APPENDICES

APPENDIX A Validity of Permeate Pressure Linear Increase with Time

for Constant Volume Permeation Method

Figure A.1 represents the linearity of permeate pressure and the time in the

range of 0-0.1 bar for the dense PMP sample. Most of the data was concentrated on

the ideal line with minimal deviation from some of the samples. The same linearity

can also be extended for other samples utilizing constant volume permeation method.

0.10

0.09 N2
CO2
0.08 CH4
Permeate pressure (bar)

0.07 Ideal line

0.06

0.05

0.04

0.03

0.02

0.01

0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Normalized time
Figure A.1: Permeate pressure increase (0.00-0.10 bar) in constant downstream
volume with normalized time for CO2, N2, and CH4 across dense PMP
membrane

APPENDIX B Viscosity of P84 PI Solution in NMP in Literature

Figure B.1 represents the relationship between P84 PI’s solution

concentration with its viscosity. Viscosity data was taken from several literature, all

was taken at around 25 °C (Ismail et al., 2017, Langan and Salmon, 1987, Vanherck

et al., 2010, Flick, 1998). R2 of the plot fitting is at 0.91.

Figure B.1: Relationship between P84 PI’s solution concentration with its viscosity

APPENDIX C Validation of Image Processing in Determining Surface

Pore Size Distribution

Image processing is a powerful tool to obtain data from images. However,

it is limited by the degree of manipulation of the user and the quality of the original

images as significantly altered data could be obtained during the processing. Figure

C.1 represent the comparison of the threshold images with the original SEM images.

For pristine substrate, some low intensity structures were omitted during the threshold

setup. Nevertheless, these structures might be the deeper pores recorded due to the

very thin skin layer of pristine samples (at 0.21 ± 0.02 µm). Hence, only highly shaded

structures were measured. On the other hand, some background pixels were recorded

when using ‘Yen’ threshold for LiCl_4% substrate due to its dispersed pore structures.

In all samples, edge structures were excluded, while the threshold includes

the silhouette noted around the pores which in some instances merged pores that are

very close to each other. This is further confirmed by adjusting the lookup table (LUT)
of the original images from greyscale to ‘fire’ which gave a better contrast overall and

compared to the outlines of the threshold pores. As a result, some of the pore sizes

were exaggerated. However, as the pore counts are very high (at least 500 pores for

LiCl_4% substrate and more than 3 000 for pristine substrate), the reported pore size

distribution and values would be averaged over and gave a comparable result to

manual pore size measurement (for pristine substrate, pore diameter was at 429 ± 1

nm for image analysis against 411 ± 7 nm for manual measurement of 15 random

pores).

(a) (b) (c)

(i)

(ii)

(iii)

Figure C.1: Validation of image processing in determining surface pore size

distribution. (a) Original surface SEM at 5kx magnification, (b)
Threshold image after applying bandpass filter between 3-100 pixels and
suppressed horizontal stripes (c) Overlay of the particle outline with the
original SEM adjusted to ‘fire’ LUT for comparison
APPENDIX D Pristine PES Substrate N2 Gas Flux Comparison between
Test Rig and Porometer

The use of test rig to determine the N2 flux across the substrate was limited

for pristine sample for transmembrane pressure above 3 bar as the permeate flow rate

was too fast to be measured using the bubble flow setup (50 ml maximum). Hence, the

N2 flux for pristine sample was verified using the gas permeation function of Porolux

1000 porometer. Figure D.1 represent the comparison between these two methods. Gas

flux by porometer was recorded slower than the test rig. Nevertheless, the

measurement by porometer is increasing linearly with transmembrane pressure up to

10 bar, suggesting no compaction occurred during the short testing period.

100
Pristine (porometer)
Pristine (test rig)

80
N2 gas flux, cm3 (STP)/(cm2.s)

60

40

20

0
0 2 4 6 8 10
Transmembrane pressure, bar

Figure D.1: Pristine PES substrate layer’s N2 gas flux over different transmembrane
pressures for test conducted using test rig and porometer
APPENDIX E Preliminary Study on the Mass of PDMS Coating on
Pristine PES Substrate

Thickness of PDMS coating on the PES substrate can be controlled through

various methodologies. In this work however, the thickness was mainly controlled

through the mass of the solution layered onto the substrates. In this sense, a minimum

solution mass can be found before the resulting coating became unreliable due to

defect formation. Figure E.1 represents the N2 & CO2 gas permeance and CO2/N2

selectivity across PES-PDMS composite membrane made from various PDMS

solution mass between 0.25 to 4.0g. Both N2 and CO2 permeances were increasing

with decreased PDMS coating solution’s mass (hence its thickness), which is expected

and comparable to the full permeance results in Figure 4.15. However, CO2/N2

selectivity were lost for 0.25g PDMS solution, down to 5.38 against the PDMS-only

layer average at 10.20 ± 0.12. This was observed over the 1 to 4 bar transmembrane

pressure tested (with selectivity ranging from 5.19 to 6.65), suggesting that significant

defects formed at this thickness. Despite that the PDMS layer would later act as a

gutter and thus, less selective, more permeable gutter layer would be preferred, the

formation of defects at 0.25g might introduce further complications when a new

selective layer (hence much less permeable) is introduced and intruded into the defects.

Hence, the subsequent studies and repeats were only done for 0.5g-4.0g PDMS

solutions.
 0
0 1 2 3 4 5

PDMS coating solution mass (g)

600 16
N2 permeance
500 CO2 permeance 14

12
400

CO2/N2 selectivity
Permeance (GPU)
10
300
8
200
6
100
4

0 2

0
0 1 2 3 4 5 0 1 2 3 4 5
PDMS coating solution mass (g) PDMS coating solution mass (g)

Figure E.1: Preliminary N2 & CO2 permeances and CO2/N2 selectivity of pristine
substrate samples, coated with PDMS
600 of various solution’s mass between
0.25g to 4.0g. Dashed line represents the base CO2/N2 Nselectivity
2 permeance of
500 CO2 permeance
standalone PDMS layer
400
Permeance (GPU)

APPENDIX F Estimation of Capillary

300 Pressure by n-Heptane on the
Porous PES Substrates
200

Capillary pressure, 𝑝𝑐 can be estimated through

100
Young-Laplace equation:

2𝛾ℓ 𝑐𝑜𝑠𝜃 0
𝑝𝑐 = (F.1)
𝑟𝑐
0 1 2 3 4 5
Where 𝛾ℓ is the surface tension of the liquid, 𝜃 is the contact anglesolution
PDMS coating of themass
liquid
(g) with

the capillary’s surface, and 𝑟𝑐 is the radius of the pore. Meanwhile, surface energy of

the solid, 𝛾𝑠 can be approximated to be equal to the critical surface tension, 𝛾𝑐𝑟 of the

solid for polymer, and hence any liquid with 𝛾ℓ < 𝛾𝑐𝑟 will wet the surface (Agrawal,

2005, Fox and Zisman, 1952). With PES surface energy at 44.8 mJ/m2 (BASF, 2019)

and n-heptane surface tension at 20.14 mJ/m2, this suggested that complete wetting of

PES will occur for n-heptane (𝜃 = 0° and 𝑐𝑜𝑠𝜃 = 1). Fitting into Equation (F.1) gives

the corresponding 𝑃𝑐 . For 𝑟𝑐 between 100-500 nm (approximated range of PES pore

size from the surface pore data available), resulting 𝑃𝑐 would hence be between 0.81-

4.03 bar.
APPENDIX G Molecular Weight of PDMS against its Kinematic viscosity
and Hydrodynamic Radius

Molecular weight of PDMS (𝑀𝑤𝑃𝐷𝑀𝑆 ) can be approximated from its dynamic

viscosity (𝜂) based on the equation (Moretto et al., 2005):

464. (𝜂25 )0.825

𝑀𝑤𝑃𝐷𝑀𝑆 = (G.1)
2 + 0.0905. (𝜂25 )0.555

Where the subscript under 𝜂25 represents the measurement temperature. By comparing

this equation with data obtained from other literatures (Mojsiewicz-Pienkowska, 2008,

Mojsiewicz-Pienkowska, 2012, Kuo, 1999, Merck, 2021b), the equation can be

verified for application in this work, as presented in Figure G.1. Do note that the

literature data were presented in kinematic viscosity (𝜈) instead of 𝜂, where 𝜈 = 𝜂⁄𝜌

(𝜌 is the density of the fluid). By taking the average density at 0.96 g/mL, the resulting

correlation as compared to the literature data is as follows.

 106

Molecular weight, MW (Da) 105

104

103

102

101
10-1 100 101 102 103 104 105 106
Kinematic viscosity, u (cSt)

Kuo (1999) Mojsiewicz (2012) Merck (2021)

Mojsiewicz (2008a) Moretto et al. (2005)
Mojsiewicz (2008b) Expected Mw (this work)

Figure G.1: Average molecular weight of PDMS against its kinetic viscosity
compared to the correlation by Moretto et al. (2005)

The data distribution from literature corresponds well with the correlation

proposed by Moretto et al. (2005). Significant deviation can be seen however for data

from Merck (2021b) for 25 cSt hydroxy terminated PDMS, which is for hydroxy

terminated PDMS. However, the data was taken in terms of number averaged

molecular weight, which can be lower than the weight averaged molecular weight used

in the graph. Taking this into account, hence, for this work, the hydroxy terminated

PDMS used (viscosity 2550-3570 cSt) should have molecular weight of around 33.2-

37.8 kDa.

Using the correlation by Armstrong et al. (2004), the equivalent hydrodynamic

radius, 𝑅ℎ (in cm) can be approximated.

 3[𝜂]𝑀𝑊 1/3
𝑅ℎ = ( ) (G.2)
10𝜋𝑁

Where [𝜂] is the intrinsic viscosity assumed at 1.04 (for PDMS MW 250 kDa) in

heptane (Filiatrault et al., 1981) and 𝑁 is Avogadro’s number (6.022×1023 mol−1). This

gives the pre-crosslinked PDMS 𝑅ℎ in the range of 1.76-1.84 nm, which is comparable

to the molecular weight cut off-pore size relation noted for ultrafiltration membrane at

about 2.0-2.3 nm (Guo and Santschi, 2007).

APPENDIX H Calibration of Casted Dense PMP Thickness

Thickness of the fabricated dense PMP membrane was measured using

micrometre screw gauge, as presented in Figure H.1. PMP solution could only be

flattened into the petri dish down to the solution’s mass of 2g, which corresponds to

the thickness of 10.9 ± 0.6 µm. Nevertheless, the two thickness values obtained is quite

linear passing through the origin with an R2 of 0.9997. Comparison with the theoretical

thickness calculated found that the measured thickness is a bit higher than expected.

Back calculating the PMP solution’s concentration suggested that the theoretical

thickness would be equal to the measured value when PMP solution is at 2.4-2.5 wt.%.

This discrepancy could be due to the high volatility of cyclohexane which evaporates

during solution preparation, increasing the final solution concentration prior to the

pouring into the petri dishes.

 18
Measured
16 Theoretical

14

12
Thickness (mm)

10

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Solution's mass (g)
Figure H.1: Comparison of calculated and measured thickness of casted dense PMP
membrane