simulation results are given in Table 5. It was found that the
and 13 in Table 2). The simulation results are given in Table 5. It was found that the utility of Case 2 wasutility of Case
2 was larger
larger thanofthat
than that Case of 1Case 1 across
across the three
the three response
response variables
variables (PVP(PVP content,
content, cross-linking
cross-linking degree,
degree, and
and membrane
membrane morphology)
morphology) instudy.
in this this study. All three
All three models—maximum
models—maximum utility,utility, Bradley–Terry–Luce
Bradley–Terry–Luce (BTL),
(BTL),
and and logit—indicated
logit—indicated that simulation
that simulation Case 2bewould
Case 2 would be preferred.
preferred. In order toInvalidate
order to validate
this simulationthis
simulation
result, result,fiber
CA hollow CAmembranes
hollow fiber weremembranes
spun under were spunconditions
these two under these two conditions
and characterized and
by FTIR,
characterized
as shown in Figureby FTIR, as shown
6. The in Figure
PVP content 6. Thecm
(A1665 −1 /A1030
PVP content cm −1 ) was
(A1665 cm−1higher
/A1030and cm−1cross-linking
) was higher
and cross-linking
degree was weakerdegree
for Case was2, weaker for Case 2,
which indicated which
that indicated that
the membrane spunthefrommembrane
Case 2 was spun from
better forCase
use
2 was
as better forfor
a precursor use as a precursor
carbonization. forpredicted
The carbonization.
score The predicted
of Case scorehigher
2 was also of Case 2 was also
compared higher
to Case 1.
compared to Case 1. Thus, the generated model from the conjoint analysis can
Thus, the generated model from the conjoint analysis can be used to guide the spinning of hollow be used to guide the
spinning
fibers withofthe
hollow fibers
desired with the
structure anddesired structure and properties.
properties.

Figure 6.
Figure 6. FTIR
FTIR spectra
spectra for
for simulation
simulation cases,
cases, which
which has
has been
been mentioned
mentioned in
in Table
Table2.
2

Table 5. Simulation results by conjoint analysis. BTL: Bradley–Terry–Luce.

Card Number Score Maximum Utility(a) BTL Logit

1 5.667 33.3% 43.7% 26.9%
2 7.500 66.7% 56.3% 73.1%

7
Membranes 2017, 7, 27 8 of 9

Table 5. Simulation results by conjoint analysis. BTL: Bradley–Terry–Luce.

Card Number Score Maximum Utility(a) BTL Logit

1 5.667 33.3% 43.7% 26.9%
2 7.500 66.7% 56.3% 73.1%

4. Conclusions
CA hollow fiber membranes were spun from a dope solution containing (CA + PVP)/NMP at
different conditions using a dry-jet wet spinning process. The orthogonal experimental design method
was firstly used for the optimization of spinning conditions. The experimental results showed that the
importance of these four factors was sorted as:
air gap > bore fluid composition > flow rate of bore fluid > quench bath temperature.
The spinning parameter air gap was identified as the most important factor during the spinning
process, which kept the consistency with the results reported in the literature. Moreover, the optimal
spinning condition with a bore fluid composition (water + NMP (85%)), an air gap (25 mm), a flow
rate of bore fluid (40% of dope flow rate), and a temperature of quench bath (50 ◦ C) was obtained for
making cellulose acetate hollow fibers with high PVP content, symmetrical cross-section, and high
cross-linking degree. The proposed OED method can be well used for the optimization of spinning
conditions, and the simulation results based on conjoint analysis can be applied to guide the spinning
of hollow fibers with desired structure and properties.

Acknowledgments: The author would like to acknowledge the Research Council of Norway for the funding of
this work in the CO2 Hing project (PETROMAKS2 #267615).
Conflicts of Interest: The author declares no conflict of interest.

References
1. Barsema, J.N. Carbon Membranes Precursor, Preparation And Functionalization. Ph.D. Thesis, University of
Twente, Enschede, The Netherlands, 2007.
2. Favvas, E.P.; Kapantaidakis, G.C.; Nolan, J.W.; Mitropoulos, A.C.; Kanellopoulos, N.K. Preparation,
characterization and gas permeation properties of carbon hollow fiber membranes based on Matrimid(R)
5218 precursor. J. Mater. Process. Technol. 2007, 186, 102–110. [CrossRef]
3. Lie, J.A. Synthesis, Performance and Regeneration of Carbon Membranes for Biogas Upgrading—A Future
Energy Carrier. Ph.D. Thesis, Norwegian University of Science and Technology, Trondheim, Norway, 2005.
4. Saufi, S.M.; Ismail, A.F. Development and characterization of polyacrylonitrile (PAN) based carbon hollow
fiber membrane. Songklanakarin J. Sci. Technol. 2002, 24, 843–854.
5. Vu, D.Q.; Koros, W.J.; Miller, S.J. High Pressure CO2 /CH4 Separation Using Carbon Molecular Sieve Hollow
Fiber Membranes. Ind. Eng. Chem. Res. 2002, 41, 367–380. [CrossRef]
6. David, L.I.B.; Ismail, A.F. Influence of the thermastabilization process and soak time during pyrolysis
process on the polyacrylonitrile carbon membranes for O2 /N2 separation. J. Membr. Sci. 2003, 213, 285–291.
[CrossRef]
7. Soffer, A.; Koresh, J.; Saggy, S. Separation Device. U.S. Patent 4,685,940, 11 August 1987.
8. Sun, N.; Swatloski, R.P.; Maxim, M.L.; Rahman, M.; Harland, A.G.; Haque, A.; Spear, S.K.; Daly, D.T.;
Rogers, R.D. Magnetite-embedded cellulose fibers prepared from ionic liquid. J. Mater. Chem. 2008, 18, 1–9.
[CrossRef]
9. Rosenau, T.; Potthast, A.; Sixta, H.; Kosma, P. The chemistry of side reactions and byproduct formation in
the system NMMO/cellulose (Lyocell process). Prog. Polym. Sci. 2001, 26, 1763–1837. [CrossRef]
10. Swatloski, R.P.; Spear, S.K.; Holbrey, J.D.; Rogers, R.D. Dissolution of Cellose with Ionic Liquids. J. Am.
Chem. Soc. 2002, 124, 4974–4975. [CrossRef] [PubMed]
11. Zhang, H.; Wu, J.; Zhang, J.; He, J. 1-Allyl-3-methylimidazolium Chloride Room Temperature Ionic Liquid:
A New and Powerful Nonderivatizing Solvent for Cellulose. Macromolecules 2005, 38, 8272–8277. [CrossRef]
Membranes 2017, 7, 27 9 of 9

12. Jie, X.; Cao, Y.; Qin, J.-J.; Liu, J.; Yuan, Q. Influence of drying method on morphology and properties of
asymmetric cellulose hollow fiber membrane. J. Membr. Sci. 2005, 246, 157–165. [CrossRef]
13. Qin, J.-J.; Li, Y.; Lee, L.-S.; Lee, H. Cellulose acetate hollow fiber ultrafiltration membranes made from
CA/PVP 360 K/NMP/water. J. Membr. Sci. 2003, 218, 173–183. [CrossRef]
14. Liu, H.; Hsieh, Y.-L. Ultrafine fibrous cellulose membranes from electrospinning of cellulose acetate. J. Polym.
Sci. Part B Polym. Phys. 2002, 40, 2119–2129. [CrossRef]
15. Son, W.K.; Youk, J.H.; Lee, T.S.; Park, W.H. Electrospinning of Ultrafine Cellulose Acetate Fibers: Studies of a
New Solvent System and Deacetylation of Ultrafine Cellulose Acetate Fibers. J. Polym. Sci. Part B Polym. Phys.
2004, 42, 5–11. [CrossRef]
16. Hao, J.-H.; Dai, H.-P.; Yang, P.-C.; Wei, J.-M.; Wang, Z. Cellulose acetate hollow fiber performance for
ultra-low pressure reverse osmosis. Desalination 1996, 107, 217–221. [CrossRef]
17. He, X.; Lie, J.A.; Sheridan, E.; Hägg, M.-B. Preparation and Characterization of Hollow Fiber Carbon
Membranes from Cellulose Acetate Precursors. Ind. Eng. Chem. Res. 2011, 50, 2080–2087. [CrossRef]
18. He, X.; Hägg, M.-B. Structural, kinetic and performance characterization of hollow fiber carbon membranes.
J. Membr. Sci. 2012, 390–391, 23–31. [CrossRef]
19. Chung, T.-S.; Hu, X. Effect of air-gap distance on the morphology and thermal properties of polyethersulfone
hollow fibers. J. Appl. Polym. Sci. 1997, 66, 1067–1077. [CrossRef]
20. Qin, J.-J.; Wong, F.-S.; Li, Y.; Liu, Y.-T. A high flux ultrafiltration membrane spun from PSU/PVP
(K90)/DMF/1,2-propanediol. J. Membr. Sci. 2003, 211, 139–147. [CrossRef]
21. Tang, H.T. Spectroscopic Identification for Organic Compounds; Beijing University Press: Beijing, China, 1992.
22. SPSS Conjoint™ 17.0. Available online: http://www.sussex.ac.uk/its/pdfs/SPSS_Conjoint_17.0.pdf
(accessed on 15 May 2017).
23. He, X.; Hagg, M.-B. Optimization of Carbonization Process for Preparation of High Performance Hollow
Fiber Carbon Membranes. Ind. Eng. Chem. Res. 2011, 50, 8065–8072. [CrossRef]

© 2017 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).