DESALINATION BY MEMBRANE DISTILLATION:

FABRICATION OF HIGH PERFORMANCE MEMBRANES

SINA BONYADI

(B. Eng. (Chemical) (Hons.), Amirkabir University of Technology)

A THESIS SUBMITED FOR

THE DEGREE OF MASTER OF ENGINEERING

DEPARTMENT OF CHEMICAL & BIOMOLECUALR

ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

2008

Acknowledgement

First of all, I would like to express my deepest heartfelt appreciation to my supervisor Prof. Tai-

Shung Chung Neal, in Dept. of Chemical & Biomolecular Engineering of NUS, for his excellent

guidance, enthusiastic encouragements and invaluable suggestions throughout my two year

master study. From him, I have learnt a great deal on both research knowledge and active work

spirits.

I am especially grateful to Prof. William B. Krantz Isac mayor professor in Dept. of Chemical &

Biomolecular Engineering of NUS for his wise guidance regarding my first paper in the journal

of membrane science.

Special thanks are also due to Mr. NG Kim Po the workshop chief at NUS Chemical Engineering

department for the fabrication of membrane distillation experimental set up.

I also want to take this opportunity to give my sincere thanks to all the colleagues in my research

group for their kind assistance. My association with them is a memorable part of my experience

and study at NUS.

I gratefully acknowledge A*STAR for providing me an opportunity to pursue my Master degree

and research scholarship.

i

Table of Contents

Acknowledgement………………………………………………………………………………....i

Summary…………………………………………………………………………………………..vi

List of Tables…………………………………………………………………………………….viii

List of Figures…………………………......……………………………………………………...ix

List of symbols…………………………………………………………………………………...xii

Chapter 1. Background Review and Objectives

1.1 Introduction…………………………………………………………….………………...1

1.2 Desalination Processes………………………………………………….………………..3

1.3 Alternative desalination processes………………………………….................................6

1.4 Introduction to membrane distillation as an alternative desalination approach…….........7

1.5 Membrane distillation configurations……………………………………………………8

1.6 Temperature polarization phenomenon…………………….............................................9

1.7 Wetting phenomenon………………………………………………………………….....9

1.8 Applications of membrane distillation……………………………………………….....10

1.9 Membrane distillation advantages and drawbacks…………………………………..….10

Chapter 2. Literature Review

2.1 Overview on MD literature……………………………………………………………...12

2.2 Literature review on membrane fabrication for MD…………………………………….13

2.3 Research objectives……………………………………………………………………...17

Chapter 3. Theory and model development

3.1 Mass transfer……………………………………………………………………………..19

ii

3.2 Heat transfer……………………………………………………………………………...22

3.3 Characteristics of a high performance MD membrane…………………………………....24

3.3.1 High membrane permeability……………………………………………………..24

3.3.2 High membrane wetting resistance and long term stability……………………....26

3.3.3 Suitable membrane geometry and dimensions……………………………………26

3.3.4 Hydrophilic layer porosity and thermal conductivity…………………………….27

Chapter 4. Experimental

4.1 Materials………………………………………………………………………………...28

4.2 Dope preparation………………………………………………………………………...29

4.3 Fabrication of flat sheet membranes…………………………………….........................30

4.4 Fiber spinning……………………………………………………………………………30

4.5 Morphology study of hollow fibers by SEM…………………………….........................31

4.6 Contact angle measurements……………………………………………..........................31

4.7 Porosity measurement……………………………………………………………………32

4.8 Pore-size distribution measurement……………………………………………………...33

4.9 Gas permeation test………………………………………………………………………33

4.10 Polymer flow observation by a high magnification camera……………………………..34

4.11 Video microscopy flow visualization……………………………………........................34

4.12 Module Fabrication………………………………………………………........................34

4.13 DCMD experiments……………………………………………………………………...35

Chapter 5. Fabrication of Dual Layer Hydrophilic-Hydrophobic Hollow Fibers

5.1 Effect of coagulant on the surface morphology of PVDF membranes…………………..37

5.2 First batch of fiber spinning……………………………………………………………...39

5.2.1 Membrane morphology…………………………………………………………...40

iii

5.2.2 DCMD performance……………………………………………………………...41

5.3 Second batch of hollow fiber spinning…………………………………………………..41

5.3.1 Hollow fibers morphology ………………………………………........................43

5.3.2 Pore-size distribution…………………………………………….........................45

5.3.3 Gas permeation and porosity measurement tests…………………........................45

5.3.4 Contact angle measurements……………………………………………………...45

5.3.5 DCMD results…………………………………………………………………….46

5.4 Summary………………………………………………………………………………...50

Chapter 6. Investigation of Corrugation Phenomenon in the Inner Contour of Hollow

Fibers during the Non-solvent Induced Phase-Separation Process

6.1 Introduction………………………………………………………………………………51

6.2 Experimental observations……………………………………………….........................53

6.3 Discussion………………………………………………………………………………..58

6.3.1 System description…………………………………………………………………...59

6.3.1.1 Phases I1, I2 or O1, O2…………………………………….……………………60

6.3.1.2. Phase I3 or O3…………………………………………………………………60

6.4 Possible instability mechanisms………………………………………………………...61

6.4.1 Hypothesis 1. (Mass transfer and hydrodynamic instability)………………………..62

6.4.2 Hypothesis 2. (Elastic and Buckling Instability)………………….............................64

6.5 Effect of air-gap distance………………………………………………………………...66

6.6 Effect of bore fluid composition…………………………………….…………………...68

6.7 Effect of external coagulant………………………………………….…………………..68

6.8 Effect of take-up speed………………………………………………..............................68

6.9 Effect of dope concentration…………………………………………..............................69

6.10 Summary………………………………………………………………………………...69

iv

Chapter 7. Conclusion…………………………………………………………………………..70

Bibliography……………………………………………………………………………………..72

Appendix…………………………………………………………………………………………84

v

Summary

For the first time, co-extrusion was applied for the fabrication of dual layer hydrophilic-

hydrophobic hollow fibers especially for the direct contact membrane distillation (DCMD)

process. The effect of different non-solvents on the morphology of the PVDF membranes was

investigated and it was found that weak coagulants such as water/methanol (20/80 wt%) can

induce a 3-dimensional porous structure on PVDF membranes with high surface and bulk

porosities, big pore size, sharp pore size distribution, high surface contact angle and high

permeability but rather weak mechanical properties. Hydrophobic and hydrophilic clay particles

were incorporated into the outer and inner layer dope solutions, respectively, in order to enhance

mechanical properties and modify the surface tension properties in the membrane inner and outer

layers. Different membrane characterizations such as pore size distribution, gas permeation test,

porosity and contact angle measurements were carried out as well. Ultimately, the fabricated

hollow fibers were tested for the DCMD process and flux as high as 55 kg/m2hr and energy

efficiency of 83% at 90 °C was achieved in the test. The obtained flux is much higher than most

of the previous reports, indicating that the application of dual layer hydrophilic-hydrophobic

hollow fibers may be a promising approach for MD.

In the second part of this research, by proposing a novel mechanism, we revealed one of the most

controversial issues in the hollow fiber fabrication process regarding the instability leading to the

deformed cross-section of fibers fabricated through nonsolvent induced phase separation. We

analyzed possible instability mechanisms based on our experimental observations and then

postulated that the principal instability occurs in the external coagulation bath where the rigid

precipitated polymer shell in the dope and bore fluid interface is buckled due to a generated

pressure. The pressure is postulated to be induced in the nascent fiber outer layer as a result of

vi

diffusion/convection, precipitation, densification and shrinkage In addition, the effect of some

spinning conditions such as air-gap distance, bore fluid composition, take-up speed, external

coagulant and dope concentration on the final shape of the fiber cross-section have been

investigated. The proposed mechanism was in good qualitative agreement with all our

observations.

vii

List of Tables

Table.2.1 Summary of commercial membranes applied by some studies in the literature

Table.4.1 Specifications of Cloisite particles

Table.5.1 Total solubility parameter (δt) of water, methanol, NMP and PVDF

Table.5.2 Spinning conditions applied for the first batch of spinning

Table.5.3 DCMD operating conditions and the obtained flux for the fabricated fibers

Table.5.4. Comparison of the maximum flux obtained in this study with the literature for DCMD processes with a hollow fiber configuration

Table.6.1 Spinning conditions of hollow fiber membrane fabrication

viii

List of Figures

Fig.1.1 Distribution of Earth’s water

Fig.1.2 Fresh water resources

Fig.1.3 Schematic presentation of a Multi-Stage Flash desalination plant

Fig.1.4 Schematic diagram of world’s desalination plants capacity percentage by 1998

Fig.1.5 Typical Costs for a Reverse-Osmosis Desalination Plant

Fig.1.6 Schematic diagram representing the separation mechanism involved in MD

Fig.1.7 Schematic diagrams representing different configurations of the MD process

Fig.1.8 Schematic diagram of temperature polarization phenomenon in DCMD

Fig.2.1 Schematic diagram representing a typical melt spinning process

Fig.2.2 A composite hydrophilic-hydrophobic membrane before the MD test (left) A composite hydrophilic-hydrophobic membrane during the MD test (right)

Fig.2.3 Schematic picture of a dual layer hydrophilic-hydrophobic fiber

Fig.3.1 Schematic DCMD process with dual layer hydrophilic-hydrophobic hollow fibers

Fig.4.1 Chemical structure of PVDF polymer

Fig.4.2 Chemical structure of PAN polymer

Fig. 4.3 Schematic representation of the fiber spinning line

Fig.4.4 Schematic diagram showing the steps involved in fabrication of lab scale membrane modules

Fig.4.5 Schematic experimental set-up applied for direct contact membrane distillation process

Fig.5.1 SEM pictures from the top surface (facing the coagulant) of the PVDF flat sheet membranes for different coagulant compositions

Fig.5.2 SEM pictures representing the structure of first spinning batch fibers

Fig.5.3 Delamination phenomenon during the DCMD process using the first batch of fibers

Fig.5.4 The spinning conditions applied in the second batch of spinning process

Fig.5.5 Cross section morphology of fibers fabricated through the second batch of spinning

ix

Fig.5.6 SEM micrographs showing the surface morphology of the dual layer fibers

Fig.5.7 Pore size distribution of the fabricated fibers

Fig.5.8 Contact angle measurements of the fabricated fibers

Fig.5.9 Flux and energy efficiency obtained in the DCMD process using the fabricated fibers and flux comparison with the literature data using hollow fiber membranes

Fig.6.1 Irregular shape in the cross section of the PAN 17% (left) and PVDF 20% (right) hollow fibers fabricated through wet spinning, Bore fluid composition of 40% NMP/water (NMP wt.%), free fall take up rate Fig.6.2 Cross section of spun fibers from 17 wt.% PAN solution, 40 wt.% NMP Bore fluid with different air-gap distances

Fig.6.3 Cross section of wet spun fibers from 17 wt.% PAN solution with different solvent amount in the bore fluid

Fig.6.4 Cross section of wet spun fibers from 17 wt.% PAN solution and 40 wt.% NMP bore fluid using different external coagulants: (A) IPA (B) water

Fig.6.5 Cross section of spun fibers with different take-up rates from 17 wt.% PAN solution, 40 wt.% NMP bore fluid, 7 cm air-gap

Fig.6.6 Cross section of spun fibers from different concentrations of PAN solution, 40 wt.% NMP aqueous mixture as bore fluid, with 3.5 cm air-gap and free fall (A) 13 wt.% PAN solution (B) 17 wt.% PAN solution (C) 22 wt.% PAN solution

Fig.6.7 Cross section of spun fibers from 20 wt.% PVDF solution, 40 wt.% NMP Bore fluid with different air-gap distances

Fig.6.8 Cross section of wet as-spun hollow fibers from 17 wt.% PAN solution and bore fluid containing 40 wt.% NMP

Fig.6.9 Magnified die swell pictures of spinning with 17 wt.% PAN solution, Bore fluid containing 40 wt.% NMP and 7 cm air-gap distance, Different bore flow rates

Fig.6.10 (A) Penetration of coagulant into the casting solution after 0.2 second using flow visualization photos taken on flat membrane inversion as an example, (B) Schematic regions in the extruded nascent fiber

Fig.6.11 Instability associated with proposed hydrodynamic and mass transfer mechanism

Fig.6.12 Schematic picture showing the exertion of inward radial forces generated by the shrinkage of nascent fiber outer layer in the external coagulation bath

x

Fig.6.13 Close agreement between the spun fibers cross sectional geometry (A) and the predicted postbuckling shapes of a long elastic cylindrical shell (B)

Fig.6.14 Energy as a function of the dimensionless group including the external pressure for various modes of deformation

Fig.6.15 Equilibrium state transitions of the buckled elastic shell as a function of spinning conditions

xi

List of Symbols

A, B, C Antoine Equation Constants

Afo hollow fiber outer surface area (m2)

Afi hollow fiber inner surface area (m2)

Ce Membrane lumped diffusive coefficient (m2/s)

Cf feed heat capacity (kJ/kg ºK)

Cp permeate heat capacity (kJ/kg ºK)

Cv water vapor heat capacity at the constant volume (kJ/kg ºK)

Dwk water vapor Knudsen diffusion coefficient (m2/s)

Dw-a water vapor-air molecular diffusion coefficient (m2/s)

di fiber inner diameter (m)

F flexural rigidity

G Young’s modulus of elasticity (Pa)

Hp permeate side overall heat transfer coefficient (J/m2°C)

h Shell thickness (m)

hf feed-side heat transfer coefficient (J/m2°C)

hp permeate-side heat transfer coefficient (J/m2°C)

kair thermal conductivity of air (m2/s)

ki thermal conductivity of the fiber inner layer (m2/s)

k thermal conductivity of the fiber outer layer (m2/s)

kB Boltzmann constant

km thermal conductivity of the membrane matrix (m2/s)

kw water thermal conductivity (m2/s)

L fiber length (m)

xii

Ma air molecular weight

Mw water molecular weight

fm feed mass flow rate (kg/hr)

pm permeate mass flow rate (kg/hr)

Nw water vapor mass flux (kg/m2 hr)

Nu Nusselt number

n number of hollow fibers in the membrane module

P Dynamic pressure (Pa)

Pa air partial pressure (Pa)

Pw water vapor partial pressure (Pa)

*P Critical buckling pressure (Pa)

Pr Prandtl number

q Buckling mode

r radial coordinate (m)

rp pore radius

R universal gas constant

Rh cylinder radius (m)

Ri fiber inner radius (m)

Rm hydrophilic layer outer radius (m)

Ro fiber outer radius (m)

Re Reynolds number

Red Reynolds number based on fiber inner diameter as the characteristic length

T temperature (C)

Tfi feed inlet temperature (C)

Tpi permeate inlet temperature (C)

xiii

t Time (s)

U Potential Energy (kg m/s2)

u

Velocity vector (m/s)

V fiber empty volume (m3)

v Poisson’s ratio

w1 weight of the sample before immersion in kerosene (kg)

w2 weight of the sample after immersion in kerosene (kg)

xw mole fraction of water in the feed

z axial coordinate (m)

σ collision diameter of the molecule (m)

δm membrane thickness (m)

τ membrane tortuosity

viscous flux (Pa)

density (kg/m3)

ε membrane bulk porosity

εi inner layer bulk porosity

α constant

γ water activity coefficient

γL liquid-vapor surface tension (J/m2)

θ angular coordinate

θef effective contact angle

ρk kerosene density (kg/m3)

ΔHv water latent heat of vaporization (J)

ΔP pressure difference at liquid-vapor interface (Pa)

ΔTp difference between the permeate inlet and outlet temperatures (ºC)

xiv

Chapter 1. Background Review and Objectives

1.1 Introduction

As time goes by, the fresh water shortage becomes a more troublesome issue for human being.

According to the Global Environmental Outlook report by United Nations in 2000, water

shortage, besides the global warming, has been considered as the most worrying problem for the

new millennium [1]. The fresh water shortage is probably related to the rapid increase of human

population and scarcity of the fresh water resources. The following figure schematically shows

the distribution of water on earth.

Fig.1.1 Distribution of Earth’s water [2]

According to the above figure, oceans contain 97% of the world’s water by volume, which is too

salty for drinking. The salinity of oceans is due to the gradual accumulation of dissolved

chemicals eroded from the earth’s crust and washed into the sea by rivers. The average salinity of

sea water is 3.5 wt%, but concentrations as high as 4.0 wt% are observed in the Red Sea and the

Persian Gulf [3]. From the remaining 3% of earth water resources which is considered fresh

1

water, 98.8% is trapped in polar ice caps and ground water. Thus, only about 0.03 % of earth’s

total available water by volume is available for human use. Water shortages are expected to

become even more sever in the future due to increasing demand of water and water pollution

issues. However, the water shortage problem is not uniform among different regions and

countries. Depending on the geographical conditions some regions are already facing a water

crisis. The following map shows the distribution of countries with respective shortage of water in

1995 and 2025.

Fig.1.2 Fresh water stress [4]

The above map indicates that a large area in Middle East, North Africa, India and some parts of

Europe are going to suffer from the water shortage problem in the next twenty years. Therefore,

finding alternative water resources seems to be vital to alleviate the water shortages. In this

respect, water desalination has been considered as an important possible fresh water resource for

a long time.

1.2 Desalination Processes

2

The main challenge against desalination is to deal with the required energy imposed by basic

thermodynamics for separating salt and water. Theoretically, the minimum energy for

desalinating seawater is around 2.7kJ/kg [5, 6]. In practice, it is not possible to operate processes

at the theoretical minimum energy. Therefore, there is always a trade-off between energy use,

production capacity, and capital costs for optimized industrial processes [7, 8]. Three basic

approaches have been considered for separating water from salt [9]. The first approach is to

induce a phase change in water (to vapor or solid) through heating or cooling, separate the pure

new phase from the remaining salt solution, and then recover the thermal energy for reuse as the

vapor or solid transform to liquid. The second approach is based on the application of semi-

permeable membranes as filters to retain the salt, as saline water is forced through the membranes

due to an externally applied chemical potential gradient. The chemical potential driving force

may be induced by pressure, concentration or electric fields. Finally, there are chemical

approaches to desalination, including ion exchange, liquid-liquid extraction, and gas hydrate or

other precipitation methods [9]. All desalination processes developed or proposed to date have

employed one or more of these three approaches.

Within the three basic approaches to desalination, some of methods have been commercialized on

a large scale; which are mainly distillation and membrane processes. Distillation processes,

including multi-stage flash (MSF), multiple effect evaporation (MEE), and thermal or mechanical

vapor compression (VC), have been used on a large scale for decades, particularly in the Middle

East, and by 1998 accounted for almost 48% of the world’s desalination capacity [10].

Distillation suffers from the inherent drawback of the very large latent heat of vaporization of

water (2200 kJ/kg). The evaporation and condensation steps in these processes are often coupled

so that the latent heat is recovered for reuse by preheating the feed water. Normally, large

distillation units are coupled with steam or gas turbine power plants for better utilization of the

fuel energy. The efficiency and energy consumption of these processes depends largely on the

3

effectiveness of recovery of this latent heat in the condensation step. The following figure

schematically demonstrates an MSF desalination plant [10].

Steam heater

Vacuum System

Primary steam source

Flash chambers

Concentrate Out Heat rejection stages

Product Out

Condensate collection trays

Fig.1.3 Schematic presentation of a Multi-Stage Flash desalination plant [11]

In large scale thermal processes, energy consumption has been reported from about 20kJ/kg (VC,

MEE) to more than 200kJ/kg (MSF) of fresh water produced, actually independent of salt

concentration [12]. Such high energy consumption rates have made thermal processes too costly

and energy inefficient.

On the other hand, large scale membrane desalination plants have found great applications during

the last 30-40 years. Three different water resources including seawater, brackish water

(groundwater and surface water), and tertiary treated wastewater can be treated by membranes for

desalination purposes. The most commonly used membrane process for desalination is reverse

osmosis. Reverse osmosis is a separation process that applies pressure higher than a liquid

osmotic pressure to force it through a membrane that retains the solute on one side and permits

the pure liquid to pass to the other side of the membrane. This is the reverse of the normal

osmosis process in which solvent from an area of low solute concentration naturally moves

4

through a membrane, to an area of high solute concentration, when no external pressure is

applied. The pressure exerted on the high concentration side of the membrane, is usually in the

range of 2–17 bar (30–250 psi) for fresh and brackish water, and 40–70 bar (600–1000 psi) for

seawater, which has around 24 bar (350 psi) natural osmotic pressure which must be overcome.

Energy requirements for the reverse osmosis process are highly dependent on concentration, and

range from about 10kJ/kg for brackish water and to about 20kJ/kg for seawater; however, new

low energy seawater RO membranes are expected to consume as little as 7.2kJ/kg [10]. The

following figure schematically demonstrates the capacity percentage of desalination plants

worldwide by 1998 [13].

Reverse Osmosis

42%

Electrodialysis 6%

Multi-Effect Distillation

4%

Vapor Compression

4%

Multi-Stage Distillation

44%

Fig.1.4 Schematic diagram of world’s desalination plants capacity percentage by 1998

Compared to energy consumption figures, it is more difficult to estimate cost figures for a

desalination plant. This is mainly due to the fact that many costs, particularly energy costs,

depend greatly on factors such as time, water quality and geography. Also, the transportation cost

of the water to the treatment or distribution point highly depends on location, similar to the cost

of disposing of the resulting concentrate solution [14]. Taking all these limitations into account,

total water costs of $0.75 - $1.5/m3 for thermal processes, vs. $0.25 - $0.70/m3 for brackish water

RO and $0.45 - $1.25 m-3 for seawater RO have been estimated based on a recent review of the

5

major current desalination processes [12]. Mainly because of the high energy consumption of

thermal processes compared to RO, membrane processes currently have an enormous economic

advantage.

Fig.1.5 Typical Costs for a Reverse-Osmosis Desalination Plant [15]

1.3 Alternative desalination processes

Despite the success of current technologies, a considerable energy and time is devoted to research

and development of alternative desalination processes. The main driving forces for these efforts

are the need to improve the productivity and the efficiency of a process as well as the desire to

decrease costs. For example, membrane fouling is becoming widely accepted as the largest cause

of permeate flux decline at normal operating pressures and temperatures in brackish water

systems [9, 12]. Furthermore, as the oil price goes up, the operation cost of RO systems increase

very fast. Therefore, alternative desalination approaches offering lower energy consumption or at

least lower energy costs of capital equipment or operation and maintenance are of great interest

among researchers.

Many of the alternate processes proposed in the literature focus on utilization of low-grade

alternative energy sources such as waste heat from other industrial processes like power

6

generation plants, solar energy and thermal or mechanical energy from ocean [12]. Among these

approaches, membrane distillation is considered as an emerging desalination alternative providing

several promising advantages over the conventional RO and distillation technologies. Although

energy consumption in this process is quite high, the process is typically run at relatively low

temperature (40 ºC-50 ºC) and thus can make use of waste heat or other relatively low grade heat

sources [12].

1.4 Introduction to Membrane Distillation as an alternative desalination approach

For the first time, membrane distillation was introduced through a US patent by Bodel in 1966

[16]. In this process, a micro-porous hydrophobic membrane is brought into contact with an

aqueous heated solution on one side of the membrane (feed side). The hydrophobic nature of the

membrane prevents penetration of liquid stream through the membrane and creates a vapor liquid

interface at the entrance of each pore. Here, volatile compounds (typically water) evaporate,

diffuse and/or convect across the membrane and condense or remove on the opposite side

(permeate side) of the system [17-20].

Fig.1.6 Schematic diagram representing the separation mechanism involved in MD [19]

1.5 Membrane Distillation Configurations

7

Depending on the method used to induce the vapor pressure gradient across the membrane, MD

can be classified into four configurations [17-19]. The most common arrangement is known as

Direct Contact Membrane Distillation (DCMD) in which a condensing fluid stream (often pure

water) is applied in the permeate side of the membrane that is directly in contact with membrane.

Therefore, the temperature difference across the membrane induces a vapor pressure gradient

which is considered as the required mass transfer driving force. In another configuration called

Air-Gap Membrane Distillation (AGMD), a gap of stagnant air is introduced between the

membrane and a condensation surface. Therefore, water vapor molecules have to penetrate

through both the membrane and the air gap to finally condense over a cold surface inside the

membrane module. In Sweeping Gas Membrane Distillation (SGMD) arrangement, the water

vapors diffused to the membrane permeate side are entrained into a condenser by a sweeping gas

such as air and finally pure water is collected through the condenser. In Vacuum Membrane

Distillation (VMD), vacuum is applied in the permeate side of the module by means of a vacuum

pump. The applied vacuum pressure is lower than the saturation pressure of volatile molecules to

be separated from the feed solution. Similar to SGMD, condensation occurs in a separate

condenser outside the membrane module [17-19]. The following pictures schematically illustrate

these arrangements.

Fig.1.7 Schematic diagrams representing different configurations of the MD process [19]

1.6 Temperature Polarization Phenomenon

8

As a result of water evaporation and heat conduction and convection from the hot feed side of the

membrane to the cold permeate side, the temperature on the membrane feed side surface is lower

than the feed bulk temperature. The similar phenomenon occurs in the permeate side in case of a

DCMD configuration in which the temperature on the membrane surface will be higher than the

permeate bulk temperature as a result of water vapor condensation; heat conduction and

convection from to the membrane permeate side. Therefore, the mass transfer driving force

(water vapor partial pressure difference at two sides of the membrane) which depends on the

membrane surface temperatures at two sides will be lower. This phenomenon is known as

temperature polarization and leads to flux reduction across the membrane [17-19].

Fig.1.8 Schematic diagram of temperature polarization phenomenon in DCMD

1.7 Wetting Phenomenon

According to the described MD mechanism in the previous section, the hydrophobic membrane

pores are supposed to be dry or in other words occupied by water vapors only. However, if the

liquid stream penetrates into the membrane pores or vapor condensation occurs in the membrane

matrix wetting phenomenon occurs. Depending on the membrane material, pore size and surface

morphology, there is a critical penetration pressure above which the liquid stream on the feed or

9

permeate side will penetrate into the membrane [17-19]. This pressure is known as the liquid

entry pressure of water (LEPW) and usually calculated through the following Laplace-Young

equation:

(1.1)

where ΔP is the pressure difference at the liquid-vapor interface, γL is the liquid-vapor surface

tension, θef is the effective contact angle and r is the pore radius.

efp

L

rP

cos

2

1.8 Applications of Membrane Distillation

Based on the mechanism governing the separation process in MD, this process is mostly suitable

for separation of non-volatile components from more volatile liquid solvents such as water [17-

19]. In this respect, desalination has been considered as the mostly used application for MD.

However, MD has also been applied for the environmental waste treatment such as removal of

trace organic compounds from water such as benzene, chloroform and thrichloroethylene [21,

22]. Food processing such as milk and juice concentration is the other application area for MD in

the literature [23, 24]. Other applications of MD include separation of azeotropic mixtures such as

alcohol-water [25, 26], treatment of humic acid solutions [27, 28] and treatment of waste water

contaminated with dyes [29].

1.9 Membrane Distillation Advantages and Drawbacks

The evaporative nature of membrane distillation process makes it possible to achieve an almost

complete rejection of non-volatile solutes, such as macromolecules, colloidal species and ions

through this process. Compared to conventional distillation columns, membrane distillation units

offer a much larger mass transfer area per unit volume which is the characteristic of membrane

based processes. This translates into more compact units with similar fresh water production

10

11

capacities. In addition, the operating temperatures in MD can be maintained as low as 30-50 °C,

thus permitting the efficient recycles of low grade or waste heat streams, as well as the use of

alternative energy resources such as solar power. Furthermore compared to reverse osmosis

process, this process does not suffer limitations of concentration polarization and high operating

pressures. As a result, this process is less susceptible to fouling compared to RO [17-19].

However, despite the distinct advantages MD offers, this process has not been yet implemented in

industry. The main barriers to commercial implementation of this process include [18]:

1- Relatively low permeate flux compared to other separation techniques such as reverse

osmosis

2- Permeate flux decay due to membrane fouling and total or partial pore wetting.

3- Uncertain energy and economic costs for each MD configuration and application

Chapter 2. Literature Review

2.1 Overview on MD literature

The research in MD literature can be divided into different categories based on the area of the

research conducted by researchers during the last 30 years. The initial attempts in 1980s mainly

involved experimental studies using commercial membranes combined with modeling the process

[30-33] for the following goals: 1) to better understand the physics underlying the process 2) to

elaborate the complex heat and mass transfer involved and 3) to estimate the water vapor flux in

the process and investigate the effect of different operating parameters such as permeate and feed

flow rates on the obtained flux. After 30-40 years, it seems that this area of MD research has

become mature and a good understanding of the physics involved in the process has been

achieved. In addition, the theoretical models proposed in the literature can predict the flux with a

high certainty. Another research trend can be identified as the attempt to suppress the undesirable

temperature polarization in the process by fabrication of novel membrane modules, applying

spacers and static mixers in order to induce a better flow geometry inside the module. However,

most of the studies in this area have been related to MD using flat sheet membranes [34-38] and

there are relatively fewer investigations using hollow fibers [39]. Finally, some published reports

can be grouped as the attempts to fabricate novel membranes with higher flux and more

durability. However, the number of published reports on fabrication of MD membranes is much

fewer compared to the previous research trends addressed previously [18]; while many facts

regarding the effects of membrane properties such as morphology, pore size distribution and

thickness on overall MD performance are still unknown. As a result, to the author’s opinion, it is

the time to devote more experimental efforts on fabrication of high performance membranes

optimally designed for the MD process. In this respect, the trends of this research have been

chosen to cover the gaps identified above to some extent.

12

As a result, it is the main objective of this research to enhance the flux in MD process by

fabrication of novel high performance membranes. In this respect, we have devoted this chapter

to review membrane distillation literature regarding membranes applied in this process and their

characteristics.

2.2 Literature review on membrane fabrication for MD

The membranes commonly applied in membrane distillation literature have been the

commercially available membranes actually fabricated for other membrane applications such as

micro-filtration and ultra-filtration [40]. Since the hydrophobic character of the membrane

represents a crucial requirement in MD, membranes have to be made of hydrophobic polymers

having a low surface energy. The most popular hydrophobic polymers applied in the MD include

Polytetrafluoroethelyne (PTFE), Polypropelene (PP) and Polyvinyldene flouride (PVDF).

PTFE having a low surface energy of 9.1 × 10-3 is a highly crystalline polymer which shows

excellent thermal stability and chemical resistance. PP similar to PTFE has a highly crystalline

structure but has higher surface energy compared to PTFE (30.0 × 10-3 N/m). These two polymers

have very low solubility practically in all common solvents. Therefore, micro-porous membranes

are fabricated out of these polymers usually by sintering and stretching methods rather than phase

inversion approaches [41].

In a sintering approach, powders of polymeric particles are pressed into a film or plate and

sintered just below their melting point. This process yields to micro-porous structures having

porosities in the range of 10-40% and a rather irregular pore size distribution. The typical pore

size determined by the particle size of the sintered powder ranges from 0.2 to 20μm [41].

13

In a stretching approach, films are obtained by extrusion of polymeric materials at temperature

close to their melting point coupled with a rapid draw down. Crystallites in the polymer are

aligned in the direction of drawing; after annealing and cooling a mechanical stress is applied

perpendicularly to direction of drawing. This manufacturing process gives a relatively uniform

porous structure with pore size distribution in the range of 0.2-20 μm but it is difficult to obtain

porosities higher than 80% [41].

Fig.2.1 Schematic diagram representing a typical melt spinning process

On the other hand, PVDF is a semi-crystalline polymer with a surface energy of 30.3 × 10-3 N/m

and very good thermal and chemical resistance. This polymer can easily be dissolved in common

solvents such as N-Methyl-2-Pyrrolidone (NMP), Dimethylacetamide (DMAC) and

Dimethylformamide (DMF). Therefore, PVDF membranes can be fabricated through phase

inversion process through which by controlling the fabrication parameters such as polymer

solution concentration, type of coagulants and additives, membranes with higher porosity and

14

narrower pore size distribution can be achieved compared to sintering and stretching approaches

[41].

Although most of the studies in MD (as shown in Table1) have used commercial membranes

originally fabricated for micro-filtration, recently a number of studies have produced their own

membranes in order to improve the flux and separation in this process. There is considerable

number of studies in the literature regarding the fabrication of PVDF membranes applied for

ultra-filtration process [42-45]. However, during the last 10 years, researchers have started to

investigate on the fabrication of PVDF membranes specifically applied for membrane contactor

processes for CO2 capture [46-48], H2S capture [49, 50], VOC removal [51] and desalination

through vacuum and direct contact membrane distillation processes. Oritz de Zarate et al. [52]

fabricated asymmetric flat sheet PVDF membranes through phase inversion approach using

DMAC and DMF as the solvents. They observed that porosity and pore diameter increased as the

PVDF concentration decreased. In another study, Tomaszewska [53] fabricated flat sheet PVDF

membranes using lithium chloride (LiCl) as an additive to the casting solution. The additive LiCl

increased the porosity and pore size of the fabricated membranes. This led to some flux

enhancement in the consequent MD experiments. In a similar study, Khayet and Matsuura [54]

found that the pore size and porosity of the membrane increase with increasing the concentrations

of pure water as a non-solvent into the dope solution. Also, they concluded that the MD flux

increases exponentially with the water content in the PVDF casting solution.

Fabrication of hydrophilic-hydrophobic membranes for MD has been another area of interest for

researchers. For the first time, Cheng and Wiersma [55] described the use of composite

membranes in MD in a series of patents. They modified a cellulose acetate membrane via

radiation graft polymerization of styrene onto the membrane surface, and a cellulose nitrate

membrane via plasma polymerization of vinyltrimethylsilicon/carbon tetrafluoride and

15

octafluoro-cyclobutane. Wu et al. [56] applied hydrophilic porous supports such as cellulose

acetate, and treated the membrane surface via radiation graft polymerization of styrene to enhance

the hydrophobicity. In a similar way, Kong et al. [57] modified a cellulose nitrate membrane via

plasma polymerization of both vinyltrimethylsilicon/carbon tetrafluoride and octafluoro-

cyclobutane.

Such composite membranes have been considered to be applied in MD in two different

configurations. In the first arrangement, a thin hydrophobic functional layer is supported by a

rather thick hydrophilic layer so that the water vapor mass transfer resistance through the

hydrophobic layer is minimized and a greater flux can be obtained. Fig.2.2 demonstrates this

concept schematically.

In the literature, Khayet and Matsuura, and their coworker [58, 59] fabricated flat sheet

hydrophilic - hydrophobic membranes for this purpose. They fabricated such composite

membranes based on the migration of hydrophobic macromolecules (SMM) to the membrane

surface. However, the flux obtained through their experiments was very low. This might be due

to the difficulty to optimize the membrane surface porosity and pore-size distribution in this

process.

Hydrophobic layer

Hydrophilic layer

Fig. 2.2 A composite hydrophilic-hydrophobic membrane before the MD test (left) A composite hydrophilic-hydrophobic membrane during the MD test (right)

16

The second area that the application of composite hydrophilic-hydrophobic membranes may be

promising is the fouling and wetting prevention of the hydrophobic functional layer by coating a

very thin hydrophilic layer in contact with the feed salt solution. The hydrophilic layer is thought

to be less susceptible to fouling and scaling phenomena. In this respect, Peng et al. [60] tested the

desalination through DCMD process by applying a composite PVA/PEG hydrophilic layer on a

hydrophobic PVDF substrate. The authors investigated the effects of brine temperature, salt

concentration, running time and the addition of ethanol on the flux of composite membranes.

They observed that durability of the membranes greatly improved; while the flux obtained

through the process did not change greatly compared to the flux in case of the single layer

hydrophobic membrane.

Among the fabrication methods applied in the literature, both radiation graft and plasma

polymerizations are expensive processes that limit their applications. On the other hand, as

pointed out above, fabricating these types of membranes based on the hydrophobic

macromolecules migration to the surface suffers from the difficulty to optimize the hydrophobic

layer thickness and morphology. In addition, all the related reports in the literature so far have

been limited to the flat sheet membranes while hollow fiber is the most preferable membrane

configuration for the industry because of providing high surface area per unit volume as well as

ease of module fabrication.

2.3 Research objectives

In overall, despite the attempts that have been initiated to fabricate high performance membranes

for the MD process, the obtained water vapor fluxes through the reported studies scatter greatly

(0.2-40 kg/m2hr) and are quite low in average. As a result, the objective of this research is to

develop novel high performance membranes and optimize their different separation properties in

17

order to enhance the MD flux and at the same time maintain a high salt rejection rate and stable

long term performance of the membranes. It is also worthy to note that the focus of this research

will be on hollow fiber membranes mainly because they have been considered as the most

suitable membrane configuration for the industrial applications. Hollow fibers provide a high

surface area per unit volume and ease of module fabrication. In the following chapter, we

demonstrate the fabrication of dual layer hydrophilic-hydrophobic hollow fibers through a novel

co-extrusion approach.

Fig.2.3 Schematic picture of a dual layer hydrophilic-hydrophobic fiber

18

Chapter 3. Theory and model development

In this chapter we review the concepts of heat and mass transport phenomena involved in the MD

process and later we develop a mathematical model for the DCMD process in a hydrophobic-

hydrophilic hollow fiber membrane to fit our experimental flux data in the next chapters. Finally,

we discuss the characteristics of a high performance dual layer hydrophilic-hydrophobic fiber

based on the theoretical concepts.

Fig.3.1 Schematic DCMD process with dual layer hydrophilic-hydrophobic hollow fibers

3.1 Mass transfer

Water vapor transport through hydrophobic membranes in MD process can be considered as the

mass transfer within the gas-phase of a porous medium. Depending on the pore size of the porous

medium, continuum and Knudsen regions can be identified in the membrane matrix [19, 61]. In

the continuum region, the mean free path of the gas is small compared with the pore diameter. In

19

this region, molecule-molecule collisions are dominant over molecule-wall collisions and the

mass transfer flux can be described by the Fick’s law. On the other hand, in a Knudsen region the

mean free path of the gas is larger than the pore diameter. As a result, molecule-wall collisions

are dominant over molecule-molecule collisions and the mass transport can be described by

Knudsen’s law. Transition region between continuum and Knudsen regions may also be

identified when the mean free path of gas molecules is in the same order as the membrane pore

diameter. Dominance, coexistence, or transition between all of these different mechanisms can be

quantitatively estimated by dimensionless Knudsen number which is defined as the ratio of mean

free path of diffusing molecules to the mean pore size of the membrane. According to the kinetic

theory of gases, the mean free path of molecules can be calculated through [19, 61]:

22

P

TkB (3.1)

where kB is the Boltzmann constant (1.380 × 10-23 JK-1), and σ is the collision diameter of the

molecule (σ =2.7Aº for water). For the binary mixture of water vapor in air, the average free

mean path of air-water molecules mixture can be evaluated at an average operating temperature

through the following equation [19]:

a

wwa

Bwa

M

MP

Tk

12

2,

(3.2)

where σa (=3.7 A° ) and σw are the collision diameters, and Ma and Mw the molecular weight for

air and water, respectively. For an average temperature T=60 ºC, Phattaranawik et al. [62]

reported a mean free path of 0.11µm that is in the same order as the pore size of the micro-porous

membranes commonly applied in the MD process. As a result, the transition region between

Knudsen and ordinary molecular diffusion is normally considered as the governing mass transfer

mechanism for the MD membranes.

20

Accordingly, the flux relationship in a DCMD process has been given in the literature as follows

[19, 61].

wa

ww

w

aw

a

wk

ww PP

PN

dr

dP

D

P

DRT

MN

11/

(3.3)

where Nw is the water vapor mass flux, Mw is the water molecular weight, R is the universal gas

constant, Pw and Pa are the water vapor and air partial pressures, respectively, r is the radial

coordinate, ε and τ are the membrane porosity and tortuosity, respectively, T is the temperature,

Dwk and Dw-a are the water vapor Knudsen and molecular diffusion coefficients, respectively.

According to this equation, the flux depends on the partial pressure of air trapped inside the

membrane, which brings some uncertainty into the modeling practice as the air partial pressure

profile depends on the initial amount of air trapped in the membrane matrix, which itself varies

upon the operating conditions such as feed and permeate hydrodynamic pressure and also the start

up procedure as well. In addition, there is no practical approach to determine the tortuosity factor

exactly. In order to avoid these complexities, researchers have often applied the following

simplified flux equation with a lumped parameter Ce which is fitted into the experimental data.

[19]

dr

dP

RT

MCN ww

ew (3.4)

Using a similar approach and assuming a steady state and one dimensional mass transfer, the

following equations and boundary conditions can be applied:

01

wrNdr

d

r ,

dr

dP

RT

MCN ww

ew (3.5)

ww xTC

BAP

exp at oRr (3.6)

TC

BAPw exp at mRr (3.7)

21

where A, B, C are Antoine constants, γ is the water activity coefficient in an aqueous sodium

chloride solution, Ro and Rm are the outer and inner radii of the hydrophobic layer, respectively

and xw is the mole fraction of water in the feed.

3.2 Heat transfer

Beside mass, heat is also transferred from the feed to the permeate side by conduction through the

membrane matrix as well as convection of water vapors across the membrane. As a result of water

vaporization and heat conduction at the membrane feed side, the membrane surface temperature at

this side will be lower than the feed bulk temperature. On the other hand, because of water vapor

condensation and heat conduction at the permeate side, the membrane surface temperature at this

side will be higher than the permeate bulk temperature. This phenomenon is known as temperature

polarization which lowers the temperature and vapor pressure driving forces across the membrane

and consequently leads to a lower flux. The following equations describe the explained heat

transfer for a steady state and one dimensional process mathematically [63].

dr

dTrk

dr

d

rdr

dTCN vw

1 (3.8)

ffvw TThd

r

dTkHN at oRr (3.9)

ppvw TThd

r

dTkHN at mRr (3.10)

where k is the membrane hydrophobic layer thermal conductivity, Cv is the water vapor heat

capacity at the constant volume, hf and hp are the feed-side and permeate-side heat transfer

coefficients, respectively and ΔHv is the water enthalpy of vaporization. As heat is transferred

between two sides of the membrane, the feed temperature decreases while permeate temperature

increases along the membrane module. This can be expressed in terms of equations as follows

[63]:

22

TThnATCmdz

dfffofff at r at oRoRr (3.11)

fif TT at 0z (3.12)

TTHnATCmdz

dpppoppp at mRr (3.13)

pip TT at 0z (3.14)

(3.15)

i

mi

im

p

p

R

Rk

RR

h

H

ln

11

ph

1= convection resistance at permeate side

i

mi

im

R

Rk

RR

ln

=hydrophilic layer resistance (3.16)

where Hp is the overall heat transfer coefficient in the permeate side, ki is the thermal conductivity

in the hydrophilic layer,

and are the feed and permeate mass flow rates, respectively,

while Cf and Cp are the feed and permeate heat capacities, respectively. n is the number of hollow

fibers in the membrane module, Afo and Afi are the hollow fiber outer and inner surface areas,

respectively. Tfi and Tpi are the feed and permeate inlet temperatures, respectively.

fm pm

In order to estimate the flux obtained through the process, the following procedure was adopted:

First, the feed and permeate heat-transfer coefficients were estimated using the existing

correlations in the literature. For the laminar permeate flow inside the fibers the following

equation was applied [64].

Nu=1.86 (di/L Red Pr)0.33 (3.17)

Nu=hp di/kw (3.18)

23

where di is the fiber inner diameter, L is the fiber length, Red is the Reynolds number based on

fiber inner diameter as the characteristic length and Pr is the Prandtl number. The heat-transfer

coefficient in the module shell side was estimated using the following equation, which is

correlated by Mengual et al. [65] for an external parallel flow along the fibers.

Nu=0.042 Re0.59 Pr0.33 (3.19)

Nu=hf L/kw (3.20)

The membrane thermal conductivity was estimated using the following equation which has

commonly been applied by researchers in membrane distillation literature [17, 19].

k = kair ε + km (1-ε) (3.21)

where k is the effective membrane thermal conductivity, kair and km are the air and polymer matrix

thermal conductivities, respectively The describing equations (3.3 - 3.21) were solved using an

iterative finite difference scheme and the membrane diffusivity coefficient Ce was obtained by

fitting the calculated flux to the experimental data at different feed inlet temperatures.

3.3 Characteristics of a high performance MD membrane

3.3.1 High membrane permeability

Based on the mass transfer mechanism in MD membranes that was described in the previous

section, highly gas permeable membranes are essential to achieve a high flux through the process.

In this respect, membrane porosity has been known to be a key parameter influencing the

membrane gas permeability [17-19]. According to the equations (4) and (5), as the membrane

bulk porosity increases the molecular and Knudson diffusivities of the vaporized molecules

diffusing through the membrane increase in a proportional manner. Furthermore, as the

membrane bulk porosity increases, the membrane thermally conductivity decreases and leads to a

24

lower heat loss and a higher energy efficiency of the membranes. The energy efficiency is usually

defined as the fraction of feed brine thermal energy that has been used to evaporate water. Energy

efficiency can be easily calculated using the following equation [64]:

(3.22)

Where J is the water vapor flux, ΔHv is water vapor latent heat of condensation, Ao is the total

membrane area based on the fibers’ outer diameter, is the permeate mass flow rate, Cp is the

specific heat capacity of the permeate stream and ΔTp is the difference between permeate inlet

and outlet temperatures. This is due to the fact that a higher volume percentage in the membrane

will be occupied by the less thermal conductive air molecules (kair=0.024 W/m°K) compared to

the polymer matrix (km=0.1-0.3 W/m°K).

pm

ppp

ov AHJEfficiencyEnergy

TCm

It is also worthy to note that besides membrane bulk porosity, membrane surfaces must be highly

porous as well, so that a high liquid-vapor interfacial area is formed at the membrane pore

mouths. In other words, a membrane with a higher surface porosity provides a higher effective

mass transfer area which translates into a greater trans-membrane flux [66].

In addition to the membrane porosity and pore size, membrane morphology and pores

interconnectivity are other determining factors that have been never discussed in the MD

literature. However, there are a number of studies regarding the influence of these parameters in

other membrane separation processes. Li and Chung [67] identified pore interconnectivity to have

a great influence on the pure water permeability (PWP) of SPES hollow fibers applied in ultra-

filtration. In other words, between two membranes with similar porosities, the membrane with an

open cell structure or a high degree of pores interconnectivity will possess a higher permeability.

In another study, Widjojo et al. [68] found that high pores interconnectivity in the supporting

25

layer of the 6FDA-ODA-NDA/PSF dual-layer hollow fiber membranes could greatly reduce the

substructure resistance in gas separation applications. Therefore, high porosity, big pore size and

high degree of pores interconnectivity should be maintained in order to maximize the membrane

permeability.

3.3.2 High membrane wetting resistance and long term stability

As discussed in chapter 1, membrane wetting is an undesirable phenomenon in MD that may lead

to flux decay or lower separation efficiency. In order to prevent this phenomenon in long term

MD operations, high membranes liquid entry pressure, proper pore size distribution (0.2-0.5 µm)

[69] and a reasonable degree of membrane tortuosity are essential.

In term of membrane permeability, macrovoids may seem to form desirable membrane structures

with low tortuosity and high porosity. But from a long term performance point of view,

macrovoids might be undesirable structures that enhance the wetting phenomenon in membranes,

consequently reduce the membranes long term stability. Therefore, fabrication of macrovoid-free

and fully sponge-like membranes with sharp pore size distributions is considered as the other

characteristics of an ideal MD membrane.

3.3.3 Suitable membrane geometry and dimensions

Similar to other membrane processes, membrane thickness has a significant effect on the overall

mass transfer resistance in the MD process [18]. However, in contrast to other membrane

approaches the membrane flux does not have a monotonically increasing trend as function of

membrane thickness [70, 71]. This is due to the fact that flux in MD depends on the coupled

effect of mass and heat transport phenomena. On one hand, a smaller membrane thickness

decreases the mass transfer resistance introduced by the membrane and leads to a higher flux. On

the other hand, as the membrane thickness decreases, heat loss through the membrane increases

26

which translates into a lower temperature gradient across the membrane, a lower driving force

and ultimately a lower flux [70, 71]. Therefore, there should be an optimum thickness for the

membranes depending on their other properties as well as the flow geometry. One way to tailor

the membrane thickness to the optimum range is to fabricate hydrophilic-hydrophobic

membranes in which a thin functional hydrophobic layer is supported by a hydrophilic layer. In

this way, as discussed above, the water vapor mass transfer resistance by the functional

hydrophobic layer thickness will be minimized while the hydrophilic layer will act as a support

for the thin functional layer so that the whole membrane is mechanically strong.

For the fabrication of hydrophilic-hydrophobic membranes with hollow fiber geometries some

other considerations come into practice, which are discussed in the following sections:

3.3.4 Hydrophilic layer porosity and thermal conductivity

In hydrophilic-hydrophobic membranes, the hydrophilic layer introduces an additional heat

transfer resistance that amplifies the undesirable temperature polarization. Based on equation

(3.13), in order to minimize the heat transfer resistance induced by a fixed hydrophilic layer

thickness, one possible way is to maximize the thermal conductivity of the hydrophilic layer ki by

fabricating the layer as porous a possible. This can be understood according to the following

equation describing the inner layer thermal conductivity.

27

imiwi kkk 1 (3.24)

Therefore, as the porosity of the hydrophilic layer increases, a higher volume percentage of the

hydrophilic layer will be occupied by the more conductive liquid water (kw = 0.58 W/m°K > km =

0.1-0.3 W/m°K) during the DCMD operation. Hence, the thermal conductivity of the layer

increases accordingly.

Chapter 4. Experimental

4.1 Materials

PVDF polymer was chosen to be applied as the main polymeric material to form the functional

membrane applied in the MD process. PVDF is a highly thermal and chemical resistant polymer

with a high hydrophobicity that corresponds very well to the properties required for a high

performance MD membrane. The highly hydrophobic nature of this polymer can be attributed to

the presence of two Fluor atoms in its monomer chemical structure that lowers the surface tension

of the polymer to about 33.7mN/m (25dynes/cm) [72]. PVDF Kynar HSV 900 was used in our

experiments and was purchased from Arkema Inc.

Fig.4.1 Chemical structure of PVDF polymer

Polyacrylonitrile (PAN) was applied in the inner layer of dual layer fibers in order to form a

hydrophilic support layer for the PVDF functional outer layer. PAN is a highly hydrophilic

rubbery polymer which is usually used in a fiber form as the chemical precursor of high quality

carbon fiber. Homopolymers of PAN have applications as fibers in reinforced concrete, hot gas

filtration systems and outdoor awnings. But mostly PAN copolymers are used as fibers to make

knitted clothing as well as outdoor products such as tents [73]. PAN was kindly supplied by

Profs. J.Y. Lai and H.A. Tsai at Chung Yuan Christian University of Taiwan. N-Methyl

Pyrrolidone (NMP) and methanol were used as solvent and nonsolvent respectively and supplied

by Merck.

28

Fig.4.2 Chemical structure of PAN polymer

Hydrophobic cloisite 15A and hydrophilic cloisite NA+ montmorillonite clay particles were

purchased from Southern Clay Products. These particles were mainly applied as additive to

polymeric dope solutions in order to enhance their physical properties and processability.

Furthermore, they were targeted to amend surface tension properties of the inner and outer layers

of membranes and also to create more free volume and pores in the membrane structures. Table.1

shows some properties of these particles provided by the supplier.

Typical dry particle size Treatment-

properties

Organic

Modifier

Modifier

Concentration 10% less

than

50% less

than

90% less

than

Density Hydrophilicity

Hydrophobicity

Cloisite 15A 2M2HT 125 meq/

100 g clay 2 μm 6 μm 13 μm 1.66 g/cc Hydrophobic

Cloisite NA+ None None 2 μm 6 μm 13 μm 2.86 g/cc Hydrophilic

Table.4.1 Specificcations of Cloisite particles

4.2 Dope preparation

The polymers were dried at 60 °C under vacuum overnight before they were used for dope

preparation. Different concentrations of PAN/NMP, PVDF/NMP and mixtures of these solutions

with clay particles were prepared by stirring the solutions at 65 °C for solutions containing PAN

and ambient temperature for other solutions, over a period of 12 hours. The prepared solutions

were degassed overnight before they were used for spinning.

29

4.3 Fabrication of flat sheet membranes

In order to investigate the effect of different coagulants on the surface porosity and roughness of

the PVDF membranes, PVDF flat sheet membranes were cast with a thickness of 250 μm using a

knife casting approach on a glass plate. Afterwards, the films were soaked into coagulation bath

containing pure water, 40/60 and 80/20 wt. % methanol and water for 24 hours. Consequently,

the cast films were freeze-dried. The fabricated films were used for SEM surface morphology and

contact angle measurement tests

4.4 Fiber spinning

Single and dual layer hollow fibers were spun through the spinning line schematically shown in

the following figure.

Fig. 4.3 Schematic representation of the fiber spinning line

At least 8 hours before starting the spinning, polymer solutions were doped into syringe pumps so

that a complete degassing of dope solutions is achieved. On the other hand, another syringe pump

was filled with bore fluid and the syringe pumps were connected into specific channels of the

spinning device called spinneret. Spinning initiated by turning on the pumps and choosing

desirable dope and bore fluid flow rates. The discharged flow form the spinneret went through an

30

air-gap distance ranging from 0-20 cm depending on the experimental conditions and entered an

external coagulant bath where the main phase inversion process occurred. The nascent fibers were

collected using a drum which induced a drawing force and consequently take-up speed on the

fibers. The take up rate ranged from free fall to higher take up rates depending on the

experimental conditions. The as-spun fibers were immersed in water for approximately two days

for solvent exchange. Two different drying procedures were initially applied for each spinning

batch. In the first procedure, fibers were immersed in methanol once for 30 minutes. Afterwards,

the fibers were dried by air at room temperature. In the second drying procedure, fibers were

frozen in a freezer for about 5 hours and then freeze dried using a Thermo Electron Corporation,

Modulyod freeze-dryer for about 8 hours. By comparing the two procedures, we found out that

freeze drying prevents the sever polymer shrinkage, which occurs during air drying. It was

observed that the fiber shortening as much as 50% occurred in case of the fibers dried in air.

Therefore, the freeze drying was adopted as the standard drying procedure for all the experiments.

4.5 Morphology study of hollow fibers by SEM

Hollow fiber cross-sections were observed by taking SEM pictures using a JEOL JSM-5600LV

scanning electron microscope. Fiber samples were immersed in liquid nitrogen, fractured and

then coated with platinum using a JEOL JFC-1300 coater.

4.6 Contact angle measurements

A Sigma 701 Tensiometer from KSV Instruments Limited was used to measure the contact angle

of a hollow fiber as follows: The hollow fiber was immersed into distilled water and the

advancing contact angle was calculated with the aid of computer software. Ten readings were

measured and an average was obtained from the results. In case of hydrophilic-hydrophobic

hollow fibers, in order to prevent the effects of the inner hydrophilic layer on the contact angle

measurements for the outer layer, the lumen side of the fibers was sealed by inserting a metallic

31

bar into the fiber lumen side. Furthermore, in order to estimate the contact angle of the inner

hydrophilic layer, flat sheet membranes were cast out of the same dope solution used for the inner

layer and immersed in a coagulant composition similar to the fiber bore fluid. The same

procedure was performed for the outer layer dope solution in order to confirm whether the contact

angle measurements of flat sheets can give a reliable estimate of the measurements performed for

hollow fibers. The fabricated flat sheet membranes were tested for contact angle measurement

using a Rame-Hart contact angle goniometer (model 100-22) by the sessile drop method at 25 °C.

A built-in image system provided by Rame-Hart was able to acquire the image, transmit to a

computer, and perform the image analysis.

4.7 Porosity measurement

Two different approaches were adopted for hollow fibers porosity measurements. In the first

approach, porosity measurements were carried out using an Autopore III 9420 mercury

porosimeter. A hollow fiber sample was cut into about 20 pieces, each piece of about 2 cm length.

Then, the fiber pieces were loaded into a penetrometer and the penetrometer was loaded and

installed in a low pressure port. At the first phase of the low pressure analysis, the penetrometer

was evacuated from gases. Later it was automatically backfilled with mercury and data were

collected at pressures up to about 50 psia. When the low pressure analysis was complete the

penetrometer was removed from the low pressure port and installed in a high pressure port. At the

high pressure analysis, data were collected at pressures up to 60,000 psia.

In the second approach, the sample hollow fibers were first weighed with a beam balance (w1),

following by immersion in 33% LIX54 kerosene solution for one week. An assumption was made

where all the empty voids were filled with the liquid membrane phase. Next, the fully

impregnated fibers were removed from the kerosene and any excess kerosene in the lumen side

32

and on the outer surface of the fibers was wiped away. Afterwards, the fibers were weighed

again. The empty volume (V) can be calculated by the following relation:

k

wwV

12 (4.1)

where ρk is the kerosene density. The porosity of the hollow fibers membrane (ε) was estimated

by the ratio of empty voids (V) to the total volume of the membrane samples [74]. An average

value of the two tests was considered for the fibers porosity.

4.8 Pore-size distribution measurement

Pore size distribution measurements were carried out using a PMI capillary flow poromter. The

sample modules were fabricated in a similar way as we did for gas permeation rest. The

fabricated modules were immersed in a low surface tension liquid such as isopropyl alcohol (IPA)

for one day so that the pores are fully saturated with the liquid. Afterwards, the Wet Up/Dry

Down test was performed using the porometer. In the Wet Up portion of the test, the sample was

wet (saturated with a wetting fluid) and the porometer acquired data points with increasing N2 gas

pressure. On the Dry Down portion of the test, the sample was dry and data points were acquired

as the pressure was gradually decreased.

4.9 Gas permeation test

Pure gas N2 permeation measurements were carried out to test the permeance of the fibers

according to the procedure described in [75]. Gas permeation modules were composed of 2

hollow fibers with a length of around 7 cm. The permeate flow rate was measured at different

pressures using a bubble flow meter. The permeance P/L was measured by the following equation

[76]:

(4.2)

PnDl

Q

PA

Q

L

P

33

where P is the permeability of separating layer (Barrer), L the thickness of the fibers (cm), Q the

pure gas flux (cm3/s), n the number of fibers in one testing module, D the outer diameter of the

testing fibers (cm), l the effective length of the modules (cm), and ΔP the gas pressure difference

cross the membrane (cmHg).

4.10 Polymer flow observation by a high magnification camera

By making use of the transparent nature of PAN dope solution we were able to take high

magnification pictures of the polymer dope and bore fluid flow geometries in the air-gap region.

An EOS 350 digital Canon camera with an MP-E65mm high magnification lens was utilized for

the digital visualization.

4.11 Video microscopy flow visualization

A drop of dope solution was sandwiched between two microscope slides. Afterwards, a drop of

water as a coagulant was introduced to the gap between the two slides by a syringe. The

diffusion, precipitation and solidification fronts of water into the dope solution were observed and

video-recorded under an Olympus BX50 polarizing optical microscope. These experiments may

provide the basic morphological evolution of the phase inversion process.

4.12 Module Fabrication

In order to fabricate membrane modules, hollow fibers were cut in about 25 cm long and placed

parallel to each other and were put together and bundled using a Teflon strip in two ends. The

prepared bundle was inserted inside the fitting and tube assembly as shown in Fig. 3.2.

34

Fig.4.4 Schematic diagram showing the steps involved in fabrication of lab scale membrane modules [77]

In the next step, the space between the fiber bundle and the fittings at the two ends of the module

were filled with hydrophilic cotton. The cotton is applied as a substrate to absorb the epoxy and

form a sealing agent among the fibers in order to separate the shell and module side of the fibers.

Later, the module was placed vertically on a holder and a proper mixture of epoxy resin and

hardener (1 to 2 in this study) was added to the higher section of the module standing on the

holder. After a few hours that the epoxy is completely hardened the same epoxy potting

procedure was carried out for the other side of the module.

4.13 DCMD experiments

DCMD experiments were carried out using the experimental set-up shown in the following

figure.

35

Fig.4.5 Schematic experimental set-up applied for direct contact membrane distillation process

There were two separate feed and permeate cycles in this set-up, which have been highlighted by

red and blue colors, respectively. A centrifugal pump circulated the preheated saline feed water

with 3.5 wt% sodium chloride (NaCl) concentration from the feed tank through the shell side of a

vertical membrane module. A saline solution of 3.5 wt% was used to represent the ocean water

which has an average salinity of 3.5 wt%. On the other hand, a peristaltic pump circulated

permeate pure water from a conical flask on a balance through a cooler and consequently through

the lumen side of the membrane module. The feed and distillate streams flowed co-currently from

the bottom to the upper part of the MD module. The ionic conductivity of the permeate stream

was measured before and after the test using a Radiometer analytical conductivity meter (model

Pionneer30) in order to calculate the separation factor.

36

Chapter 5. Fabrication of dual layer hydrophobic-hydrophilic hollow fibers

5.1 Effect of coagulant on the surface morphology of PVDF membranes

Fig.5.1 shows the SEM pictures from the top surface (facing the coagulant) of the PVDF flat

sheet membranes for different coagulant compositions.

Water/Methanol (20/80 wt.%) Water/Methanol (60/40 wt.%) Water

Fig.5.1 SEM pictures from the top surface (facing the coagulant) of the PVDF flat sheet membranes for different coagulant compositions

It can be clearly observed from Fig. 5.1-A that when a strong coagulant such as water was used

for the fabrication of the membranes, dense and smooth surfaces with no obvious pores was

formed. However, when we introduced methanol (a weaker coagulant compared to water) to the

coagulant up to 40 wt.%, the roughness of the membrane surface increased considerably while the

number and the size of the surface pores increased as well (Fig. 5.1-B). Finally, as it is shown in

Fig. 5.1-C, a coagulant containing 80 wt% methanol induced a porous 3-dimensional fiber-like

surface morphology with some degrees of spherulitical structure. The pore sizes are in the range

of 0.4 μm on the membrane outer surface. The formation of a 3-dimensional fiber-like structure is

probably due to the delayed demixng, spinnodal decomposition and then coarsening [78].

37

Spinnodal decomposition is a phase separation process which occurs in unstable systems. In other

words, the small perturbations in the system are amplified and any small change in system

composition lowers the free energy of the system. This process differs from nucleation and

growth phase separation in a sense that in spinnodal decomposition phase separation

occurs throughout the material rather than nucleation sites only [78].

The difference in morphology may be explainable from the solubility parameters of PVDF, water

and methanol as summarized in Table 3.

Solvent Water

[79]

Water/methanol

(60/40 wt.%)

Water/methanol

(20/80 wt.%)

Methanol

[79]

NMP

[80]

PVDF

[81]

δt (MPa 0.5) 47.8 41.6 35.5 32.4 11.2 23.2

δt (Cal 0.5 /cm1.5) 23.4 20.3 17.4 15.9 5.5 11.3

Table 5.1 Total solubility parameter (δt) of water, methanol, NMP and PVDF

Water is the strongest non-solvent toward PVDF compared to 60/40 wt.% and 80/20 wt.%

water/methanol mixtures. Therefore, the liquid-liquid miscibility gap for PVDF-NMP-water will

be the biggest, while this gap for PVDF-NMP-water/methanol (20/80 wt.%) will be the smallest.

As a result, for the PVDF-NMP-water system, the solution may go through rapid phase separation

via nucleation growth and thus produce a rough dense surface. However, in case of PVDF-NMP-

water/methanol (20/80 wt.%) system, the liquid-liquid miscibility gap is small that favors a delay

demixing (i.e., a slow phase separation). With an increase in methanol and water content diffused

from the external coagulant into the nascent fiber, the precipitate path may go through spinodal

decomposition and form a 3-directional fiber-like network [78]. The apparent spherulitical

structure upon the fiber network may be due to coarsening effects. Therefore, in order to induce

high porosity and roughness on the inner and outer surfaces of the PVDF hollow fibers, a mixture

38

of methanol/water with 80 wt% methanol concentration was adopted as the standard coagulant to

be used as the bore fluid and external coagulant in the spinning line.

5.2 First batch of fiber spinning

In our first batch of spinning, we adopted PVDF and PAN polymers as the hydrophobic outer

layer and the hydrophilic inner layer of the spun fibers respectively. It is known that a dope

solution with a low polymer concentration undergoing a non-solvent induced phase separation

tends to result in more porous membranes [82, 83]. Therefore, low concentration polymer

solutions of PVDF/NMP and PAN/NMP were used for outer and inner layer dope solutions

respectively. The following Table shows the composition of the inner and outer layer dope

solutions used for the spinning of this batch with other spinning conditions.

Table.5.2 Spinning conditions applied for the first batch of spinning

39

5.2.1 Membrane morphology

The following SEM pictures show the cross section and surface morphology of the spun dual

layer fibers.

Fig.5.2 SEM pictures representing the structure of first spinning batch fibers

As can be observed in Fig.5.2, the fabricated fibers showed desirable membrane morphology and

structures without any delamination. The outer hydrophobic PVDF layer was 30 µm thick with a

fully sponge-like structure and high surface porosity. The inner support layer was 150 µm thick

with an asymmetric structure with some macrovoids observable in the inner contour of the fibers.

40

5.2.2 DCMD performance

Although the fibers in the first spinning batch were fabricated in a desirable structure, when they

were applied in the DCMD separation process, the interface between the two layers was pealed

off and no separation could be achieved. This might be due to different thermal expansion

coefficients of these two polymers or the repulsive force acted on the interface between

hydrophilic and hydrophobic layers in an aqueous environment.

Fig.5.3 Delamination phenomenon during the DCMD process using the first batch of fibers

In addition, in other attempts to fabricate single layer PVDF hollow fibers using weak coagulants

such as a methanol-water mixture containing 80 wt.% methanol, we noticed that the nascent

fibers formed in the spinning line were mechanically weak and the spinning process was

accompanied with instabilities, fiber breakage and fiber collapse in the collecting wheel.

5.3 Second batch of hollow fiber spinning

In order to overcome the obstacles we faced during the first batch o spinning, we formulated new

dope solutions containing PVDF as the major component in both layers. In this way, we expected

to overcome the problem regarding the interface separation of the two layers containing polymers

with different surface tension and thermal expansion characteristics. However, in order to induce

different surface tension properties in the two layers and simultaneously increase both the

41

processability and mechanical strength of nascent fibers, hydrophilic and hydrophobic clay

particles were added to the inner and outer layer dope solutions, respectively. Clay particles have

been extensively used for polymer reinforcement [84]. Furthermore, some amounts of highly

hydrophilic PAN polymer was added to the inner dope solution in order to enhance the water

transfer from the fiber permeate side into the interface between the two layers. We postulate that

the PAN polymer chains could act as bridges among the hydrophilic clay particles which can be

considered as islands in the polymer matrix.

As a result, modified dope solutions made of 12.5/87.5 (in wt. %) PVDF/NMP containing 30

wt% hydrophobic cloisite 15A loading in the total solid was chosen for the outer layer, and a

dope solution made of 8.5/4/87.5 (in wt%) PVDF/PAN/NMP containing 50 wt% hydrophilic

cloisite NA+ loading in the total solid was selected for the inner layer. The maximum amount of

30 wt% and 50 wt% loadings in total solids were adopted because in higher loadings the

fabricated membranes became rather brittle and the spinneret blocked easily. Less loading was

adopted for the outer layer because of the smaller spinneret discharge gap for the outer layer

compared to the inner layer. The following schematic figure summarizes the spinning conditions

used in the second batch of spinning.

42

Fig.5.4 The spinning conditions applied in the second batch of spinning process

5.3.1. Hollow fibers morphology

The SEM and EDX micrographs in Fig.5.5 show the cross section morphology and clay particle

distribution in the fabricated fibers. It can be observed that a desirable sponge like structure with

rather uniform distribution of particles has been achieved. , the thickness of the outer hydrophobic

layer is about 50μm and the hydrophilic support layer is about 330 µm. On the other hand, as can

be observed in Fig.5.6 a high outer surface porosity has been formed which is in agreement with

our earlier observations for the flat sheet membranes as shown in Fig.5.1, fabricated using a weak

coagulant.

43

Fig.5.5 Cross section morphology of fibers fabricated through the second batch of spinning

Fig.5.6 SEM micrographs showing the surface morphology of the dual layer fibers

44

5.3.2. Pore-size distribution

The pore size distribution of the newly developed hollow fibers was also obtained. As can be

observed in Fig.5.7, a very sharp pore size distribution with an average pore size of 0.41μm has

been achieved. Such relatively big pore sizes with a sharp distribution may lead to water vapor

mass transfer enhancement in the MD process.

Pore Size (µm)

Fig.5.7 Pore size distribution of the fabricated fibers

5.3.3. Gas permeationand porosity measurement tests

A high permeance of 1.25 cm3 (STP)/(cm2)s.cmHg was achieved for the fibers tested for the N2

gas permeation test. Such a high permeance is in agreement with a measured porosity of about

80% and big pore size of the fibers.

5.3.4. Contact angle measurements

The graph in Fig.5.8-A represents the measured force per unit length of the hollow fiber versus

the immersion depth obtained from the hollow fiber contact angle measurements corresponding to

an advancing contact angle of 136°, while Fig.5.8-B shows the equilibrium state of a water

45

droplet on the fabricated flat sheet membrane cast out of the outer layer dope solution using the

same coagulant, which corresponds to a contact angle of 140°. The high contact angles obtained

can be attributed to the high roughness induced on the membrane surface caused by the porous 3-

dimensional fiber-like network with some degrees of spherulitical structure and clay particles

used in the membrane matrix. Since the values obtained from the two methods were quite

comparable, we estimated that the contact angle of the fiber inner layer should be quite similar to

the equilibrium contact angle of the flat sheet film cast out of the inner layer dope solution. A

very low contact angle of 50° was obtained from this experiment, which exhibits the hydrophilic

character of the inner layer (Fig.5.8-C).

Fig.5.8 Contact angle measurements of the fabricated fibers

5.3.5 DCMD results

The fabricated fibers were tested for the DCMD process. Table 5.3 shows the obtained flux as a

function of feed inlet temperature and other operating conditions.

46

Table 5.3 DCMD operating conditions and the obtained flux for the fabricated fibers

During the experiment we observed that the flux increased rapidly as we increased the permeate

inlet pressure to about 8 psig but it remained almost constant for further pressure increase. This

observation is probably due to the fact that as we increased the permeate pressure, a higher

volume of the inner hydrophilic layer occupied by water and the effective water vapor mass

transfer thickness reduced accordingly; but at permeate pressure about 8 psig the inner layer was

probably fully wet and the flux remained constant. Therefore, all the flux measurements were

carried out at permeate flow rates with corresponding pressures of about 8 psig. It can be

observed from Table 5 that water vapor flux as high as 55 kg/m2hr based on the fiber outer

diameter (it is equivalent to 115 kg/m2hr based on the fiber inner diameter) has been achieved at

about 90 C. The reproducibility of data was also tested and confirmed. It is worthy to note that

keeping the permeate side pressure higher than the feed side pressure can insure that the obtained

flux could not be attributed to the leaking of the feed side to the permeate side. A separation

47

factor as high as 99.8% during the experiments confirms this conclusion as well. The separation

factor was calculated based on the following equation [59].

10011001 12

f

pp

f

p

C

C

where Cp is the salt concentration in the permeate side, Cf is the salt concentration in the feed

side, ξp1 and ξp2 are conductivities of permeate stream before and after the test, respectively and ξf

is the conductivity of the feed stream. It is worthy to note that the ionic conductivity of a 3.5 wt%

salt water and deionized water which used for the xpreiments were around 53000 μS/cm and 10

μS/cm respectively. On the other hand, the ionic conductivity of the permeate maintained below

100 μS/cm corresponding to a separation factor of 99.8%.

[85]

[63]

[64]

[86]

[87]

[88]

Table 5.4. Comparison of the maximum flux obtained in this study with the literature for DCMD processes with a hollow fiber configuration **The original data reported in the references [64, 86]

are based on the fibers inner diameter, while we have converted them based on the fibers outer diameter, which is a more realistic and meaningful way to report flux

Fig.5.9 compares the obtained DCMD flux in this study with the maximum flux reported by the

previous investigations on hollow fiber configuration summarized in Table 5.2.

48

[64]

[86]

[63]

[85]

[86] [87]

Fig.5.9 Flux and energy efficiency obtained in the DCMD process using the fabricated fibers and flux comparison with the literature data using hollow fiber membranes

It can be observed that the obtained data in this study are much higher than the most of the

previous reports and there is only one data reported by Li and Sirkar [64] which is quite close to

data obtained in this study. However, as it can be inferred from Table 6, the length of the module

used in [64] is too short (6.4 cm) which actually ignores the limiting effect of permeate

temperature increase along the fibers. This argument is in agreement with the later data reported

by Song et al. [86] using similar conditions but a longer module (25.4cm) that resulted in a lower

flux. The high flux obtained in this study can be mainly attributed to the optimization of different

membrane characteristics such as functional layer thickness, pore size, pore size distribution,

surface and bulk porosities and high contact angle associated with the functional layer without

any advanced module design optimizations. It is believed that if the membrane fabrication

optimizations which were attempted in this work are accompanied with more efficient module

designs to suppress the temperature polarization phenomenon, even higher flux are achievable in

the MD process.

49

50

5.4 Summary

In summary, very different membrane morphologies can be formed in PVDF membranes by

tailoring the strength of coagulation bath. Highly porous membrane structure with three

dimensional networks is formed in PVDF membranes using weak coagulants such as

methanol/water mixtures. Such membrane morphologies possess a very high gas permeability,

leading to a high flux in membrane distillation process. Furthermore, co-extrusion looks to be an

efficient approach to fabricate hydrophilic-hydrophobic hollow fiber membranes to be applied in

MD process. This process provides the freedom to easily tailor membrane properties such as

porosity, pore size distribution and thickness. Finally, high flux and high energy efficiency can be

obtained using the dual layer hydrophilic-hydrophobic fibers.

Chapter 6. Investigation of corrugation phenomenon in the inner contour of hollow

fibers during the non-solvent induced phase-separation process

During the spinning of PVDF and PAN hollow fiber membranes, we came across an instability

that led to the formation of fibers with irregular cross section. In this chapter, we are going to

elaborate the observed phenomenon that here after we call it “corrugation phenomenon”.

6.1 Introduction

As it was discussed in the previous chapters, among different existing geometrical shapes of

membranes, hollow fibers are of great interest because of their high surface area per unit volume

and ease of module fabrication. Comprehensive reviews on hollow fiber membrane fabrication

process can be found in references [89-92]. However, similar to conventional spinning processes,

the fabrication of hollow fibers is inherently accompanied by instabilities that can lead to

technical problems in the production line or undesirable final products that can increase

production costs. Therefore, understanding the nature of these instabilities as well as methods for

controlling them is extremely important. However, these subjects have been often treated as

proprietary know-how in the membrane industry and rarely openly discussed in the membrane

literature.

The major instabilities encountered in fiber spinning are draw-resonance, necking, capillary

break-up and irregular fiber cross-section. The first three phenomena are rather similar in nature

and lead to fiber breakage during the spinning or a non-uniform cross-sectional diameter along

the spun fibers. These instabilities have been investigated theoretically and experimentally by

many researchers and a reasonably good understanding of their mechanism and controlling

methods has been outlined in the literature. They are mainly attributed to the amplification of

51

small fluctuations in polymer jet flow as a result of drawing or capillary forces. Larson, Petrie and

Lipscomb [93-95] have given comprehensive reviews on this subject.

The last type of spinning instability is related to the non-uniform wall thickness and the irregular

cross-section of the as-spun hollow fibers. The severe non-uniformity of the fiber wall thickness

is considered to be major impediment for some membrane applications. For example, in high

pressure-driven membrane processes, the thin regions in the fiber cross-section are considered as

mechanically weak points that are easily collapsed. In addition, the reduced inner cross-sectional

area may lead to high pressure drop in treatment of liquid streams flowing through the fiber

lumen. On the other hand, if a good control of the phenomenon is achieved, it might be possible

to purposely create favorable geometries to the spun fibers that can highly enhance the separation

performance of the membranes. For example, hollow fibers with wavy inner or outer wall could

be preferable because of their high mass transfer area per unit volume. Recently, both Nijdam et

al. [96] and Widjojo and Chung [97] fabricated hollow fibers with wavy geometries at their outer

surface, while Koros et al. [89] and Santoso et al. [98] observed hollow fibers with a wavy inner

skin.

Fibers with wavy geometries might potentially help to increase the effective membrane area and

may induce flow turbulence (in the shell side of the modules) and reduce concentration and

temperature polarizations. Therefore, a more thorough understanding of the governing

mechanisms and approaches to controlling the wavy phenomenon are of great interest. However,

as far as we know, there are very few studies on this subject. Van’t Hof [99] observed the

deformed inner layer in the polyethersulfone (PES) fibers fabricated for gas- separation

applications. He noticed that the addition of solvent or a weaker coagulant to the bore fluid makes

the irregularity more pronounced. Borges et al. [100] observed the same phenomenon in spun

polyetherimide (PEI) and PES fibers. They found that delayed precipitation induced by the

52

addition of a solvent to the bore fluid and increased air-gap distance impeded the deformation of

the fiber cross-section. Both of these studies [99, 100] attributed the instability to the competitive

effects caused by die-swell (which tried to reduce the bore volume) and by bore liquid (which

tried to fill out or expand the bore volume). However, this mechanism does not explain some the

observations reported in the present paper.

In this chapter, we intend to report the results of our experiments on the corrugation phenomenon

to investigate the factors influencing the underlying instability that causes these patterns. We also

propose a novel mechanism that is in good qualitative agreement with our new observations. It is

believed that the proposed mechanism can be potentially applied to mathematically model and

predict the instability.

6.2 Experimental Observations

The following Table summarizes the spinning conditions applied for spinning single layer PAN

hollow fibers. In order to get a better understanding of the phenomenon, we varied the five

spinning parameters namely bore fluid composition, air-gap distance, external coagulant, take-up

speed and dope concentration.

Parameters Value Dope flow rate (ml/min) 2.0

Bore fluid NMP-water (NMP wt. %) 0, 20, 40, 60, 80

Bore flow rate (ml/min) 0.5

Length of air-gap (cm) 0, 3.5, 7, 12, 17, 20

External Coagulant water, Isopropyl alcohol

External coagulant temperature (°C) 25

Spinneret dimensions (mm) 0.85/0.5 (OD/ID)

Spinneret temperature (°C) 25

Take-up speed (ml/min) free fall, 404, 515, 626, 737, 848, 958

Table.6.1 Spinning conditions of hollow fiber membrane fabrication

53

Fig.6.1 shows the cross-section morphology of PAN and PVDF fibers spun through a wet

spinning (no air-gap) process. It is clearly observable that the inner contour of the fibers is

deformed in a way that the wall thickness of the fibers is not uniform circumferentially. In

addition, the cross-sectional view of different sections along the fiber shows that the deformation

had a similar shape all along the fiber. This shows that in contrast to draw-resonance, the

instability is not axially periodic and occurs consistently throughout the spinning process.

Fig.6.1 Irregular shape in the cross section of the PAN 17% (left) and PVDF 20% (right) hollow fibers fabricated through wet spinning, Bore fluid composition of 40% NMP/water

(NMP wt.%), free fall take up rate

The effect of the air-gap distance on macroscopic morphology of the fibers was investigated. The

SEM micrographs in Fig.6.2 show the cross-section of the spun PAN hollow fibers for different

air-gap distances ranging from 0 to 20 cm. It is observed that by increasing the air-gap distance,

the number of corrugations in the inner contour of the fibers decreases from 5 (in the zero air-

gap) to 1 corresponding to a circle (in the 20 cm air-gap).

54

Air gap=17 cm Air gap=20 cm Air gap=12 cm

Air gap=7 cmAir gap=3.5 cmAir gap=1 cmAir gap=0 cm

Fig.6.2 Cross section of spun fibers from 17 wt.% PAN solution, 40 wt.% NMP Bore fluid with different air-gap distances

In another experiment with no air-gap, we observed that by increasing the solvent concentration

in the bore fluid from 0% to 60% the number of corrugations in the inner layer increases from 2

to 8. However, by further increasing the solvent concentration to 80% no deformation is

observable (Fig.6.3).

40% NMP

60% NMP

20% NMP

80% NMP

Pure Water

Fig.6.3 Cross section of wet spun fibers from 17 wt.% PAN solution with different solvent amount in the bore fluid

55

The same trend was observed for higher air-gaps. In addition, we observed that by replacing a

strong external coagulant such as water with a weaker one such as IPA, while keeping other

conditions unchanged, no deformation was observed in the cross section of fabricated fibers. The

SEM micrographs in Fig.6.4 show this behavior.

(A) IPA (B) Water

Fig.6.4 Cross section of wet spun fibers from 17 wt.% PAN solution and 40 wt.% NMP bore fluid using different external coagulants: (A) IPA (B) water

The effect of take-up speed was also investigated. The SEM micrographs in Fig.6.5 show that by

increasing the take-up speed at a constant air-gap distance, the number of waves increased at first

from three to four. However, by further increasing the take-up rate, it decreased so that a circular

cross-section appeared ultimately.

958 cm/min737 cm/min 848 cm/min

626 cm/min 515 cm/min404 cm/minFree Fall

Fig.6.5 Cross section of spun fibers with different take-up rates from 17 wt.% PAN solution, 40 wt.% NMP bore fluid, 7 cm air-gap

56

Finally, we observed that by increasing the dope concentration from 13 wt. % to 17 wt. % and 22

wt. %, the number of waves decreased from 6 to 4 and 1 respectively. The SEM micrographs in

Fig.6.6 show this trend.

Air gap=1 cm 13 wt.% PAN 17 wt.% PAN 25 wt.% PAN

Fig.6.6 Cross section of spun fibers from different concentrations of PAN solution, 40 wt.% NMP aqueous mixture as bore fluid, with 3.5 cm air-gap and free fall

(A) 13 wt.% PAN solution (B) 17 wt.% PAN solution (C) 22 wt.% PAN solution

The same experiments were carried out using a PVDF dope solution and very similar trends but

with different number of waves were observed in these experiments. Fig.7 shows the effect of air-

gap distance on the corrugation of PVDF fibers as an example.

Air gap=7 cmAir gap=7 cm Air gap=0 cm Air gap=3.5 cm

Air gap=20 cmAir gap=17 cm

Fig.6.7 Cross section of spun fibers from 20 wt.% PVDF solution, 40 wt.% NMP Bore fluid with different air-gap distances

57

6.3 Discussion

To propose a suitable mechanism for our observations, we tried to find out in which step of the

fiber-fabrication process the instability arises. The cross-sectional micrographs of as-spun fibers

in Fig.6.8 show that the deformation happens in the fiber-spinning line. Therefore, the

deformation cannot be attributed to the drying or post-treatment stages of the fiber fabrication.

Fig.6.8 Cross section of wet as-spun hollow fibers from 17 wt.% PAN solution and bore fluid containing 40 wt.% NMP

Previous studies [99, 100] postulated that the competitive flow direction between dope solution

and bore fluid in the die-swell region is the responsible for this phenomenon. In order to evaluate

this hypothesis, we took pictures with high magnifications at the die-swell region. Because of the

translucent nature of the PAN solution, we could observe the flow interface between the dope and

the bore fluid in the die-swell region. The pictures in Fig.6.9 show the two-phase flow of the PAN

solution and the bore fluid after discharging from the spinneret. The location of the interface

between the two streams agrees with the spinneret dimensions used in the spinning. Therefore, we

confirm that the interface observed in the pictures is not affected by light scattering. These

pictures show that the interface between the two streams is not corrugated. Therefore, other

factors might be responsible for the instability rather than the competitive force between the dope

and the bore fluid.

58

Bore fluid Rate=0.5 ml/minBore fluid Rate=0 ml/min Bore fluid Rate=2 ml/min

Fig.6.9 Magnified die swell pictures of spinning with 17 wt.% PAN solution, Bore fluid containing 40 wt.% NMP and 7 cm air-gap distance, Different bore flow rates

6.3.1 System description

As polymer dope discharges from the spinneret and contacts with the bore fluid in the die-swell

region, the solvent-nonsolvent exchange between the bore fluid and dope flow leads to the

formation of three phases in the nascent fiber during the dry-jet wet-spinning. Using flow

visualization photos taken from the video microscopy on a flat membrane inversion as an

example, the innermost phase, shown as I1 in Fig.6.10, is formed as a result of vitrification

induced by the bore fluid. The intermediate phase I2 lies between the solidification and

precipitation fronts. In this region, the solution divides into polymer-lean phase and polymer-rich

phases either by nucleation and growth of the polymer-lean phase or by spinodal decomposition.

The third region is between the precipitation and diffusion/convective fronts where macrovoids

propagate and grow [101, 102]. The pictures in Fig.6.10 (A) clearly show the solidification,

precipitation and diffusion/convective fronts and the regions associated with them.

59

O1

I1

O3

I2 I3

O2

Coagulant phase

Precipitation front

Solidification front

I3 or O3 I2 or O2

I1 or O1

(A) (B)

Fig.6.10 (A) Penetration of coagulant into the casting solution after 0.2 second using flow visualization photos taken on flat membrane inversion as an example, (B) Schematic regions in

the extruded nascent fiber

6.3.1.1 Phases I1, I2 or O1, O2

Because of the rapid precipitation of polymer in I1 and I2 (or O1 and O2) phases, they can be

approximated as one pseudo homogeneous elastic cylindrical shell, Is. Therefore, Von Karman

elasticity theory will govern the mechanical behavior of this phase [103].

6.3.1.2. Phase I3 or O3

Since this phase is not fully solidified, we treat it as a viscoelastic fluid with varying density and

rheology that progress with the solvent-nonsolvent exchange between the dope solution and the

bore fluid (or the external coagulant for O3 phase), and the consequent phase separation. For a

detailed description of mass transfer in membrane formation one can refer to the references [104-

106].

60

Since the total density is changing in this phase the following equations of continuity and motion

can be applied. In writing the following equations, we have assumed a convected coordinate

system at the velocity as the interface between the dope and the bore fluid with the positive axial

coordinate (z) in the flow direction.

0.

ut

(6.1)

.... Puut

u

(6.2)

where ρ is the total density, u is the velocity vector ( , , ), t is time, P is the hydrodynamic

pressure and τ is the stress tensor. According to the continuity equation (5.1), velocities are

induced in the r,

ru u zu

and z directions as a result of density change, die swell, non-uniform

progresses of diffusion/convective fronts, and rapid precipitation and densification. Therefore, in

the corresponding equation of motion (5.2), a pressure term has to be introduced in order to

balance the momentum convection and accumulation terms in the left-hand side of this equation.

In a simplified explanation, pressure is generated in this phase mainly as a result of density

change.

6.4 Possible instability mechanisms

As it was mentioned before, our system involves elastic, hydrodynamic, mass transfer and

solidification phenomena. Therefore, the instability could arise from any of these processes. In

the following we analyze the possibility of these mechanisms according to our experimental

observations.

A pure hydrodynamic instability such as those observed in polymeric flows, which mostly initiate

from nonlinear viscoelastic behavior of polymer solutions or viscosity stratifications in two-phase

61

flows is probably not responsible for the observed instability shown in Figs.(6.2 - 6.7). The strong

evidence for this argument is the important effect of the bore fluid composition on the instability.

The results section indicated that different deformations of spun fibers were observed depending

on the solvent amount in the bore fluid, which translates into a different mass-transfer driving

force. On the other hand, it is known that the density and viscosity of the NMP as the solvent are

very close to those of water. Therefore, it seems that the effect of mass transfer, convective flow

and solidification must be taken into account in the stability analysis.

Hence, one possibility is the initiation of an instability that arises from the combined

hydrodynamic, mass-transfer, convective flow and phase-separation processes. In the following,

we describe this possible mechanism qualitatively.

6.4.1 Hypothesis 1. (Mass transfer and hydrodynamic instability)

In this hypothesis, the principal instability is described by the equations of continuity, motion and

mass transfer in phase I3, assuming negligible resistance by the elastic cylindrical shell in phase

Is. This assumption is probably valid if the instability occurs very fast before the rigidity of the

elastic shell in the interface between the bore fluid and the dope becomes so high. In this case the

perturbations occurring in the interface will divide the polymer solution matrix into some regions

with different penetration and contact area with the bore fluid. As shown in Fig.6.11, the regions

associated with the inward movement of the interface have a deeper penetration and increased

contact area with the bore fluid. As a result the solvent-nonsolvent exchange rate between the

dope and bore fluid will be faster and the pressure induced by precipitation/densification in these

regions will be higher compared to the other regions. Subsequently, the generated pressure in

regions associated with inward movement would favor the deformation until a new stable state is

reached.

62

Regions with relatively higher diffusion/convective/solidification rates and higher induced pressure

Perturbed interfaceDeformation progress initiated by the interface perturbations

Fig.6.11 Instability associated with proposed hydrodynamic and mass transfer mechanism

This hypothesis provides a good explanation for the instability initiation. However, it is believed

that without a strong hoop shrinkage force resulting from the solvent-nonsolvent exchange

between the dope flow and the external coagulant, and if the dominant motion and force are in

gravity direction, the initial instabilities may not be able to grow significantly. This argument is in

agreement with our observations in Fig.6.4 and Fig.6.9, which show no corrugation occurrence in

case of a weak external coagulant and in the air-gap region respectively. Therefore, we postulate

that the initial instabilities occurring as a result of densification induced pressure in phase I3 will

be magnified in the external coagulation bath by polymer shrinkage and exponentially growth of

the dope solution density, while a higher viscosity can be considered as a stabilizing factor.

Another possibility is that the elasticity governs the instability, which will be discussed in the

next section as our second hypothesis.

63

6.4.2 Hypothesis 2. (Elastic and Buckling Instability)

In this hypothesis, the principal instability is attributed to the buckling of the elastic cylindrical

shell in phase Is, rather to a hydrodynamic instability. We postulate that the instability occurs in

the external coagulation bath where the inward radial force resulting from the shrinkage of the

dope outer layer and the density-induced pressure in regions I3 and O3 exceeds the critical

buckling pressure of the elastic cylindrical shell in phase Is (defined in 6.3.1.1). Therefore, the

instability can be reduced to the buckling of a thin elastic cylindrical shell in the interface

between the bore fluid and dope solution under a uniform external pressure. This assumption is

probably valid if the resistance by the elastic polymeric shell in phase Is becomes dominant.

Fig.6.12 shows this mechanism schematically.

Elastic inner cylindrical Shell influenced by the bore fluid

Spinneret Dope

Bore fluid

External coagulant

Nascent fiber outer Layer influenced by the external coagulant

Air gap Region

Fig.6.12 Schematic picture showing the exertion of inward radial forces generated by theshrinkage of nascent fiber outer layer in the external coagulation bath

The stability of a long elastic cylinder under external pressure was first studied by Levy [107]. He

reduced the stability problem to an algebraic equation involving elliptic integrals and described

the critical pressure required for buckling by the following equation:

64

3

2* 1R

FqP

(6.3)

2

3

112 v

GhF

(6.4)

In the above equation *P is the buckling critical pressure, B is the flexural rigidity, G is the

Young’s modulus, h is the thickness of phase Is, v is the Poisson’s ratio and k is the buckling

mode or the number of circumferential waves observed in the postbuckling shape. Later, using

elastic theory, Greenhill [108] integrated the post-buckling shapes corresponding to each

equilibrium state. Fig.6.13 shows these shapes corresponding to buckling modes 2 to 5. It is

observable that these shapes agree very well with the cross-section of our spun fibers.

(A)

(B)

Fig.6.13 Close agreement between the spun fibers cross sectional geometry (A) and thepredicted postbuckling shapes of a long elastic cylindrical shell (B) [108]

Fig.6.14 shows the energy of the elastic cylinder as a function of system variables for various

buckling modes as proposed by Tadjbakhsh and Odeh. [109]. This figure shows that as the

pressure increases, the energy of the system increases until it reaches a critical bifurcation point at

which the equilibrium path changes its stability and divides into some stable and unstable

branches.

65

q=4

q=3

q=2

P Rh3 / F

U

Fig.6.14 Energy as a function of the dimensionless group including the external pressure for various modes of deformation [105]

It is worthy to note that the spinneret eccentricity, nonuniform rigidity of the inner precipitated

shell as a result of nonuniform formation of macrovoids and the initial small deformations by the

mass transfer-hydrodynamic instability in the air-gap region may lead to the deviation of the

fibers cross sectional shapes from those theoretically predicted in Fig.6.13 (B).

In the following sections, we explain our observations through the proposed hypotheses.

6.5. Effect of air-gap distance

One possible reason for this trend is that by increasing the air-gap distance, there will be a longer

contact time as well as solvent exchange between the bore fluid and the dope flow before entering

the coagulation bath. As a result, the elastic cylindrical shell in phase Is may become more rigid,

phase I3 may become more viscous and the initial instabilities may be damped out during the air-

gap region at a point in the external coagulation bath where the radial inward shrinkage force is

66

exerted. In addition, for higher air-gaps the cross sectional area of the dope flow decreases before

it enters the coagulation bath. Therefore, the precipitated shell radius, R, in the inner contour of

the fiber decreases accordingly. Each of these factors decreases the tendency for the instability to

occur. The combination of the mentioned factors could be responsible for the decreasing trend in

the corrugation numbers or buckling modes associated with the elastic shell as the air-gap

distance increases. Fig.6.15 illustrates the relationship between P*R3 and the flexural rigidity B as

a function of buckling mode k and shows the stable and unstable regions corresponding to

equation (5.3). The transitional path of the shell equilibrium states as a function of air-gap

distance is shown qualitatively by the symbol of circles in this figure.

q=6

F

Stable

q=2

Rh q=5 q=4

q=3

From low to high P*Rh

3

Bore fluid Composition (NMP/water)

External coagulant (solubility

Air-gap distance

Take-up speed

Fig.6.15 Equilibrium state transitions of the buckled elastic shell as a function of spinning conditions

67

6.6 Effect of bore fluid composition

As the solvent amount in the bore fluid increases, the inner nascent fiber has a lower viscosity and

the flexural rigidity of the inner elastic shell decreases accordingly. Therefore, assuming a

constant external pressure, the corrugation mode is expected to increase. However, in spinning

with a bore fluid containing about 80 wt.% NMP, no nascent solidified inner layer is formed. In

addition, the mass transfer driving force may not be high enough to create the initial instabilities.

As a result corrugation does not happen for this condition and the cross-section maintains a

circular shape.

6.7 Effect of external coagulant

The fact that using a weak external coagulant such as IPA inhibits the instability might be due to

the fact that a weak coagulant induces a slower solvent exchange, diffusion/convection flows,

densification rate and shrinkage which lead to a lower induced external pressure. Therefore, the

induced inward radial force could be lower than the critical force required to buckle the elastic

shell in phase Is or amplify the initial instabilities in phase I3.

6.8 Effect of take-up speed

The increasing and subsequent decreasing trend in corrugation mode as a function of take-up

speed can be explained as follows: A higher take-up rate translates into a shorter contact time and

solvent exchange between the bore fluid and the dope flow in the air-gap region that leads to a

less rigid inner elastic shell. This might be the reason for the initial increase of the buckling.

However, as the take-up rate increases further, the cross-sectional diameter of the dope flow

decreases consequently. Therefore, the rigid elastic shell will have a smaller radius, greater

molecular orientation, and consequently greater resistance to buckling. The smaller radius (i.e.,

smaller R value) and higher molecular orientation (i.e., higher B value) of the elastic shell

resulting from higher take-up rates compensates for the shorter contact time between the bore

68

69

fluid and the dope flow and produces similar and comparable effects that dominate for lower

take-up speeds (i.e., mainly higher B value). The square symbol in Fig.6.15 shows the qualitative

instability path associated with the inner elastic shell as a function of take-up rate.

6.9 Effect of dope concentration

The stabilizing effect of higher concentration or higher viscosity of the spinning solution may be

due to the fact that the rigidity and chain orientation of the inner precipitated shell could be higher

for a concentrated dope compared to a dope with a low concentration. As a result there is less

mass transfer-hydrodynamic instability and the inner shell with a higher rigidity will be more

resistant to buckling.

6.10 Summary

In summary, the non-uniform cross section of hollow fibers can be attributed to an instability

occurring during the fabrication of hollow fibers. It was postulated that the main instability

happens in the coagulation bath where the pressure induced in the nascent fiber outer layer as a

result of diffusion/convection, precipitation, densification and shrinkage will buckle the rigid

elastic shell formed in the interface between the bore fluid and the dope solution. It is also

inferred that since it takes only a fraction of second for a nascent fiber traveling through the air

gap region, the hydrodynamic instability may not have enough time to magnify its effects.

Therefore, it is postulated that the hydrodynamic instability may be the initial stage of causing

instability and the elastic and buckling instability is the magnifying stage for the development of

non-uniform cross section of hollow fibers.

Chapter 7. Conclusion

The following conclusions can be drawn out of this study:

1- Very different membrane morphologies can be formed in PVDF membranes by tailoring

the strength of coagulation bath. Highly porous membrane structure with three

dimensional networks is formed in PVDF membranes using weak coagulants such as

methanol/water mixtures. Such membrane morphologies possess a very high gas

permeability , leading to a high flux in membrane distillation process

2- PVDF surface tension and mechanical properties can be modified by blending the PVDF

polymer with hydrophobic or hydrophilic clay particles. The addition of hydrophobic

particles to the hydrophobic layer further increases the contact angle of the functional

hydrophobic layer by increasing the membrane surface roughness.

3- The surface tension of PVDF polymer can be reduced significantly by blending the

PVDF polymer with hydrophilic PAN as well as the hydrophilic lay particles.

4- Co-extrusion looks to be an efficient approach to fabricate hydrophilic-hydrophobic

hollow fiber membranes to be applied in MD process. This process provides the freedom

to easily tailor membrane properties such as porosity, pore size distribution and thickness.

5- Delamination between the two hydrophilic and hydrophobic layers of the fibers can be

prevented by carefully formulating the dope composition of the two layer in a way that

the two layers are compatible with each other

  

70

6- High flux and high energy efficiency can be obtained using the dual layer hydrophilic-

hydrophobic fibers.

7- The non-uniform cross section of hollow fibers can be attributed to an instability

occurring during the fabrication of hollow fibers. It was postulated that the main

instability happens in the coagulation bath where the pressure induced in the nascent fiber

outer layer as a result of diffusion/convection, precipitation, densification and shrinkage

will buckle the rigid elastic shell formed in the interface between the bore fluid and the

dope solution. It is also inferred that since it takes only a fraction of second for a nascent

fiber traveling through the air gap region, the hydrodynamic instability may not have

enough time to magnify its effects. Therefore, it is postulated that the hydrodynamic

instability may be the initial stage of causing instability and the elastic and buckling

instability is the magnifying stage for the development of non-uniform cross section of

hollow fibers.

8- The proposed buckling mechanism can effectively explain all our experimental

observations regarding the effect of different spinning parameters on the final shape of

the hollow fibers

9- The proposed qualitative mechanism can be applied to quantitatively model the

corrugation phenomenon in hollow fiber spinning by phase inversion process

In summary, this thesis provides a contribution to tackle one of the major barriers against

commercialization of membrane distillation process regarding the relatively low membrane flux

in this process, by fabricating high flux PVDF composite membranes. Furthermore, this study

provides an explanation for the irregular cross section phenomenon in hollow fiber membranes

fabricated through non-solvent induced phase inversion process.

  

71

Bibliography

[1] Global Environmental Outlook (GEO) 2000

[2] http://ga.water.usgs.gov/edu/watercyclefreshstorage.html

[3] http://www.palomar.edu/oceanography/salty_ocean.htm

[4] http://www.grida.no/climate/vital/37.htm

[5] J. H. Van’t Hoff, “The role of osmotic pressure in the analogy between solutions and gases,”

Z. Phys. Chem. 1, (1887) 481–508

[6] F. P. Chinard and T. Enns, “Osmotic pressure,” Science 124, (1956) 472–474

[7] K.S. Spiegler, Y.M. El-Sayed, The energetics of desalination processes, Desalination 134

(2001) 109

[8] Y.M. El-Sayed, Designing desalination systems for higher productivity, Desalination 134

(2001) 129

[9] K. S. Spiegler, M. A. El-Sayed, A Desalination Primer Balaban Desalination Publications,

Santa Maria Imbaro, Italy (1994)

[10] R. Trussell, State of the Science of Membrane Technologies, JWR&DTF Workshop, San

Diego, (2005)

[11] http://people.uwec.edu/piercech/desalination/MSF.htm

[12] J. E. Miller, Review of Water Resources and Desalination Technologies, SAND 2003-0800,

Sandia National Laboratories, (2003)

[13] http://pls.atu.edu/biology/people/elovely/Water/new_page_1_files/image004.jpg

  

72

[14] G. F. Leitner, Desalination and Water Reuse Quarterly 5, (1995) 24.

[15] http://www.pacinst.org/images/desal_typical_costs.gif

[16] B.R. Bodell, Silicone rubber vapor diffusion in saline water distillation, 691 United States

Patent

Serial No. 285,032 (1963)

[17] K. W. Lawson, D. R. Lloyd, Membrane distillation (Review), J. Membr. Sci. 124 (1997) 1

[18] M.S. El-Bourawi, Z. Ding, R. Ma, M. Khayet, A framework for better understanding

membrane

distillation separation process, J. Membr. Sci. 285 (2006) 4

[19] E. Curcio, E. Drioli, Membrane Distillation and Related Operations (A Review), Sep. Pur.

Rev., 34: 35–86, 2005

[20] K. W. Lawson, D.R. Lloyd, Membrane distillation. II. Direct contact membrane distillation,

J.

Membr. Sci. 120 (1996) 123 [21] S.H. Duan, A. Ito, A. Ohkawa, Removal of trichloroethylene from water by aeration,

pervaporation and membrane distillation, J. Chem. Eng. Jpn. 34 (8) (2001) 1069–1073

[22] F.A. Banat, J. Simandl, Removal of benzene traces from contaminated water by vacuum

membrane distillation, Chem. Eng. Sci. 51 (8) (1996) 1257–1265.

[23] S. Nene, S. Kaur, K. Sumod, B. Joshi, K.S.M.S. Raghavarao, Membrane distillation for the

concentration of raw cane-sugar syrup and membrane clarified sugarcane juice, Desalination 147

(2002) 157–160.

  

73

[24] C. Varming, M.L. Andersen, L. Poll, Influence of thermal treatment on black currant (Ribes

nigrum L.) juice aroma, J. Agric. Food Chem. 52 (25) (2004) 7628–7636.

[25] N. Qureshi, M.M. Meagher, R.W. Hutkins, Recovery of 2,3-butanediol by vacuum

membrane distillation, Sep. Sci. Technol. 29 (13) (1994) 1733–1748.

[26] F. A. Banat, M. Al-Shannag, Recovery of dilute acetone–butanol–ethanol (ABE) solvents

from aqueous solutions via membrane distillation, Bioprocess Eng. 23 (6) (2000) 643–649.

[27] M. Khayet, A. Vel´azquez, J.I. Mengual, Direct contact membrane distillation of humic acid

solutions, J. Membr. Sci. 240 (2004) 123–128

[28] M. Khayet, J.I. Mengual, Effect of salt concentration during the treatment of humic acid

solutions by membrane distillation, Desalination 168 (2004) 373–381

[29] F. Banat, S. Al-Asheh, M. Qtaishat, Treatment of waters colored with methylene blue dye by

vacuum membrane distillation, Desalination 174 (2005) 87–96.

[30] R. W. Schofield, A. G. Fane, C. J. D. Fell, Gas and vapour transport through microporous

membranes II. Membrane distillation, J. Membr. Sci., 53 (1990) 173-185

[31] R. W. Schofield, A. G. Fane, C. J. D. Fell, Heat and mass transfer in membrane distillation,

Journal of Membrane Science, 33 (1987) 299-313

[32] S. Kimura, S.I. Nakao, S.I. Shimatani, Transport phenomena in membrane distillation, J.

Membr. Sci. 33 (3) (1987) 285–298

[33] K. Schnieder, T.J. van Gassel, Membrane distillation, Chem. Eng. Technol. 56 (1984) 514–

521

  

74

[34] K.W. Lawson, D.R. Lloyd, Membrane distillation. I. Module design and performance

evaluation using vacuum membrane distillation, J. Membr. Sci. 120 (1996) 111–121

[35] C. Zhu, G.L. Liu, C.S. Cheung, C.W. Leung, Z.C. Zhu, Ultrasonic stimulation on

enhancement of air gap membrane distillation, J. Membr. Sci. 161 (1999) 85–93

[36] J. Phattaranawik, R. Jiraratananon, A.G. Fane b, C. Halim, Mass flux enhancement using

spacer filled channels in direct contact membrane distillation, J. Membr. Sci 187 (2001) 193–201

[37] J. Phattaranawik, R. Jiraratananon, A. G. Fane, Heat transport and membrane distillation

coefficients in direct contact membrane distillation, J. Membr. Sci. 212 (2003) 177–193

[38] L. Martinez-Diez, M.I. Vazquez-Gonzalez, F.J. Florido-Diaz, Study of membrane distillation

using channel spacers, J. Membr. Sci. 144 (1998) 45-46

[39] M. M. Teoh, S. Bonyadi, T. S. Chung, Investigation of different hollow fiber module designs

for flux enhancement in the membrane distillation process, J. Membr. Sci. 311 (2008) 371-379

[40] A. Burgoyne, M.M. Vahdati, Direct contact membrane distillation, Sep. 693 Sci. Technol. 35

(2000) 1257

[41] E. Drioli, A. Criscuoli, E. Curcio, Membrane Contactors: Fundamentals, Applications and

Potentials, 11, Elsevier (Chapter 2)

[42] M. Khayet, C. Y. Feng, K. C. Kulbe, T. Matsuura, Preparation and characterization of

polyvinylidene fluoride hollow fiber membranes for ultrafiltration, Polymer 43 (2002) 3879-3890

[43] W.Z. Langa, Z.L. Xua,b, H. Yanga, W. Tong, Preparation and characterization of PVDF-

PFSA blend hollow fiber UF membrane, J. Membr. Sci. 288 (2007) 123–131

  

75

[44] M. Khayet, C.Y. Feng, K.C. Khulbe, T. Matsuura, Study on the effect of a non-solvent

additive on the morphology and performance of ultrafiltration hollow-fiber membranes,

Desalination 148 (2002) 321-327

[45] M. Khayet, The effects of air gap length on the internal and external morphology of hollow

fiber membranes, Chemical Engineering Science 58 (2003) 3091 – 3104

[46] J. Ren, R. Wang, H. Y. Zhang, Z. Li, D. T. Liang, J. H. Tay, Effect of PVDF dope rheology

on the structure of hollow fiber membranes used for CO2 capture, J. Membr. Sci. 281 (2006)

334–344

[47] J. Ren, R. Wang, H. Y. Zhang, Z. Li, D. T. Liang, J. H. Tay Effect of PVDF dope rheology

on the structure of hollow fiber membranes used for CO2 capture, J. Membr. Sci. 281 (2006)

334–344

[48] S. Atchariyawut, C. Feng, R. Wang, R. Jiraratananon, D.T. Liang, Effect of membrane

structure on mass-transfer in the membrane gas–liquid contacting process using microporous

PVDF hollow fibers, Journal of Membrane Science 285 (2006) 272–281

[49] K. Li, J. F. Kong, Dongliang Wang, and W. K. Teo, Tailor-Made Asymmetric PVDF Hollow

Fibers for Soluble Gas Removal, AIChE Journal (1999) Vol. 45, No. 6 1211

[50] D. Wang, W.K. Teo, K. Li, Removal of H2S to ultra-low concentrations using an

asymmetric hollow fibre membrane module, Sep. Puri. Tech. 27 (2002) 33–40

[51] H. Zhen, S. M. J. Jang, W. K. Teo, K. Li, Modified Silicone–PVDF Composite Hollow-

Fiber

Membrane Preparation and Its Application in VOC Separation, Journal of Applied Polymer

Science, Vol. 99, 2497–2503 (2006)

  

76

[52] M. Ortiz de Zfirate, L. Pefia, J.I. Mengual, Characterization of membrane distillation

membranes prepared by phase inversion, Desalination 100 (1995) 139-148

[53] Tomaszewska, M., 1996. Preparation and properties of flat-sheet membranes from

poly(vinylidene fluoride) for membrane distillation. Desalination 104, 1

[54] M. Khayet, T. Matsuura, Preparation and characterization of polyvinylidene fluoride

membranes for membrane distillation, Ind. Eng. Chem. Res. 40 724 (2001) 5710

[55] (a) D.Y. Cheng, S.J. Wiersma, Composite membranes for a membrane distillation system,

U.S. Patents 4,316,772 (1982);

(b) D.Y. Cheng, S.J. Wiersma, Composite membranes for a membrane distillation system, U.S.

Patents 4,419,242 (1983).

[56] Y. Wu, Y. Kong, X. Lin, W. Liu, J. Xu, Surface-modified hydrophilic membranes in

membrane distillation, J. Membr. Sci. 72 (1992) 189.

[57] Y. Kong, X. Lin, Y.Wu, J. Cheng, J. Xu, Plasma polymerization of octaflu-orocyclobutane

and hydrophobic microporous composite membranes for membrane distillation, J. Appl. Polym.

Sci. 46 (1992) 191.

[58] M. Khayet, T. Matsuura, Application of surface modifying macromolecules for the

preparation of membranes for membrane distillation, Desalination 158 (2003) 51

[59] M. Khayet, J. I. Mengual, T. Matsuura, Porous hydrophobic/hydrophilic composite

membranes application in desalination using direct contact membrane distillation, J. Membr. Sci.

252 (2005) 101

  

77

[60] P. Peng, A.G. Fane, X. Li, Desalination by membrane distillation adopting a hydrophilic

membrane, Desalination 173 (2005) 45-54

[61] W. Kasta, C. R. Hohenthanner, Mass transfer within the gas-phase of porous media, Int. J.

Heat Mass Trans. 43 (2000) 807-823

[62] J. Phattaranawik, R. Jiraratananon, and A. G. Fane, Effect of Pore Size Distribution and Air

Flux on Mass Transport in Direct Contact Membrane Distillation. J. Membrane Sci., 215 (2003)

75–85.

[63] M. Gryta, M. Tomaszewska, Heat transport in the membrane distillation process, J. Membr.

Sci. 144 (1998) 211

[64] B.A. Li, K.K. Sirkar, Novel Membrane and Device for Direct Contact, Membrane

Distillation-Based Desalination Process, Ind. Eng. Chem. Res. 43 (2004) 5300

[65] J. I. Mengual, M. Khayet, M. P. Godino, Heat and mass transfer in vacuum membrane

distillation, Int. J. Heat. Mass Tra. 47 (2004) 865.

[66] M. Gryta, Influence of polypropylene membrane surface porosity on the performance of

membrane distillation process, Journal of Membrane Science 287 (2007) 67–78

[67] Li Y, Chung TS, Exploration of highly sulfonated polyethersulfone (SPES) as a membrane

material with the aid of dual-layer hollow fiber fabrication technology for protein separation, J.

Membr. Sci. 2008; 309:45-55.

[68] Widjojo N, Zhang SD, Chung TS, Liu Y, Enhanced gas separation performance of dual-layer

hollow fiber membranes via substructure resistance reduction using mixed matrix materials, J.

Membr. Sci. 2007; 306:147-158.

  

78

[69] R. W. Schofield, P. A. Hogan, A. G. Fane, C. J. D. Fell, Developments in membrane

distillation, Desalination, 62 (1987) 728.

[70] F. Lagana, G. Barbieri, E. Drioli, Direct contact membrane distillation: modeling and

concentration experiments, J. Membr. Sci. 166 (2000) 1.

[71] L. Mart´ınez, J.M. Rodr´ıguez-Maroto, Membrane thickness reduction effects on direct

contact membrane distillation performance, Journal of Membrane Science 312 (2008) 143–156

[72] Zhang, Q. M., Bharti, V., Kavarnos, G., Schwartz, M. (Ed.), (2002). "Poly (Vinylidene

Fluoride) (PVDF) and its Copolymers", Encyclopedia of Smart Materials, Volumes 1-2, John

Wiley & Sons, 807-825

[73] D. W. Chae, B. C. Kim, Effects of Zinc Oxide Nanoparticles on the Physical Properties of

Polyacrylonitrile, Journal of Applied Polymer Science, Vol. 99, 1854–1858 (2006)

[74] L. Palacio, P. Pradanos, J.I. Calvo, A. Hernandez, Porosity measurements by a gas

penetration method and other techniques applied to membrane characterization, Thin Solid Films

348 (1–2) (1999) 22–29.

[75] D. L. Wang, W.K. Teo, K. Li, Preparation and characterization of high-flux polysulfone

hollow fibre gas separation membranes, J. Membr. Sci. 204 (2002) 247

[76] D.F. Li, T.S. Chung, R. Wang, Y. Liu, Fabrication of fluo768ropolyimide/polyethersulfone

(PES) dual-layer asymmetric hollow fiber membranes for gas separation, J. Membr. Sci. 198

(2002) 211

[77] D. Li, R. Wang, T.S. Chung, Fabrication of lab-scale hollow fiber membrane modules with

high packing density, Sep. Pur. Tech. 40 (2004) 15–30

  

79

[78] O. Olabisi, L.M. Robeson, M.T. Shaw, Polymer–Polymer Miscibility, Academic Press, New

York, 1979 (Chapter 2)

[79] T. Yamaguchi, S. Nakao, S. Kimura, Design of pervaporation membrane of organic-liquid

separation based on solubility control by plasma-graft filling polymerization technique, Ind. Eng.

Chem. Res. 32 (1993) 848–853.

[80] J.W. van Dyk, H.L. Frisch, D.T. Wis, Solubility, solvency and solubility 782 parameters,

Ind. Eng. Chem. Prod. Res. Dev. 24 (1985) 473

[81] A. Bowino, G. Capannelli, S. Munari, A. Turturro, Solubility parameters of poly(vinylidene

fluoride), J. Polym. Sci. B: Polym. Phys. 26 (1988) 785

[82] A.J. Reuvers, C.A. Smolders, Formation of membranes by means of immersion

precipitation. Part II, J. Membr. Sci. 34 (1987) 67

[83] P. Radovanovic, S.W. Thiel, S.T. Hwang, Formation of asymmetric polysulfone membranes

by immersion precipitation. Part II, J. Membr. Sci. 65 (1992) 231

[84] F. Hussain, M. Hojjati, M. Okamoto, R.E. Gorga, Polymer–matrix nanocomposites,

processing, manufacturing, and application: an overview, J. Compos. Mater. 40 (2006) 1511

[85] M. Gryta, M. Tomaszewska, J. Grzechulska, A.W. Morawski, Membrane distillation of

NaCl solution containing natural organic matter, J. Membr. Sci. 181 (2001) 279

[86] L. Song, B. Li, K.K. Sirkar, J.L. Gilron, Direct contact membrane distillation-based

desalination: novel membranes, devices, larger-scale studies and a model, Ind. Eng. Chem. Res.

46 (2007) 2307

  

80

[87] J. Li, Z. Xu, Z. Liu, W. Yuan, H. Xiang, S. Wang, Y. Xu, Microporous polypropelene and

polyethelene hollow fiber membranes. Part 3. Experimental studies on membrane distillation for

desalination, Desalination 155 (2003) 153

[88] Y. Fujii, S. Kigoshi, H. Iwatani, M. Aoyama, Selectivity and characteristics of direct contact

membrane distillation type experiment. I. Permeablity and selectivity through dried hydrophobic

fine porous membranes, J. Membr. Sci. 72 (1992) 53

[89] S. A. McKelvey, D. T. Clausi, W. J. Koros, A guide to establishing hollow fiber

macroscopic properties for membrane applications, J. Membr. Sci. 124 (1997) 223

[90] K Y. Wang, T. Matsuura, T. S. Chung, W. F. Guo, The effects of flow angle and shear rate

within the spinneret on the separation performance of poly(ethersulfone) (PES) ultrafiltration

hollow fiber membranes, J. Membr. Sci. 240 (2004) 67

[91] H.A. Tsai, C.Y. Kuo, J.H. Lin, D.M. Wang, A. Deratani, C. Pochat-Bohatier, K.R. Lee, J.Y.

Lai, Morphology control of polysulfone hollow fiber membranes via water vapor induced phase

separation, J. Membr. Sci. 278 (2006) 390

[92] Y. Su , G. G. Lipscomb , H. Balasubramanian , D. R. Lloyd, Observations of recirculation in

the bore fluid during hollow fiber spinning, AICHE, 52 (2006), 2072

[93] R.G. Larson, Instabilities in viscoelastic flows, Rheol Acta, 31 (1992) 213

[94] C. J. S. Petrie, M. M. Denn, Instabilities in polymer processing, AICHE Journal, 22 (1976),

209

[95] G. G. Lipscomb, The melt hollow fiber spinning process: steady-state behavior, sensitivity

and stability, polymers for advanced technologies, 5 (1994) 745.

  

81

[96] W. Nijdam, J. de Jong, C.J.M. van Rijn, T. Visser, L. Versteeg, G. Kapantaidakis, G.H.

Koops, M. Wessling, High performance micro-engineered hollow fiber membranes by smart

spinneret design, J. Membr. Sci., 256 (2005) 209

[97] N. Widjojo, T.S. Chung, The thickness and air-gap dependence of macrovoid evolution in

phase-inversion asymmetric hollow fiber membranes, Ind. Chem. Eng. Res. 45 (2006) 7618.

[98] Y. E. Santoso, T. S. Chung, K. Y. Wang, M. Weber, The investigation of irregular inner-skin

morphology of hollow fiber membranes at high speed spinning and the solutions to overcome it,

J. Membr. Sci., 282 (2006), 383.

[99] J. P. Van’t Hoff, Wet spinning of polyethersulfone gas separation membranes, Ph.D. Thesis,

Twente University, 1988 (Chapter 2).

[100] C.C.Pereira, R.Nobrega, C.P.Borges, Spinning process variables and polymer solution

effects in the die swell phenomenon during hollow fiber membrane formation, Brazil. J. Chem.

Eng. 17 (2000), 599

[101] H. Strathmann, K. Kock, P. Amar, R.W. Baker, The formation mechanism of asymmetric

membranes, Desalination, 16 (1975), 179

[102] P.S.T. Machado, A.C. Habert, C.P. Borges, Membrane formation mechanism based on

precipitation kinetics and membrane morphology: flat and hollow fiber polysulfone membranes,

J. Membr. Sci., 155 (1999), 171

[103] F. Bloom, D. Coffin, Handbook of Thin Plate Buckling and Postbuckling, CHAPMAN &

HALL/CRC, 2000 (Chapter 1)

[104] L. Yilmaz, A. J. McHugh, Analysis of Nonsolvent -Solvent -Polymer Phase Diagrams and

Their Relevance to Membrane Formation Modeling, J. App. Poly. Sci. 31 (1986), 997

  

82

[105] L. Yilmaz, A. J. McHugh, Modeling of asymmetric membrane formation II. The effects of

surface boundary conditions, J. App. Poly. Sci. 35 (1988), 1967

[106] S. S. Shojaie, W. B. Krantz, A. R. Greenberg, Membrane formation via the dry-cast process

Part I. Model development, J. Membr. Sci., 94 (1994), 255

[107] M. Levy, Memoire sur un nouveau cas integrable du probleme de l’elastique et l’une de Ses

applications. J. de Math. (Liouville), Series 3, 10 (1884), 5.

[108] A. G. Greenhill, The elastic curve under uniform normal pressure, Math Ann, 52 (1899),

465.

[109] I. Tadjbakhsh, F. Odeh, Equilibrium states of elastic rings, J. Math. Ann. App., 18 (1967),

59.

  

83

  

84

Appendix

List of papers prepared and published during M. Eng Study

1. S. Bonyadi, T.S. Chung, Flux Enhancement in Membrane Distillation by Fabrication of

Dual Layer Hydrophilic-Hydrophobic Hollow Fiber Membranes, Journal of Membrane

Science, 306 (2007) 134–146

2. S. Bonyadi, T.S. Chung, W.B. Krantz, Investigation of the Corrugation Phenomenon

in the Inner Contour of Hollow Fibers during Non-solvent Induced Phase Separation

Process, Journal of Membrane Science, 299 (2007) 200-210

3. M.M. Teoh, S. Bonyadi, T.S. Chung, J.J. Shieh, Investigation of different module

designs for flux enhancement in the membrane distillation process, Journal of

Membrane Science, 311 (2008) 371–379

4. S. Bonyadi, T.S. Chung, K.Y. Wang, Fabrication of High Performance Dual Layer

Hydrophilic-Hydrophobic Hollow Fibers for Membrane Distillation Process, a US Patent

filed, March 2007

5. S. Bonyadi, T.S. Chung, W.B. Krantz, Investigation of the Corrugation Phenomenon in

the Inner Contour of Hollow Fibers during Non-solvent Induced Phase Separation

AIChE Annual Meeting (2007) Salt Lake City, Utah

 

6. S. Bonyadi, T.S. Chung, Flux Enhancement in Membrane Distillation by Fabrication of

Dual Layer Hydrophilic-Hydrophobic Hollow Fiber Membranes AIChE Annual Meeting

(2007) Salt Lake City, Utah