Celulosa Nanotubos

THE AMERICAN UNIVERSITY IN CAIRO
Department of Chemistry
Carbon Nanotubes-Cellulose
Acetate Nanocomposites
Membranes for Water Desalination
By
Nouran Ashraf
BS, Chemistry, The American University in Cairo, 2006
Advisors:
Dr. Adham Ramadan

Department of Chemistry
Dr. Amal Esawi
Department of Mechanical Engineering
Submitted to the School of Sciences and Engineering, Department of

Chemistry, the American University in Cairo
In partial fulfillment of the requirements of the Degree of Master of

Science in Chemistry
Fall 2012
To the martyrs of January 25th revolution,
May you all rest in heaven
Abstract
Cellulose acetate (CA) (Mw = 52,000 Da) membranes were prepared by phase inversion (PI)
using acetone as a solvent. Investigation of different preparation conditions were carried out.
The effect of membrane casting thickness, CA content, coagulation bath temperature (PI
temperature), solvent evaporation, addition of a non-solvent (deionized water), and addition
of multi-walled carbon nanotubes (MWCNTs) on membrane morphology and performance
(permeation rates and salt rejection rates) were investigated. Membranes morphologies were
studied using scanning electron microscopy (SEM). Membranes permeations rates and salt
rejection rates were investigated using 1000 ppm NaCl solution. Optimum conditions for
developing a CA based nanocomposites were attained, entailing 15 wt% CA content, 20 wt%
H2O non-solvent additive, low functionalized CNTs contents (0.0005, 0.005, and 0.01 wt%),
PI at room temperature, and sonication time for CNTs proper dispersion less than 1 minute.
MWCNTs/CA nanocomposites membranes were prepared. MWCNTs were first

functionalized by oxidation purification in a strong acidic medium to enhance their dispersion
within the polymer matrix, and the success of the functionalization of MWCNTs was
characterized using Fourier transform infrared (FTIR) spectroscopy. The nanocomposites
morphologies were characterized by several methods by SEM and nitrogen adsorption. SEM
images showed large networks of MWCNTs, randomly oriented and properly dispersed, with
a significant decrease in the number of macrovoids development with CNT content increase
at the same final thickness of the nanocomposites. This was verified by analysis of pore sizes
(differential volumes and surface areas), which were found to decrease with the increase in
CNT content. Nanocomposites permeations rates and salt rejection rates were investigated
using 1000 ppm NaCl solution, and it was found that permeation improved significantly with
the addition of CNTs, with the improvement being highest for lowest CNT content. Salt
rejection was found to decrease with the presence of CNTs. However the decrease was
minimal for low CNT contents. In this respect, it was possible to prepare CA-CNT
nanocomposite membranes with improved permeation of 19.57 L/m2h, together with a
minimal decrease of salt retention performance of 69.4% at 24 bars operating pressure. The
membranes performance could be explained by membrane morphology (surface areas and
porosity).
ii
Acknowledgment
I would like to deeply thank my advisor and mentor Dr. Adham Ramadan and my co-advisor
Dr. Amal Esawi for giving me the time, effort, and guidance to carry out this project. Their
encouragement, patience, and constructive feedbacks were such a positive energy that kept
me going until seeing this through. Their moral support on all matters, academic and personal
significantly affected my life as a whole, not only my research skills. No words will ever
fulfill my gratitude, and I am really honored to have them as my advisors. Thank you.
I would also like to express my deep appreciation for the Youssef Jameel Science and
Technology Research Center (YJSTRC) for their financial support, through a full fellowship,
over my years of study for the Masters degree in Chemistry. I cannot fail to mention Dr.
Sherif Sedky, the former director of YJSTRC, who always believed in me, and Dr. Ehab
Abdel Rahman, the current director of YJSTRC, who always encouraged me. This has been a
wonderful research experience in a well renowned facility. I also want to thank Eng. Rami
Wasfi, Dr. Omar Zaki, and Eng. Ehab Salama, for training me on how to use some of the
equipment in the research center, for, without them, I wouldn’t have been able to carry on the
work done.
I also wish to thank the faculty and staff members of the Chemistry Department who have
always had their doors wide open for all students who needed help in any matter. I would
specially like to thank Dr. Nahed Yaqoub for her time in aiding me with many of the
calculations carried out. The Chemistry Department has been my second home for more than
eleven years. It is the place which expanded my horizon to new possibilities, and offered me
new chances of education and achievements. It is the place where I have a real sense of
belonging.
I would also like to thank Dr. Mohammed Morsi, Mechanical Engineering Department, for
helping me carry out the permeation measurements. His time and effort are greatly
appreciated.
On more personal bases, I would love to thank my parents for being my lighthouse when life
darkened the most. Their unconditional love and devotion humble me. No daughter can ever
have better parents, for they are, and forever will be my comfort zone.
I would also love to thank my best friends Haidy Youssef, Noha Khattab, Samaa Nawaar,
Dina Nemr, and Ayaat Mahmoud for being there for me all the time and never letting me
down.
Last, I wish to express my gratefulness, love, and admiration to my Tahrir buddies: Hany el
Kady, Ayaat Mahmoud, Dina Nemr, Mahmoud Bizzari, Samaa Nawaar, Essam Omar, Nada
Nashaat, Nadeem Gawad, Ramy el Gohari, Haidy Youssef, Hani Tawfeek, and Amr Zeid.
These people are treasure. They had my back, cherished my life over theirs, and helped me
survive the past two years both psychologically and physically. I owe them more than I can
possibly pay back, and I am really blessed to have them in my life.
Thank you all for an overwhelming wonderful and challenging experience.
iii
Contents
Abstract ......................................................................................................................................ii
Acknowledgment ......................................................................................................................iii
List of Figures ......................................................................................................................vii
List of Tables .......................................................................................................................xiii
List of Abbreviations ........................................................................................................... xiv
1 Introduction ................................................................................................................. 1
1.1 History ......................................................................................................................... 2
1.2 Desalination Technologies .......................................................................................... 2
1.2.1 Multi-Stage Flash (MSF)..................................................................................... 3
1.2.2 Multiple-Effect Distillation (MED)..................................................................... 4
1.2.3 Reverse Osmosis (RO) ........................................................................................ 5
1.2.4 Other Processes ................................................................................................... 7
2 Literature Background ................................................................................................ 9
2.1 Introduction ............................................................................................................... 10
2.2 Effect of Different Additives..................................................................................... 10
2.2.1 Effect of Solvents and Non-Solvents ................................................................ 10
2.2.2 Effect of Pore Formers ...................................................................................... 13
2.2.3 Effect of Using Polymer Blends ........................................................................ 17
2.2.4 Effect of Inorganic Additives ............................................................................ 21
2.2.5 Effect of Using Carbon Nanotubes ................................................................... 22
2.3 Research Aim ............................................................................................................ 26
3 Theoretical Background ............................................................................................ 27
3.1 Cellulose Acetate in Desalination ............................................................................. 28
3.1.1 CA Structure and Properties .............................................................................. 28
3.1.2 CA Membrane Preparation ................................................................................ 29
3.2 Permeation and Retention Rates Calculation ............................................................ 32
3.3 Nanocomposites ........................................................................................................ 34
3.4 Carbon Nanotubes as Nanofillers .............................................................................. 35
3.4.1 CNTs Structure and Properties .......................................................................... 35
3.4.2 Functionalization of CNTs by Oxidation Purification ...................................... 37
iv
3.4.3 CNTs in Water Desalination and Treatment ..................................................... 38
3.5 Inside CNTs Molecular Dynamics (MD) .................................................................. 38
3.6 Membrane Characterization ...................................................................................... 41
3.6.1 Scanning Electron Microscope (SEM) .............................................................. 41
3.6.2 Surface Analysis by Gas Sorption ..................................................................... 42
4 Experimental ............................................................................................................. 49
4.1 Materials Used........................................................................................................... 50
4.2 Stock Solutions Preparation ...................................................................................... 50
4.2.1 CA Stock Solution ............................................................................................. 50
4.2.2 CA Stock Solution with Pristine (non-functionalized) Carbon Nanotubes
(pNTs) ............................................................................................................... 50
4.2.3 CA Stock Solution with Functionalized CNTs (fNTs) ..................................... 51
4.3 Methods and Instrumentation .................................................................................... 51
4.3.1 Functionalization of Carbon Nanotubes ............................................................ 51
4.3.2 Membrane Casting ............................................................................................ 52
4.3.3 Sample Surface Area and Porosity Determination ............................................ 52
4.3.4 Infrared Analysis ............................................................................................... 53
4.3.5 Scanning Electron Microscopy ......................................................................... 53
4.3.6 Liquid Test Cell ................................................................................................. 53
4.4 Investigating Preparation Conditions ........................................................................ 54
4.4.1 Membrane Casting Thickness ........................................................................... 55
4.4.2 Coagulation Bath Temperature (CBT) (PI Temperature) ................................. 57
4.4.3 CA Content ........................................................................................................ 57
4.4.4 Solvent Evaporation .......................................................................................... 57
4.4.5 Addition of H2O (Non-Solvent) ........................................................................ 58
4.4.6 Addition of CNTs .............................................................................................. 59
4.5 Optimal Preparation Conditions Samples ................................................................. 60
5 Results and Discussion ............................................................................................. 61
5.1 Effect of Preparation Conditions on Membranes Morphology ................................. 62
5.1.1 Membrane Casting Thickness ........................................................................... 63
5.1.2 Coagulation Bath Temperature (CBT) .............................................................. 67
5.1.3 CA Content ........................................................................................................ 70
v
5.1.4 Solvent Evaporation .......................................................................................... 73
5.1.5 Addition of H2O (Non-Solvent) ........................................................................ 77
5.1.6 Addition of CNTs .............................................................................................. 82
5.2 Optimal Preparation Conditions ................................................................................ 91
5.3 Membrane Surface Properties ................................................................................... 92
5.3.1 Differential Pore Sizes....................................................................................... 92
5.3.2 Adsorption Isotherms ........................................................................................ 97
5.3.3 The t-Plots ......................................................................................................... 99
5.4 Membranes Performance......................................................................................... 102
5.4.1 Permeation Rates ............................................................................................. 102
5.4.2 Salt Retention Rates ........................................................................................ 103
5.4.3 BET Surface Areas .......................................................................................... 105
5.4.4 Membrane Morphology and Performance: General Discussion ..................... 107
6 Conclusions ............................................................................................................. 112
6.1 Conclusion............................................................................................................... 113
6.2 Future Work ............................................................................................................ 114
References .............................................................................................................................. 116
Appendix I ............................................................................................................................. 124
vi
List of Figures
Image on the front page is the cover page of nanotoday, vol.6, no. 2, December 2007
Figure ‎1.1 Desalination saline sources % worldwide [11] ........................................................ 3
Figure ‎1.2 Desalination processes % worldwide [11] ............................................................... 3
Figure ‎1.3 Diagram of Multi Stage Flash Process [13] ............................................................. 4
Figure ‎1.4 Diagram of Multiple Effect Distillation Process [13] .............................................. 5
Figure ‎1.5 Osmosis and Reverse Osmosis flow [17] ................................................................. 6
Figure ‎1.6 Pore size distribution of each membrane type [14] .................................................. 7
Figure ‎2.1 Flux rates vs. Rejection rates of 2000 ppm NaCl solution for temperature ranges of
60,65, 70, 75, and 80oC, (a) data of 20 wt% cellulose acetate membranes, (b) data
of 22 wt% cellulose acetate membranes [23] .......................................................... 12
Figure ‎2.2 Pure water flux and BSA rejection rates vs. Pluronic F127 content [26]............... 14
Figure ‎2.3 (a) SEM of blank cellulose acetate membrane in pure water gelation medium, (b)
SEM of cellulose acetate membrane with 30 wt% silica nanoparticles prepared in
HCl gelation medium (pH 1) [33] ........................................................................... 17
Figure ‎2.4 Effect of PVP wt% on metal ions rejection: (a) blank cellulose acetate, (b) 75/25
wt% CA/PSF [31] ................................................................................................... 18
Figure ‎2.5 SEM of the effect of PVP on morphology; (a) 2.5 wt% PVP in 100/0 wt% CA/PSF
respectively, (b) 7.5 wt% PVP in 100/0 wt% CA/PSF respectively, (c) 2.5 wt%
PVP in 75/15 wt% CA/PSF respectively, (d) 7.5 wt% PVP in 75/15 wt% CA/PSF
respectively [31] ...................................................................................................... 19
Figure ‎2.6 31P NMR of (HPO4)2- vs. Zr(HPO4)2-. (a) singlet of phosphorous of (HPO4)2-, (b)
singlet of phosphorous of Zr(HPO4)2- [36] ............................................................ 21
Figure ‎2.7 SEM of the cross section of CNTs/PSF membrane, (a) 0 wt% CNTs blank PSF
membrane, (b) 2 wt% CNTs/PSF membrane [38] .................................................. 23
Figure ‎2.8 MWCNTs contents effect on PWF and PEG rejection rates [39] ......................... 24
Figure ‎2.9 A schematic representation of how CNTs concentration can block pores, thus
reduce pore sizes [40]............................................................................................. 25
Figure ‎3.1 Membrane lifetime vs. pH [48] .............................................................................. 29
Figure ‎3.2 Schematic diagram of a film applicator and a casting knife .................................. 29
vii
Figure ‎3.3 Interpretation of the demixing process ................................................................... 30
Figure ‎3.4 Chart elaborating the effect of solvents on membrane morphology [51] ............... 31
Figure ‎3.5 Schematic representation of an asymmetric membrane developed via PI, modified
from [52] ................................................................................................................. 32
Figure ‎3.6 Permeation cell (modified from [29])..................................................................... 33
Figure ‎3.7 Illustration of how the particle number increase as their sizes decrease in a given
volume [61] ............................................................................................................. 34
Figure ‎3.8 (a), (b), (c) SWCNTs different structures with highlight on C-C bond orientation
vs. the nanotube axis [67], (d) MWCNTS structure [63] ........................................ 36
Figure ‎3.9 (a) (n, m) vector indication [64], (b) Different conformation of CNTs, their chiral
angles, and indices [68] ........................................................................................... 36
Figure ‎3.10 Section of an oxidized CNT reflecting terminal and wall oxidation [71] ............ 37
Figure ‎3.11 1D wired mode of water molecules inside (8, 2) CNT [68] ................................. 39
Figure ‎3.12 Water molecules structure inside CNTs (a) Spiral-like layered mode, (b) bulky
mode [68] ................................................................................................................ 40
Figure ‎3.13 Schematic diagram showing the main components of SEM [80] ........................ 41
Figure ‎3.14 The dimension of a MWCNT ............................................................................... 42
Figure ‎3.15 Different stages of gas physisorption on sample surface with pressure increase
[85] .......................................................................................................................... 43
Figure ‎3.16 Different types of isotherm linear plots [84] ........................................................ 44
Figure ‎3.17 BET plot to evaluate Vm [91]................................................................................ 46
Figure ‎3.18 t-plot comparison between porous and nonporous surface [92] ........................... 47
Figure ‎4.1 Elcometer 4040 Automatic Film Applicator .......................................................... 52
Figure ‎4.2 Micrometrics ASAP 2020 instrument [94]............................................................. 53
Figure ‎4.3 Sterlitech HP4750 Stirred Cell ............................................................................... 54
Figure ‎4.4 Different preparation conditions............................................................................. 55
Figure ‎5.1 SEM of an asymetric membrane ............................................................................ 62
viii
Figure ‎5.2 SEM of 17 wt% CA membranes, PI carried out at room temperature: (a) ICT =
500 µm, final thickness = ~55 µm, (b) ICT = 600 µm, final thickness = ~75 µm,
(c) ICT = 800 µm, final thickness = ~122 µm ........................................................ 63
Figure ‎5.3 SEM of 15 wt% CA membranes, PI carried out at room temperature: (a) ICT =
(c) ICT = 800 µm, final thickness = ~122 µm ........................................................ 64
Figure ‎5.4 SEM of 17 wt% CA + 10 wt% H2O membranes, PI at room temperature: (a) ICT =
700 µm, final thickness = ~137 µm, (b) ICT = 800 µm, final thickness = ~226 µm
................................................................................................................................. 65
(c) ICT = 500 µm, final thickness = ~55.4 µm, (d) ICT = 600 µm, final thickness =
~110 µm .................................................................................................................. 65
Figure ‎5.6 A plot showing the non-linear change in final thickness of 17 wt% CA membranes
with the increase in ICT. PI was done at room temperature ................................... 67
Figure ‎5.7 SEM of 17% CA content, ICT = 1000 µm: (a) FT = ~180 µm, PI at room
temperature, (b) Voids necks at high magnification for PI at room temperature (c)
FT = ~172 µm, PI at 40oC, (d) Voids necks at high magnification for PI at 40oC . 68
temperature, (b) Voids necks at a high magnification, PI at room temperature, (c)
FT = ~138 µm, PI at 40oC , (d) Voids necks at a high magnification, PI at 40oC .. 69
Figure ‎5.9 SEM of three 13 wt% CA membranes prepared with ICT = 1200 µm: (a) final
thickness = ~173 µm, (b) final thickness = ~186 µm, (c) final thickness = ~216 µm
................................................................................................................................. 71
Figure ‎5.10 SEM of 13 wt% CA membranes, ICT = 1000 µm: (a) final thickness= ~155 µm,
(b) final thickness= ~185 µm .................................................................................. 72
Figure ‎5.11 SEM of membranes prepared with PI at room temperature, ICT = 800 µm, final
thickness of ~122 µm: (a) 17 wt% CA, (b) 15 wt% CA ......................................... 72
Figure ‎5.12 SEM of 17 wt% CA membranes: (a) complete PI for 7.5 min, final thickness =
~180 µm, (b) PI for 5.5 min, followed by solvent evaporation for 3.5 min, final
thickness = ~138 µm, (c) PI for 2.5 min followed by solvent evaporation for 3.5
min, final thickness = ~ 135 µm.............................................................................. 74
Figure ‎5.13 SEM of 17 wt% CA membranes, PI at room temperature for 7.5 minutes: (a) ISE
= 0 sec, final thickness = ~180 µm, (b) ISE = 60 sec, final thickness = ~77 µm ... 75
ix
Figure ‎5.14 SEM of 13 wt% CA membranes, ICT = 1000 µm, and PI at room temperature:
(a) and (b) are two different locations on the same sample with ISE = 30 sec, final
thickness range = 217-161 µm, (c) uniform sample prepared with ISE = 60 sec.,
final thickness = 123.5 µm ...................................................................................... 77
Figure ‎5.15 SEM of 17 wt% CA, PI at room temperature, ICT = 700 µm: (a) 0 wt% H2O,
final thickness = ~80 µm, (b) 10 wt% H2O, final thickness = ~137 µm, (c) 15 wt%
H2O, final thickness = ~153 µm, (d) 20 wt% H2O, final thickness = ~246 µm...... 78
final thickness = ~30 µm, (b) 20 wt% H2O, final thickness = ~80 µm ................... 79
final thickness = ~138 µm, (b) 5 wt% H2O, final thickness = ~199 µm, (c) 10 wt%
H2O, final thickness = ~264 µm (non-uniform) ...................................................... 80
Figure ‎5.18 SEM displaying compaction difference in the middle sections of 13 wt% CA
membranes, PI at room temperature: (a) 0 wt% pNTs at 5KX, (b) 0 wt% pNTs at
10KX, (c) 0.5 wt% pNTs at 5KX, (d) 0.5 wt% pNTs at 10KX .............................. 82
membranes, PI at room temperature: (a) 0 wt% pNTs at 5KX, (b) 0.5 wt% pNTs at
5KX ......................................................................................................................... 83
Figure ‎5.20 SEM of agglomerates of pNTs in 17 wt% CA membranes at 25KX in the middle
section of the membranes: (a) 0.5 wt% pNTs at 25KX, (b) 1 wt% pNTs at 25KX 83
Figure ‎5.21 SEM of the pores structures in the middle section of 17 wt% CA membranes at
25KX: (a) 0 wt% pNTs membrane, (b) 0.5 wt% pNTs membrane, (c) 1 wt% pNTs
membrane ................................................................................................................ 84
Figure ‎5.22 SEM image of both pNTs and fNTs: (a) an agglomerate of pNTs at 700X, (b) a
bundle of pNTs at 100 KX, (c) agglomerates of fNTs at a small magnification
(700X), (d) arrays of fNTs at 100 KX .................................................................... 85
Figure ‎5.23 FTIR spectrum of non-functionalized MWCNTs ................................................ 86
Figure ‎5.24 FTIR spectrum of functionalized MWCNTs ........................................................ 86
Figure ‎5.25 Beakers filled with 0.01 g CNTs, sonication time for 1 minute in 20 ml deionized
water, (a) fNTs fully dispersed giving opaque black solution, (b) pNTs poorly
dispersed forming agglomerates throughout the solution and on the bottom of the
beaker ...................................................................................................................... 87
Figure ‎5.26 MWCNTs networks in 0.0005/15/20 wt% fNT/CA/H2O nanocomposite at

different SEM magnifications, (a) at 25KX, (b) at 50KX....................................... 88
x
Figure ‎5.27 MWCNTs networks in 0.005/15/20 wt% fNT/CA/H2O nanocomposite at
different SEM magnifications, (a) at 50KX, (b) at different location at 50KX, (c) at
100KX, (d) at different location at 100KX ............................................................. 88
Figure ‎5.28 MWCNTs networks in 0.01/15/20 wt% fNT/CA/H2O nanocomposite at different

SEM magnifications, (a) at 25 KX, (b) at 50KX, (c) at 100KX, (d) at 300KX ...... 89
Figure ‎5.29 SEM of morphology at same final thickness (100±10 µm): (a) 0/15/20 wt%
fNT/CA/H2O, (b) 0.0005/15/20 wt% fNT/CA/H2O, (c) 0.005/15/20 wt%
fNT/CA/H2O, (d) 0.01/15/20 wt% fNT/CA/H2O ................................................... 90
Figure ‎5.30 Plot of differential pore volumes .......................................................................... 92
Figure ‎5.31 Plot of differential pore volumes for micro and mesopores ................................. 93
Figure ‎5.32 Plot of differential pore volumes for macropores................................................. 93
Figure ‎5.33 Plot of differential pore areas ............................................................................... 94
Figure ‎5.34 Plot of differential pore areas for micro and mesopores ...................................... 94
Figure ‎5.35 Plot of differential pore areas for macropores ...................................................... 95
Figure ‎5.36 Adsorption isotherm of 0/15/20 wt% fNT/CA/H2O at ICT = 350 µm and FT =
100±10 µm .............................................................................................................. 97
120±10 µm .............................................................................................................. 97
Figure ‎5.38 Adsorption isotherm of 0.0005/15/20 wt% fNT/CA/H2O at ICT = 400 µm and FT
= 100±10 µm ........................................................................................................... 98
Figure ‎5.39 Adsorption isotherm of 0.005/15/20 wt% fNT/CA/H2O at ICT = 400 µm and FT
= 100±10 µm ........................................................................................................... 98
Figure ‎5.40 Adsorption isotherm of 0.01/15/20 wt% fNT/CA/H2O at ICT = 400 µm and FT =
100±10 µm .............................................................................................................. 99
Figure ‎5.41 t- plot of 0/15/20 wt% fNT/CA/H2O at ICT = 350 µm and FT = 100±10 µm ..... 99
Figure ‎5.42 t- plot of 0/15/20 wt% fNT/CA/H2O at ICT = 400 µm and FT = 120±10 µm ... 100
Figure ‎5.43 t- plot of 0.0005/15/20 wt% fNT/CA/H2O at ICT = 400 µm and FT = 100±10 µm
............................................................................................................................... 100
............................................................................................................................... 101
xi
............................................................................................................................... 101
Figure ‎5.46 Plot showing the effect of fNTs on CA membrane permeation ......................... 103
Figure ‎5.47 Plot showing the effect of fNTs on CA membrane salt retention rates .............. 104
Figure ‎5.48 Representation of the effect of functionalized fNTs addition on the membrane
overall performance at same FT ............................................................................ 104
Figure ‎5.49 Plot showing the effect of different wt% of fNTs on CA membranes surface area
............................................................................................................................... 105
Figure ‎5.50 Representation of the effect of surface area on the membranes permeation rates at
same FT ................................................................................................................. 106
Figure ‎5.51 Representation of the effect of surface area on the membranes salt retentions at
same FT ................................................................................................................. 106
xii
List of Tables
Table ‎2.1 Preparation conditions vs. membrane performance of the top 10 membranes.
Measurements done at 40 bars, Evaporation time (Ev time), Annealing time (An
time), Annealing temperature (T), Permeation rate (P in L/m2h.bar), Retention (R
in %) [25] ................................................................................................................ 13
Table ‎2.2 PWF rates at different CA/silica contents [32] ........................................................ 16
Table ‎2.3 Permeation rates of metal ions using CA/PSF/PVP blend membranes [31] ........... 18
Table ‎2.4 PWF rates of CA/PES blend membranes with different PEG 600 additive
concentrations [34] ................................................................................................. 20
Table ‎2.5 Pore size and density and pure water flux of the CA/ZrO2 UF membranes [37] ... 22
Table ‎2.6 Permeation and rejection rates as a function of CNTs concentration for 10 wt% PA
at 39 bars and 25oC [41] ......................................................................................... 26
Table ‎5.1 Macrovoids numbers and dimensions relative to CA content, non-solvent content,
and membrane thickness ......................................................................................... 66
Table ‎5.2 Macrovoids numbers and dimensions relative to CA content and PI temperature .. 70
Table ‎5.3 Macrovoids numbers and dimensions relative to CA content ................................. 73
Table ‎5.4 Macrovoids numbers and dimensions relative to solvent evaporation .................... 75
Table ‎5.5 Macrovoids numbers and dimensions relative to non-solvent addition .................. 81
Table ‎5.6 Macrovoids numbers and dimensions relative to fNTs addition ............................. 90
Table ‎5.7 Summary of surface area measurements, and micropores volumes ...................... 102
Table ‎5.8 Effect of fNTs wt% on permeation, salt retention, and surface area, rate
measurements are for 15/20 wt% CA/H2O membranes using 1000 ppm NaCl
solution at 24 bars and room temperature ............................................................. 107
Table ‎5.9 Calculated change in permeation rates and salt retention rates at same membrane
final thickness........................................................................................................ 108
initial casting thickness ......................................................................................... 108
xiii
List of Abbreviations
Abbreviation Meaning
BSA Bovine Serum Albumin
CA Cellulose Acetate
CBT Coagulation Bath Temperature
CNTs Carbon Nanotubes
DMAc N,N-Dimethylacetamide
DMF N,N-Dimethylformamide
ED Electrodialysis
fNTs Functionalized Carbon Nanotubes
FT Final Thickness
FTIR Fourier Transform Infrared Spectroscopy
GA Glutaraldehyde
GBA Ɣ-Butyrolactone
ICIC Isocyanate and Isophthaloyl Chloride
ICT Initial Casting Thickness
IPA Isopropanol
ISE Initial Solvent Evaporation
MD Molecular Dynamics
MED Multiple-Effect Distillation
MF Microfiltration
MSF Multi-Stage Flash
MWCNTs Multi-Walled Carbon Nanotubes
NF Nanofiltration
NMP 1-Methyl-2-Pyrrolidone
xiv
Abbreviation Meaning
NMR Nuclear Magnetic Resonance
NOCC N,O-Carboxymethyl Chitosan
PA Aromatic Polyamide
PEG Polyethylene Glycol
PEO Polyethylene Oxide
PES Polyethersulfone
PI Phase Inversion
pNTs Pristine (Non-Functionalized) Carbon Nanotubes
PSF Polysulfone
PVP Polyvinylpyrrolidone
PWF Pure Water Flux
RO Reverse Osmosis
SEM Scanning Electron Microscope
SLS Sodium Lauryl Sulphate
SWCNTs Single-Walled Carbon Nanotubes
TFC Thin Film Composite
UF Ultrafiltration
v/v Volume by Volume
VCD Vapor Compression Distillation
wt% Weight Percent
xv
1 Introduction
1
1.1 History
Water is the basic ingredient in life, and man knew this since dawn age. He looked for
sources of drinkable water and settled right next to them. As history tells, the search of good
tasting water was recorded on the tombs of the Pharaohs [1]. In the ancient Greek books,
sailors used the classical distillation method of boiling seawater and using wool condensation
to get fresh one [2]. But it was not until the early 17th century that water desalination
laboratory experiments were suggested in the hope of removing salts from saline water (sea
and ocean sources) and converting them into drinkable water using salt filters. Then in 1685,
the 1st recorded experiment was carried out in which the Italian physician Lucas Portius used
multiple sand filters with different grain sizes distributed along the filter system in order to
produce the best tasting water [1]. As time passed by, man depended on either sand filters or
distillation methods to purify water. The alternative was carrying large amounts of freshly
stored water in tanks during trips. It wasn’t until WWII that the need of ‘light’ portable
desalination filters was the focus of governments to aid military troops in different arid
regions. Later in the mid 20th century, different countries established research and
development programs funded by their governments to improve water desalination
technologies to produce drinkable and irrigation water at the lowest possible costs [3-4].
Since then, thousands of research investigations focusing on water treatment were carried out
worldwide and large numbers of desalination plants were established to reach 14,451 plants
producing 59.9 million cubic meters per day in 2009 with an annual growth rate of ~12%
each year according to Lisa Henthorne, the president of the International Desalination
Association [5].
However, this is still not enough especially that natural fresh water is nothing permanent
and is possibly depleting due to the enormous annual population increase, extensive water
consumption, and water sources pollution [6]. Locally, Egypt is expected to face a grave
water scarcity problem in less than 12 years from today according to the Water Research
Center of Egypt where 60% of the farmers won’t be able to retrieve water for irrigation due
to the shortage of fresh water supplies [7]. In 2010, the UN announced that Egypt is below
the water poverty line with only 1000 m3 of water per person per year, while the world
average is 7000 m3 of water per person per year [8], even though the country has ~2,500 km
worth of coastline [9], as well as 35,000 km of brackish water canals [10]. Accordingly, with
sufficient number of desalination plants established, the government could meet the large
water demands especially for coastal cities, and irrigation areas.
1.2 Desalination Technologies
Desalination technologies are concerned with finding alternative sources for drinkable
and irrigation water, as well as water used for industrial purposes, other than the ground and
the underground fresh water. Sources usually include seawater, brackish water, and
wastewater. Figure 1.1 shows the shares of different saline water sources used in desalination
industries [11]. Typical seawater has salt content ranging between 30,000 to 50,000 ppm, but
on average it contains 35,000 ppm. Brackish water has 500 to 30,000 ppm salinity, while
fresh water contains less than 500 ppm salinity, yet, saline content in drinkable water
2
shouldn’t exceed 250 ppm [12]. As for wastewater, its desalination and purification for
industrial purposes and even to be used as a drinkable water source is a relatively new
concept. Yet indeed, it can’t be denied that it is an excellent recycling technology since
wastewater is usually discharged without further use [13].
Figure ‎1.1 Desalination saline sources % worldwide [11]
To date, among the many types of desalination processes, only few were converted into
large industries of fresh water production because commercializing a desalination process is
limited by the amount of energy required for production, the price of cubic meter of
freshwater produced by different processes, and environmental impact especially for brine
discharge [11, 13]. Figure 1.2 shows the different shares of each process globally [11], where
reverse osmosis (RO), multi-stage flash (MSF), and multiple-effect distillation (MED) take
the lead. Other desalination processes include electrodialysis (ED), vapor compression
distillation (VCD), and micro, ultra, and nanofiltration (MF, UF, NF respectively). In the next
sections, a briefing about all mentioned processes is given.
Figure ‎1.2 Desalination processes % worldwide [11]
1.2.1 Multi-Stage Flash (MSF)
MSF, a thermal desalination process, was first introduced in the early 1960’s, and then it
conquered the desalination market in 1980’s and 1990’s because it proved to be a simple
3
reliable process [11, 13-14]. Today, MSF represents more than 93% of all the thermal
desalination processes’ share [15] and 25% of the global desalination processes’ share [11].
MSF is a simple evaporation and condensation process, where the saline feedwater is first
preheated by passing over a series of closed pipes. It is there heated in a brine heater. Next,
the hot saline enters a series of flashing chambers (called vessels or stages) of progressively
low pressures. The lower pressure of each stage compared to the one before it results in an
immediate boiling (flashing) of the heated saline and evaporation of the fresh water. The
steam is collected and directed inside the closed pipes in which heat exchange takes place
between the vapor and the feedwater (that becomes preheated in the first step), while the
steam condensates. The distillate is gathered by collector trays, and directed as an output of
fresh water. The remaining feedwater (now becoming brine) is partially disposed, and
partially recirculated into the stages to recover more fresh water. Finally, the remaining brine
now with 15-20% more salt concentration than the initial saline feedwater is cooled and then
discharged.
Figure ‎1.3 Diagram of Multi Stage Flash Process [13]
It is worth mentioning that in each stage, only small percentages of saline feedwater
vaporize, due to the limited pressure drop from one stage to the next. That is why in order to
retrieve large percentage of fresh water, up to 40 stages would be used, with an average of 18
to 25 stages in regular MSF plants [11].
The main advantages of MSF are simplicity, reliability, irrelevance of the saline
feedwater concentrations or existence of suspended particle, and very low salt concentration
in the retrieved fresh water. However the main disadvantage of this process is its electrical
and thermal energy requirements that make it an expensive technology [11, 13-14].
1.2.2 Multiple-Effect Distillation (MED)
MED, also a thermal desalination process was first introduced in the mid 19th century to
be classified as the oldest industrial desalination technique. Yet, after the introduction of
MSF and RO, the demand for MED significantly decreased over the years and represents
only 8% of today’s world desalination industrial share [11, 14].
The process, similarly to MSF, depends on evaporation and condensation at progressively

lower pressures, and temperatures within 8 to 16 consecutive effects. Yet, the structure of
4
MED setup is different. The saline feedwater is first preheated, and then introduced into the
first effect where it passes over a series of tubes filled with hot steam, and through heat
exchange, the feedwater is heated until boiling, while the steam inside the tubes condensates.
The resultant steam is introduced into the second effect condensation tubes, while the
remaining feedwater is introduced outside the tubes inside the effect at lower pressure and
temperature. Then, the process of evaporation and condensation continues inside different
successive effects. The final steam is directed into the primary condenser where preheating of
the feedwater takes place. As for the brine formed, it is cooled and discharged [11, 13-14].
Figure ‎1.4 Diagram of Multiple Effect Distillation Process [13]
The advantage of MED over MSF is that it requires lower thermal and electrical energy,
making it more environmentally friendly since the lower energy required means lower fuel
used and lower exhaust. However, due to its more complex setup, MED faces corrosion and
scaling inside the condensation tubes that are difficult to remove, which is a major
disadvantage.
1.2.3 Reverse Osmosis (RO)
RO, a membrane desalination process, is the global leading technology in fresh water
production by desalting with a 53% share, and its use is rapidly increasing due to the large
ongoing research and development in the membrane technology [11, 16]. This process
depends on reversing the osmotic flow through a semipermeable membrane by applying
pressure as indicated in Figure 1.5.
5
Figure ‎1.5 Osmosis and Reverse Osmosis flow [17]
The nominal pore size of RO membranes is 1 to 10 Å, which is capable of filtering out

colloidal particles, bacteria, viruses, and ions including monovalent ones, while allowing only
pure water to pass through [17]. In practice, the saline water is introduced into a pressurized
sealed vessel, and pumped against the semipermeable membrane, which is capable of
filtering up to 70% of the salts with a target final salt concentration of 250 ppm or less from a
single run [11, 14]. The required pressure depends on the degree of water salinity. For
example, seawater desalination requires 56 to 105 bars, while brackish water desalination
requires 14 to 42 bars [18]. The pressure values, (thus the amount of energy consumed),
could be reduced by developing membranes with nanopores, large enough to allow the
passage of large volumes of water, but small enough to block the passage of monovalent ions
[16]. A good example of low pressure developed membranes was created by Loeb and
Sourirajan in 1963 using cellulose acetate, which is an asymmetric porous membrane with
high permeation rates and good salt rejection. Another material used in RO membranes is
polysulfone mixed with aromatic polyamides to form the so called thin film composite
membrane [14].
The major disadvantage of RO process is its membrane fouling which is a key problem,
requiring good pretreatment processes to minimize its effects. Fouling happens due to the
precipitation of a foulant on, in, and/or near the membrane surface, blocking the nanopores,
reducing the water flux, possibly increasing salt passage, significantly increasing the
operational pressure, and consequently the energy involved in the process [19]. There are 4
types of fouling [14, 19]:
1. Inorganic fouling: (also called scaling) it is the precipitation of dissolved salts from
iron, aluminum, calcium, and sulfates under the concentration polarization effect.
That is the accumulation of salts in high concentrations at the membrane boundary
compared to their concentration in the remaining feedwater.
2. Colloidal fouling: it is the precipitation of colloidal solids of 5µm or little less, such as
clays or silica, resulting in the formation of a secondary layer, resistant to water flow.
3. Organic fouling: it is the development of a thin film of organic compounds like
proteins, polysaccharides, polyphenolic molecules, or hydrocarbons on the membrane.
6
4. Biofouling: it is the development of a biofilm on the membrane surface from bacterial
colonies, fungi, or algae that are able to reduce water flux, increase transmembrane
pressure, and/or degrade the membrane either by enzymatic biochemical degradation
or by creating a local pH at the membrane surface, resulting in membrane hydrolysis.
Fouling pretreatment plans are usually prepared according to the initial constituents of the
feedwater, i.e. they are directly related to the water source [19]. For example, feedwater from
sources located near sea ports need to consider organic fouling from types of organic
compounds that don’t normally exist with such high concentrations in seawater (ships’ fuel
wastes).
1.2.4 Other Processes
Electrodialysis (ED): it is an electrical desalination process in which salt ions from saline
water are transported toward two membranes (a cation and an anion) by applying a potential
difference between them, thus the saline water becomes fresh. It is a very useful technique for
desalting brackish low saline concentration water [11, 13].
Vapor compression distillation (VCD): it is a mechanical desalination process that resembles

MSF and MED in using steam to convert the feedwater into vapors, yet it uses mechanical
compressors to condensate the distillate into fresh water instead of heat exchange. This
technique is utilized in small scale desalination industries [13-14].
Microfiltration, Ultrafiltration, and Nanofiltration (MF, UF, NF): those are semipermeable
membranes, used mainly as pretreatment steps in desalination plants. Figure 1.3 shows the
pore size distribution of each membrane type [11, 14, 17].
Figure ‎1.6 Pore size distribution of each membrane type [14]
Microfiltration (MF) membranes have the largest pore diameter, thus they could be used in
removal of large suspended particles, some microorganisms like large bacteria and algae, and
reduction of turbidity. MF membranes materials include poly (vinylidene fluoride),
polysulfone, poly (acrylonitrile), poly (acrylonitrile)- poly (vinyl chloride) copolymers,
cellulose acetate-cellulose nitrate blends, nylons, and poly (tetrafluoroethylene) [14, 17].
Ultrafiltration (UF) membranes come second in pore size diameters where it can reject large
colloidal particles, bacteria, some viruses, organic compounds of high molecular weight, and
7
macromolecules such as proteins. UF membranes are made of some of the same materials of
MF membranes such as poly (vinylidene fluoride), polysulfone, and poly (acrylonitrile), yet
under different preparation conditions to have smaller pores. UF membranes are also
manufactured from poly (ether sulfone), which is the most common material used [14, 17].
Nanofiltration (NF) membranes come third in pore size diameters between UF and RO
membranes. They are used in the removal of everything that MF and UF membranes can
remove, in addition to organic compounds, multi-valent ions, and viral particles. NF
membranes are manufactured from materials similar to RO and UF membranes, using
different preparation conditions. These include cellulose acetate blends, polyamide
composites, or sulfonated polysulfone [14, 17].
8
2 Literature Background
9
2.1 Introduction
For membrane research, different investigations targeted different directions for

membrane development, however, the final aim is the same: developing a membrane that at
the lowest possible pressures gives high water permeation rates and high salt retention, while
being resistant to different fouling factors with long durability. Such a difficult to reach
combination of requirements opened lots of areas of variation in membranes’ preparation
conditions. Typically, work focused on material types, concentrations, solvent and non-
solvent types, presence of additives, preparation temperature, and/or a blend of different
materials to achieve the requirements.
The first ever reported RO membrane with effective salt retention (98%) was
manufactured in 1959 by Reid and Breton. It was a hand cast symmetrical thin film of
cellulose acetate (CA) with a water permeation rate less than 0.01 L/m2h. In less than 5 years,
Loeb and Sourirajan developed the famous CA asymmetric membrane with a little higher salt
retention (99%) than Reid’s, and much higher water permeation rate of 14.6 L/m2h to become
the first high flux asymmetric RO membrane [20]. Then, a new introduction of materials took
place in 1970’s when Cadotte et al. prepared a multiple layered membrane using in situ
polymerization of branched polyethyleneimine and 2,4-diisocyanate to form a polyamide
deposited on the surface of a porous polysulfone membrane. They called it thin film
composite (TFC), and when tested, it gave water flux of 3.5 L/m2h and salt retention of 94.5
% [19]. Since then, RO membranes were either manufactured from CA or TFC while varying
preparation conditions and additives to get better performance.
The literature background will address CA membranes in filtration applications, as well

as the usage of carbon nanotubes in water desalination since these are the materials used in
the current study.
2.2 Effect of Different Additives
Additives refer to the addition of an extra material (a second solvent, a non-solvent, a

nanofiller, or a surface modifier) during the membrane preparation with the aim of enhancing
permeation rates and salt retention. However, permeation rates and retention are inversely
proportional to one another since increasing pore sizes across the membrane to enhance the
flux, usually leads to allowing more solutes to pass through. Thus, experiments target the
highest possible salt rejection at acceptable rates. Achieving this is controlled by the
concentration and type of the additive, which directly affect the macrovoids sizes and
location across the membrane. This also affects the porosity of the top dense layer, changing
the membrane type from ultra to nano or to RO filter membranes.
2.2.1 Effect of Solvents and Non-Solvents
Experimentally, Li et al. [21], who used different concentrations of ɣ-butyrolactone

(GBA) (secondary solvent) added to 11 wt% CA dissolved in 1-methyl-2-pyrrolidone (NMP)
to convert an UF CA membrane to a NF one, and to check how this affects its pure water flux
10
(PWF) and salt retention. The membranes were prepared by phase inversion (PI) process at
room temperature, and water as non-solvent in the coagulation bath. When GBA
concentration was 50 wt%, PWF was as high as ~150 L/m2h at 10 bars, yet, with 0% salt
retention. Once GBA concentration increased to 80%, salt retention increased (though by
only 10%), which means that the membrane became a NF, yet a poor one. At the same time,
PWF decreased dramatically to only 0.2 L/m2h. The paper concluded that by using good
solvent mixtures, membrane porosity could be modified to serve a designated purpose.
Ye et al. [22] went further in trying to mix different solvents, where they compared
membranes formed from 20 wt% CA at 20oC and dissolved in one solvent with CA
membranes dissolved in a mixture of two solvents. Both sets of membranes were prepared by
PI at room temperature and water as non-solvent in the coagulation bath. Then they compared
the two sets to CA membranes formed from a mixture of two solvents and a non-solvent, also
prepared with PI at room temperature, and water as non-solvent in the coagulation bath.
Membranes formed from CA dissolved in only N,N-dimethylformamide (DMF) gave a
permeation rate of 7.5 L/m2h. On adding 40 wt% acetone (secondary solvent), the rate
decreased to 1.7 L/m2h. Further addition of isopropanol (IPA) non-solvent to give 30/20 wt%
of acetone/IPA respectively increased PWF to 15.8 L/m2h, which is almost double the initial
rate. All permeation tests were carried out at the same temperature and pressure. The results
indicate that the presence of a non-solvent could enhance water permeation rates, more than
simply mixing two solvents together. It is worth mentioning though, that the paper focused on
how these membranes could be used in blood purification for the removal of macromolecules
such as Methyl orange, Cytochrome C, Albumin, and ɣ-Globulin. This means that the
manufactured membranes can’t be used in desalination since salt ions’ sizes are much smaller
than the mentioned molecules.
Haddad et al. [23] succeeded in creating nanofilter membranes from 20 and 22 wt% CA
using a mixture of acetone solvent and formamide non-solvent with a ratio of 2:1. This was
made possible via manipulation of the preparation conditions, where in the PI step, the
gelation medium was distilled water at a temperature of 4oC, and the casted solution
remained in the medium for 1 hour. Then the membranes were immersed individually in
annealing water baths at temperature ranges from 60 to 80oC for 10 minutes to study the
effect of annealing temperature on membrane performance. The final membranes’
thicknesses were 70-90 µm. PWF measurements were done at different pressure ranges (4-16
bars), and the study showed that PWF is directly proportional to pressure increase which
made perfect sense, and inversely proportional to temperature increase, which was expected
because the idea behind annealing is reduction of pore sizes by creating denser membranes
under the effect of temperature. This was further supported by scanning electron microscope
(SEM) images. Salt rejection and permeation rates were tested using a 2000 ppm NaCl
solution at transmembrane pressure of 16 bars. At the same salt rejection rates, membranes
with less CA content (20 wt%) showed higher flux rates. On comparing flux rates with
rejection rates, the results showed that the higher the flux rates the better the rejection rates,
which is a trend supported in the literature [24]. Finally, on comparing rejection rates to
annealing temperature, as expected, higher temperature of 75 and 80oC resulted in higher
11
rejection rates, the highest of which was ~87% with permeation rate of ~10 L/m2h for the 20
wt% CA, and the best conditions were found to be for membranes with 20 wt% CA, annealed
at 75oC to have permeation rate of ~75 L/m2h and rejection rate of ~85%. Figure 2.1 displays
the results found.
Figure ‎2.1 Flux rates vs. Rejection rates of 2000 ppm NaCl solution for temperature ranges of
60,65, 70, 75, and 80oC, (a) data of 20 wt% cellulose acetate membranes, (b) data of 22 wt%
cellulose acetate membranes [23]
Odena et al. [25] also managed to create nanofilter CA membranes by studying the effect
of a very wide variation of preparation conditions. CA wt% ranged from 12 to 21 wt%, a
mixture of acetone and dioxane solvents were used, and further mixed with methanol non-
solvent. PI coagulation medium was distilled water kept at constant temperature of 4oC, while
the annealing temperature ranged from 65 to 85oC. The initial solvent evaporation (ISE)
before gelation was tried for 30, 60, 90, or 120 seconds, and the membranes’ performances
were tested with 5000 ppm NaCl solution at operational pressure of 40 bars, at room
temperature. Of all these conditions, the highest salt rejection rates were for two membranes,
the first with 19 wt% CA, 7.5 wt% methanol prepared with ISE of 60 seconds, and annealing
temperature of 75oC for 2 minutes, and gave rejection rate of 82.55% and permeation rate of
48 L/m2h at 40 bars. The second was prepared with 21 wt% CA, 12 wt% methanol, with ISE
of 60 seconds, and annealing temperature of 75oC for 6 minutes, and it gave a salt rejection
rate of 79.9% and permeation rate of only 2.4 L/m2h at 40 bars. This significant drop in
permeation rate from membrane 1 to 2 could be explained by the increase in both CA wt%
and annealing time, each of which contribute to the formation of denser membranes. On the
other hand, the increase in non-solvent content was expected to increase the macrovoids and
thus the permeation rate, but it seems that the effect of densing membrane 2 over-ruled the
effect of increasing macrovoids numbers or size via the non-solvent. As for the rest of the
membranes prepared, their performance was tested using ibuprofen, a small micro-pollutant
used in drugs manufacturing that is commonly found in drinking water. The best ten
membranes that gave high permeation and retention rates are displayed in Table 2.1.
12
Table ‎2.1 Preparation conditions vs. membrane performance of the top 10 membranes.
Measurements done at 40 bars, Evaporation time (Ev time), Annealing time (An time),
Annealing temperature (T), Permeation rate (P in L/m2h.bar), Retention (R in %) [25]
According to the results, the highest rejection rate (90.35%) is found for the 20 wt% CA
membrane, since it had more polymer content than the rest. As for the highest permeation
rate (52 L/m2h at 40 bars) is for the 19 wt% CA with the shortest ISE and annealing periods.
The data above is a clear example of how manipulation of conditions can give the same
results though the polymer contents are different.
2.2.2 Effect of Pore Formers
Lv et al. [26] thought of adding to CA a copolymer that acts as a pore former, and see
how this would affect the membrane behavior. A pore former is a term used to describe an
additive, solid or liquid, organic or inorganic, that is more polar and less volatile than the
solvent. At the same time, a pore former is soluble/miscible in the non-solvent used in PI
such that when the main polymer matrix starts gelation, the pore former leachability creates
macrovoids that enhances permeation rates [27]. Putting this idea in practice, membranes of
16 wt% of CA mixed with different concentrations of Pluronic F127, dissolved in DMF were
tested for PWF and rejection rates of bovine serum albumin (BSA), since they were expected
to be ultrafiltration membranes. Operational pressure was 1 bar and temperature was 25±1oC.
Prior to the tests, membranes were subjected first to 1.5 bars for 30 minutes to overcome the
compaction effect of pressure. At 0 wt% Pluronic F127, the blank CA membrane gave a very
low PWF rate of only 3 L/m2h, which increased with the increase of Pluronic F127 content to
give a PWF of 93.24 L/m2h at 20 wt%. BSA rejection tests showed that using the additive,
the membrane is able to reject higher BSA contents compared to blank membranes.
Interpreting membranes morphology using SEM at different Pluronic F127 wt% showed that
the more additive used, the more porous the membrane becomes, which justifies the PWF
increase by an order of magnitude from the blank to the CA membranes prepared with the
additive. Yet, the rejection rates didn’t show a linear increase with Pluronic F127 wt%
increase, instead, the rate increased to a maximum of 77% at additive content of 8 wt%, then
it started decreasing again. A suggesting interpretation to this is that addition of small
contents of Pluronic F127 decreased the pore size on the membrane surface (while increasing
the size of the macrovoids) until a flipping point at which further addition would only
increase the pore sizes. Figure 2.2 shows the membrane performance as a function of
Pluronic F127 wt%.
13
Figure ‎2.2 Pure water flux and BSA rejection rates vs. Pluronic F127 content [26]
Saljoughi et al. [28] tried using a different pore forming agent, polyethylene glycol 600
(PEG polymer), on CA membranes. Small different concentrations were added to 15.5 wt%
CA dissolved in NMP at 23oC and transmembrane pressure of 0.2 bars. Initially, with 0 wt%
PEG 600, the PWF rate was 7.75 L/m2h. By adding 5 wt% of PEG 600, the PWF increased
more than 11 times the initial rate to give 90 L/m2h. Further increase of PEG 600 (10 wt%)
increased the rate again to give 200 L/m2h. This shows that the presence of a pore former
could enhance the formation of macrovoids across the membrane, which in turn enhances the
PWF rates. It is worth mentioning, though, that when Lv et al. [26] tried using 4 and 8 wt%
of PEG 2000 instead of PEG 600 with a 16 wt% CA, the membranes didn’t allow permeation
of water since it interacted with the CA too strongly to leach out during PI. Thus, the
molecular weight of the additive is important. As for rejection rates, Saljoughi et al. didn’t
mention anything about the molecular weight cut off of the developed membranes. They only
suggested that the membranes could be used in ultrafiltration applications.
Idris et al. [29] used a lower molecular weight PEG 400 as a pore former, and added to it
distilled water as a swelling agent to have narrow pores to maximize rejection rates. Different
contents of PEG 400, water, and CA were dissolved in acetic acid, and permeation
experiments for 2000 ppm BSA solution were carried out at 25 oC and transmembrane
pressure of 2 bars. PWF was not measured though. The highest permeation rate of the 2000
ppm BSA solution was 56.67 L/m2h and was found to be for the membrane with the lowest
CA and water contents of 15 and 5 wt% respectively, and 10 wt% PEG. However, on
increasing CA and water contents to 20 and 10 wt% respectively, permeation rate dropped
more than half its value to be 25.24 L/m2h. On the other hand, this decrease was accompanied
by an increase in BSA rejection rate from 90.19 to 96.19%. The team tried one more contents
combination of 20/5/5 wt% for CA/PEG/H2O respectively. This resulted in the lowest
permeation rate of 17.54 L/m2h and rejection rate of 42.54% for the 2000 ppm BSA solution.
A possible explanation to these values was at lower PEG and water contents, there is more of
the acetic acid solvent that reduced the polymer packing density, creating larger pores, but
not macrovoids. Thus, rejection rates decreased since the pores are too large to prevent the
14
protein from passing through, yet, they are too small to allow high flux rates. On increasing
PEG and water to 10 wt% each at 20 wt% CA, more macrovoids were created to increase the
permeation rate, and at the same time, the packing density of the polymer increased to
maximize the protein rejection rate. Finally, decreasing only the swelling agent (water to 5
wt%) and CA (15 wt%) gave the maximum earlier mentioned permeation rate, since the pore
forming agent PEG created the necessary macrovoids, and at the same time, less polymer and
less swelling agent existed.
Saljoughi et al. [30] studied the effect of using a different pore former
polyvinylpyrrolidone (PVP). They used different concentrations added to 15.5 wt% CA
dissolved in NMP, and prepared the membranes using PI at 25oC. Permeation tests were
carried out at room temperature and transmembrane pressure of 0.35 bars. They found that
the addition of 3 wt% PVP increased macrovoids formation, consequently PWF reached 65
L/m2h vs. only 18 L/m2h for CA membrane prepared without PVP at the mentioned
temperature and pressure. Further increase in PVP concentration (6 wt%) led to decrease in
PWF to give a rate of 39 L/m2h. The team concluded that increasing the second polymer
concentration doesn’t necessarily mean increasing water flux rates, since there could be a
flipping point at which too much of the polymers suppress the macrovoids formation, instead
of enhancing it. Yet, the team failed to mention how the addition of PVP affects salt or
protein rejection.
Sivakumar et al. [31] also studied the effect of adding different concentrations of PVP to
17.5 wt% CA dissolved in DMF on PWF, macrovoids formation, and heavy metals filtration.
Membranes were prepared using PI at 10oC in a coagulation bath filled with water, DMF, and
sodium lauryl sulphate (SLS) surfactant, used in reducing the surface tension at the
developing membrane non-solvent interface. Tests were carried out at transmembrane
pressure of 4.14 bars and room temperature. An increase in PWF from 14.1 to 78.8 L/m2h
with the increase of PVP weight content from 0% to 7.5% was observed. Comparing such
result with Saljoughi et al. [30] consequently shows that the solvent type and its interaction
with PVP is playing a role in enhancing/suppressing macrovoids formation, thus PWF rates,
since Sivakumar et al. used higher PVP and CA wt% and still got higher PWF rates.
Arthanareeswaran et al. [32] used an inorganic pore former added to 17.5 wt% CA
membranes to test their performance. Tetraethyl orthosilicate (aka silica) nanoparticles with
different concentrations (0-40%) were mixed with CA dissolved in DMF, to have a fixed
CA/silica concentration of 17.5 wt%. PI was carried out after an ISE for 30 seconds. Gelation
medium was a mixture of water (non-solvent), DMF (solvent), and SLS surfactant. PI lasted
for 30 minutes to ensure a complete removal of the solvent and the pore former from the
membrane, and membranes’ final thicknesses were 200±20 µm. Flux rates were measured at
operational pressure of 3.45 bars. PWF rate for 0% silica membranes was 15.58 L/m2h that
increased to 46.74 L/m2h for the 60/40 % CA/silica content as shown in Table 2.2.
15
Table ‎2.2 PWF rates at different CA/silica contents [32]
According to tabulated results, the increase in silica content had a direct effect on PWF
that increased 3 times from 0% silica to 40% silica content. This was directly reflected in the
membrane performance in the proteins’ permeation and rejection rates. BSA, Trypsin,
Pepsin, and albumen permeation increased with the increase in silica contents. For example,
BSA permeation rate increased from ~5.2 to ~19.7 L/m2h. However, all proteins’ rejection
rates decreased with silica increase. For example, BSA rejection decreased from ~94 to ~81%
for the 60/40% CA/silica contents, indicating larger macrovoids formation with silica
contents increase.
Chen et al. [33] also used different concentrations of silica nanoparticles (0-30 wt%)
added to 16 wt% CA dissolved in DMF to test the membranes’ performance. PI gelation
medium varied from pure water to acidic medium (pH 1) and basic medium (pH 13) to try to
understand the effect of different pHs on the efficiency of the nanoparticles leachability from
the polymer matrix, as well as on the membranes’ permeability and rejection rates. In both
the acidic and basic mediums, SiO2 nanoparticles were unable to effectively leave the
polymer matrix. This was characterized by Fourier transform infrared (FTIR) spectroscopy
that gave clear absorption peaks for Si-OH and Si-O-Si bonds, the hydrolyzed and poly-
condensated forms of the silica nanoparticles, after being catalyzed by HCl or NaOH in both
mediums. Further analysis using transmission electron microscopy (TEM) proved the
existence of circular shaped beads imbedded into the matrix. Energy dispersive X-ray (EDX)
spectroscopy mapping for Si element showed a homogenous distribution of the nanoparticles
throughout the membrane. This could be due to the cross-linkage formed between the
nanoparticles and the polymeric chains of CA. In pure water medium, the story was different,
where silica nanoparticles totally leached out of the CA membranes, and no absorption peaks
could be seen. Flux rates were studied for all membranes at operational pressure of 1 bar,
after initially leaving the membrane for 30 minutes at 1.5 bars to eliminate compaction effect.
For blank (0 wt% silica) CA membranes, PWF was very low (approximately 1.7 L/m2h),
which didn’t change regardless of the gelation medium type. The addition of silica enhanced
PWF significantly, especially in the acidic medium, where 30 wt% silica in pH 1 gave PWF
of 436.6 L/m2h, indicating drastic increase in the number and size of pores as shown in
Figure 2.3.
16
Figure ‎2.3 (a) SEM of blank cellulose acetate membrane in pure water gelation medium, (b)
SEM of cellulose acetate membrane with 30 wt% silica nanoparticles prepared in HCl gelation
medium (pH 1) [33]
With regards to rejection rates, the membranes were tested using BSA solution. The
rejection rates decreased from 49% for pure CA membranes to 10% and 13% for CA-30 wt%
SiO2 in HCl and CA-30 wt% SiO2 in NaOH gelation medium respectively. The membranes
were also tested for oil separation in a 900 ppm oil/water emulation solution. All membranes
gave the same rejection rate of 99.8% regardless of the presence or absence of silica
nanoparticles, or gelation medium type. Permeation rates for BSA solutions were not
investigated. However, oil/water flux rates were studied. The pure water flux rates over the
first 30 minutes were as high as ~400 L/m2h that decreased to ~100 L/m2h on the introduction
of the oil/water emulsion. When the membrane was washed to be reused for the second, third,
and fourth times, it gave the same pattern of flux change regardless of the gelation medium
and silica concentration for all CA-silica membranes (except for 0 wt% silica). This indicated
that the CA-silica membranes have antifouling effect for organic compounds, i.e. the oil
droplets adsorbed and desorbed easily on and from the membrane surface and the membrane
could be easily washed with pure water to be reused for further cycles.
2.2.3 Effect of Using Polymer Blends
Attempting to enhance the membrane performance furthermore, Sivakumar et al. [31]

added polysulfone (PSF), a polymer also used in manufacturing RO membranes, to the CA-
PVP blend, such that the final CA/PSF content is always 17.5 wt% of the total stock weight.
At 0 wt% PVP, the CA/PSF blend with composition 75/25 wt% respectively gave PWF of
48.1 L/m2h, which was about 3 times more than what blank CA membranes gave.
Incorporating the three polymers together with compositions 7.5 wt% PVP and CA/PSF of
75/25 wt% gave PWF of 115 L/m2h, more than twice the PWF of CA-PVP membranes, and
more than 8 times of what blank CA membranes gave. The research addressed the effect of
enhancing water permeation on salt rejection rates. Table 2.3 shows the significant increase
in permeation rates for individual 1000 ppm saline solutions of each metal ion with the
increase in PVP content at constant CA/PSF composition.
17
Table ‎2.3 Permeation rates of metal ions using CA/PSF/PVP blend membranes [31]
However, the existence of large voids created by the addition of PVP reduced the
rejection rates remarkably as illustrated in Figure 2.4.
Figure ‎2.4 Effect of PVP wt% on metal ions rejection: (a) blank cellulose acetate, (b) 75/25 wt%
CA/PSF [31]
A plausible explanation to the high permeation rates was that the incompatibility of
cellulose acetate and polysulfone created repulsive forces between the polymers, which led to
a macrophase separation during PI. This was further enhanced by the rapid percolation of
PVP that speeded up the solvent/non-solvent demixing process, resulting in large numbers of
macrovoids. Through visual inspection using SEM, it was clear that the more PVP wt%
added, the larger the pore size especially the long finger like voids, for both the CA
membranes and the CA/PSF blend membranes as shown in Figure 2.5. On the other hand,
metal ions retention rates decreased significantly with both the increase in PVP and PSF
contents since the pose sizes increased.
18
Figure ‎2.5 SEM of the effect of PVP on morphology; (a) 2.5 wt% PVP in 100/0 wt% CA/PSF
respectively, (b) 7.5 wt% PVP in 100/0 wt% CA/PSF respectively, (c) 2.5 wt% PVP in 75/15
wt% CA/PSF respectively, (d) 7.5 wt% PVP in 75/15 wt% CA/PSF respectively [31]
Mahendran et al. [34] tried mixing CA with polyethersulfone (PES) to have a final
polymer concentration of 17.5 wt% dissolved in DMF. They also used PEG 600 as an
additive with different concentrations to study the resultant membrane performance. PI
gelation medium contained distilled water as a non-solvent, 2.5 % (v/v) DMF that reduced
the rate of the demixing process to reduce macrovoids volume, as well as 0.2 wt% SLS
surfactant. Permeation rates were carried out at 3.45 bars transmembrane pressure, after
subjecting the membrane to an initial transmembrane pressure of 4.14 bars for 4-5 hours until
a steady flow rate is attained to eliminate any measurements errors resulting from membrane
compaction. Trying different combinations for the concentration of the two polymers and the
additive showed that 75/25 wt% CA/PES with 10 wt% PEG 600 gave the highest PWF rate
of 275.7 L/m2h when compared to blank CA membranes that gave only 12.9 L/m2h as shown
in Table 2.4. Trials of adding more than 10 wt% of PEG 600 to the blend failed since the
blend solution became highly incompatible with the additive.
19
Table ‎2.4 PWF rates of CA/PES blend membranes with different PEG 600 additive
concentrations [34]
Such high PWF rates were suggested to be primarily due to the fact that PES is more
hydrophilic than CA. Thus in PI, PES water’s affinity accelerated the solvent/non-solvent
demixing and created macrovoids large enough to allow water permeation with such high
rates. The research suggested that the pore former PEG 600 solubility in water had no effect
on the PI process, and that the low molecular weight PES (in comparison to that of CA) is
mainly responsible for the morphology of the dense top layer of the membrane since PES
precipitates faster than CA. The researchers didn’t address, though, the reasons for using 2%
(v/v) DMF in the gelation medium that was expected to slow down the demixing rate, even
though they used a surfactant and a highly hydrophilic secondary polymer, both of which are
used to increase the rate of the demixing. Furthermore, they didn’t study the performance on
the membrane in nanofiltration, but rather for ultrafiltration applications on proteins with
different molecular weights that ranges from 20 to 60 kDa.
Boricha et al. [24] studied the performance of N,O-Carboxymethyl Chitosan (NOCC)/CA

blend membranes with a compositions of 4/6 NOCC/CA dissolved in acetone. The solution
was casted and left for complete solvent evaporation until a clear transparent membrane was
formed. Since NOCC is water soluble, keeping it within the blend requires that it is cross-
linked with the CA polymer. This was carried out by immersing the membrane in a solution
of glutaraldehyde (GA) at 60oC for 1 hour. The membrane was then left to stand vertically for
GA to drain, and any excess was washed out using deionized water. The membrane’s
performance was tested for permeation and rejection of a solution of 168 ppm and 74 ppm of
chromium and copper respectively at different applied pressures, feed flow rates,
temperatures, and feed pH. The highest permeation rate was found to be ~168 L/m2h, with
ions rejection rates of 83.4% and 72.6% for chromium and copper ions respectively at 10 bars
and feed flow rate of 960 L/h. It was also found that the permeation and rejection rates were
directly proportional to the applied pressure and feed flow rate, thus, decreasing the pressure
below 10 bars gave less cations rejection and solution permeation, which is a trend that was
reported earlier by Haddad et al. [23]. As for the temperature effect on the membrane
performance, it was found that with temperature increase, the permeation rate increased, yet,
rejection rates decreased which is expected since temperature was expected to affect the
membrane morphology and pore sizes. It could also affect the viscosity of the feed solution.
Finally, pH effects were investigated and it was found that with pH increase, solution
20
permeation decreased, which is thought to be due to the shrinkage of the nanofilter membrane
under the influence of pH increase, which is a behavior that was reported for different
nanofilters in the literature [35]. As chromium and copper ions rejection rates, they were not
significantly influenced by pH change, where chromium rejection rate increased with pH
increase, and copper rejection rate decreased with pH increase.
2.2.4 Effect of Inorganic Additives
Inorganic additives include silica nanoparticles as mentioned in an earlier section that act
as pore formers. Another example of inorganic additives is zirconium dioxide that was
studied in filtration applications.
Filho et al. [36] prepared two membranes S1 and S2 with 13 and 11 wt% CA dissolved in
acetic acid and acetone using PI with an ISE of 30 seconds. After having the membranes
ready, ZrO2 particles were developed into the membranes by hydrolysis through immersing
the membranes in a solution of 2.2 wt% Zr(PrO)4-propanol for 6 minutes, and then
immersing them in 0.0009M HNO3 solution, individually. Membranes’ final thicknesses
were 90 and 150 µm, and zirconia contents were 0.59±0.5 and 0.51±0.5 wt% for S1 and S2
respectively. Permeation and rejection rates were carried out for 2.5 ppm phosphate ions
solution at operational pressure of 1.72 bars. S1 and S2 showed flux rates of 22±10 and
78±12.6 L/m2h, and rejection rates of 90% and 75% respectively. A blank membrane of S1
composition (without zirconia) was developed, and when tested for phosphate rejection, it
gave 0% rejection. From this finding, it was assumed that the presence of zirconia is the main
factor in phosphate retention. This was further supported by investigating the presence of
phosphorous adsorbing on zirconia in the membrane using P31 nuclear magnetic resonance
(NMR) technique. The characterization was done for Zr(HPO4)2- versus (HPO4)2-, where both
gave clear singlets due to phosphorous but the singlet of the Zr(HPO4)2- was shifted to the
right due to the presence of the Zr attached to the phosphate group, thus affecting the 31P
resonance energy as shown in Figure 2.6.
Figure ‎2.6 31P NMR of (HPO4)2- vs. Zr(HPO4)2-. (a) singlet of phosphorous of (HPO4)2-, (b)
singlet of phosphorous of Zr(HPO4)2- [36]
Arthanareeswaran et al. [37] incorporated 0 to 7 wt% ZrO2 microparticles in CA

membranes, with a CA/ZrO2 ratio that is equivalent to 17.5 wt% of the total solution weight,
21
dissolved in DMF. PWF experiments were carried out at transmembrane pressure of 3.45
bars. The lowest PWF rate was found to be 15.6 L/m2h for the pure CA membrane. As ZrO2
weight content increased, PWF increased to reach 46.7 L/m2h with the 7 wt%, which
indicates an increase in the membranes pore sizes. This was supported by the pore radius and
pore density values that were calculated using the membranes rejection rates for different
molecular weight proteins like Trypsin, Pepsin, Egg albumin, and BSA, with average solute
radius of 21.5, 28.5, 33, and 45 Å respectively. Table 2.5 is a summary of the CA/ZrO2 ratios,
PWF, pore radii, and pore density.
Table ‎2.5 Pore size and density and pure water flux of the CA/ZrO2 UF membranes [37]
This work did not report the rejection rates of any of the membranes. However, one can
deduce by comparing the pore sizes and average solute radii of the proteins that some
membranes like the CA-4 is less likely to efficiently reject Trypsin or Pepsin. This could also
be supported by the permeation rates of the proteins reported in the investigation, where the
more the wt% of ZrO2, the higher the permeation rates of protein solutions. For example BSA
permeation increased from 5.19 to 19.79 L/m2h from CA-0 to CA-4, which applied for the
rest of the proteins with different values. This increase can also be explained by the increase
in the hydrophilicity of the membrane due to the inorganic additive, which allowed more
BSA solution to pass through.
2.2.5 Effect of Using Carbon Nanotubes
Carbon nanotubes (CNTs) as additives have started to draw researchers’ interest over the
past ten years, especially with all the molecular dynamics simulations that strongly suggest
that water passage through the nanotubes is expected to be exceptionally high as explained in
the previous chapter. CNTs were added to polymers like polysulfones (PSF), aromatic
polyamides, and chitosan in polymer based nanocomposites used in filtration applications,
and there are reports in the literature that suggest changes in the produced membranes
performance in terms of permeation and rejection rates. However, the use of CNTs/CA
nanocomposites was not reported previously in the literature.
Choi et al. [38] prepared multi-walled carbon nanotubes (MWCNTs)/PSF blend

membranes in the same exact manner used to prepared ultra and nano CA filtration
membranes using PI. A known volume of NMP was divided into two beakers. PSF was
completely dissolved in one. At the same time, CNTs were first functionalized by oxidation
purification, and then fully dispersed into the other beaker of NMP in the other via sonication
22
for 80 seconds. Once the two stocks were ready, they were mixed thoroughly to have a final
composition of CNTs-PSF/NMP of 15/85. As for the CNTs contents, they were varied from 0
to 4 wt%. Then the solution was cast at thickness of 150 µm and inserted in a coagulation
bath of distilled water at room temperature for 24 hours to have a complete PI. Membrane
performance was evaluated by measuring the PWF rates, and permeability and rejection rates
for different solutions of 1000 ppm PVP and polyethylene oxide (PEO) at 1-4 bars
operational pressure at room temperature. PWF increased with the increase in applied
pressure, which was expected. PWF also increased with MWCNTs wt% until a flipping point
at which the rate decreased. This was explained by the fact that the increase in CNT wt%
increased the membrane hydrophilicity and pore sizes until a point at which too much of the
CNTs resulted in small pore sizes that reduced permeation rate. This was also supported by
SEM images of the pores and macrovoids of the membranes as in Figure 2.7.
Figure ‎2.7 SEM of the cross section of CNTs/PSF membrane, (a) 0 wt% CNTs blank PSF
membrane, (b) 2 wt% CNTs/PSF membrane [38]
The flipping point is thought to be due to the increase in CNTs/PSF stock solution
viscosity, which affects the PI process. In viscous cast membranes undergoing PI, the
solvents diffusion out of the cast is more favorable than the non-solvent diffusion into the
cast, resulting in smaller pore sizes. Permeation measurements of PEO agreed with the results
of the PWF when studying the effect of CNTs on the membrane, yet permeation rates were
smaller than PWF due to the fouling effect from the accumulation of the polymer into the
small pores. Rejection rates were found to be inversely proportional to the permeation rates,
which is common for ultrafilteration membranes, and both permeation and rejection rates
increased with the applied pressure. PVP showed a similar trend in permeation and rejection
rates to PEO, but had higher values since PVP is smaller than PEO. However, PVP rejection
rates decreased with the applied pressure since the size of the PVP molecules were small
enough to pass through the membrane under the effect of increasing pressure. The paper
concluded that CNTs could be a good improvement to the hydrophilicity of membranes used
in filtration applications.
Qiu et al. [39] also embedded different wt% of modified CNTs into PSF membranes via
PI. Here, CNTs were functionalized by oxidation purification first, then further by the
23
addition of isocyanate and isophthaloyl chloride (ICIC) functional groups to their surface.
DMF solvent was used for both the dissolution of PSF and for CNTs dispersion. After having
a homogenous CNTs/PSF stock solution, cast solution was immersed in distilled water
coagulation bath for PI to take place. Ultra filtration tests were carried out for 50 ppm PEG
solution at operational pressure of 1 bar and room temperature. PWF were also measured, and
the results showed an increase in rates with the increase in CNTs content until a flipping
point at which the rate started decreasing again, which agrees with Choi et al. [38] in trend,
not in values. On the other hand, rejection rate trends for PEG didn’t match the PWF trend as
shown in Figure 2.8.
Figure ‎2.8 MWCNTs contents effect on PWF and PEG rejection rates [39]
This decrease in rejection rates was explained by measuring the average pore diameter for
different CNTs/PSF membranes. For 0, 0.12, 0.15, 0.19, 0.32, 0.5 wt% CNTs, the pore sizes
were 5.2, 18.6, 25.7, 38.9, 34.1, and 30.3 nm respectively. Though the trend of pore sizes
matches the rotation of PWF rates, it could also explain the reason behind the continuous
decrease in rejection rates, since the pore sizes changed from 5.2 to 18.6 nm and above. This
possibly meant that for better rejection to take place, the pores had to stay smaller than 18.6
nm. From Figure 2.8, the steepness of the rejection rate decrease was not large (within ~10%
change), i.e. all pores resulted from CNTs addition were too large to reject PEG. Thus, CNTs
managed to increase pore sizes in PSF membranes, and with the proper conditions and CNTs
contents, membrane performance could be enhanced.
Tang et al. [40] prepared CNTs/chitosan nanocomposites with different concentrations of

CNTs (ranging from 0 to 13%) added to 2% (w/v) chitosan. They also incorporated two
different molecular weights pore formers PEG 6000 and PEG 10,000 to different solution
blends, and used acetic acid solvent to dissolve/disperse chitosan, PEG, and CNTs. Instead of
using PI, complete solvent evaporation took place, and followed by placing the resultant solid
membranes in a water bath for the PEG to dissolve leaving a 130-140 µm porous membrane
behind. PWF rates were measured to test the membrane performance at 1 bar and room
temperature. For the membranes prepared with PEG 6K, PWF increased with the increase in
24
MWCNTs content until a flipping point at which the permeation rate started decreasing
again. This agrees with the pattern found in the literature since the membrane pore sizes
increase until a point at which too much of CNTs reduces the pore sizes. Interesting enough,
the maximum PWF as a result of CNTs presence was 128.1 L/m2h, which was 4.6 times that
of blank chitosan membrane PWF (27.6 L/m2h). Figure 2.9 illustrates how CNTs could block
pores.
Figure ‎2.9 A schematic representation of how CNTs concentration can block pores, thus reduce
pore sizes [40]
From the figure above, though CNTs could create alternative channels for water passage,
thus increase flux rates, the increasing amounts of CNTs could block the existing pores thus
reduce pore sizes. As for membranes prepared with PEG 10K, PWF values decreased with
the increase in MWCNTs concentration, unlike PWF trend of membranes prepared with PEG
6K. This could be explained by the fact that PEG 10K created bigger pores than those by
PEG 6K due to its larger size and its poor compatibility with chitosan. Thus, the further
addition of CNTs to the membrane didn’t significantly contribute to increasing pore sizes. On
the contrary, addition of CNTs decreased the pore sizes, thus PWF rates.
Shawky et al. [41] prepared CNTs/aromatic polyamide (PA) nanocomposites using

different concentrations of CNTs added to 10 wt% PA. The team didn’t use the PI method,
but rather complete solvent evaporation. N,N-dimethylacetamide (DMAc) was used for
MWCNTs dispersion and PA dissolution. To prevent agglomeration, benzoyl peroxide was
used to create free radicals on the polymer and the CNTs surface to maximize CNTs
dispersion. After having a homogeneous mixture, the solution was cast and left in the oven
for complete solvent evaporation, and a final membrane thickness of 200 µm. Membrane
performance, flux and rejection rates, was measured using a 4000 ppm NaCl solution at
operational pressure of 39 bars at room temperature. The results showed that permeation rates
decreased with the addition of CNTs. However, salt rejection increased significantly from
24% to 76%. Table 2.6 displays the results.
25
Table ‎2.6 Permeation and rejection rates as a function of CNTs concentration for 10 wt% PA at
39 bars and 25oC [41]
This large increase in rejection rates was believed to be due to the membrane
compactness resulting from the strong interaction between the CNTs and the polymer matrix,
thus creating a network structure responsible for enhancing salt rejection. On testing the same
membranes with larger molecules like humic acid, an organic molecule that is larger than
salts but much smaller than proteins and polymers, the 10 mg/g CNTs/PA membrane was
able to reject up to 90% of the acid. That is to say, CNTs addition was able to enhance the
membranes’ performance significantly.
2.3 Research Aim
The thesis project aims at studying the effect of adding functionalized CNTs as
nanofillers to cellulose acetate membranes, as well as studying the effect of varying the
nanocomposite preparation conditions on the membrane morphology, and its performance in
terms of permeability and selectivity in water desalination applications.
26
3 Theoretical Background
27
3.1 Cellulose Acetate in Desalination
3.1.1 CA Structure and Properties
CA is a relatively cheap, mechanically tough, chlorine resistant, hydrophilic abundant

polymer [28, 42-44] that results from the acetylation of cellulose as in the reaction below
[14]:
As mentioned earlier, CA membranes were first developed in the 1960’s to become the
first high flux asymmetric (anisotropic) membranes, used in a large number of applications
including reverse osmosis, micro, ultra and nanofiltration, gas separation, water desalination,
and wastewater treatment [21, 44-46].
Water permeation and salt rejection rates are affected by the acetylation degree of CA. In
general, cellulose structure has three OH groups that could be partially or fully substituted.
Completely substituted cellulose by three acetate groups (equivalent to 44.2 wt% of acetyl
groups), known as cellulose triacetate, was found to have high salt rejection rates of 99.5%
from seawater feed, while water flux is the lowest possible. On the other hand, lower
acetylation percentage gave higher permeation rates but with lower salt rejection. The
commercial cellulose acetate membranes used in RO desalination for example have 40 wt%
acetate, equivalent to acetyl content of 2.7 on the scale of 0 to 3, where 0 means the 3 OH
groups were not substituted, while 3 means fully substituted. Those membranes can yield 98-
99% salt rejection rates at reasonable water fluxes [47].
Regarding membrane fouling, CA membranes are relatively protected from biofoulants

due to their hydrophilic smooth surface, which doesn’t leave inter-membrane gaps or
interfacial water-resistant areas for foulants to grow. Being chlorine resistant (up to 1 ppm of
continuous exposure to free chlorine) helps in using them in high bacterial count water
sources, pretreated with chlorine [21, 45, 47-48]. CA membrane surfaces are also neutrally
charged, thus reducing the charged macromolecules fouling.
On the other hand, CA membrane surfaces change with the increase in temperature due to
the formation of a dense layer that requires high pressure to diffuse water through. Typical
operational temperature range is 30-50oC maximum. Beyond that, the membrane may be
hydrolyzed and degraded [45, 47-48]. Furthermore, CA membranes could also be hydrolyzed
by pH changes in the medium. This limits their usage in treatment of water with organic
foulants since the tolerance pH range is narrow (only 4-6) [45, 47-48] as shown in Figure 3.1.
28
Figure ‎3.1 Membrane lifetime vs. pH [48]
CA membranes’ main application is for brackish and seawater desalination at an

operational pressure of 15-30 bars, above which the membrane surface is literally “crushed”,
forming a compact dense layer that prevents water or salt passage. This range is significantly
smaller than the earlier mentioned in section 1.2.3, making the RO process using CA
membranes cost effective and energy saving [21, 45, 47-48].
3.1.2 CA Membrane Preparation
The most common preparation method of CA membranes is known as PI or wet phase

separation process. In this method, CA is first dissolved in a solvent until a clear solution is
attained. Then, on a substrate, usually glass, the solution is homogeneously spread using a
casting knife to maintain a uniform casting thickness, as shown in Figure 3.2.
Figure ‎3.2 Schematic diagram of a film applicator and a casting knife
29
The cast film is then immersed into a coagulation bath filled with a CA non-solvent
solution, usually water that is highly miscible with the solvent used. At this point, liquid-
liquid demixing takes place, where the cast solution starts developing a dense nascent skin
layer which is either nonporous or has very small pores (nanometers in diameter) on its top
surface. This is the result of an immediate polymer precipitation at the top layer once the cast
solution is in contact with the non-solvent. Below that layer, a phase separation process takes
place due to the invading non-solvent, resulting in a polymer rich and polymer lean phases.
The polymer rich phase consists of the solidifying CA polymer that is being drained out of its
solvent. The polymer lean phase consists of the extracted solvent mixed with the non-solvent
solutions [30, 49]. The polymer rich phase starts shrinking since more solvent is being
expelled into the polymer lean phase via diffusion. As for the polymer lean phase, nucleation
(pores) develops as a result of solvent-non-solvent exchange within the membrane, where
nucleation grows as long as demixing continues [30, 49]. Figure 3.3 gives a schematic
interpretation of how demixing takes place.
Figure ‎3.3 Interpretation of the demixing process
30
A post treatment step could be used in modifying the membrane morphology, known as
annealing. At this step, the membrane is placed for a short period of time (few minutes) in a
hot water bath after PI. This is used to compact the nascent top layer and narrow the pores
size so that the membranes’ retention ability is enhanced [18], where high temperature gives
the top layer the ability to reorganize its crystalline arrangements to treat any defects on the
membrane surface [50]. The membrane is finally placed in storage medium (usually distilled
water) to remove the excess remaining solvent [30, 49].
There are two types of demixing that are controlled by the coagulation bath temperature,
solvent types, and coagulation time [30, 49]. The first is slow (delayed) demixing at which
the water/solvent diffusion and exchange is slow giving a symmetrical porous membrane.
The slow diffusion requires several conditions to be present, most commonly are low
coagulation bath temperature and solvents with low solubility property for the polymer.
Figure 3.4 shows the different types of solvents and their effect on membrane porosity
evaluated by membrane’s water content.
Figure ‎3.4 Chart elaborating the effect of solvents on membrane morphology [51]
The second type of demixing is instantaneous demixing at which long macrovoids are
formed throughout the membrane since the precipitation process takes place in a very short
time, leading to classical drop-like shaped voids appearing.
The mechanism of macrovoids formation has been explained by several ways. The main
factors studied to suppress the development of macrovoids were found to be: a choice of a
solvent-non-solvent with low tendency of demixing, the addition of the solvent to the non-
solvent bath before cast solution immersion, an ISE before PI, and the increase in the
polymer content in the cast solution which increases cast solution viscosity. Moreover, the
31
addition of the non-solvent to the polymer-solvent stock solution can suppress macrovoids
formation if its content exceeds a certain minimal limit that could be evaluated
experimentally [52]. As just mentioned, macrovoids are formed during the instantaneous
demixing PI, where the non-solvent is trapped in the polymer lean phase. At the interface of
the non-solvent at the boundaries of the nuclei, however, the type of demixing taking place
was found to be different from that taking place throughout the rest of the cast solution. It
was suggested, and later experimentally verified, to be delayed demixing. What happens is
the solvent defusing into the non-solvent droplets is greater than the non-solvent defusing
into the polymer rich phase, thus the size of the nuclei increase [52-53]. This can be
explained by osmosis phenomenon, where inside the non-solvent droplets, the concentration
of the expelled solvent keeps increasing to a point at which the rate of exchange of the
solvent-non-solvent decreases, thus delayed demixing occurs at the same time as the
surrounding polymer rich phase is shrinking and solidifying [54]. Figure 3.5 shows the
different layers of an asymmetric membrane and the difference in the type of demixing
occurring within the layers.
Figure ‎3.5 Schematic representation of an asymmetric membrane developed via PI, modified
from [52]
3.2 Permeation and Retention Rates Calculation
To evaluate membrane performance in filtration applications, measurements of

permeability and selectivity are carried out under applied pressure in a specially designed
liquid cell as shown in Figure 3.6 [29]. In such a cell, the membrane is placed on a highly
porous metal support at the bottom of the cell, with its top layer (the metallic side) facing the
solution, since it is the layer that is expected to filter off the solute and allow the solvent
through. Then a known quantity of the solution is placed on top of the membrane, and the cell
is tightly closed. The inlet on the top of the cell allows gas in (usually nitrogen gas) to apply
pressure on the system. Once a steady flow rate is attained from the outlet part at the bottom
of the cell, the permeation rate is measured. The rejection rate can be measured as well by
analyzing the permeate.
32
Figure ‎3.6 Permeation cell (modified from [29])
Mathematically, to evaluate membranes’ performance under different pressures and

facilitate their comparison with similar or even different membranes, liquid permeability P
(L/m2 h bar) is calculated as such:
where J is the permeation flux (L/m2 h), and Appl. Prs is the applied pressure (bars). As for
the permeation flux rate, J, it is calculated using the following equation:
where V is permeate volume that passed through the membrane (L), A is the total area of the
membrane on which pressure is applied (m2), while t is the time taken for the permeate to fill
a certain volume V (h) [40].
On the other hand, salt retention rates, R, (%) is calculated from the following equation:
where R (%) is an evaluation of how much solute is rejected by the membrane, Cp is the
solute concentration of the permeate, Cf is the solute concentration in the feed solution [21].
Membranes’ morphology can be manipulated by varying the preparation conditions such

as CA content, initial casting thickness (ICT), coagulation bath temperature, and presence of
organic or inorganic additives [28, 49]. It could also be manipulated by mixing CA with other
33
polymers to form a blend [22, 30], adding different types of inorganic compounds [42, 55], or
by adding nanofillers like alkoxysilane [56] to form a nanocomposite.
3.3 Nanocomposites
A nanocomposite is a synthesized material made up of mixing two or more components,

different in their physical and/or chemical properties to give a final structure with a preserved
phase difference between its constituents. One or more of those components, called
nanofillers, has one or more of its dimensions (length, width, or thickness) on the nanoscale,
ranging in size from 1 to 100 nm [57-58]. Bones, for instance, are the natural example of
nanocomposites since they are designed from organized layers of ceramic tablets and organic
binders, giving a final structure of a remarkable difference in properties from its constituting
components [59].
The exceptional high surface to volume ratio is the main factor controlling the
structure/property relationship of nanocomposites, since particle size is directly related to
particle number in a given volume [60] as shown in Figure 3.7.
Figure ‎3.7 Illustration of how the particle number increase as their sizes decrease in a given
volume [61]
In a fixed volume, say 3% as in the figure, nanofillers occupy more space than larger
particles of the same material, leading to more surface exposure at the interface, i.e. an
enormous increase of the interfacial area [61]. This directly affects the nanocomposites’
properties to become dependent on the size of the nanofiller, since properties are dominated
by the strong interactions at the interface between the nanofiller and the corresponding filled
matrix without even a significant change in the density of the latter [57, 60-61].
In order to maximize the use of the interfacial properties, designing nanocomposites is

controlled by two important morphological conditions. First, nanofillers must be
homogenously distributed within the matrix without any agglomerating particles in order to
avoid reduction in interfacial area and surface to volume ratio. Second, the surfaces of the
nanofiller and the matrix have to be compatible in order to achieve the best possible
interaction between them to assure an effective load transfer across the nanocomposite, since
under the effect of external forces, like an applied RO pressure on a non-properly interacted
nanofiller-matrix nanocomposite, the stress will not be uniformly distributed, leading to the
rupture of the nanocomposite. Furthermore, without proper interactions at the interface, the
34
matrix properties wouldn’t be modified since they are dominated by the interactions at the
interface as just explained [61].
Nanocomposites membranes can be prepared by several methods, including PI. Yet, there
are two additional steps in the previously explained procedure. The nanofiller is first
functionalized to create the best possible interaction at the interface between them and the
matrix. The second step is having the nanofiller homogeneously mixed with the polymer
solution before casting takes place, to guarantee the formation of a nanocomposite without
any bulky agglomerating particles [61].
Polymers offer a great option for matrices that can be reinforced by the different types of
nanofillers such as clays, ceramics, nanowires, and nanotubes. CNTs in particular have been
of a major interest for researchers due to their unique structure, novel properties, and wide
applications, as will be explained below.
3.4 Carbon Nanotubes as Nanofillers
3.4.1 CNTs Structure and Properties
In 1991, Sumio Ijima attempted to vaporize graphene sheets using the high current arc
discharge process, but instead, he observed using a high-resolution electron microscope that
the graphene sheets wrapped up to give a hollow tubular shape of elongated fullerenes. This
novel arrangement of carbons is capped at each end, and the walls are made of hexagonal
bonded carbons with a single crystalline conformation. The nanotube diameter is around 10
nm, its length is a few micrometers long, which makes it 1000 times more than that of the
diameter [62-65].
There are two classes of CNTs, classified according to their structure. The first class is
Single Walled CNT (SWCNT), which are wrapped single layers of graphene sheets, and
which usually exist in bundles or ropes, with an outer diameter of 10-30 nm. The orientation
of the carbon-carbon bonds compared to the tubular axis divides SWCNTs into three
subgroups: i. armchair nanotubes whose carbon-carbon bond is perpendicular to the nanotube
axis, ii. zigzag nanotubes, whose carbon-carbon bond is parallel to the nanotube axis, iii.
chiral (helical) nanotubes, whose hexagon bonding carbons are wrapped up in a helical shape
around the axis of the nanotube [63, 66].
The second class of CNTs entails MWCNTs. Those are what was originally observed by
Iijima before catalyzing the process two years later to manufacture SWCNTs [63]. A
MWCNT is a group of concentric coinciding SWCNTs whose diameters increase
successively outwards, yet, remain within the nanoscale dimensions. The number of shells
could be as large as a dozen [63]. Figure 3.8 gives an idea of the shapes of SWCNTs and
MWCNTs.
35
Figure ‎3.8 (a), (b), (c) SWCNTs different structures with highlight on C-C bond orientation vs.
the nanotube axis [67], (d) MWCNTS structure [63]
Mathematically, CNTs conformations are specified by their two integer rolled-up vectors
(n, m) [64] which are used to calculate both the diameter (d) and the chiral angle (θ) of the
nanotube from the following equations [68]:
where a is the distance between two bonding carbons and it is equal to 1.421 Å [68], and
where θ ranges from 0o to 30o. For instance, zigzag nanotubes have m = 0 at θ = 0o, armchair
nanotubes have n = m at θ = 30o, and helical nanotubes have 0o< θ <30o [68] as shown in
Figure 3.9.
Figure ‎3.9 (a) (n, m) vector indication [64], (b) Different conformation of CNTs, their chiral
angles, and indices [68]
36
CNT structures are considered to be one of the strongest in nature due to the “covalent in-
plane” bonding between carbons, similar to the bonding type between carbons of the
unfolded graphene sheets [65]. This strongly influences CNTs thermal, electrical, and elastic
properties in exceptional ways, where for example SWCNTs stiffness is comparable to
diamonds, tensile strength is much higher than that of steel, density is half that of aluminum,
electrical conduction is comparable to copper wires, thermal conductivity is nearly double
that of diamonds, and last, but not least, SWCNTs are thermally stable up to 2800oC in
vacuum and 750oC in air, while metal wires (for example in the ones used in microchips)
would melt at 600–1000oC [62, 69-70].
Ideal nanofillers for polymer matrices as they may seem, CNTs are highly stable, that it is
extremely difficult for them to interact chemically or physically with other materials such as
solvents, polymers matrices, or other nanofillers. They are insoluble in water or organic
solvents. They are wet resistant and tend to agglomerate in clusters due to the high van der
Waal interaction between them, making their dispersion into solvents very challenging. As
for penetrating their bonding system, it requires very high energy to rapture due to its high
stability. Their ends caps are closed so they cannot be used as channels for fluids
transportation [62, 70]. For these reasons, CNT functionalization is crucial for creating
functional groups on CNTs surfaces to allow them to interact with other materials, as well as
for opening their ends caps to allow them to act as transportation channels for fluids.
3.4.2 Functionalization of CNTs by Oxidation Purification
Oxidation purification is the most common method used for functionalizing CNTs with
the purpose of homogeneously mixing them with polymer matrices. In this process, carbons
on the defected sites of the crystalline structure of CNTs, or on the sites of misaligned π
bonds, or usually on both, get oxidized giving COOH hydrophilic groups [63, 70]. The
process involves using concentrated nitric acid, sulfuric acid, or a mixture of both to reflux
CNTs for a period of time defined according to the reaction conditions [63, 70]. Then the
functionalized CNTs are washed several times until they are neutralized at pH 7. They are
then washed with drying agents such as acetone and THF to remove as much water as
possible. Finally, they are dried under vacuum to be ready to use [71]. Figure 3.10 shows how
the CNTs look like after oxidation.
Figure ‎3.10 Section of an oxidized CNT reflecting terminal and wall oxidation [71]
37
3.4.3 CNTs in Water Desalination and Treatment
In water treatment and desalination, CNTs continue to surprise scientists, where a

nanocomposite filled with open ended CNTs was found to pass water at flux rates
significantly higher than using the same matrix without the nanofiller. Additionally, CNTs-
polymer nanocomposites can stand high transmembrane pressures up to 1 atm, indicating that
the membranes don’t easily crush due to macrovoids or surface flaws under the effect of
applied pressures [16, 38, 41, 72]. As for salt rejection rates, they could be improved as well
[16]. CNTs were also found to exhibit cytotoxic properties for bacteria, which could be used
as a resort for RO membrane biofouling problems [72].
A major drawback in using CNTs for water desalination is the fear of environmental
contamination with CNTs waste disposed from production plants, which would eventually
enter into the human food chain. This raises many questions since studies on CNTs’ effects
on mammalian cells include pulmonary inflammation (inflammation of lung tissues), cellular
proliferation (multiplication of cell numbers by division), heart growth inhibition, and
toxicity [72]. CNTs are also classified as ‘hard’ biopersistant, hence, they are not cleared
from the human body through excretion [73]. However, having strictly controlled disposable
management system would reduce their release in nature. Moreover, studies showed that if
CNTs are homogeneously dispersed within membranes, they are more likely to have minimal
toxic effect on humans, since toxicity is achieved by agglomerated CNTs more than by the
dispersed. So it all pours into proper functionalization of CNTs to prevent their
agglomeration with the membrane matrix in which they are impeded [72].
Though water transport through hydrophobic channels is still not well understood, the
experimental results were conclusive: water pass at exceptional high rates through CNTs.
Attempting to clarify the transport mechanism, scientists resorted to molecular dynamics
simulations and compared them to the experimental results.
3.5 Inside CNTs Molecular Dynamics (MD)
MD are computer simulations methods that study molecular transport mechanisms, since
short time scale and small particle sizes (nanoscale or smaller) obstruct real time
experimental measures. Using Newton’s laws of motion, initial molecular configurations, and
Lennard-Jones potential energy functions (functions that calculate intermolecular potential
energies and forces), molecules’ location and momentum trajectories in space could be
estimated [74]. In other words, MD could be used to predict how water molecules would
behave inside a nanotube in terms of position, arrangement, bonding properties, and
interaction with the hydrophobic CNTs walls, and how these behaviors compare to entering a
wider diameter microscale or even larger tubes.
MD simulations estimate that the smallest possible CNT for water molecules to enter into
is the (5, 5) with d = 6.78 Å, and any smaller nanotubes, like the (7, 1) and the (5, 4) with d =
5.91 and 6.12 Å respectively don’t allow water passage [68]. On filling CNTs, water
molecules are accommodated in one of three ways, either wire mode, layered mode, or bulk
38
mode. The type of mode is determined by the diameter of the nanotube, since molecular
conformation transition takes place from bulky-like to layered-like when water molecules are
squeezed in a solid structure of approximately 1 nm or less [68, 74]. As for water density, it
changes significantly according to the size of the nanotube, but it is independent of the CNTs’
chirality. For example, a nanotube with diameter of 7.8 Å is expected to affect water density
to become as small as 0.2 g/cm3compared to the bulk water density of 1 g/cm3 [68].
Wire mode is a one dimensional ordered conformation of water molecules in which the
probability of finding hydrogen and oxygen atoms in the center of the CNT is maximal, and
the hydrogen-bonded molecules rearrange to form a water wire along the nanotube axis
(Figure 3.11), hence the name [64, 68].
Figure ‎3.11 1D wired mode of water molecules inside (8, 2) CNT [68]
What happens is when the tetrahedral hydrogen bonded molecules approach the nanotube,
two of the hydrogen bonds are broken, while the remaining two rearrange in a highly oriented
linear chain like. The average lifetime of those H-bonds is estimated to be 5.6 ps, which is
five times longer than the lifetime of H-bonds in bulky water conformations. The OH groups
engaged in the H-bonds are estimated to be nearly in the center of the nanotube, and they flip
directions approximately every 2-3 ns [75-77].
Despite the fact that losing two hydrogen bonds is energetically unfavorable (~10
kcaL/mol), some of this energy is retained by the van der Waal’s interaction between the
water molecules and the carbons of the nanotube wall (~4 kcaL/mol). To compensate for the
rest of the remaining energy, MD simulations predict that entropy plays an important role
since water molecules were found to have the ability to rotate freely around the H-bonds
chain, creating a “degenerate energetic ground state” that is more occupied than the higher
dominant free energy state [77].
In wider CNTs, a second layer of water molecules could squeeze in (layered mode),
forming helical like conformations such as in the (10, 5) CNTs. Increasing the diameter of the
CNTs creates more space inside the nanotube to form a ring like layer with a wired water
chain trapped in the middle of the ring [68]. Further increasing the diameter like in the (10,
10) nanotubes creates more spiral confirmations that MD suggest to be due to the presence of
π electrons of the carbon-carbon bonds along the nanotube [78].
As the diameter increases, more layers are added until water molecules reach a point of
not sensing the surrounding CNT walls, i.e. the bonding carbons become too far away to have
39
an effect on the orientations of the water molecules. At this point, water starts acting as a
bulk-structure (bulk mode) [68]. Figure 3.12 shows different water orientations inside CNTs.
Figure ‎3.12 Water molecules structure inside CNTs (a) Spiral-like layered mode, (b) bulky
mode [68]
Experimentally, water fluxes though CNTs were found to be exceptionally high as

mentioned before. MD suggest that this is due to the significantly smooth chemically inert
graphene inner walls forming the nanotubes [64, 79]. Moreover, the fact that water molecules
have the capability of fluctuation inside the nanotubes results in a “drift-like motion” which
highly accelerates the water flow across the nanotubes [77]. MD also suggest that the flow
rates are not governed by the length of the nanotube, contrary to what classical macroscopic
hydrodynamics models suggest [64, 79].
As for salt retention rates, MD simulations suggest that in order for a cation such as
sodium or potassium ions, or an anion such as chlorine ions to be 100% prevented from
passing into a nanotube, there is a minimal diameter required of ~0.4 nm, equivalent to the
hydration shell around the ion. At this diameter, the ions have to lose parts of their hydration
shell, which creates an energy barrier, high enough for ions not to favor passing through the
nanotube (approximately ~120 kJ/mole) [64]. Since water cannot pass through this diameter
either, MD simulations examined salts’ behavior in larger diameter nanotubes such as the (5,
5) vs. the (8, 8). MD postulates that 100% salt rejection is expected from the (5, 5) since its
diameter (6.87 Å) is narrow enough to prevent salts from passing through. However, only
~60% salt rejection is found for the (8, 8) due to its relatively large diameter (10.86 Å). Any
CNTs with diameters range between the (5, 5) and (8, 8) show different rejection rates.
Furthermore, MD speculate that CNTs with diameters larger than the (8, 8) would allow free
movement of small ions since the energy barrier is no longer there. Yet, there aren’t enough
studies or experimental data to support these claims. There is one, however, that opposes
them using (10, 10) CNTs with d = 13.57Å, which postulates that there exist an energy
barrier for Na+ ions trying to pass through these CNTs [16, 64, 68].
40
3.6 Membrane Characterization
In order to evaluate whether a membrane could be used in filtration applications, several

techniques are used to study its performance and morphology. For membrane performance,
desalination measurements using a pressurized cell are carried out, and salt content is
quantified using several techniques, the simplest of which is titration. As for morphology
characterization, it can be done using scanning electron microscope (SEM), and porosimetric
analysis via gas sorption.
3.6.1 Scanning Electron Microscope (SEM)
SEM is a powerful technique in topography analysis on the nanoscale, which has major
applications in material sciences and nanotechnology. This is due to using an electron beam,
of very short wavelengths that could identify small details on a sample surface and give
images of features as small as 200 nm.
Basically, the sample is subjected to a ray of electrons and an image is being created
from their elastic and inelastic scattering, as well as, the electromagnetic radiations (X-rays)
generated by the sample. As shown in the Figure 3.13, the electrons are ejected from an
electron gun which is usually a tungsten filament to reach an anode with an acceleration
energy that could reach 1-30 keV. Then the beam is focused through three lens systems, till it
hits the specimen, and gets scattered [80].
Figure ‎3.13 Schematic diagram showing the main components of SEM [80]
There are two different types of scattering: the back scattering, which is due to the
electron beam hitting the electrons on the surface of the specimen and elastically bouncing
back, with no exchange of momentum between the electrons on the sample and the electrons
from the gun. Thus, the scattered electrons are reflected into discrete locations on a detector
located right above the sample, and from these locations, the surface features of the specimen
41
could be formulated. The second type of scattering is the secondary scattering, which is due
to the inelastic interaction between the electron beam and the electrons of the elements on the
surface where exchange of momentum takes place, resulting in the ejection of one of the
electrons of the K shell from the atoms on the surface. These secondary electrons are usually
very weak (10 to 50 eV), and they are traditionally collected by Everhart-Thornley detector,
where the secondary electrons hit a scintillator. The flashed light signals hit a photomultiplier
which converts them into an electrical signal. This is then digitally output as a topographic
image of the surface [80-82]. The third type of signal detected is the x-rays generated from
the material which are collected using EDX detectors [80]. This type of signals is very useful
in the qualitative analysis of elements found on the surface of a sample or within a mixture of
compounds, since the generated x-rays are element specific [83].
Samples have to be conductive because, otherwise, bombarding them with electrons

could result in static charges on a nonconductive surface, leading to noise in the sample
image. In this respect, nonconductive materials are sprayed with a thin layer of a conductive
coat, which is so thin that it takes the shape of the surface and doesn’t modify its topography,
hence, prevents charging. Figure 3.14 shows a single carbon nanotube with diameter ~12 nm
at 400,000 X magnification.
Figure ‎3.14 The dimension of a MWCNT
The huge magnification power of SEM makes it a very useful tool in topography analysis
that allows a wide range of applications in nano-science.
3.6.2 Surface Analysis by Gas Sorption
Identifying pores sizes, quantity, and overall surface area of a porous sample is important
in analyzing and interpreting its behavior, for example, in filtration applications, which
depend on pore sizes to screen out dissolved particles. The most common method in porosity
analysis is physisorption of a gas onto the solid surface, and by knowing the quantity
adsorbed and the corresponding pressures, analysis is carried out.
Before explaining the process in details, some important terms have to be identified [84]:
42
 Micropores are pores smaller than 2 nm
 Mesopores are pores between 2-50 nm
 Macropores are pores larger than 50 nm
 Adsorption means the attachment of an element or a molecule, gas or liquid, onto a
solid surface either physically via van der Waal forces (physisorption), or chemically
via a chemical bond (chemisorption)
 Adsorbent surface is the solid substrate on which molecules adsorb or from which
they desorb
The surface atoms of the solid have the ability to attract atoms or molecules from the
surrounding medium via van der Waal forces [85-86]. Based on this, measuring the surface
area of the solid could be carried out. The first step in the process is emptying the top layer on
the sample surface from any adsorbate molecules, the most common of which are water
vapor and carbon dioxide. This is done by a vigorous degassing process in which the sample
is heated up (temperature selected depending on the sample type) under vacuum for several
hours. Once the surface is free of adsorbate molecules, known increments of an inert gas, like
N2, Ar, or Kr are introduced, and the sample in contact with the gas is left to equilibrate so
that some of the gas adsorbs on the sample surface. At equilibrium, the pressure in the sample
vessel is measured, and from the original pressure and equilibrium pressure, the quantity of
the gas adsorbed is obtained. Knowing the cross-sectional area of the gas particles allows the
calculation of the sample surface area assuming monolayer formation. Figure 3.15 shows the
stages of gas adsorption on the solid surface. It’s important to note that this process is
temperature dependent, and in this respect, the process is carried out at constant temperature
[85-86].
Figure ‎3.15 Different stages of gas physisorption on sample surface with pressure increase [85]
Translating the process into usable data, the first chart produced is called the isotherm,
which represents the quantity adsorbed as a function of the measured pressure [87]. Yet, the
pressure value is represented in the form of relative pressure P/Po where Po is the saturation
43
pressure at which the gas and liquid phases of the adsorbate molecules can coexist together,
at a specific temperature. There are 6 types of isotherms as shown in Figure 3.16.
Figure ‎3.16 Different types of isotherm linear plots [84]
Type I, also known as “pseudo-Langmuir”, can be seen in chemisorption since the

adsorbent surface becomes completely covered with only few molecules thick monolayer that
formed a chemical bond with the surface, thus saturating the possible adsorption sites on the
surface. As for physisorption, type I isotherms are common for porous solids with
micropores. It is worth mentioning that the reason why it is also called pseudo-Langmuir is
because Langmuir built his theory about sorbing gases while assuming that solid surfaces are
covered with a monolayer of the adsorbate molecules. This is close to the truth for
micropores since they are too narrow to be filled with multilayers, while it is pretty much the
case in chemisorption, where type I isotherms have wider applications and used in more
calculations [84, 87-89].
Type II isotherms represent adsorption on nonporous surfaces, or solids with macropores.

The most important feature they have is the knee, which is assumed to be the point at which a
total formation of a monolayer is complete, and the rest of the curve represents the different
multilayers adsorbing [84, 87-89].
Type III isotherms represent a weak interaction between the adsorbate molecules and the
adsorbent surface, which is clearly reflected by the disappearance of the knee. This takes
place when the adsorbent surface has a weak potential to attract surrounding molecules to
adsorb. In this case, the adsorbate molecules prefer to adsorb over one another in the form of
44
multilayers rather than adsorbing over the adsorbent surface. Example of such surfaces
includes organic polymers [84, 87-89].
Type IV isotherms represent porous solids usually with mesopores. The plot has two
characteristic features: the first is the presence of the knee as in type II isotherms, which
serves the same purpose of identifying the starting point of multilayers development. The
second feature is the presence of hysteresis which represents a desorption deviation from the
adsorption curve. This was suggested to be due to the change in the mechanism of mesopores
emptying from the adsorbate molecules as compared to the mechanism of their filling up [84,
87-89].
Type V isotherms represent porous solids also with mesopores which accounts for the
hysteresis on the curve. However, the shape of the curve is very similar to type III isotherms
because of the type of the adsorbent surface which also happen to have weak affinity towards
the surrounding molecules. Thus, the development of a monolayer is not detectable since
multilayers tend to develop at very low relative pressures [84, 87-89].
Type VI isotherms are very rare ones that represent adsorption on highly uniform
homogenous nonporous surfaces. The characteristic feature on the plot is the presence of
steps. Each step is thought to represent a new developing monolayer, and the height of the
step is used to calculate the capacity of the developing monolayer (i.e. the quantity of
adsorbate molecules forming the new layer). A very popular example of such isotherms is the
one for the adsorption of argon on pyrolytic graphite [84, 87-89].
The data of isotherms are used to deduce surface areas, and pore size analysis is carried
out by different models based on the type of the isotherm and the types of pores present. For
this current research: the BET theory was used for adsorption isotherms interpretation and
sample surface determination; the density functional theory (DFT) was used to profile the
pore volumes and areas and develop pore size distribution; and the de Boer t-plot method was
used to calculate the micropores volume and external surface areas (i.e. areas which
micropores do not contribute in, like meso and macropores areas, as well as the area of the
surface).
3.6.2.1 BET Method
The main contribution of BET is the idea of multiple layer adsorption. The rate of
adsorption suggested to be equal to the rate of desorption was introduced by Langmuir for
monolayer adsorption whereas BET made the same assumption for multilayers [87-88, 90].
The BET model succeeded in explaining isotherms II and IV that Langmuir’s failed to
explain. From their equation, a constant c could be derived and used mathematically in
calculating the inflection point at which multilayers start forming. Furthermore, the c value
was found to resemble the strength of interaction between the adsorbate molecules and the
adsorbent surface, where the stronger the affinity between them, the higher the c value. The
linear form of the equation is [91]:
45
where V is the total volume of the adsorbed molecules, Vm is the volume of the monolayer,
and c is related to the adsorption energy. To calculate Vm, a plot of P/V(Po-P) vs. P/Po is
constructed such that its slope is (c-1)/(Vmc) and its intercept is 1/Vmc [91].
Figure ‎3.17 BET plot to evaluate Vm [91]
Vm is used in evaluating the monolayer’s volume, the number of molecules occupying the
monolayer, and multiplying this by the surface area occupied by each molecule yields the
total surface area of the surface. Though the linear part of the plot is in the range of 0.05-0.30
P/Po, it is enough to enable Vm evaluation [87-88, 90].
It is important to note that although the BET model explained isotherms II and IV, it
failed to explain isotherms I and III since there is no inflection point to linearize its range.
Furthermore, the c values for both isotherms were found to be higher than normal, hence,
studying such isotherms require other models [90].
3.6.2.2 de Boer t-plot Method
Interpretations of multilayers adsorbing on nonporous surfaces showed that there exists a

direct relation between the thickness of the adsorbate molecules layers and their volume. This
allowed thickness equations to be able to predict the adsorbate molecules volumes at different
pressures. de Boer and Lippens made use of this observation and compared the isotherms of
microporous materials with isotherms type II of the nonporous surfaces. By taking into
account comparable c values of both the porous and nonporous surfaces, resembling
similarity in the nature of the absorbent surface, the gas volume adsorbed on the porous
surface is plotted against the statistical thickness of the reference nonporous material (at the
same P/Po experimental values). This necessarily results in a deviation from linearity
whenever pores are present as shown in Figure 3.18.
46
Figure ‎3.18 t-plot comparison between porous and nonporous surface [92]
In the absence of micropores, the extrapolation of the curve passes by the origin, yet,
when micropores exist, the Y-intercept (i) of the linear parts of the curve (B) is used to
calculate micropores volume (VMP) by the following equation:
Furthermore, the slope of the (A) part of the t-plot could be used in calculating the
external surface area (St) of the porous material, excluding the micropores surface areas, by
the following equation1:
As for the slope of the (B) part of the t-plot, it could be used in calculating the surface
area of the mesopores of the porous sample [92].
3.6.2.3 DFT Method
DFT method is a complex mathematical interpretation of the adsorbate-adsorbent

interactions with geometrical consideration of the pores shapes and sizes. These
interpretations create a real density profile of the adsorbate molecules as a function of
temperature and pressure, from which the number of molecules adsorbed could be calculated.
DFT method depends on several assumptions: (1) there exists attraction between the
adsorbate molecules and the adsorbent surface (2) The organization of the adsorbate
molecules on the surface maintains minimal overall free energy for the system, which is done
by finding the molecules rearrangements with the highest probability distribution, which
usually reflects minimal energy [93-94].
1
The factors multiplied by both the intercept value and the slope value are derived from different
equations.
47
As for the pore size distribution, the experimental isotherms are compared to hundred
different isotherms, each with an individual pore dimension. In this respect, pore geometry
could be analyzed, and pore size distribution compared to pore width is constructed [93-94].
There are other different interpretation models, and each can provide a piece of
information about samples porosity (if any), surface areas, and can even evaluate samples
adhesion property. Using this, membranes morphology and porosity can be analyzed, and
their behavior in desalination and filtration applications can be interpreted.
48
4 Experimental
49
4.1 Materials Used
CA (avg. molecular weight 50,000 Da, 39.7 wt% acetyl content) purchased from Sigma-
Aldrich Co. was used as a polymer matrix. Acetone (density 0.791 g/mL at 25oC, purity ≥
99.8%) purchased from Sigma-Aldrich Co. was used as a solvent. Deionized water was used
as a non-solvent. Multi-walled carbon nanotubes (MWCNTs) Baytubes® C150P (C-purity ≥
95 wt%, inner diameter of ~4 nm, outer diameter of ~13 nm, and length > 1 µm) provided by
Bayer Material Science AG. were used as nanofillers. H2SO4 (purity = 99.999%) purchased
from Sigma-Aldrich Co. and HNO3 (purity = 69%) purchased from Patel Group, India were
used in oxidation purification of the CNTs. Sodium Chloride (molecular weight 58.44 g/mol,
density 2.165 g/mL at 25oC, purity ≥ 99.5%) purchased from Sigma-Aldrich Co. was used in
desalination measurements. Sodium Hydroxide (molecular weight 40.00 g/mol, density 2.130
g/mL at 25oC, purity ≥ 98%) purchased from Sigma-Aldrich Co. was used in the
neutralization step of CNTs functionalization. Silver Nitrate (molecular weight 169.88 g/mol,
density 4.35 g/mL at 25oC, purity ≥ 99.8%) purchased from Sigma-Aldrich Co. was used in
salt retention measurements.
4.2 Stock Solutions Preparation
4.2.1 CA Stock Solution
1. X grams of CA powder were dissolved in Y ml of acetone, X and Y determined

according to the wt% required. A Z ml of deionized water was added for solutions
containing non-solvent, Z determined according to the wt% of non-solvent required;
2. The mixture was left overnight under continuous stirring until the CA completely
dissolved forming a clear homogenous solution;
3. The solution was then sonicated for 15 minutes and then left to rest so that all trapped
air bubbles (resulting from sonication) are released. This typically required between 1
and 2 hours.
4.2.2 CA Stock Solution with Pristine (non-functionalized) Carbon Nanotubes (pNTs)
1. X grams of CA powder were dissolved in Y ml of acetone, X and Y determined

according to the wt% required. L grams of pNTs were added, L determined according
to the wt% of pristine CNTs needed. A Z ml of deionized water was added for
solutions containing non-solvent, Z determined according to the wt% of non-solvent
required;
2. The mixture was left overnight under continuous stirring until CA completely
3. The solution was then sonicated for 2 hours in order to ensure complete dispersion of
the pNTs, then left to rest so that all trapped air bubbles (resulting from sonication)
are released. This typically required between 1 and 2 hours
50
4.2.3 CA Stock Solution with Functionalized CNTs (fNTs)
1. 15 grams of CA powder were dissolved in Y ml of acetone determined according to

the wt% required;
2. The mixture was left overnight under continuous stirring until CA completely
3. L grams of fNTs were dispersed in 20 ml deionized water (L determined according to
the wt% of functionalized CNTs needed) and divided into 4 equal portions. Each
portion was sonicated for 1 minute before adding it to the CA stock solution. After the
addition of each portion to the stock solution, vigorous stirring of the mixture was
carried out for a few seconds. The 4 portions were thus added sequentially to the CA
stock. This ensured proper fNTs dispersion in the CA stock solution;
4. The mixture was then sonicated for 2 minutes, then left to rest overnight so that all
trapped air bubbles (resulting from sonication) are released.
The CA and H2O contents (15 wt% and 20 wt% respectively) were fixed in the
nanocomposites preparation.
4.3 Methods and Instrumentation
4.3.1 Functionalization of Carbon Nanotubes
MWCNTs functionalization was carried out as described in the literature [71]:
1. 12 g of pristine MWNTs were added to 300 mL H2SO4 and 100 mL HNO3 (3:1 by
volume) in a round bottom flask.
2. The flask was immersed in an ultrasonic bath for 10 minutes to disperse the CNTs.
3. The flask was then mounted to a reflux with continuous water circulation to minimize
acid evaporation. The top of the reflux was connected to a bottle filled with
concentrated solution of NaOH.
4. The flask was immersed in a boiling water bath for 100 minutes, after which the flask
was cooled down under tap water.
5. Filtering the CNTs was carried out using Whatman 0.2 µm pore size Teflon filter
membranes placed in a microfiltration system connected to a pump.
6. The collected CNTs were washed with deionized water until the pH of the filtrate
solution was neutral.
7. The collected CNTs were then washed with 50 mL of acetone to ensure the complete
removal of water. Then they were placed in a desiccator under vacuum for 24 hours to
complete drying.
8. The CNTs were then ground in a ceramic mortar, and the particles passed through a
180 µm pores sieve to have fine powdered CNTs.2
The functionalized CNTs were characterized by FTIR spectroscopy as well as SEM imaging.
2
This step was added to the procedure obtained from the literature as it proved useful in achieving
proper dispersion of the CNTs in the polymer matrix.
51
4.3.2 Membrane Casting
Membrane casting was carried out using an Elcometer 4040 Automatic Film Applicator
using stock solutions prepared as specified in section 4.2 above, according to the following:
1. The stock CA solutions were poured into the casting equipment feed container;
2. The solution was then spread over a glass substrate by the equipment moving casting
knife, adjusted in height to the needed membrane ICT;
3. The glass substrate with the cast solution was then immersed in a deionized water
bath at the selected coagulation bath temperature for PI;
4. The resulting membranes were then stored in deionized water overnight to ensure
complete solvent (acetone) removal.
For membranes where solvent evaporation was carried out, this took place either after
solution casting and prior to PI, or post PI, as was necessary.
Figure ‎4.1 Elcometer 4040 Automatic Film Applicator
4.3.3 Sample Surface Area and Porosity Determination
Sample surface area and porosity determinations by nitrogen adsorption were carried out
using a Micrometrics ASAP 2020 instrument. Membrane samples were prepared for analysis
according to the following procedure:
1. A 5x5 cm2 section of the sample membrane was dried for an hour in a furnace at
100oC, then left to cool down to room temperature in air;
2. The sample, cut into very small pieces and weighed, was inserted in the ASAP 2020
sample holder glass tube;
3. The tube was mounted in the ASAP 2020 equipment and degassed below 50 µmHg at
30ºC for 30 min then heated (10 ºC/min) to 80 ºC for 360 min.
52
Adsorption/desorption isotherms were obtained by nitrogen adsorption at -196 ºC. Specific
surface areas were calculated by the Brunauer-Emmett-Teller (BET) equation and the t-plot
method which was also used to calculate micropore volumes. The pore size distribution was
determined using DFT model.
Figure ‎4.2 Micrometrics ASAP 2020 instrument [95]
4.3.4 Infrared Analysis
Infrared spectra were obtained using a Thermo Scientific NICOLET 380 FTIR. Solid
samples were prepared as KBr pellets, where 2 mg of the sample were mixed with
approximately 200 mg of KBr spectroscopic grade. The mixture was then subjected to a
pressure of about 1400 kPa in a hydraulic press.
4.3.5 Scanning Electron Microscopy
Sample images were obtained using a Leo Supra 55 (ZEISS) field emission electron
microscope. SEM images were recorded without sample coating.
4.3.6 Liquid Test Cell
Permeation and salt retention determinations were carried out using a Sterlitech HP4750
Stirred Cell. The measurements were carried out as follows:
1. A 5 cm diameter membrane disc was placed within the test cell on a porous metal
support, and the cell was filled with the feed solution;
2. Pressure was slowly increased to 24 bars using compressed nitrogen gas;
53
3. The sample was then given 5 to 15 minutes to have a steady liquid flow rate;
4. If after 15 minutes no liquid flow occurred, pressure was increased (with a rate of
about 0.7 bars/minute) until a flow was obtained (the maximum pressure reached was
55 bars). For salt retention determinations, a 1000 ppm NaCl solution was used as
feed. The amount of NaCl retained was determined by volumetric analysis
determinations (titration with standardized silver nitrate) of the permeate solution.
Permeation and salt retention measurements were carried out in triplicates and the average
obtained.
Figure ‎4.3 Sterlitech HP4750 Stirred Cell
4.4 Investigating Preparation Conditions
On developing CA membranes to use in filtration applications, it was important to

investigate different preparation conditions on the CA membrane morphology before adding
CNTs, so as to establish the best possible combinations that affect the presence/absence of
macrovoids, their sizes, as well as the porosity of the surrounding matrix. This was important
because macrovoids and matrix porosity play the major role in membranes’ permeability and
salt retention. The larger the voids, the higher the permeability; and the smaller the pores size
of the surrounding matrix, the better the salt retention [51]. In this respect, a number of
parameters were investigated, with the resulting membrane morphology characterized by
SEM imaging. These parameters included:
54
 Membrane casting thickness;
 The temperature at which PI was carried out, “coagulation bath temperature”;
 CA content in the solvent;
 Time for solvent evaporation;
 Non-solvent (deionized water) content in the CA/acetone mixture;
 Content of carbon nanotubes, both pNTs and fNTs in the membrane
Once these different parameters were investigated, and a set of optimum values
established, membrane morphology (surface area and porosity) and performance were
established as a function of varying content of fNTs.
Figure ‎4.4 Different preparation conditions
4.4.1 Membrane Casting Thickness
Membrane casting thickness, which is the thickness of the cast solution adjusted using the
casting knife, is a primary parameter in the development of membrane porosity and the
formation of macrovoids [49]. Membrane casting thickness was found to change during the
preparation process. ICT is the membrane thickness once it is cast prior to PI, while final
casting thickness is the membrane thickness once PI and coagulation has been carried out. In
this respect, the following trials were carried out:
A. Effect of the membrane ICT on membrane morphology, using CA in acetone solvent:
Two percentage weight values of CA in acetone were used: 15 wt% and 17 wt%. These
values were chosen based on approximately similar values in the literature [28, 52]. For the
15 wt% CA in acetone the following ICT values were used:
 600 µm
55
 700 µm
 800 µm
 For the 17 wt% CA in acetone the following ICT values were used:
 500 µm
 600 µm
 800 µm
These values were selected to scan a range of thicknesses to understand their effect on
morphology. PI was carried out at room temperature, and morphologies were investigated
using SEM.
B. Effect of the membrane ICT on membrane morphology, using CA in acetone solvent

together with a non-solvent (deionized water):
Two sets of compositions were used: 15 wt% CA in acetone with 5% non-solvent, and 17
wt% with 10% non-solvent. These values were chosen also to scan a range of thicknesses to
understand their effect on morphology.
 For the 15 wt% CA in acetone with 5% non-solvent the following ICT values were used:
 300 µm
 400 µm
 500 µm
 600 µm
 For the 17 wt% CA in acetone with 10% non-solvent the following ICT values were
used:
 700 µm
 800 µm
PI was carried out at room temperature, and morphologies were investigated using SEM.
C. Effect of the membrane ICT on membrane final casting thickness:
17 wt% CA in acetone were cast with a wide range of ICT values in order to check the
linearity of the variation of ICT with the final membrane thickness. ICT values used were:
 180 µm
 500 µm
 600 µm
 700 µm
 800 µm
 1000 µm
56
The corresponding final thickness for each membrane after PI was measured using SEM, and
a plot of ICT vs. final thickness was constructed. PI was carried out at room temperature.
4.4.2 Coagulation Bath Temperature (CBT) (PI Temperature)
CBT was investigated using two different contents of CA, 14 wt%, and 17 wt%.
Membranes were prepared at the same ICT of 1000 µm, at room temperature and 40oC. SEM
imaging was used to compare the final thicknesses and macrovoids sizes for the resulting
membranes.
4.4.3 CA Content
Different CA contents of 13, 15, and 17 wt% in acetone were investigated for effect on
membrane morphology and reproducibility:
 13 wt% CA was cast at 1000 µm and 1200 µm;

 15 wt% CA was cast at 800 µm;
 17 wt% CA was cast at 800 µm.
These values are chosen based on similar values in the literature [21, 25, 28]. PI was carried
out at room temperature, and morphologies were investigated using SEM.
4.4.4 Solvent Evaporation
Solvent evaporation is exposing the cast solution to air for some time, resulting in some
of its solvent evaporating, and is used for decreasing pore sizes throughout the membranes.
Two different sets of trials were carried out for investigating the effect of time of solvent
evaporation on membrane morphology:
A. Incomplete PI with post-PI solvent evaporation:
Incomplete PI with post-PI solvent evaporation for different timings were carried out. 17
wt% CA in acetone was used as in the literature [25]. Typically, PI takes 7.5 minutes to
complete. In this respect, two membranes were prepared as follows:
 PI for 5.5 minutes, followed by solvent evaporation for 3.5 minutes;

 PI for 2.5 minutes, followed by solvent evaporation for 3.5 minutes.
ICT was 1000 µm for both membranes and PI was carried out at room temperature.
Morphologies were investigated using SEM.
B. Initial Solvent Evaporation (ISE):
ISE entails allowing the acetone solvent to evaporate after the membrane is cast and prior
to coagulation by PI. In this respect, membranes were prepared as follows:
57
 17 wt% CA in acetone was used for a membrane for which PI was carried out at room
temperature for 7.5 minutes;
 17 wt% CA in acetone was used for a second membrane which was subjected to ISE 60
seconds, followed by PI at room temperature for 7.5 minutes;
 13 wt% CA in acetone was used for a membrane which was subjected to ISE for 30
seconds, followed by PI at room temperature for 7.5 minutes;
 13 wt% CA in acetone was used for another membrane which was subjected to ISE for
60 seconds, followed by PI at room temperature for 7.5 minutes.
ICT was 1000 µm for both membranes, and morphologies were investigated using SEM. CA,
as well as ISE values were used similar to the values used in the literature [25].
4.4.5 Addition of H2O (Non-Solvent)
The addition of a non-solvent (deionized water) to the CA/acetone solution used for
membrane casting affects membrane final thickness and morphology [52]. In this respect, the
following membranes were prepared:
 17 wt% CA in acetone was used for a set of membranes of ICT of 700 µm, and the
following contents of deionized water as non-solvent:
 0 wt%
 10 wt%
 15 wt%
 20 wt%
 25 wt%
 5 wt%
 20 wt%
 0 wt%
 5 wt%
 20 wt%
These different values of CA content, ICT, and non-solvent content were selected so as to
develop a clear idea on the effect of non-solvent addition on macrovoids formation regardless
of the CA content used. For all membranes, PI was carried out at room temperature, and
morphologies investigated using SEM.
58
4.4.6 Addition of CNTs
The effect of the addition of pristine (non-functionalized) CNTs on the morphology of the
membranes was investigated in terms of membrane compaction, and pNTs dispersion within
the polymer matrix. In this regard, the following trials were carried out:
A. Effect of presence of pNTs on membrane morphology:
Two values of CA wt% in acetone were used, with membrane ICT of 1000 µm and
different contents of pNTs, as follows:
 17 wt% CA in acetone with:

 0 wt% pNTs
 0.5wt% pNTs
 1 wt% pNTs
 13wt% CA in acetone with:

 0 wt% pNTs
 0.5wt% pNTs
These values are based on similar CNTs contents used with other polymer matrices in the
literature [39]. PI was carried out at room temperature and morphologies were investigated
using SEM.
B. Effect of the content of fNTs on their dispersion within the CA membranes:
The effect of the addition of fNTs on the morphology of membranes was investigated in
terms of membrane compaction, and fNTs dispersion within the polymer matrix. In this
respect, 15 wt% CA in acetone was used with 20 wt% water (non-solvent) and a membrane
ICT of 400 µm with different contents of fNTs as follows:
 0 wt% fNTs
 0.0005 wt% fNTs
 0.005 wt% fNTs
 0.01 wt% fNTs
Significantly low fNTs content values were selected in order to minimize CNT agglomeration
within the membrane matrix, and to ensure optimal dispersion. PI was carried out at room
temperature, and morphologies were investigated using SEM.
59
4.5 Optimal Preparation Conditions Samples
Based on the findings of the variation of the six parameters membrane samples were
prepared according to the following:
 Final membrane thickness = 100±10 µm

 Coagulation Bath Temperature = room temperature
 15% CA content in acetone
 No solvent evaporation
 20% water (non-solvent) content
 0, 0.0005, 0.005, 0.01 %wt fNTs
 For the 0 %wt fNTs, two samples were prepared: one with an ICT of 400 µm, and a
final thickness of 120±10 µm; and a second with an ICT of 350 µm and a final
thickness of 100±10 µm. Each of these samples represented a “blank”.
These samples’ morphologies were characterized by SEM, as well as nitrogen adsorption for
surface area and porosity determination, together with solution permeation and salt retention
performance. All measurements and determinations were carried out in triplicates and the
averages obtained.
60
5 Results and Discussion
61
5.1 Effect of Preparation Conditions on Membranes Morphology
Different preparation conditions have different effects on membrane morphology. Yet, for
filtration applications, the asymmetric membrane needs to have a dense top layer to screen
out salts, macrovoids to enhance permeation rate, and a porous structure throughout the rest
of the membrane [23, 26, 28, 30, 38, 51]. An illustrative figure for such morphology is
displayed in Figure 5.1.
Figure ‎5.1 SEM of an asymetric membrane
In this chapter, the results of different preparation conditions and their combinations on
membrane morphology and performance are first reported followed by results and
interpretation for membrane characterization and performance for those membranes prepared
based on the optimal preparation conditions results. The conditions investigated were:
 Membrane Casting thickness

 Coagulation Bath Temperature
 CA content
 Solvent Evaporation
 Addition of H2O (additive) in the CA/Acetone Stock Solution
 Addition of MWCNTs
62
5.1.1 Membrane Casting Thickness
A. Effect of the membrane ICT on membrane morphology, using CA in acetone solvent:
Membrane thickness variation was found to have an effect on membrane morphology:

thicker membranes exhibit larger well-developed macrovoids. It was also found that below a
certain critical thickness (whose values depends on the CA and non-solvent content),
macrovoids didn’t develop. Figure 5.2, 5.3, 5.4, and 5.5 show the effect of casting thickness
on macrovoid development. In Figures 5.2 and 5.3, membranes were prepared with 17 wt%
CA and 15wt% respectively. PI was carried out for both experiments at room temperature.
The Figures show that the decrease in membrane final thickness reduced the size and
numbers of macrovoids. Figure 5.2 (a) shows a membrane thickness small enough to be
devoid of macrovoids.
Small macrovoids
Macrovoids almost disappeared
Long macrovoids
Figure ‎5.2 SEM of 17 wt% CA membranes, PI carried out at room temperature: (a) ICT = 500
µm, final thickness = ~55 µm, (b) ICT = 600 µm, final thickness = ~75 µm, (c) ICT = 800 µm,
final thickness = ~122 µm
63
Short macrovoids
Macrovoids size increased
Long macrovoids
Figure ‎5.3 SEM of 15 wt% CA membranes, PI carried out at room temperature: (a) ICT = 600
µm, final thickness = ~79 µm, (b) ICT = 700 µm, final thickness = ~114 µm, (c) ICT = 800 µm,
final thickness = ~122 µm
B. Effect of the membrane ICT on membrane morphology, using CA in acetone solvent

together with a non-solvent (deionized water):
In Figure 5.4 and 5.5, membranes were prepared with the addition of water as non-solvent
content expected to enhance macrovoids development3. The effect of casting thickness was
investigated on the macrovoids formation and membrane shrinkage. Figure 5.4 shows typical
membrane morphology for a sample with 17 wt% CA content, 10 wt% non-solvent, and PI
carried out at room temperature. Figure 5.5 shows typical membrane morphology for a
sample with 15 wt% CA content, 5 wt% addition of non-solvent, and PI was carried out room
temperature. Both Figures 5.4 and 5.5 show the significance of casting thickness on
macrovoids formation. In Figure 5.4 (a) the macrovoids shape, size, and depth within the
membrane cross section changed when the membrane casting thickness was increased by
~100 µm only as compared to Figure 5.4 (b). In Figure 5.5 (a) and (b), membrane thickness
3
The addition of water as non-solvent on membrane morphology will be discussed in details in
section 5.1.5, while here, the ICT effect on the final thickness is the main focus
64
was not enough to allow the development of macrovoids. As for Figure 5.5 (c) and (d), the
increase in casting thickness allowed macrovoids formation.
Very large macrovoids

Short macrovoids
700 µm, final thickness = ~137 µm, (b) ICT = 800 µm, final thickness = ~226 µm
Macrovoids demolished Macrovoids demolished
Large macrovoids relative

to the final thickness
Macrovoids started developing
300 µm, final thickness = ~30 µm, (b) ICT = 400 µm, final thickness = ~44 µm, (c) ICT = 500 µm,
final thickness = ~55.4 µm, (d) ICT = 600 µm, final thickness = ~110 µm
65
From these data, the following general trend could be established:
 CA content variation did not lead to a significant variation in membrane final

thickness relative to ICT for the 15 wt% and 17 wt% CA membranes.
 The presence of non-solvent did reduce the decrease in the membrane thickness as a
result of PI with a noticeable enhancement effect on macrovoid formation.
 The maximum critical thickness for macrovoids development seemed to be about 50
µm, and was dependent on CA and nonsolvent contents as supported in the literature
[49, 52].
Table ‎5.1 Macrovoids numbers and dimensions relative to CA content, non-solvent content, and
membrane thickness
17 wt% CA membrane
0 wt% H2O 10 wt% H2O
Macrovoids Macrovoids
ICT FT Macrovoids dimensions ICT FT Macrovoids dimensions
(µm) (µm) number (length (µm) X (µm) (µm) number (length (µm) X
width (µm)) width (µm))
500 ~55 1 28 X 20 700 ~137 15 50 X 30
6 19 X 18 10 140 X 68
600 ~75 800 ~226
6 24 X 20 7 30 X 20
2 52 X 30
800 ~122
12 28 X 18
15 wt% CA membrane
0 wt% H2O 5 wt% H2O
600 ~79 12 32 X 20 300 ~30 0 0
3 52 X 24 400 ~44 0 0
700 ~114
16 34 X 18 500 ~55.4 12 30 X 16
3 80 X 40 600 ~110 8 48 X 20
800 ~122
14 36 X 20
This behavior can be explained from understanding the demixing process occurring for
the thick and thin membranes. Typically when the combination of preparation conditions are
selected to form instantaneous demixing with macrovoids formation, as elaborated in section
3.3.2, the polymer rich phase gives up its solvent into the polymer lean phase, resulting in
nucleation that continues to grow as demixing continues (instantaneous demixing). On the
other hand, a delayed demixing occurs at the interface of the nuclei since the non-solvent
becomes filled with the solvent. This expands the non-solvent droplets size concurrently with
66
the solidification of the polymer rich phase because the polymer is being depleted from its
solvent. For lower ICT, the solvent that typically increases voids sizes wouldn’t have enough
time to expand the size of the non-solvent droplets by delaying the demixing process at the
nucleus boarders. This would consequently affect the macrovoids size in relation to the ICT.
Below the critical thickness, the membrane is thin enough for instantaneous demixing to
prevail throughout the membrane.
C. Effect of the membrane ICT on membrane final casting thickness:
Investigating the linearity between ICT and the corresponding final thickness (FT) was
done for the 17 wt% CA membrane with ICT range of 180-1000 µm. The relation between
the ICT and the FT is presented in Figure 5.6. It was non-linear showing a rising increase in
FT with ICT increase.
Initial Casting Thickness vs. Final Thickness for 17

wt% CA
200
Final thickness (µm)
150
100
50
0
0 200 400 600 800 1000 1200
Initial casting thickness (µm)
Figure ‎5.6 A plot showing the non-linear change in final thickness of 17 wt% CA membranes
with the increase in ICT. PI was done at room temperature
The change in the final thickness as ICT changes is expected since during PI, the cast
CA/acetone solution exchanges the acetone with water as non-solvent, and the CA starts
shrinking until the polymer rich layer is depleted from the solvent and CA precipitates. The
variation between ICT and FT was found to be non-linear as displayed in the representation
in Figure 5.6. This could be explained by the fact that as the membrane final thickness
increase, the macrovoids formation is enhanced, and their shape changes, where the thicker
the membrane, the longer drop-like the macrovoids become. Thus, it is fair to assume that
macrovoids formation affects the final thickness of the membrane in such way that it takes
smaller than predicted ICTs to reach a desired final thickness.
5.1.2 Coagulation Bath Temperature (CBT)
The results of CBT effect on membrane morphology using 17 wt% and 14 wt% CA
contents are displayed in Figures 5.7 and 5.8 respectively. In Figure 5.7, two 17 wt% CA
67
membranes were cast at the same ICT (1000 µm), and at two different CBT temperatures
(room temperature and 40oC). Figure 5.7 (a) shows the overall membrane morphology with
different macrovoid sizes that ranged from small to large. The large ones are distributed away
from one another, separated by the small ones. Figure 5.7 (b) shows the top layer (high
magnification) with the macrovoids necks developed at about 10 µm from membrane surface.
Figure 5.7 (c) shows the 17 wt% membrane prepared using PI at 40oC, where the macrovoid
sizes changed, producing medium size macrovoids more closely packed together. Figure 5.7
(d) shows the macrovoids necks developed at a much smaller distance (<1 µm) from the
membrane surface.
Small macrovoids Macrovoids necks separated by

porous polymer skin
Large macrovoids
Small macrovoids
Medium macrovoids
temperature, (b) Voids necks at high magnification for PI at room temperature (c) FT = ~172
µm, PI at 40oC, (d) Voids necks at high magnification for PI at 40oC
Membranes obtained from 14 wt% CA, cast at the same ICT (1000 µm), and at the same
CBT temperatures (room temperature and 40oC) showed a different trend for the sizes and
shapes of the macrovoids. Figure 5.8 (a) shows large macrovoids separated from each other
by smaller macrovoids (similar to Figure 5.7 (a)), yet, the voids necks (Figure 5.8 (b)) are at a
shorter distance (about 3 µm) from the membrane surface. For PI at 40oC, mostly medium
68
macrovoids developed, with occasional large ones (Figure 5.8 (c)). The macrovoids necks
were approximately located close to the membrane surface (about 3 µm) as in the room
temperature PI 14 wt% CA samples (Figure 5.8 (d)).
Small macrovoids Macrovoids necks separated

by thin porous polymer skin
Large macrovoids
Large macrovoid Macrovoids necks separated by

thick porous polymer skin
Medium macrovoid
temperature, (b) Voids necks at a high magnification, PI at room temperature, (c) FT = ~138
µm, PI at 40oC , (d) Voids necks at a high magnification, PI at 40oC
69
Table ‎5.2 Macrovoids numbers and dimensions relative to CA content and PI temperature
17 wt% CA membrane
PI at room temperature PI at 40°C
(µm) (µm) number (length (µm) (µm) (µm) number (length (µm) X
X width (µm)) width (µm))
7 140 X 100 8 120 X 40
1000 ~180 1000 ~172
23 76 X 40 24 100 X 40
14 wt% CA membrane
PI at room temperature PI at 40°C
(µm) (µm) number (length (µm) (µm) (µm) number (length (µm) X
X width (µm)) width (µm))
3 60 X 40 1 55 X 50
1000 ~138 1000 ~138
22 40 X 20 14 80 X 25
The results indicate that generally, higher PI temperature reduced the size distribution of
macrovoids, allowing them to form close to the membrane surface, and increased the sizes of
the small macrovoids formed. This latter effect seems to be somewhat impacted by the CA
content, with the higher CA content membranes showing a more pronounced effect.
Increasing the temperature leads to enhancing the demixing process, which would lead to
rapid membrane shrinking and solidification, promoting the formation of macrovoids [28,
30]. For the higher CA content of 17 wt%, an increased coagulation temperature resulted in a
decrease of the size of larger macrovoids. This decrease was compensated by an increase in
the size of smaller macrovoids as displayed in Table 5.2. For the lower CA content of 14
wt%, the increase in the size of the smaller macrovoids at high coagulation temperature was
more obvious. Though the overall number of the macrovoids (both small and large) for 14
wt% membranes prepared at 40oC is smaller than those prepared at room temperature,
calculating their overall dimensions roughly (length X width X macrovoids number) shows
that membranes prepared at 40oC have an overall macrovoids dimension of 30,750 µm2 vs.
only 24,800 µm2 for macrovoids of membranes prepared at room temperature.
5.1.3 CA Content
Results of the 13 wt% CA membranes showed that reproducibility was found to be a

major problem for these membranes where the same ICT at the same PI temperature
produced different final thicknesses. Figures 5.9 and 5.10 show two samples prepared at the
same temperature, one with ICT of 1200 µm, and the second with ICT of 1000 µm
respectively. Figure 5.9 shows three different locations on the 1200 µm membrane with three
70
different final thicknesses, ~173 µm, ~186 µm, and ~217 µm. Figure 5.10 shows two
different locations on the 1000 µm membrane with two different final thicknesses, 155 µm
and 185 µm. This is believed to be due to the low viscosity of the cast solution. When this is
immersed in the coagulation bath, the lower viscosity leads to more intensive demixing
process, probably due to the presence of high solvent content. This is believed to have lead to
a rippling effect during the formation of the macrovoids, leading to the variation in the
thickness. Although this has not been reported in the literature, it seems to be an experimental
artifact.
Figure ‎5.9 SEM of three 13 wt% CA membranes prepared with ICT = 1200 µm: (a) final
thickness = ~173 µm, (b) final thickness = ~186 µm, (c) final thickness = ~217 µm
71
Figure ‎5.10 SEM of 13 wt% CA membranes, ICT = 1000 µm: (a) final thickness= ~155 µm, (b)
final thickness= ~185 µm
17 wt% and 15 wt% CA content gave reproducible final thicknesses. Using the same ICT
of 800 µm, the two membranes gave comparable final thicknesses of approximately 80 µm.
Figure 5.11 shows the morphology of both. The change in size of the macrovoids could be
easily spotted where the 17 wt% CA membrane (Figure 5.11 (a)) has smaller macrovoids
than the 15 wt% CA membrane (Figure 5.11 (b))
Small macrovoids
Large macrovoids
Figure ‎5.11 SEM of membranes prepared with PI at room temperature, ICT = 800 µm, final
thickness of ~122 µm: (a) 17 wt% CA, (b) 15 wt% CA
72
Table ‎5.3 Macrovoids numbers and dimensions relative to CA content
Cellulose Acetate Content

17 wt% CA 15 wt% CA
2 52 X 30 3 80 X 40
800 ~122 800 ~122
12 28 X 18 14 36 X 20
From Figure 5.11 and Table 5.3, it is clear that at the same final thickness, the smaller the
CA content, the larger the size of the macrovoids. This is expected and matches the
observations in the literature [51-52, 54]. At the same ICT, the low CA content is
compensated by high solvent volume, and as earlier explained, the higher the solvent content,
the more delayed the demixing process on the borders of the growing pores until they become
macrovoids. This takes place at the same time as the rest of the membrane is shrinking and
solidifying, thus creating macrovoids, as explained in details in section 3.3.2.
5.1.4 Solvent Evaporation
A. Incomplete PI with post-PI solvent evaporation:
Solvent evaporation prior to complete PI is not a usual method. Figure 5.12 clearly shows
its effect on membranes morphologies and final thicknesses, for membranes with 17 wt% CA
membranes, ICT = 1000 µm and PI at room temperature. A membrane prepared with
complete PI for 7.5 min had a final thickness = ~180 µm, with the typical drop-like large
macrovoids (Figure 5.12 (a)). Figure 5.12 (b) shows a membrane prepared with PI for 5.5
min, followed by solvent evaporation for 3.5 min. The final thickness of the membrane was
reduced to be ~138 µm, while the shape of the macrovoids wasn’t affected much in light of
the final thickness of the membrane. Figure 5.12 (c) shows a membrane prepared with PI for
2.5 min, followed by solvent evaporation for 3.5 min. The final thickness was reduced to ~
135 µm (which is very close to that of Figure 5.12 (b)), however, the shape and size of the
macrovoids had a significant change: they shrunk into small macrovoids, with a
disappearance of the large macrovoids.
73
Figure ‎5.12 SEM of 17 wt% CA membranes: (a) complete PI for 7.5 min, final thickness = ~180
µm, (b) PI for 5.5 min, followed by solvent evaporation for 3.5 min, final thickness = ~138 µm,
(c) PI for 2.5 min followed by solvent evaporation for 3.5 min, final thickness = ~ 135 µm
B. Initial Solvent Evaporation (ISE):
Figure 5.13 shows a comparison between two 17 wt% CA membranes, prepared with ICT
of 1000 µm, and complete PI at room temperature for 7.5 minutes. The first membrane was
prepared without ISE, while the second was subjected to 60 seconds ISE prior to PI. From the
figure, the thickness of the latter membrane final thickness was half that of the former (180
µm vs. 77 µm) (Figure 5.13 (a) vs. Figure 5.13 (b)). At the same time, the macrovoid sizes
changed significantly, where those subjected to ISE for 60 seconds were much smaller than
the normally prepared (Figure 5.13 (b)).
74
Figure ‎5.13 SEM of 17 wt% CA membranes, PI at room temperature for 7.5 minutes: (a) ISE =
0 sec, final thickness = ~180 µm, (b) ISE = 60 sec, final thickness = ~77 µm
Table ‎5.4 Macrovoids numbers and dimensions relative to solvent evaporation
17 wt% CA membrane
post-PI solvent ICT FT Macrovoids Macrovoids dimensions
evaporation (µm) (µm) number (length (µm) X width (µm))
4 104 X 50
after zero minutes 1000 ~180
10 40 X 26
5 98 X 42
after 3.5 minutes 1000 ~138
10 34 X 20
after 2.5 minutes 1000 ~135 20 30 X 16
17 wt% CA membrane
ICT FT Macrovoids Macrovoids dimensions
ISE
(µm) (µm) number (length (µm) X width (µm))
4 104 X 50
Zero seconds 1000 ~180
10 40 X 26
60 seconds 1000 ~77 14 26 X 20
Figures 5.12 and 5.13, and Table 5.4 show that solvent evaporation, whether done before
PI or after incomplete PI had a significant effect on macrovoids development and size, as
well as on membrane final thickness. Solvent evaporation after incomplete PI was not
recorded in the literature, while ISE is commonly used to condense the nascent top layer to
enhance the membranes’ retention ability [24, 32, 36]. However, none of the reported work in
the literature had such a dramatic shrinkage in macrovoids. Going back to the solvent types
used in the reported work, DMF, acetone mixed with dioxane, and acetone mixed with acetic
acid were used, which are all less volatile than acetone alone. Moreover, different additives
75
were used to enhance pore formation and inorganic particles to produce the desired
membrane morphology, leading to different results.
As for why macrovoid size was reduced because of solvent evaporation, as mentioned
before, macrovoid formation requires enough solvent to fill in the developing nuclei filled
with the non-solvent in the polymer lean phase, thus expanding the nuclei sizes and creating
macrovoids. For solvent evaporation prior to complete PI, the solvent has not been
completely exchanged with the non-solvent, and the polymer rich phase has not completely
solidified. Thus, instead of filling in the non-solvent droplets, the solvent evaporates since the
developing membrane has been removed from the coagulation bath, and PI process is
interrupted, leading to smaller macrovoids for shorter PI times. For ISE for 60 seconds,
acetone, being highly volatile solvent is thought to have evaporated with a quantity large
enough that what was left in the cast solution wasn’t enough to expand the macrovoids.
In an attempt to produce uniform membranes using 13 wt% CA content, ISE was carried
out for such membrane. Figure 5.14 shows the morphology of two 13 wt% CA membranes
prepared with ISE 30, and 60 seconds, ICT of 1000 µm, and PI at room temperature. Figure
5.14 (a) and (b) are for the same membrane at two different locations that still showed non
uniform final thickness after ISE for 30 seconds. Figure 5.14 (c) shows a uniform membrane
after ISE for 60 seconds, however, the ISE reduced the macrovoids sizes significantly, as
well as the overall membrane final thickness to reach ~123.5 µm. In this respect, it seems that
30 seconds ISE was not enough to solidify the membrane (increasing its viscosity) so that the
“rippling” effect doesn’t result in varying the membrane thickness. On the other hand, the 60
seconds of ISE seemed to lead to an increase in membrane viscosity significant enough to
overcome this effect. However, this came at the cost of a small volume of the solvent
remaining leading to the reduction of the macrovoids.
76
Short macrovoids
Figure ‎5.14 SEM of 13 wt% CA membranes, ICT = 1000 µm, and PI at room temperature: (a)
and (b) are two different locations on the same sample with ISE = 30 sec, final thickness range =
217-161 µm, (c) uniform sample prepared with ISE = 60 sec., final thickness = 123.5 µm
5.1.5 Addition of H2O (Non-Solvent)
The results of the effect of added H2O as non-solvent on membrane morphology are
displayed in Figures 5.15, 5.16, and 5.17. In Figure 5.15, 17 wt% CA membranes prepared
with ICT of 700 µm, PI carried out at room temperature, and different H2O non-solvent
contents: 0 wt%, 10 wt%, 15 wt%, and 20 wt% (Figures 5.15 (a), (b), (c), and (d)
respectively) showed a clear change in membrane final thickness and macrovoids shapes and
sizes. In Figure 5.15 (a), the macrovoids sizes were small and the membrane final thickness
was ~80 µm. As water % increased by 5 wt%, the membrane final thickness increased to
~137 µm and medium size macrovoids started developing with a finger like shape as shown
in Figure 5.15 (b). Further increase in water content to 15 wt% increased the final thickness
further to reach ~153 µm and large drop-like macrovoids developed as shown in Figure 5.15
(c). With 20 wt% water addition, the membrane final thickness increased to reach ~246 µm
and the shape of the macrovoids differed, where some had the typical drop-like shape, while
others had a spherical like shape as shown in Figure 5.15 (d).
77
It is noteworthy that the further increase in the water content to 25 wt% failed to give a
homogenous CA stock solution since the CA wasn’t able to dissolve in the presence of such
high percentage of non-solvent additive to CA-acetone solution.
Finger like medium macrovoids

Small macrovoids
Large drop-like macrovoids
Large drop-like macrovoids

Large spherical like macrovoids
Figure ‎5.15 SEM of 17 wt% CA, PI at room temperature, ICT = 700 µm: (a) 0 wt% H2O, final
thickness = ~80 µm, (b) 10 wt% H2O, final thickness = ~137 µm, (c) 15 wt% H2O, final thickness
= ~153 µm, (d) 20 wt% H2O, final thickness = ~246 µm
The results of the second set entailing 15 wt% CA membranes, ICT of 300 µm, PI at
room temperature, and H2O contents of 5 wt%, and 20 wt% showed a similar trend. The
increase in water content generally enhanced macrovoids formation. Figure 5.16 (a) shows
that though 5 wt% H2O wasn’t able to promote macrovoids formation in the small final
thickness of the membrane (~30 µm). The increase in water content to 20 wt% resulted in the
development of large macrovoids drop-like in shape, and a membrane final thickness increase
to ~80 µm (Figure 5.16 (b)).
78
thickness = ~30 µm, (b) 20 wt% H2O, final thickness = ~80 µm
The results of the third set entailing 14 wt% CA content, ICT of 1000 µm, PI at room
temperature, and H2O water contents of 0 wt%, 5 wt%, and 20 wt% are displayed in Figure
5.17 (a), (b), and (c) respectively. In Figure 5.17 (a), the final thickness of the membrane was
~138 µm with large drop-like macrovoids, as well as small ones. When 5 w% water was
added, the final thickness of the membrane increased to be ~199 µm and the size and number
of the large macrovoids increased as well as shown in Figure 5.17 (b). With the further
increase in water content to 10 wt%, the final thickness of the membrane increased more to
~264 µm, associated with the increase in macrovoids size as shown in Figure 5.17 (c).
79
thickness = ~138 µm, (b) 5 wt% H2O, final thickness = ~199 µm, (c) 10 wt% H2O, final thickness
= ~264 µm (non-uniform)
Table 5.5 summarizes the macrovoids dimensions of Figures 5.15, 5.16, and 5.17. The
results indicate a general increase in macrovoids formation with increased water content (as
non-solvent) in membrane solution. This is also shown to lead to increase in membranes final
thickness.
80
Table ‎5.5 Macrovoids numbers and dimensions relative to non-solvent addition
17 wt% CA membrane
Non-solvent ICT FT Macrovoids Macrovoids dimensions

addition (H₂O) (µm) (µm) number (length (µm) X width (µm))
0 wt% 700 ~80 15 20 X 15

10 wt% 700 ~137 15 60 X 26
2 74 X 70
15 wt% 700 ~153
28 24 X 20
20 wt% 700 ~256 8 100 X 96
15 wt% CA membrane
5 wt% 300 ~30 0 0
20 wt% 300 ~80 7 44 X 38
14 wt% CA membrane
3 60 X 40
0 wt% 1000 ~138
22 40 X 20
5 98 X 36
5 wt% 1000 ~199
14 60 X 32
6 120 X 56
10 wt% 1000 ~264
16 50 X 30
The increase in macrovoids size due to the addition of a non-solvent to the casting
solution had been reported in the literature [21, 23, 25, 52]. Macrovoids size increase with the
content of non-solvent added to the casting solution was explained by Smolders et al. [52]. It
is believed that this is the result of a local induced nucleation under the dense layer once the
cast solution is immersed in the non-solvent bath. This is due to the presence of the solvent in
high contents in this area, thus delaying the liquid-liquid demixing, and expanding the nuclei
sizes. In other words, when the cast solution is immersed in the non-solvent bath,
instantaneous demixing happens and the solvent heads upwards into the non-solvent bath.
This increases, the solvent’s, concentration under the nascent dense skin. At the same time,
there exists a quantity of non-solvent in the cast solution that started nucleation without
waiting for the non-solvent of the coagulation bath to enter and replace the solvent present.
At that point, a local induced nucleation takes place, which is accompanied by delayed
demixing at the boarders of the nuclei under the effect of the increased solvent concentration
below the top layer. This would give enough time for the solvent to enter into the non-solvent
81
droplets, and expand their size, and this time would be longer than the usual time taken for
this process to complete without the presence of the non-solvent in the cast solution.
5.1.6 Addition of CNTs
A. Effect of presence of pNTs on membrane morphology:
Pristine CNTs had a compacting effect on CA membranes. Figure 5.18 and 5.19 show the
effect of compaction on the middle sections4 of membranes prepared with 13 wt% CA
content in acetone. Figure 5.18 shows the morphological difference between 0 wt% pNT
membrane (Figures 5.18 (a) and (b) at two different magnifications), as compared to a 0.5
wt% pNT membrane (Figures 5.18 (c) and (d)). The 0.5 wt% pNT was found to have denser
polymer areas between pores and smaller pore sizes.
Smaller pores Dense areas
membranes, PI at room temperature: (a) 0 wt% pNTs at 5KX, (b) 0 wt% pNTs at 10KX, (c) 0.5
wt% pNTs at 5KX, (d) 0.5 wt% pNTs at 10KX
4
The location of the middle section of a membrane was illustrated in Figure 5.1.
82
Further investigation was carried out using 17 wt% CA content, which showed the same
effect. A 0.5 wt% pNTs lead to a denser membrane with smaller pore sizes (Figure 5.19).
Smaller pores
Dense areas
membranes, PI at room temperature: (a) 0 wt% pNTs at 5KX, (b) 0.5 wt% pNTs at 5KX
As for the effect of pNTs content, as expected, lower pNT content lead to better
dispersion as seen in Figure 5.20, where larger pNT agglomerates were clear for the higher,
1wt% content (1 agglomerate of about 9 µm width) when compared to 0.5 wt% pNT content
(3 agglomerates of about 2 µm width). When comparing the middle sections of these
membranes together with 0 wt% pNTs, there were areas in the middle sections with no
significant difference between all three membranes, which is indicative that the pNTs were
not properly dispersed within the whole membrane structure (Figure 5.21).
Small agglomerates
Large agglomerate
Figure ‎5.20 SEM of agglomerates of pNTs in 17 wt% CA membranes at 25KX in the middle
section of the membranes: (a) 0.5 wt% pNTs at 25KX, (b) 1 wt% pNTs at 25KX
83
Dense CA
polymer
Figure ‎5.21 SEM of the pores structures in the middle section of 17 wt% CA membranes at
25KX: (a) 0 wt% pNTs membrane, (b) 0.5 wt% pNTs membrane, (c) 1 wt% pNTs membrane
These results show that it was very challenging to disperse pNTs within CA membranes
due to the pNTs highly hydrophobic nature [62]. The denser pNTs membrane structure is due
to the space occupied by the pNTs that resulted in high compactness of the membrane as
explained in the literature [40]. Furthermore, the quantity of the nanotubes that were able to
disperse within the CA solution could lead to delayed demixing process. This is because the
cast solution would have a hydrophobic source, the pNTs, which would slow down the
penetration of the polar non-solvent (water) from the coagulation bath to start nucleation and
formation of the polymer lean phase, thus, slowing down the process and reducing the pore
sizes.
B. Functionalization of CNTs:
In order to enhance the dispersion of CNTs within the CA matrix, functionalization via
oxidation purification in acidic medium was carried out. SEM, as well as FTIR were used to
verify the success of functionalization. Below are the comparative SEM images for both
pNTs and fNTs.
84
Figure ‎5.22 SEM image of both pNTs and fNTs: (a) an agglomerate of pNTs at 700X, (b) a
bundle of pNTs at 100 KX, (c) agglomerates of fNTs at a small magnification (700X), (d) arrays
of fNTs at 100 KX
The SEM images show a significant change in the shape of the agglomerates where the
pNT agglomerates appear as a “thread bundles” with individual CNTs clearly discernable
(Figure 5.22 (a), and (b)). On the other hand, fNTs agglomerates are particle like with the
CNTs more orderly aligned (Figure 5.22 (c) and (d)). This is due to the strong interaction
between the COOH groups attached to the CNTs as a result of functionalization.
The FTIR spectra of both CNTs are presented in Figures 5.23 and 5.24, with noticeable
differences.
85
3450.0
1458.0
1650.8
1540.9
Figure ‎5.23 FTIR spectrum of non-functionalized MWCNTs
1701.2
1458.0
3450.0
1650.8
1081.9
1559.0
1419.8
2918.1
Figure ‎5.24 FTIR spectrum of functionalized MWCNTs
Functionalization of CNTs showed a strong absorption band at 3450 cm-1 associated with
the presence of OH groups (stretching mode) on the surface of the nanotubes from two
sources, the COOH group, as well as adsorbing humidity. The appearance of an absorption
peak at 2918.1 cm-1 that didn’t show for the pNT reflects the single bonds of the OH groups
(stretching mode) of the COOH groups. The peak at 1701 cm-1 is probably due to C=O
groups (stretching mode) of COOH, or CO of ketones/quinines if present. Strong absorption
peaks from 1650 cm-1 to 1450 cm-1 are usually characteristic for C=C of aromatic rings
(stretching mode), or in this case, aromatic like structure. These peaks appeared for both pNT
86
and fNTs, reflecting the C-C bonding structure of the nanotubes. The absorption peak at 1419
cm-1 is for single bond of C-O of the C-OH (bending mode), associated with a strong peak at
1081 cm-1 corresponds to the stretching mode of the same bond. As for the broad absorption
peak in Figure 5.23 of pNTs at 3450 cm-1, it corresponds to the stretching modes of OH the
water molecule reflecting the presence of humidity within the sample [96-97]. It is important
to note that Stobinski et al. [96] reported FTIR spectra for functionalized CNTs, identical to
the one reported here. Their CNTs were also functionalized using oxidation purification
under similar conditions. These results show that functionalization of CNTs was successful.
Dispersion trials were carried out for both fNTs and pNTs in water. Figure 5.25 shows the
difference between two dispersions of 0.1 grams of fNTs and pNTs in 20 ml water. From the
figure, it is clear that the fNTs were fully dispersed in the medium (Figure 5.25 (a)), however,
the pNTs failed to do so due to their highly hydrophobic nature, where agglomerates of pNTs
could be seen at the bottom of the beaker (Figure 5.25 (b)).
Figure ‎5.25 Beakers filled with 0.01 g CNTs, sonication time for 1 minute in 20 ml deionized
water, (a) fNTs fully dispersed giving opaque black solution, (b) pNTs poorly dispersed forming
agglomerates throughout the solution and on the bottom of the beaker
On testing fNTs dispersion in acetone vs. water, before mixing the dispersed nanotubes
solution with the CA stock solution, it was found that fNTs disperse in water much more than
in acetone, probably due to the high polarity nature of the latter, which is compatible with the
functional groups located on the surface of the functionalized nanotubes. This lead to the
usage of lower weight percentage of fNTs to develop the CA based nanocomposites, and
minimal sonication time (less than 1 minute) to achieve good dispersion.
C. Effect of the content of fNTs on their dispersion within the CA membranes:
fNTs orientation inside the CA matrix was investigated using SEM. Large networks fNTs
were easily spotted for the 0.005 and 0.01 wt% nanotubes contents. For the 0.0005 wt%
fNTs, individual nanotubes were imaged. Figures 5.26, 5.27, and 5.28 show the random
distribution and orientations of fNTs within the CA matrices of 0.0005 wt%, 0.005 wt% and
0.01 wt% fNTs respectively. The fNTs are highlighted with the red arrows.
87
SEM magnifications, (a) at 25KX, (b) at 50KX

SEM magnifications, (a) at 50KX, (b) at different location at 50KX, (c) at 100KX, (d) at
different location at 100KX
88
SEM magnifications, (a) at 25 KX, (b) at 50KX, (c) at 100KX, (d) at 300KX
The images show that fNTs were randomly oriented and properly dispersed within the
CA membranes, creating large networks that extended across the membranes, and no
agglomerates were detected.
Further investigations on the effect of fNTs on morphology included studying their effect
on macrovoids formation. SEM images in Figure 5.29 showed that the addition of fNTs to the
CA membranes resulted in a significant reduction in size and number of large macrovoids
(for the same final membrane thicknesses) associated with the increase in fNTs contents
forming 0/15/20 wt%, 0.0005/15/20 wt%, 0.005/15/20 wt%, and 0.01/15/20 wt%
fNT/CA/H2O nanocomposites (Figure 5.29 (a), (b), (c), and (d) respectively). Table 5.6
displays the macrovoids numbers and dimensions relative to the fNTs quantities used.
89
Single macrovoid
Small size macrovoids
Figure ‎5.29 SEM of morphology at same final thickness (100±10 µm): (a) 0/15/20 wt%
fNT/CA/H2O, (b) 0.0005/15/20 wt% fNT/CA/H2O, (c) 0.005/15/20 wt% fNT/CA/H2O, (d)
0.01/15/20 wt% fNT/CA/H2O
Table ‎5.6 Macrovoids numbers and dimensions relative to fNTs addition
15/20 wt% CA/H₂O membrane

fNTs ICT FT Macrovoids Macrovoids dimensions
addition (µm) (µm) number (length (µm) X width (µm))
0 wt% 350 100±10 4 46 X 38

0.0005 wt% 400 100±10 4 56 X 32
0.005 wt% 400 100±10 3 46 X 30
0.01 wt% 400 100±10 1 50 X 30
As discussed earlier, the hydrophobic nature of the pNTs decrease its interaction with the
polymer matrices. This changed under the effect of functionalization. This is explained in the
literature [38-41] to be due to the enhanced interfacial interaction between the functional
groups on the fNTs surfaces with the polymer matrices. In the fNT-CA nanocomposite, such
90
interactions are strongly suggested to be between the OH of the COOH groups on the fNTs
graphene sheets and the carbonyl groups on the ester linkages of the CA polymer [98]. These
interactions are thought to be hydrogen bonds.
The increase in the weight percentage of the fNTs suppressed macrovoids formation. A
possible explanation is thought to be related to a delay in the overall solvent-non-solvent
instantaneous demixing taking place due to the interference of the fNTs in the process.
Macrovoids formation in the presence of a percentage of a non-solvent in the cast solution is
the result of a local induced nucleation that gets expanded by the solvent in the polymer lean
phase below the nascent top layer. In the regular demixing process with the presence of a
non-solvent (let’s call it NS1) in the cast solution, NS1 is responsible for the local induced
nucleation that expands the macrovoids as explained before. The non-solvent in the
coagulation bath (let’s call it NS2) is responsible for the regular nucleation process via
demixing with the solvent and the formation of the polymer rich and polymer lean phases.
Now, the new intruders, the fNTs, have COOH functional groups on their walls, and they are
well dispersed into the cast solution of CA-acetone-NS1. Such functional groups can easily
develop hydrogen bonds with NS1, and the entering NS2 from the coagulation bath. The
hydrogen bonds that probably developed between the nanotubes and NS1 could delay the
induced local nucleation process because NS1 remained trapped in the CA-acetone-water cast
solution, and not separating as fast as it regularly would. This starts changing as NS2 enters
into the cast solution. Regularly, NS2 forces the solvent out to develop two separate phase, a
polymer lean, and polymer rich. At this point, the tendency of the NS1 to remain mixed with
the developing polymer rich phase decreases as the polymer solidifies. Thus, it gets expelled
into the polymer lean developing areas. This doesn’t necessarily mean that it wouldn’t keep
the hydrogen bonds with the functional groups on the graphene walls, yet, these bonds could
be responsible for the delayed initiation of the local nucleation process. A second possibility
entails a contribution of hydrogen bonds between fNTs and NS2 to delayed demixing process,
and nucleation, as the entering NS2 becomes restrained from its regular free motion via such
bonds, thus slowing down the entire process.
5.2 Optimal Preparation Conditions
Based on the above results, the optimal conditions for the preparation of samples to be
fully characterized for surface area and porosity, as well as performance for solution
permeation and salt retention, were identified as listed in section 4.5. The addition of fNTs to
CA membranes has lead to changes in membrane morphology, observed in the number and
dimensions of macrovoids, as well as membrane compaction. In addition it has lead to a
change in membrane surface area, porosity and performance.
In determining the effect of functionalized CNTs addition, two blank membranes were
used. The first blank entailed a membrane cast at a different ICT value, but which had the
same final thickness of 100 µm as the nanocomposite membranes containing CNTs. This
blank helped compare the effect of CNT addition in light of a constant final membrane
thickness. The blank did however present the shortcoming of having a different absolute
91
amount of CA (as a result of the different ICT values). In this respect, the second blank
entailed a membrane having the same ICT value of 400 µm as the nanocomposite membranes
with CNTs. This however had a different final thickness value, but contained the same
absolute amount of CA.
5.3 Membrane Surface Properties
The results of investigating the pore size distribution for the nanocomposites in
comparison to CA membranes included differential pore volumes vs. pore width, differential
pore areas vs. pore width, adsorption isotherms, t-plots, micropores volume calculations,
external surface areas calculations using t-plot method and BET surface area.
5.3.1 Differential Pore Sizes
The results for fNTs nanocomposites with the same final thickness 100±10 µm in
comparison to the two blanks included membranes with compositions (fNT/CA/H2O):
0/15/20 wt% (ICT = 350 µm), 0/15/20 wt% (ICT = 400 µm), 0.0005/15/20 wt%, 0.005/15/20
wt%, and 0.01/15/20 wt% membranes. Porosity results are shown in Figures 5.30, 5.31, and
5.32.
Differential Pore Volume vs. Pore Width
0/15/20 wt%
0.06 (FT=100±10µm)
Differential Pore Volume (cm³/g)
0/15/20 wt%
0.05 (FT=120±10µm)
0.0005/15/20 wt%
0.04
0.005/15/20 wt%
0.03
0.01/15/20 wt%
0.02
0.01
0
0 50 100 150 200 250
Pore Width (Nanometers)
Figure ‎5.30 Plot of differential pore volumes
92
(Display of differential volume for micro and mesopores)
0.06 0/15/20 wt%
(FT=100±10µm)
0.05 0/15/20 wt%

(FT=120±10µm)
0.04
0.0005/15/20 wt%
0.03
0.005/15/20 wt%
0.02
0.01/15/20 wt%
0.01
0
0 5 10 15 20 25 30 35 40 45 50 55
Figure ‎5.31 Plot of differential pore volumes for micro and mesopores

(Display of differential volume for macropores)
0.06 0/15/20 wt%
(FT=100±10µm)
0.05 0/15/20 wt%
(FT=120±10µm)
0.04 0.0005/15/20 wt%
0.005/15/20 wt%
0.03
0.01/15/20 wt%
0.02
0.01
0
50 75 100 125 150 175 200 225 250
Figure ‎5.32 Plot of differential pore volumes for macropores
From the plots above, the largest volumes occupied by pores are for those larger than ~20
nm width. Figure 5.31 shows that only the 0/15/20 wt% and 0.0005/15/20 wt% fNT/CA/H2O
had pores of 5nm width. The same figure shows that all membranes had micropores with
width less than 2 nm, however, the amount present varied significantly from one membrane
to another.
Figures 5.31 and 5.32 show that the 0.01/15/20 wt% fNT/CA/H2O membrane had the
least volume of pores which means the increase content of fNTs suppressed pores
93
development. The two figures also show that the pore volumes in the 0/15/20 wt%, the
0.0005/15/20 wt%, and the 0.005/15/20 wt% fNT/CA/H2O are almost neck to neck in the
range from 5nm to 70 nm. However, for pores larger than 70 nm, the 0/15/20 wt%
fNT/CA/H2O (FT = 100±10 µm) had the largest volume occupied by the 117 nm macrovoids.
Finally, the plot shows that the 0.0005/15/20 wt% fNT/CA/H2O membranes had macrovoids
larger than 180 nm wide.
Figures 5.33, 5.34, and 5.35 show the corresponding pore area distribution.
Differential Pore Area vs. Pore Width

9
0/15/20 wt%
8
(FT=100±10µm)
7 0/15/20 wt%
Differential Surface Area (m²/g)
(FT=120±10µm)
6
0.0005/15/20 wt%
5
0.005/15/20 wt%
4
3 0.01/15/20 wt%
2
1
0
0 50 100 150 200 250
Figure ‎5.33 Plot of differential pore areas

(Display of differential area for micro and mesopores)
9
0/15/20 wt%
8 (FT=100±10µm)
7 0/15/20 wt%
(FT=120±10µm)
6
0.0005/15/20 wt%
5
4 0.005/15/20 wt%
3 0.01/15/20 wt%
2
1
0
0 5 10 15 20 25 30 35 40 45 50 55
Figure ‎5.34 Plot of differential pore areas for micro and mesopores
94
(Display of differential area for macropores)
9
8 0/15/20 wt%
(FT=100±10µm)
7
0/15/20 wt%
6 (FT=120±10µm)
5 0.0005/15/20 wt%
4
0.005/15/20 wt%
3
2 0.01/15/20 wt%
1
0
50 75 100 125 150 175 200 225 250
Figure ‎5.35 Plot of differential pore areas for macropores
On comparing the plots of pore width (pore diameter) vs. pore areas and vs. pore
volumes, the results showed that that the membranes featured micropores at 1.7 nm and very
small mesopores at 2.7 nm. The total surface area of these pores is significant. It reached
maximum values of about 8.3 m2/g and 1.7 m2/g respectively for membranes with no fNT
content, as well as membranes with low fNT content (0.0005 wt%), and with the lowest
values of about 3.5 m2/g and 0.12 m2/g respectively for membranes of highest fNT content
(0.01 wt%). At the same time, these pores exhibited the smallest volumes (less than 0.008
cm3/g), only possible if their number was significant in comparison to larger mesopores (6-50
nm) and macropores (> 50 nm) in the membranes. On the other hand, the meso and
macropores exhibited very low surface area values (less than 0.9 m2/g), with significant
volumes reaching values of about 0.051 cm3/g. They generally fell with size values of about
18.5 nm, 25 nm, 40 nm, 54 nm, 68 nm, 93nm, 117 nm, 147 nm, and 185 nm with the pores of
117 nm being the most abundant.
To clarify the idea of small volume large surface area that the data showed, a calculation
could be formulated for a membrane with two different sets of spherical pores with diameters
of ~2 nm and ~100 nm, for example. Assuming that the volume of the ~2 nm pores was
approximately half that of the ~100 nm pores, similar to the case of the 0.0005/15/20 wt%
fNT/CA/H2O membranes, a calculation of the ratio of the areas derived from the volumes for
the two sets of pores would be as follows:
V1 is the total volume of the ~100 nm pores, r1 is the radius (half the
given width for spherical pores), and n1 is the total number of the ~100 nm pores
V2 is the total volume of the ~2 nm pores, r2 is the radius (half the

given width for spherical pores), and n2 is the total number of the ~2 nm pores
95
Assuming that the pore width is the diameter of the spherical pores:
Now to calculate the ratio between the areas of the two sets of pores:
From this calculation, the ~2 nm pores had ~25 times more area than the ~100 nm pores
even though the latter occupied twice as much volume. However, the large numbers of the ~2
nm pores was reflected in the total area they posses. In reality, considering the pore volume
and pore area plots of the 0.0005/15/20 wt% fNT/CA/H2O membranes for example, the area
of the ~2 nm pores is ~30 times more than that of the 100 nm pores, which was very close to
the calculated values above.
From the calculation, a comparative interpretation could be done between the different
nanocomposites’ pores sizes and numbers. For pores smaller than 2 nm, the 0/15/20 wt%,
0.0005/15/20 wt%, and the 0.0005/15/20 wt% fNT/CA/H2O had the largest quantity since
their corresponding volumes were small but had huge areas per gram of the samples. As for
the macrovoids, the areas they possessed were very small, indicating that their numbers per
gram of the samples is very small. As for the 0.01/15/20 wt% fNT/CA/H2O nanocomposite,
its pores occupied the smallest areas, meaning it had the least number of pores than the rest of
the samples most probably due to higher compaction.
Comparing the differential pore volumes and areas of the nanocomposite membranes with
fNTs to the blank membrane of similar ICT (400 µm), the overall pattern of variation was
96
found to be similar to that of the 0/15/20 wt% fNT/CA/H2O blank membrane with FT =
100±10 µm.
5.3.2 Adsorption Isotherms
The adsorption isotherms of representative samples of the sets above were done to
identify their types, and it was found that all samples could be classified as types I and IV.
Figures 5.36 to 5.40 show the isotherms of the target membranes.
0/15/20 wt% fNT/CA/H2O Isotherm at ICT = 350 µm, FT =

25 100±10 µm
Quantity Adsorbed (cm³/g STP)
20
15
10 Adsorption
Desorption
5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Relative Pressure (P/Po)
100±10 µm
0/15/20 wt% fNT/CA/H2O Isotherm at ICT = 400 µm, FT =

25 120±10 µm
20
15
10 Adsorption
Desorption
5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
120±10 µm
97
0.0005/15/20 wt% fNT/CA/H2O Isotherm at FT = 100±10 µm
25
20
15
Adsorption
10
Desorption
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
100±10 µm

25
20
15
Adsorption
10
Desorption
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
100±10 µm
98
25
20
15
Adsorption
10
Desorption
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
100±10 µm
5.3.3 The t-Plots
Using the t-plot method, the micropore volumes were calculated, and the external surface
areas were evaluated. The idea is identifying the knee on the curve which forms a straight
line that when extrapolated to the y-axis gives an intercept used in calculating the micropores
volume. As for the region before the knee, it is used in calculating the total area occupied by
the nitrogen monolayer adsorbed over the external surface area. Figures 5.41 to 5.45 show the
t-plots of the representative samples per replica.
0/15/20 wt% fNT/CA/H2O t-plot at ICT = 350 µm, FT = 100±10

µm
8
7
6
5 Series1
4 Micropores
3 y = 0.2897x + 0.6674 Series3
2 Linear (Series1)
1 Linear (Micropores)
0 y = 0.5211x
0 5 10 15 20 25
Thickness (nm)
Figure ‎5.41 t- plot of 0/15/20 wt% fNT/CA/H2O at ICT = 350 µm and FT = 100±10 µm
99
Calculated Area = 8.1 m²/g
Micropores Volume = 1.3x10-3 ml/g
0/15/20 wt% fNT/CA/H2O t-plot at ICT = 400 µm, FT = 120±10

8 µm
7
5
Series1
4
Micropores
3 y = 0.3139x + 0.5278 Series3
2 Linear (Series1)
Linear (Micropores)
1
y = 0.4717x
0
0 5 10 15 20 25
Thickness (nm)
Figure ‎5.42 t- plot of 0/15/20 wt% fNT/CA/H2O at ICT = 400 µm and FT = 120±10 µm
Calculated Area = 7.30 m2/g

Micropores Volume = 8x10-4 ml/g
0.0005/15/20 wt% fNT/CA/H2O t-plot at FT = 100±10 µm

8
7
6
5 Series1
4 Micropores
y = 0.3185x + 0.675
3 Series3
2 Linear (Series1)
Linear (Micropores)
1
y = 0.5187x
0
0 5 10 15 20 25
Thickness (nm)

100
8
7
6
5 Series1
4 Micropores
y = 0.2932x + 0.6496
3 Series3
2 Linear (Series1)
Linear (Micropores)
1
y = 0.476x
0
0 5 10 15 20 25
Thickness (nm)


8
7
6
5 Series1
4 Micropores
3 Series3
y = 0.1255x + 0.1047
2 Linear (Series1)
Linear (Micropores)
1
y = 0.1517x
0
0 5 10 15 20 25
Thickness (nm)

To summarize the data above, Table 5.7 displays the external surface areas calculated
using t-plot method, the volume of the micropores, as well as the BET surface area.
101
Table ‎5.7 Summary of surface area measurements, and micropores volumes
0/15/20 0/15/20
fNT/CA/H2O
(ICT = 350 (ICT = 400 0.0005/15/20 0.005/15/20 0.01/15/20
wt%
µm) µm)
BET Area
7.77 7.11 8.68 7.07 2.33
(m2/g)
t-plot
Calculated 8.06 7.30 8.02 7.36 2.35
Area (m2/g)
t-plot
Micropores 1.03x10-3 8.2x10-4 1.04x10-3 1.00x10-3 1.62x10-4
Volume (ml/g)
It is clear from the data that the addition of functionalized CNTs resulted in a general
decrease in membrane surface area. This decrease became significant for CNTs content of
0.01 wt%. This is in line with the expected increased membrane compaction with CNT
content and the decrease in porosity. The different absolute values of CA content in the
membranes represented by the two blank samples 0/15/20 wt% ICT = 400 µm, FT = 120±10
µm, and ICT = 350 µm, FT = 100±10 µm, didn’t seem to play any noticeable role in
determining the surface micropore volume values of the nanocomposite membranes.
5.4 Membranes Performance
5.4.1 Permeation Rates
Permeation and salt retention rates for 15/20 wt% CA/H2O membranes (ICT = 400 µm)
having 0, 0.0005, 0.005, 0.01 wt% fNTs as nanofillers were measured using 1000 ppm NaCl
solution in a pressurized water cell at 24 bars and room temperature. Figure 5.46 shows the
significant increase (54.7%) in permeation rate due to the addition of only 0.0005 wt% fNTs
vs. the 0/15/20 wt% fNT/CA/H2O membrane (ICT = 350 µm). On increasing the nanotubes
weight percentage to 0.005%, the rate slightly decreased compared to the 0.0005, yet,
compared to the 0/15/20 wt% fNT/CA/H2O membrane, the rate increased by 47.2%. The
further increase in nanotubes weight percentage resulted in a sharp decrease in the
permeation rate to become a little less than that of the 0/15/20 wt% fNT/CA/H2O membrane
(-3.8% decrease).
Figure 5.46 also shows a similar trend when comparing the nanocomposites to the blank
0/15/20 wt% fNT/CA/H2O membranes (ICT = 400 µm), where a significant increase of 39%
in permeation rate were reported for the 0.0005 wt% fNTs compared to the blank membrane.
Increasing the nanotubes content to 0.005 wt% slightly decreased the rate compared to that of
the 0.0005 wt% membranes. However, this rate compared to the blank membrane showed an
increase by 32.2%. Further increase in fNT wt% (0.01 wt%) decreased the nanocomposite
permeation rate to become a less than that of the blank membrane by 13.6 %.
102
Figure ‎5.46 Plot showing the effect of fNTs on CA membrane permeation
5.4.2 Salt Retention Rates
Contrary to what other researchers have reported regarding the decrease in salt retention
rates with the increase in permeation rates [2-3, 7-8], salt retention rates were not
significantly affected by the addition of 0.0005 and 0.005 wt% fNTs when compared to the
0/15/20 wt% fNT/CA/H2O membrane at the same final thickness, where a slight decrease in
the salt retention was noticed for both fNTs contents (-6% and -6.8% respectively). The
further addition of fNTs (0.01 wt%), however had a noticeable negative effect on the
nanocomposite retention rate by 52.9% as displayed in Figure 5.47. A similar trend can be
observed when comparing the nanocomposites to blank CA membranes prepared at the same
ICT of 400 µm.
103
Figure ‎5.47 Plot showing the effect of fNTs on CA membrane salt retention rates
On comparing the overall membranes’ performance using both permeation and salt
retention rates, Figure 5.48 displays the enhanced permeation rate with the minor drop in salt
retention rate of the 0.0005/15/20 and 0.005/15/20 wt% fNT/CA/H2O membranes when
compared to the two blank 0/15/20 wt% fNT/CA/H2O membranes and the 0.01 wt% fNT/CA
membranes. Accordingly, the 0.0005/15/20 wt% fNT/CA/H2O nanocomposite was found to
be the best performing membrane.
Salt Retention vs. Permeation Rates

90%
80% 0/15/20 wt%
70% (FT=100±10µm)
60% 0/15/20 wt%

Salt retention (%)
(FT=120±10µm)
50%
0.0005/15/20 wt%
40%
30%
0.005/15/20 wt%
20%
10% 0.01/15/20/wt%
0%
0.0 0.2 0.4 0.6 0.8 1.0
Permeation rate (l/m²h.bar)
Figure ‎5.48 Representation of the effect of functionalized fNTs addition on the membrane
overall performance at same FT
104
5.4.3 BET Surface Areas
CA membranes surface area varied according to the fNTs content as displayed in Figure
5.49. The addition of only 0.0005 wt% fNTs had an insignificant effect on the surface area
when compared to the blank 0/15/20 wt% fNT/CA/H2O membranes at the same final
thickness (8.07±1.19 m2/g vs. 7.96±0.51 m2/g respectively). Further increase in the nanofiller
content started decreasing the surface area to become 7.29±0.76 m2/g, where a sharp decrease
is observed on the addition of 0.01 wt% fNTs to become 4.06±2.11 m2/g. The same trend can
be observed when comparing the nanocomposites surface areas to the other blank 0/15/20
wt% fNT/CA/H2O membranes (ICT = 400 µm).
Figure ‎5.49 Plot showing the effect of different wt% of fNTs on CA membranes surface area
On comparing the overall membranes’ performances under the effect of pores’ surface
area, Figure 5.50 shows that for the 15 wt% CA membranes, although there is a little change
in the surface area between the two blank 0/15/20 wt% fNT/CA/H2O membranes and the
0.0005 and 0.005 wt% fNTs-CA nanocomposites, permeation rates for both nanocomposites
increased. As for the 0.01 wt% fNT-CA nanocomposite, although there is a significant
decrease in its surface area in comparison to the rest of the 15 wt% CA membranes, it still
had a permeation rate comparable to both blank 0/15/20 wt% fNT/CA/H2O membranes.
105
Permeation Rates vs. BET SA
0.9 0/15/20 wt%

0.8 (FT=100±10µm)
Permeation rate (l/m²h.bar)
0.7 0/15/20 wt%

(FT=120±10µm)
0.6
0.0005/15/20 wt%
0.5
0.4 0.005/15/20 wt%
0.3
0.2 0.01/15/20/wt%
0.1
0
0.0 2.0 4.0 6.0 8.0 10.0
Avg. BET Surface area (m²/g)
Figure ‎5.50 Representation of the effect of surface area on the membranes permeation rates
On comparing the overall membranes salt rejection rates in relation to membranes surface
area, Figure 5.51 shows that for the 15 wt% CA membranes, the little change in the surface
areas of the 0.0005 and 0.005 wt% fNTs CA nanocomposites when compared to both the
0/15/20 wt% fNT/CA/H2O membranes had a little change in salt rejection rates. The
0.01/15/20 wt% fNT/CA/H2O membranes behaved differently because the pores surface area
decreased leading to a decrease in salt retention.
Salt Retention Rates vs. BET SA
90%
80% 0/15/20 wt%
70% (FT=100±10µm)
0/15/20 wt%
Salt retention (%)
60%
(FT=120±10µm)
50%
0.0005/15/20 wt%
40%
30% 0.005/15/20 wt%
20%
10% 0.01/15/20/wt%
0%
0 2 4 6 8 10
Figure ‎5.51 Representation of the effect of surface area on the membranes salt retentions
106
5.4.4 Membrane Morphology and Performance: General Discussion
Table 5.8 summarizes the data for permeation rates, salt retentions, and surface areas with the change in CNT content. Tables 5.9 and 5.10
have the calculated variation in percentages for the same samples due to the mentioned effects to clarify the change in membrane performance.
Table ‎5.8 Effect of fNTs wt% on permeation, salt retention, and surface area, rate measurements are for 15/20 wt% CA/H2O membranes using 1000
ppm NaCl solution at 24 bars and room temperature
Avg. BET
Permeation Salt Avg. Salt Avg. BET Name
ICT FT Additive Permeation Surface
CA wt% Additive(s) rate (L/m² retention retention Surface area Displayed
(µm) (µm) % rate (L/m² area
h.bar) % % (m²/g) on Charts
h.bar) (m²/g)
0.63 73.52% 8.54
0/15/20 wt%
15% 350 ~100±10 Water 20% 0.53 0.53 ± 0.088 72.60% 73.82 ±1.39 7.57 7.96±0.51 fNT/CA/H2O
0.45 75.34% 7.77 (Blank 1)
0.62 80.56% 6.80
0/15/20 wt%
15% 400 ~120±10 Water 20% 0.68 0.59 ± 0.11 79.63% 78.82±2.26 6.97 6.96±0.16 fNT/CA/H2O
0.47 76.26% 7.11 (Blank 2)
0.83 69.40% 6.70
CNTs + 0.0005% 0.0005/15/20
15% 400 ~100±10 0.86 0.82 ± 0.057 66.50% 69.37±2.85 8.69 8.07±1.19 wt%
Water +20%
0.75 72.20% 8.83 fNT/CA/H2O
0.72 69.20% 6.66
CNTs + 0.005% + 0.005/15/20
15% 400 ~100±10 0.75 0.78 ± 0.067 69.90% 68.80±1.35 7.07 7.29±0.76 wt%
Water 20%
0.85 67.30% 8.13 fNT/CA/H2O
0.55 34.48% 6.41
CNTs + 0.01% + 0.01/15/20
15% 400 ~100±10 0.46 0.51 ± 0.047 35.07% 34.75±0.30 3.45 4.06±2.11 wt%
Water 20%
0.51 34.70% 2.33 fNT/CA/H2O
107
final thickness
Avg.
Change in Avg.
Permeation Avg. Salt Change in Avg. Salt
Name Displayed on Permeation vs.
Charts rate (L/m² retention % Retention vs. Blank 1
Blank 1
h.bar)
0/15/20 wt%
0.53 ± 0.088 ----- 73.82 ±1.39 -----
fNT/CA/H2O
(Blank 1)
0.0005/15/20 wt% 0.82 ± 0.057 54.7% 69.37±2.85 -6.0%
fNT/CA/H2O
0.005/15/20 wt% 0.78 ± 0.067 47.2% 68.80±1.35 -6.8%
fNT/CA/H2O
0.01/15/20 wt% 0.51 ± 0.047 -3.8% 34.75±0.30 -52.9%
fNT/CA/H2O
initial casting thickness
Avg.
Change in Avg.
Name Displayed on Permeation Avg. Salt Change in Avg. Salt
Permeation vs.
Charts rate (L/m² retention % Retention vs. Blank 2
Blank 2
h.bar)
0/15/20 wt%
fNT/CA/H2O 0.59 ± 0.11 ----- 78.82±2.26 -----
(Blank 2)
0.0005/15/20 wt%
0.82 ± 0.057 39.0% 69.37±2.85 -12.0%
fNT/CA/H2O
0.005/15/20 wt%
0.78 ± 0.067 32.2% 68.80±1.35 -12.7%
fNT/CA/H2O
0.01/15/20 wt% 0.51 ± 0.047 -13.6% 34.75±0.30 -55.9%
fNT/CA/H2O
The first point to consider in permeation rates was having a steady rate at an acceptable
pressure range. Many membranes with different conditions were tested to evaluate their
steady flow rate. The pressure range tested varied from 8 bars all the way to 55 bars.
According to literature, CA membranes usually can handle a range of 15-30 bars, beyond
which the membranes are damaged under the effect of compaction [47]. This was verified for
all the membranes reported in Appendix I. They were damaged due to high pressure.
Membranes that performed well, on the other hand, showed a steady flow rates at 24 bars,
which lies in the range mentioned in the literature [47].
Comparing the permeation rates of the 0/15/20 wt% fNT/CA/H2O membranes (FT =
100±10 µm) to the 0.0005/15/20 wt% fNT/CA/H2O and 0.005/15/20 wt% fNT/CA/H2O
108
membranes show that the minor addition of fNTs to CA matrix increased the solution
permeation rate even though the fNTs decreased the number of the 93 nm, 117 nm, and 143
nm pores, as demonstrated by the decrease in their surface area and volume, while having a
lesser effect on meso and micropores (Figures 5.31, 5.32, 5.34, and 5.35) This suggests that
the nanotubes could have created connection channels between the pores, thus enhancing the
flow rates. This can also be supported by the interpretation of molecular dynamics simulation
of water passage within CNTs, since the hydrophobic channels are expected to enhance
permeation rates. The fact that the channels didn’t significantly affect the pores surface area
justifies the claim that the permeate might have used the new available route, and moved
faster than within the 0/15/20 fNT/CA/H2O membranes. However, MD simulations suggested
that after a certain nanotube diameter, the permeate wouldn’t sense the effect of the
hydrophobic channels, and wouldn’t experience the drift like motion [77]. This could explain
why the increase in the NaCl solution permeation rate for the 0.0005/15/20 wt%
fNT/CA/H2O and 0.005/15/20 wt% fNT/CA/H2O membranes wasn’t several orders of
magnitude higher than usual.
Interpreting the permeation rates for the 0.005/15/20 wt% and the 0.0005/15/20 wt%
membranes, however, raises a question. Even though the former has more fNTs content, and
has higher meso and macropores volume and area than the latter, it still exhibits less
permeation rate. This could be clarified from the BET surface area and the t-plot surface area
calculations of both as in Table 5.7. According to the data, the 0.005/15/20 wt% membranes
have an overall surface area that is less than that of the 0.0005/15/20 wt% membranes, which
means that the former membrane is more compact than the latter. Though this was not
reflected in the size and number of large meso and macropores, it was reflected in the
reduction of the volume, area, and consequently the number of the mesopores with diameters
less than 12 nm, as well as in the volume, area, and number of micropores as shown in
Figures 5.31 and 5.34. This in turn affected the permeation rate with the small difference
shown in Figure 5.50.
Another factor affecting permeation rate is the fact that the fNTs are expected to have
enhanced the hydrophilic property of the two nanocomposites due to the existence of the
polar functional groups on the nanofiller walls. This is thought to play a role in improving the
permeation rates since it facilitates the polar solvent (water in this case) to move inside the
matrix faster than for the blank CA matrix. This was suggested by Choi et al. [38] on
studying the effect of the addition of fNTs to PSF membranes.
On the other hand, the noticeable decrease in permeation rate of the 0.01/15/20 wt%
fNT/CA/H2O membrane can be related to the significant decrease in the membrane porosity,
as reflected by the decrease in pore surface area, pore volume, as well as the overall
membrane surface area (Figure 5.50). This was clear in the membrane SEM image in Figure
5.29, and clear in the porosity results in Figures 5.30 and 5.33, where the high content of
fNTs compacted the overall pore surface area of the membrane such that the meso and
macropores needed to facilitate high permeation rates decreased in size and numbers. Adding
109
to this, the high content of the fNTs could have blocked the existing pores, as suggested by
Tang et al. [40].
Further interpretation of the plot representing permeation rates and the plot representing
the relation between surface area and permeation (Figures 5.46 and 5.50) again supports the
postulation that the nanotubes created alternative channels for water passage. The 0.01/15/20
wt% fNT/CA/H2O membranes permeation rate was still comparable to the blank 15/20 wt%
CA/H2O membrane (at the same final thickness) even though the pores surface area of the
former is approximately half that of the latter. This implies that water permeation through the
nanocomposite membrane was dependent on the effect of CNTs, as well as membrane
porosity.
It is worth mentioning that interpretation of the performance of the 0/15/20 wt%

fNT/CA/H2O membranes with ICT of 400 µm and final thickness of 120±10 µm in
comparison with the 0.0005/15/20 wt% and 0.005/15/20 wt% fNT/CA/H2O nanocomposites
is thought to be dependent on both the presence of the fNTs, in parallel with the available
macro, meso, and micropores. Even though the surface areas of the three membranes are very
close (Figure 5.49), the minor increase in the surface area associated with the nanotubes
channels of both nanocomposites increased their permeation rates relative to the 0/15/20 wt%
membrane as shown in Figure 50. As for the behavior of the 0.01/15/20 wt% fNT/CA/H2O
nanocomposites performance in comparison to the 0/15/20 wt% fNT/CA/H2O membrane
(ICT = 400 µm), the presence of the fNTs didn’t enhance the permeation, but rather
decreased it by only 3.8% (Table 5.10). This is because the available surface area in the
former is half the latter (Figure 5.49). Thus, one of the permeate routes is blocked (the pores)
even though the second route (the nanotubes) is used.
Interpretation of the salt retention rates of the 0.0005/15/20 wt% and 0.005/15/20 wt%
fNT/CA/H2O membranes show that the new channels created by the fNTs didn’t allow larger
quantities of NaCl molecules to pass through when compared to the salt retention rates of the
0/15/20 wt% fNT/CA/H2O membranes (FT = 100±10 µm) (Figure 5.47). This doesn’t seem
to agree with what MD simulations suggest, since the large diameter of the nanotubes used in
the experiments were expected to allow the passage of ions freely [64, 68]. The results are
promising because the addition of small amounts of functionalized CNTs had a good effect
on permeation rates without altering salt retention to a significant effect. On the other hand,
salt retention seems to be more dependent on membrane porosity, particularly small pores (<6
nm). This can be verified from the salt retention data when compared to porosity data
(Figures 5.47, 5.31 and 5.34), where salt retention exhibit small decreases with the addition
of fNTs to the membranes as long as the number of small pores is not noticeably reduced
(observed salt retention decreases are about 6% and 7% for the 0.0005 wt% CNT and the
0.005 wt% CNT respectively).
The 0.01/15/20 wt% fNT/CA/H2O nanocomposite salt retention rate, however, was the
most unexpected because on comparing it to the 0/15/20 wt% fNT/CA/H2O membranes, the
former has less small pores within the same FT (Figures 5.31 and 5.34). Furthermore, the
110
former was cast with an ICT of 400 µm to reach the same final thickness as the latter which
was cast at ICT = 350 µm, thus it is expected to have more dense polymer layers than the
latter (more absolute CA content). At the same time, the two membranes have approximately
the same permeation rates (Figure 5.46). Still, the 0.01/15/20 wt% fNT/CA/H2O
nanocomposite salt retention was reduced by more than 50% compared to the 0/15/20 wt%
fNT/CA/H2O membranes (Table 5.10 and Figure 5.47). This behavior could be explained by
considering several things. The first is by looking at the micropores and small mesopores (<6
nm) size distribution in Figures 5.31 and 5.34, their sizes and numbers decreased
significantly. This can also be supported by the calculated micropores’ volumes using the t-
plot in table 5.7. The second parameter that could have contributed to this is the possibility
that the nanotubes acted as passage routes for the entering salt since there wasn’t enough
micropores to screen off the salt, which actually agrees with what MD simulations suggests
[64]. Last, although the mesopores and macropores are less in number compared to the
0/15/20 wt% fNT/CA/H2O membranes, water permeation is more prominently taking place
through them or through the nanotubes, both of which are less effective in salt retention,
leading to the significant decrease in salt retention values obtained.
111
6 Conclusions
112
6.1 Conclusion
Having investigated the effects of membranes thickness, CA contents, PI temperature, solvent

evaporation, H2O addition, and CNTs addition, the following conclusions can be drawn.
1. ICT strongly affects voids formation: where the thicker the membrane, the larger the
voids. Below a critical thickness, macrovoids are unable to develop probably because
the membranes are too thin for nucleation expansion to take place. ICT also affects
macrovoids shapes and depth within the membranes.
2. The increase in coagulation bath temperature during membrane PI increased

macrovoids sizes, which agrees with literature. However, the shapes and sizes of the
macrovoids were found to be dependent of CA content as well.
3. The decrease in CA content increases the macrovoids size. However, decreasing CA

content was found to have a limited range because of its effect on the cast solution
viscosity. The 13 wt% CA membranes, though having large macrovoids in
comparison to the 15 and 17 wt% CA membranes, were found to be non-reproducible
in their final thicknesses or morphology. This is also thought to be due to the low
viscosity of the cast solution.
4. The main aim of solvent evaporation was to decrease the pores sizes on the top layer
of the membrane so that in salt retention tests, the membranes would perform better.
Both trials for solvent evaporation (either prior to PI or post incomplete PI) had
negatively affected the macrovoids structure, where they decreased in size. A main
factor to consider is the high volatility of the acetone solvent used, which could
explain the dramatic decrease in macrovoids sizes.
5. The addition of a non-solvent during stock preparation became essential to create

large macrovoids especially for the 15 and 17 wt% CA membranes. Different water
amounts were tried (5, 10, 15, and 20 wt%), all of which were successful in creating
large macrovoids with the increase in water content. Beyond 20 wt% though, having a
homogenous stock solutions was unsuccessful since the amount of non-solvent added
precipitated the CA in the stock solution creating gelatinous like bulky particles.
6. MWCNTs’ addition, both pristine and functionalized, had a compaction effect on

membrane morphology. The amount of compactness depended on both the quantity of
the nanotubes and how well dispersed they were throughout the membranes.
Agglomerization was a main issue for pNTs since it was very challenging to disperse
them into the hydrophilic medium of the CA matrix. Resorting to functionalization of
the nanotubes became a must.
7. MWCNTs functionalization was successful. This facilitated nanocomposites

preparation as fNTs dispersion.
113
8. The proper random dispersion of fNTs showed that networks of fNTs spread
throughout the CA matrix with no detectable agglomerates. This was due to the polar
graphene walls that enhanced the interaction between the nanofiller and the polymer
matrix at the interface. The interaction effect was clear in the decreased number of
macrovoids as the wt% of fNTs increased. This is thought to be due to the fact that
fNTs presence decreased the rate of the demixing process during PI.
9. Porosity analysis showed that for the fNTs nanocomposites, the increase in fNTs
content was associated with a general decrease in pores surface areas relative to the
15/20 wt% CA/H2O membranes with the same final thickness. This led to a small
decrease in salt retention rates. However the permeation rates increased with the
increase in the fNTs content until a flipping point at which it decreased again. This is
believed to be due to the nanotubes opening new channels for the permeate to pass
through along with the existing pores.
10. Porosity analysis showed that fNTs nanocomposites cast with the same ICT as a
15/20 wt% CA/H2O membranes don’t differ significantly, yet their permeation rates
were generally higher (except for one), and their salt retention were lower. This lead
to a more belief that the permeate passage is not only through the available pores but
rather through a parallel route via the nanotubes.
6.2 Future Work
As a continuation for this project, several aspects could be carried out in the future:
 Investigating the possibility of using MWCNTs of different dimensions (aspect ratios,

length, and diameter)
 Investigating the effect of SWCNTs addition to CA membranes and their effect on

membrane performance and morphology
 Investigating the effect of using a different solvent (like acetic acid) on the
nanocomposites performance and morphology
 Investigating the effect of using a pore former on the nanocomposites performance

and morphology
 Investigating the effect of lowering the PI coagulation bath temperature to 4oC, and
having an annealing step at high temperatures added to the preparation procedure
 Investigating different types of functionalization for MWCNTs, making them more

ions selective, thus developing more specialized nanocomposites
114
 Investigating the nanocomposites performance with real brackish water from open
water sources in Cairo or Giza premises, and measuring the nanocomposites rejection
rates for different salts
 Investigating different fouling effects on membrane performance
115
References
1. Search for Clean Water Continues

[http://www.nesc.wvu.edu/old_website/ndwc/ndwc_dwh_1.html]
2. Ridgway H: Membrane Separation…Basics. In: Advanced Membrane

Technologies. Stanford University; 2008.
3. Desalination History
[http://www.water.vic.gov.au/programs/desalination/desalination/desalination-history]
4. The History of Desalination [http://www.rpi.edu/dept/chem-eng/Biotech-

Environ/Environmental/desal/history.html]
5. Henthorne L: The Current State of Desalination. In: 2009; Dubai: International

Desalination Association; 2009.
6. Yangali-Quintanilla V, Li Z, Valladares R, Li Q, Amy G: Indirect desalination of

Red Sea water with forward osmosis and low pressure reverse osmosis for water
reuse. Desalination, In Press, Corrected Proof.
7. Water crisis hits Egypt "Country of Nile River"

[http://news.egypt.com/en/20070724293/news/-egypt-news/water-crisis-hits-egypt-
country-of-nile-river.html]
8. Egypt's farmers desperate for clean water

[http://edition.cnn.com/2010/WORLD/africa/11/09/egypt.water.shortage/index.html]
9. Fishery Country Profile, Arabic Republic of Egypt. In: Food and Agriculture
Organization of the United Nations. FAO; 2003.
10. Sustainable development and water resources in Egypt: Translated page

[http://www.sis.gov.eg/Ar/LastPage.aspx?Category_ID=1082]
11. Mezher T, Fath H, Abbas Z, Khaled A: Techno-economic assessment and

environmental impacts of desalination technologies. Desalination 2011, 266(1-
3):263-273.
12. Lentz J: Salinity. In.: School of Coast and Environment, Louisiana State University;
2010.
13. Van der Bruggen B, Vandecasteele C: Distillation vs. membrane filtration:

overview of process evolutions in seawater desalination. Desalination 2002,
143(3):207-218.
14. Alyson Sagle BF: Fundamentals of Membranes for Water Treatment. The Future
of Desalination in Texas 2004, 2(363):137-154
116
15. Ma Q, Lu H: Wind energy technologies integrated with desalination systems:
Review and state-of-the-art. Desalination 2011, 277(1-3):274-280.
16. Corry B: Designing Carbon Nanotube Membranes for Efficient Water

Desalination. The Journal of Physical Chemistry B 2007, 112(5):1427-1434.
17. Mark LeChevallier K-KA: Removal Process. In: Water Treatment and Pathogen
Control: Process Efficiency in Achieving Safe Drinking Water. World Health
Organization; 2004: 33-39.
18. Darunee Bhongsuwan TB: Preparation of Cellulose Acetate Membranes for Ultra-
Nano- Filtrations. Kasetsart Journal: Natural Sciences 2008(42):311 - 317.
19. Edward S. K. Chian JPC, Ping-Xin Sheng, Yen-Peng Ting, Lawrence K. Wang:
Reverse Osmosis Technology for Desalination. In: Handbook of Environmental
Engineering: Advanced Physicochemical Treatments Technologies. vol. 5, 1 edn:
329-366.
20. Lee KP, Arnot TC, Mattia D: A review of reverse osmosis membrane materials for
desalination—Development to date and future potential. Journal of Membrane
Science 2011, 370(1-2):1-22.
21. Li Z, Ren J, Fane AG, Li DF, Wong F-S: Influence of solvent on the structure and
performance of cellulose acetate membranes. Journal of Membrane Science 2006,
279(1-2):601-607.
22. Ye SH, Watanabe J, Iwasaki Y, Ishihara K: Novel cellulose acetate membrane

blended with phospholipid polymer for hemocompatible filtration system.
Journal of Membrane Science 2002, 210(2):411-421.
23. Haddad R, Ferjani E, Roudesli MS, Deratani A: Properties of cellulose acetate

nanofiltration membranes. Application to brackish water desalination.
Desalination 2004, 167:403-409.
24. Boricha AG, Murthy ZVP: Preparation of N,O-carboxymethyl chitosan/cellulose

acetate blend nanofiltration membrane and testing its performance in treating
industrial wastewater. Chemical Engineering Journal 2010, 157(2–3):393-400.
25. Cano-Odena A, Spilliers M, Dedroog T, De Grave K, Ramon J, Vankelecom IFJ:

Optimization of cellulose acetate nanofiltration membranes for micropollutant
removal via genetic algorithms and high throughput experimentation. Journal of
Membrane Science 2011, 366(1–2):25-32.
26. Lv C, Su Y, Wang Y, Ma X, Sun Q, Jiang Z: Enhanced permeation performance of

cellulose acetate ultrafiltration membrane by incorporation of Pluronic F127.
Journal of Membrane Science 2007, 294(1–2):68-74.
117
27. Chen J, Li J, Zhan X, Han X, Chen C: Effect of PEG additives on properties and
morphologies of polyetherimide membranes prepared by phase inversion.
Frontiers of Chemical Engineering in China 2010, 4(3):300-306.
28. Saljoughi E, Sadrzadeh M, Mohammadi T: Effect of preparation variables on

morphology and pure water permeation flux through asymmetric cellulose
acetate membranes. Journal of Membrane Science 2009, 326(2):627-634.
29. Ani Idris KYL, H. K. Hing: Preparation of cellulose acetate dialysis membrane for
separation of bovine serum albumin. Jurnal Teknologi, Universiti Teknologi
Malaysia 2005(42 F):35-46.
30. Saljoughi E, Mohammadi T: Cellulose acetate (CA)/polyvinylpyrrolidone (PVP)

blend asymmetric membranes: Preparation, morphology and performance.
Desalination 2009, 249(2):850-854.
31. Sivakumar M, Mohan DR, Rangarajan R: Studies on cellulose acetate-polysulfone

ultrafiltration membranes: II. Effect of additive concentration. Journal of
Membrane Science 2006, 268(2):208-219.
32. Arthanareeswaran G, Sriyamuna Devi TK, Raajenthiren M: Effect of silica particles

on cellulose acetate blend ultrafiltration membranes: Part I. Separation and
Purification Technology 2008, 64(1):38-47.
33. Chen W, Su Y, Zhang L, Shi Q, Peng J, Jiang Z: In situ generated silica

nanoparticles as pore-forming agent for enhanced permeability of cellulose
acetate membranes. Journal of Membrane Science 2010, 348(1–2):75-83.
34. Mahendran R, Malaisamy R, Mohan DR: Cellulose acetate and polyethersulfone

blend ultrafiltration membranes. Part I: Preparation and characterizations.
Polymers for Advanced Technologies 2004, 15(3):149-157.
35. Mänttäri M, Pihlajamäki A, Nyström M: Effect of pH on hydrophilicity and charge

and their effect on the filtration efficiency of NF membranes at different pH.
Journal of Membrane Science 2006, 280(1–2):311-320.
36. Rodrigues-Filho UP, Gushikem Y, Gonçalves MdC, Cachichi RC, de Castro SC:
Composite Membranes of Cellulose Acetate and Zirconium Dioxide:
Preparation and Study of Physicochemical Characteristics. Chemistry of
Materials 1996, 8(7):1375-1379.
37. Arthanareeswaran G, Thanikaivelan P: Fabrication of cellulose acetate–zirconia

hybrid membranes for ultrafiltration applications: Performance, structure and
fouling analysis. Separation and Purification Technology 2010, 74(2):230-235.
118
38. Choi J-H, Jegal J, Kim W-N: Fabrication and characterization of multi-walled
carbon nanotubes/polymer blend membranes. Journal of Membrane Science 2006,
284(1-2):406-415.
39. Qiu S, Wu L, Pan X, Zhang L, Chen H, Gao C: Preparation and properties of

functionalized carbon nanotube/PSF blend ultrafiltration membranes. Journal of
Membrane Science 2009, 342(1–2):165-172.
40. Tang C, Zhang Q, Wang K, Fu Q, Zhang C: Water transport behavior of chitosan

porous membranes containing multi-walled carbon nanotubes (MWNTs).
Journal of Membrane Science 2009, 337(1-2):240-247.
41. Shawky HA, Chae S-R, Lin S, Wiesner MR: Synthesis and characterization of a
carbon nanotube/polymer nanocomposite membrane for water treatment.
Desalination 2011, 272(1-3):46-50.
42. Wang Y, Yang L, Luo G, Dai Y: Preparation of cellulose acetate membrane filled
with metal oxide particles for the pervaporation separation of methanol/methyl
tert-butyl ether mixtures. Chemical Engineering Journal 2009, 146(1):6-10.
43. Haddada R, Ferjani E, Roudesli MS, Deratani A: Properties of cellulose acetate

nanofiltration membranes. Application to brackish water desalination.
Desalination 2004, 167:403-409.
44. B. McCray S, Vilker VL, Nobe K: Reverse osmosis cellulose acetate membranes II.
Dependence of transport properties on acetyl content. Journal of Membrane
Science 1991, 59(3):317-330.
45. Suzana Pereira Nunes K-VP: Membrane Technology: in the Chemical Industry, 2
edn: Wiley-VCH; 2006.
46. Valente AJM, Polishchuk AY, Burrows HD, Lobo VMM: Permeation of water as a
tool for characterizing the effect of solvent, film thickness and water solubility in
cellulose acetate membranes. European Polymer Journal 2005, 41(2):275-281.
47. Baker RW: Reverse Osmosis. In: Membrane Technology and Applications. Wiley;
2004: 197-200.
48. Kucera J: Membranes. In: Reverse Osmosis: Design, Processes, and Applications for
Engineers. Wiley; 2010: 47-51.
49. Vogrin N, Stropnik C, Musil V, Brumen M: The wet phase separation: the effect of
cast solution thickness on the appearance of macrovoids in the membrane
forming ternary cellulose acetate/acetone/water system. Journal of Membrane
Science 2002, 207(1):139-141.
119
50. Su J, Zhang S, Chen H, Chen H, Jean YC, Chung T-S: Effects of annealing on the
microstructure and performance of cellulose acetate membranes for pressure-
retarded osmosis processes. Journal of Membrane Science 2010, 364(1–2):344-353.
51. Baker R: Membrane Technology and Applications: Wiley; 2004.
52. Smolders CA, Reuvers AJ, Boom RM, Wienk IM: Microstructures in phase-
inversion membranes. Part 1. Formation of macrovoids. Journal of Membrane
Science 1992, 73(2–3):259-275.
53. Kim HJ, Tyagi RK, Fouda AE, Ionasson K: The kinetic study for asymmetric
membrane formation via phase-inversion process. Journal of Applied Polymer
Science 1996, 62(4):621-629.
54. Ren J, Wang R: Preparation of Polymeric Membranes. In: Membrane and

Desalination Technologies. Edited by Wang LK, Chen JP, Hung Y-T, Shammas NK,
vol. 13: Humana Press; 2008: 47-100.
55. Wara NM, Francis LF, Velamakanni BV: Addition of alumina to cellulose acetate
membranes. Journal of Membrane Science 1995, 104(1-2):43-49.
56. Aparecida da Silva C, Maria Favaro M, Pagotto Yoshida IV, do Carmo Gonçalves M:
Nanocomposites derived from cellulose acetate and highly branched
alkoxysilane. Journal of Applied Polymer Science 2011, 121(5):2559-2566.
57. Damme HV: Nanocomposites: The End of Compromise. In: Nanomaterials and
nanochemistry. Edited by Catherine Bréchignac PH, Marcel Lahmani, vol. 2: Springer
Berlin Heidelberg; 2008: 347-380.
58. A.E. Gash RLS, T.M. Tillofson, J.H. Safcher, L. W. Hrubesh: Making
Nanostructured Pyrotechnics in a Beaker. In.: International Pyrotechnics Seminars,
Grand Junction, CO; 2000.
59. Ajayan P: Nanocomposite Science and Technology: Wiley-VCH Verlag GmbH Co;
2003.
60. Erik T. Thostenson CL, Tsu-Wei Chou: Nanocomposites in Context: Review.

Composites Science and Technology 2005(65):491–516.
61. Michler GH: Polymer Nanocomposites. In: Electron Microscopy of Polymers vol. 3:
Springer Berlin Heidelberg; 2008: 419-428.
62. Xie X-L, Mai Y-W, Zhou X-P: Dispersion and alignment of carbon nanotubes in
polymer matrix: A review. Materials Science and Engineering: R: Reports 2005,
49(4):89-112.
120
63. Hirsch A, Vostrowsky O: Functionalization of Carbon Nanotubes. In: Functional
Molecular Nanostructures. Edited by Schlüter AD, vol. 245: Springer Berlin /
Heidelberg; 2005: 193-237.
64. Aleksandr Noya HGP, Francesco Fornasieroa, Jason K. Holta, Costas P.

Grigoropoulosb, Olgica Bakajina: Nanofluidics in carbon nanotubes: Review.
Nanotoday 2007, 2(6):22-29.
65. Ahir SV, Huang YY, Terentjev EM: Polymers with aligned carbon nanotubes:
Active composite materials. Polymer 2008, 49(18):3841-3854.
66. Maser W, Benito AM, Muñoz E, Martínez MT: Carbon Nanotubes: From
Fundamental Nanoscale Objects Towards Functional Nanocomposites and
Applications. In: Functionalized Nanoscale Materials, Devices and Systems. Edited
by Vaseashta A, Mihailescu IN: Springer Netherlands; 2008: 101-119.
67. Various Carbon Forms [http://www.nano-enhanced-wholesale-

technologies.com/faq/carbon-forms.htm]
68. Alexiadis A, Kassinos S: The density of water in carbon nanotubes. Chemical

Engineering Science 2008, 63(8):2047-2056.
69. Thostenson ET, Li C, Chou T-W: Nanocomposites in context. Composites Science

and Technology 2005, 65(3-4):491-516.
70. Mohl M, Kónya Z, Kukovecz Á, Kiricsi I: Functionalization of Multi-Walled

Carbon Nanotubes (MWCNTS). In: Functionalized Nanoscale Materials, Devices
and Systems. Edited by Vaseashta A, Mihailescu IN: Springer Netherlands; 2008:
365-368.
71. Gao C, Vo CD, Jin YZ, Li W, Armes SP: Multihydroxy Polymer-Functionalized

Carbon Nanotubes: Synthesis, Derivatization, and Metal Loading.
Macromolecules 2005, 38(21):8634-8648.
72. Upadhyayula VKK, Deng S, Mitchell MC, Smith GB: Application of carbon
nanotube technology for removal of contaminants in drinking water: A review.
Science of The Total Environment 2009, 408(1):1-13.
73. Scientific basis for the definition of the term “nanomaterial”. In. European Union:
Scientific Committee on Emerging and Newly Identified Health Risks; European
Commission; 2010.
74. John A. Thomas AJHM, Ottoleo Kuter-Arnebeck: Pressure-driven water flow

through carbon nanotubes: Insights from molecular dynamics simulation.
International Journal of Thermal Sciences 2010, 49:281–289.
121
75. Hanasaki I: Flow structure of water in carbon nanotubes: Poiseuille jour or plug-
like? J Chem Phys 2006, 124(14):144708.
76. Kotsalis EM, Walther JH, Koumoutsakos P: Multiphase water flow inside carbon
nanotubes. International Journal of Multiphase Flow 2004, 30(7-8):995-1010.
77. Hummer G, Rasaiah JC, Noworyta JP: Water conduction through the hydrophobic
channel of a carbon nanotube. Nature 2001, 414(6860):188-190.
78. Liu Y: Fluid structure and transport properties of water inside carbon
nanotubes. J Chem Phys 2005, 123(23):234701.
79. Kalra A, Garde S, Hummer G: Osmotic water transport through carbon nanotube
membranes. Proceedings of the National Academy of Sciences 2003, 100(18):10175-
10180.
80. Goodhew PJ, Humphreys FJ, Beanland R: Electron Microscopy and Analysis,
Third Edition: Taylor & Francis Group; 2001.
81. Goldstein J: Scanning Electron Microscopy and X-Ray Microanalysis: Kluwer

Academic Pub; 2003.
82. Everhart T: Wide-Band Detector For Micro-Microampere Low-Energy Electron

Currents. Journal of Scientific Instruments 1960, 37:246–248.
83. Bubert H, Jenett H: Surface and Thin Film Analysis: Wiley; 2011.
84. Allen T: Particle Size Measurement: Volume 2: Surface Area and Pore Size
Determination: Springer; 1996.
85. Gas Adsoption Theory. In. Edited by Corporation MI: Micrometrics Instrumental
Corporation.
86. Gas Sorption Analysis [http://www.micromeritics.com/]
87. Lowell S, Shields JE, Thomas MA, Thommes M: Characterization of Porous Solids
and Powders: Surface Area, Pore Size and Density: Springer; 2006.
88. Lowell S, Shields JE: Powder Surface Area and Porosity: Springer; 1991.
89. Marsh H, Rodríguez-Reinoso F: Activated Carbon: Elsevier Science & Tech; 2006.
90. Gregg SJ, Sing KSW: Adsorption, surface area, and porosity: Academic Press;
1967.
91. Strickland ML: Physical Adsorption Theory. In.: Micromeritics Instrument Corp.
122
92. Powder Tech Note 18
[http://www.atomikateknik.com/pdf/QCI_PowderTech_18.pdf]
93. Condon JB: Surface Area and Porosity Determinations by Physisorption:

Measurements and Theory: Elsevier Science; 2006.
94. Lastoskie CM, Gubbins KE: Characterization of porous materials using density
functional theory and molecular simulation. In: Studies in Surface Science and
Catalysis. Edited by K.K. Unger GK, Baselt JP, vol. Volume 128: Elsevier; 2000: 41-
50.
95. Micrometrics ASAP 2020 [http://www.micromeritics.com/product-showcase/ASAP-

2020-Physisorption.aspx]
96. Stobinski L, Lesiak B, Kövér L, Tóth J, Biniak S, Trykowski G, Judek J: Multiwall

carbon nanotubes purification and oxidation by nitric acid studied by the FTIR
and electron spectroscopy methods. Journal of Alloys and Compounds 2010,
501(1):77-84.
97. Coates J: Interpretation of Infrared Spectra, A Practical Approach. In:

Encyclopedia of Analytical Chemistry. Edited by Meyers RA: John Wiley & Sons
Ltd, Chichester; 2001: 10815–10837.
98. Li M, Kim I-H, Jeong YG: Cellulose acetate/multiwalled carbon nanotube

nanocomposites with improved mechanical, thermal, and electrical properties.
Journal of Applied Polymer Science 2010, 118(4):2475-2481.
123
Appendix I
124
Unsuccessful Membranes
Permeation and salt retention rates measurements were carried out to evaluate
membranes’ performance and to find the best possible combination of conditions. The table
below summarizes the unsuccessful membranes’ preparation conditions that didn’t allow
permeation through the membranes at the maximum operating pressure used (55 bars).
Combination of conditions that failed to allow a permeation rate
CA% Additive(s) Additive % ICT (µm) ISE (sec)
13% Water 0 1000 0
13% Water 0 1000 30
13% Water 0 1000 60
13% Water 0 1200 0
13% Water 0 1200 30
13% pNT 0.5 1000 0
13% Water 5 1000 0
13% Water 5 1200 0
13% Water 10 800 0
13% Water 10 800 30
13% Water 10 1000 0
14% Water 0 1000 0
14% Water 5 800 0
14% Water 5 800 30
14% Water 5 1000 0
14% Water 10 800 0
15% Water 0 600 0
15% Water 0 700 0
15% Water 0 800 0
15% Water 0 900 0
15% Water 5 300 0
125
CA% Additive(s) Additive % ICT (µm) ISE (sec)
15% Water 10 400 0
15% Water 10 500 0
15% fNT + Water 0.5+20 300 0
15% fNT + Water 0.05+21 300 0
15% fNT + Water 0.05+22 360 0
15% fNT + Water 0.05+23 400 0
15% fNT + Water 0.05+24 500 0
15% fNT + Water 0.05+24 550 0
17% Water 0 500 0
17% Water 0 600 0
17% Water 0 700 0
17% Water 0 800 0
17% Water 0 1000 0
17% pNT 0.5 1000 0
17% pNT 1 1000 0
17% Water 5 1000 0
17% Water 5 1000 30
17% Water 10 700 0
17% Water 10 800 0
17% Water 20 200 0
17% Water 20 260 0
17% Water 20 700 0
126

Celulosa Nanotubos

Uploaded by

Copyright:

Available Formats

Celulosa Nanotubos

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Celulosa Nanotubos

Uploaded by

Copyright:

Available Formats

THE AMERICAN UNIVERSITY IN CAIRO

Dr. Adham Ramadan

Submitted to the School of Sciences and Engineering, Department of

In partial fulfillment of the requirements of the Degree of Master of

MWCNTs/CA nanocomposites membranes were prepared. MWCNTs were first

Thank you all for an overwhelming wonderful and challenging experience.

Figure ‎1.1 Desalination saline sources % worldwide [11] ........................................................ 3

Figure ‎1.2 Desalination processes % worldwide [11] ............................................................... 3

Figure ‎1.3 Diagram of Multi Stage Flash Process [13] ............................................................. 4

Figure ‎1.4 Diagram of Multiple Effect Distillation Process [13] .............................................. 5

Figure ‎1.5 Osmosis and Reverse Osmosis flow [17] ................................................................. 6

Figure ‎3.1 Membrane lifetime vs. pH [48] .............................................................................. 29

Figure ‎3.6 Permeation cell (modified from [29])..................................................................... 33

Figure ‎3.14 The dimension of a MWCNT ............................................................................... 42

Figure ‎3.16 Different types of isotherm linear plots [84] ........................................................ 44

Figure ‎3.17 BET plot to evaluate Vm [91]................................................................................ 46

Figure ‎4.1 Elcometer 4040 Automatic Film Applicator .......................................................... 52

Figure ‎4.2 Micrometrics ASAP 2020 instrument [94]............................................................. 53

Figure ‎4.3 Sterlitech HP4750 Stirred Cell ............................................................................... 54

Figure ‎4.4 Different preparation conditions............................................................................. 55

Figure ‎5.1 SEM of an asymetric membrane ............................................................................ 62

Figure ‎5.23 FTIR spectrum of non-functionalized MWCNTs ................................................ 86

Figure ‎5.24 FTIR spectrum of functionalized MWCNTs ........................................................ 86

Figure ‎5.26 MWCNTs networks in 0.0005/15/20 wt% fNT/CA/H2O nanocomposite at

Figure ‎5.28 MWCNTs networks in 0.01/15/20 wt% fNT/CA/H2O nanocomposite at different

Figure ‎5.30 Plot of differential pore volumes .......................................................................... 92

Figure ‎5.32 Plot of differential pore volumes for macropores................................................. 93

Figure ‎5.33 Plot of differential pore areas ............................................................................... 94

Figure ‎5.35 Plot of differential pore areas for macropores ...................................................... 95

Table ‎2.2 PWF rates at different CA/silica contents [32] ........................................................ 16

Table ‎5.3 Macrovoids numbers and dimensions relative to CA content ................................. 73

BSA Bovine Serum Albumin

CBT Coagulation Bath Temperature

CNTs Carbon Nanotubes

fNTs Functionalized Carbon Nanotubes

FTIR Fourier Transform Infrared Spectroscopy

ICIC Isocyanate and Isophthaloyl Chloride

ICT Initial Casting Thickness

ISE Initial Solvent Evaporation

MED Multiple-Effect Distillation

MSF Multi-Stage Flash

MWCNTs Multi-Walled Carbon Nanotubes

NMR Nuclear Magnetic Resonance

NOCC N,O-Carboxymethyl Chitosan

PEG Polyethylene Glycol

PEO Polyethylene Oxide

pNTs Pristine (Non-Functionalized) Carbon Nanotubes

PWF Pure Water Flux

SEM Scanning Electron Microscope

SLS Sodium Lauryl Sulphate

SWCNTs Single-Walled Carbon Nanotubes

TFC Thin Film Composite

v/v Volume by Volume

VCD Vapor Compression Distillation

wt% Weight Percent

1.2 Desalination Technologies

Figure ‎1.1 Desalination saline sources % worldwide [11]

Figure ‎1.2 Desalination processes % worldwide [11]

1.2.1 Multi-Stage Flash (MSF)