 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

SI08 KzlProj/Acc8Pres/CIRC/Dateoue.indd

COMBINED FOULING OF PRESSURE-DRIVEN MEMBRANES TREATING FEED
WATERS OF COMPLEX COMPOSITION

By

Fulin Wang

A DISSERTATION

Submitted to
Michigan State University
In partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Environmental Engineering

2008

ABSTRACT
COMBINED FOULING OF PRESSURE-DRIVEN MEMBRANES TREATING FEED
WATERS OF COMPLEX COMPOSITION
By

Fulin Wang

Pressure-driven membrane ﬁltration processes can serve as a reliable barrier for
microorganisms, organics, precursors of disinfection by-products (DBPs), colloids, and
other contaminants. However, membrane fouling impedes a wider application of
membrane ﬁltration technology. Because typical feed waters have complex composition
and the pretreatment often can not guarantee a complete removal of foulants, different
fouling mechanisms act simultaneously resulting in combined fouling. Although each of
the individual fouling mechanisms has been extensively studied, there have been very
few studies examining the effect of combined fouling on membrane performance. The
focus of this work is on identifying contributions of individual fouling mechanisms in the
overall fouling process and on evaluating the synergistic effects stemming from

interactions between individual mechanisms.

In Chapter 2, concentration of rejected salt at the membrane surface when colloidal
particles were deposited on the membrane is determined experimentally based on
measured salt permeability constant. This approach allowed for a clear identiﬁcation of
individual contributions of concentration polarization and colloidal fouling to the

permeate ﬂux decline.

In Chapter 3, a method based on a simple linear regression n ﬁtting is proposed and used
to determine the type, the chronological sequence, and the relative importance of
individual fouling mechanisms in the experiments on the dead-end membrane ﬁltration of
colloidal suspensions. For all membranes, ﬂux decline was consistent with one or more
pore blocking mechanisms during the earlier stages and with the cake ﬁltration

mechanism during the later stages of ﬁltration.

In Chapter 4, a model is proposed for predicting the permeability of porous media
consisting of ﬁbers and spherical particles. The model combines Kozeny-Carman
equation and Ethier’s theoretical model to predict the permeability of the mixed media
based on the permeability of the porous bed of particles alone and on the volume fraction
of the ﬁbers. The model prediction was found to be in accordance with the experimental

observations for the ﬁber mass fractions of less than 50%.

Chapters 5 and 6 examine the effect of combined colloidal fouling and gypsum scaling on
the performance of RO membranes. In Chapter 5, the effect of silica colloids on gypsum
scaling is studied in a batch system. The results showed that silica colloids retarded the
induction of gypsum crystallization and decreased the nucleation rate. It was also found
that gypsum crystallization rate increased in the presence of silica colloids. Chapter 6
presents the ﬁrst study on the combined colloidal fouling and gypsum scaling in RO
membrane systems. A signiﬁcant synergistic effect between gypsum scaling and colloidal

fouling was observed.

To my dearest parents, my brother, and my sister.

iv

ACKNOWLEDGEMENTS

I would like to express my deepest gratitude and indebtedness to my advisor and
dissertation committee chairman, Dr. Volodymyr Tarabara, for his continuous mentoring,
support, and guidance during this work. I would also like to thank him for his generous
encouragement, infectious enthusiasm and constructive criticisms, without which the

work described in this dissertation would not have been possible.

I would also like to extend my great appreciation to the other dissertation committee
members: Dr. Phanikumar S. Mantha, Dr. Merlin L. Bruening, and Dr. Manoochehr

Koochesfahani for their valuable comments and suggestions on this work.

I would like to thank Dr. Seokjong Byun, Dr. Jeonghwan Kim, Julian Taurozzi, and
Wenqian Shan, Alla Alpatova, Adam Rogensues, Elodie Pasco, Tan Zhao, for many

ﬁ'uitful discussions and for their kind help in sharing lab equipment.

I would like to acknowledge Ms. Lori Hasse and Mr. Yang-lyang Pan for the
administrative and research supports. I also would like to thank Ms. Ewa Danielewicz, Dr.
Alicia Pastor, and Ms. Carol S. Flegler for their constant assistance with microscopy

sample preparation.

I would like to thank the Department of Civil and Environmental Engineering, Michigan
State University, and the National Water Research Institute for the ﬁnancial support of

this work.

Finally I would like to thank my parents, my brother, and my sister for their enduring

love, support, sacriﬁce, and motivation over the years.

vi

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. xi
LIST OF FIGURES .......................................................................................................... xii
ABBREVIATIONS .......................................................................................................... xv
CHAPTER ONE

COMBINED FOULING ON THE PERFORMANCE OF PRESSURE-DRIVEN
MEMBRANES TREATING WATERS OF COMPLEX COMPOSITION: CURRENT
UNDERSTANDING AND FUTURE RESEARCH NEEDS

1 .1. Introduction .................................................................................................................. 1
1.1.1. Overview of membrane ﬁltration processes ......................................................... 1
1.1.2. Combined fouling in practice ................................................................................ 3

1.2. Current research on the effect of combined fouling on membrane performance ........ 6
1.2.1. Colloidal fouling and salt concentration polarization in salt-rejecting membrane
systems.. ......................................................................................................................... 6

1.2.1.1. Colloidal fouling ............................................................................................ 6
1.2.1.2. The structure of colloidal deposit layers ........................................................ 9
1.2.1.3. Coupled effect of colloidal deposition and salt concentration polarization. 10
1.2.2. Combined colloidal and organic fouling ............................................................. 12
1.2.3. Colloidal fouling and scaling .............................................................................. 14
1.2.4. Other combined fouling ...................................................................................... 16

1.3. The future research need ............................................................................................ 17

Reference .......................................................................................................................... 18

CHAPTER TWO

COUPLED EFFECTS OF COLLOIDAL DEPOSITION AND SALT
CONCENTRATION POLARIZATION ON REVERSE OSMOSIS MEMBRANE
PERFORMANCE

ABSTRACT ...................................................................................................................... 27
2. 1. INTRODUCTION ..................................................................................................... 28
2.2. BACKGROUND ....................................................................................................... 32
2.3. EXPERIMENTAL METHODS ................................................................................. 37
2.3.1. Experimental apparatus ....................................................................................... 37
2.3.2. Reagents and chemical solutions ........................................................................ 40
2.3.3. Silica colloids and R0 membranes ..................................................................... 41
2.3.4. Experimental protocol ......................................................................................... 42
2.3.4.1. Membrane compaction ................................................................................. 42

vii

2.3.4.2. Pure water permeate ﬂux and measurement of membrane hydraulic

 

resistance ................................................................................................................... 43
2.3.4.3. Membrane conditioning ............................................................................... 43
2.3.4.4. Membrane fouling experiments ................................................................... 44
2.4. RESULTS AND DISCUSSION ................................................................................ 45
2.4.1. Characteristics of colloidal particles and the membrane .................................... 45
2.4.2. Changes in membrane resistance during conditioning. Correction factor .......... 46
2.4.3. Measurement of the membrane salt permeability constant ................................. 49
2. 4. 4. Permeate ﬂux, NaCl rejection, and structure of the deposited layer at different
ionic strengths ................................................................................ 51
2. 4. 5. Hypothesis of the concentrated ﬂowing layer of silica partibles and its
depolarizing effect on NaCl transport ........................................................................... 59
2.4.6. Evolution of relative contributions of osmotic pressure and resistance of the
colloidal deposit to the permeate ﬂux decline .............................................................. 60
2.4.7. Effect of the feed channel spacers on NaCl rejection and permeate flux ........... 61
2.5. CONCLUSIONS ........................................................................................................ 65
ACKNOWLEDGEMENTS .............................................................................................. 66
REFERENCE .................................................................................................................... 68
82. SUPPLEMENTARY DATA ..................................................................................... 72
8.2.1. Determination of the salt permeability ............................................................... 72
8.2.2. Imaging the fouling layer using transmission electron microscopy (TEM) ....... 73
CHAPTER THREE

ON THE IDENTIFICATION OF PORE BLOCKING MECHANISMS DURING
EARLY STAGES OF MEMBRANE FOULING BY COLLOIDS

ABSTRACT ...................................................................................................................... 75
3. 1 . INTRODUCTION ..................................................................................................... 76
3.2. Materials and methods ............................................................................................... 81
3.2.1. Membranes and colloids ..................................................................................... 81
3.2.2. Filtration experiments ......................................................................................... 82
3.2.3. Scanning electron microscopy (SEM) analysis .................................................. 84
3.2.4. Data ﬁtting procedure ......................................................................................... 84
3.2.5. Determination of the effective porosity of the colloidal deposit ........................ 85
3.3. Results and discussion ............................................................................................... 86
3.3.1. Flux performance of different membranes ......................................................... 86
3.3.2. Blocking mechanisms for UF membranes .......................................................... 89
3.3.3. Blocking mechanisms for salt—rejecting membranes .......................................... 91
3.4. Conclusions ................................................................................................................ 96
Acknowledgements ........................................................................................................... 97
Reference .......................................................................................................................... 98
S3. SUPPLEMENTARY DATA .................................................................................... 101
viii

CHAPTER FOUR

MODELING THE PERMEABILITY OF FIBER-FILLED POROUS MEDIA USING
KOZENY-CARMAN-ETHIER EQUATION

4. 1. INTRODUCTION ................................................................................................... 108
4.2. Theoretical background ........................................................................................... 110
4.2.1. The hydraulic resistance of a deposit layer in membrane systems ................... 110
4.2.2. The hydraulic resistance of porous media of particles ...................................... 112
4.2.3. The permeability of highly polydisperse ﬁbrous media ................................... 112
4.3. Experimental method ............................................................................................... 113
4.3.1. Experimental apparatus ..................................................................................... 1 13
4.3.2. Reagents and chemicals .................................................................................... 116
4.3.3. Ultraﬁltration (UF) membrane, Silica colloids, and carbon ﬁbers .................... 116
4.3.4. Scanning electron microscopy (SEM) analysis ................................................ 117
4.3.5. Experimental protocol ....................................................................................... 117
4.3.5.1. Preparation of the feed suspensions ........................................................... 117
4.3.5.2. Membrane compaction ............................................................................... 117
4.3.5.3. Measurement of membrane hydraulic resistance using pure water ........... 118
4.3.5.4. Membrane ﬁltration experiments ............................................................... 118
4.4. Results and discussion ............................................................................................. 119
4.4.1. Characterization of the silica colloids and carbon ﬁbers .................................. 119
4.4.2. Permeate ﬂux and hydraulic resistance of the deposit layer ............................. 122
4.4.3. SEM images of the deposit layers ..................................................................... 124
4.4.4. Modeling the hydraulic resistance of the deposit layers consisting of silica
colloids and ﬁbers ....................................................................................................... 124
4.4.4.1. Theoretical development ............................................................................ 124
4.4.4.2. The effective porosity of the porous bed of silica colloids ........................ 128
4.4.4.3. Prediction of the hydraulic resistance of the deposit layer consisting of silica
colloids and carbon ﬁbers ....................................................................................... 131
4.5. Conclusions .............................................................................................................. 134
Acknowledgments ........................................................................................................... 135
Nomenclature .................................................................................................................. 136
Reference ........................................................................................................................ 138
CHAPTER FIVE

CRYSTALLIZATION OF GYPSUM IN THE PRESENCE OF COLLOIDAL SILICA. I:
IN A STIRRED BATCH SYSTEM

ABSTRACT .................................................................................................................... 141
5. 1 . Introduction .............................................................................................................. 143
5.2. Background .............................................................................................................. 146
5.3. Material and methods ............................................................................................... 151
5.3.1. Experimental apparatus ..................................................................................... 151
5.3.2. Regents, silica colloids, and feed suspensions .................................................. 153

ix

5.3.3. Characterization of the gypsum crystals ........................................................... 155

5.4. Results and discussion ............................................................................................. 156
5.4.1. Determination of induction time ....................................................................... 156
5.4.2. Effect of silica colloids on the induction of gypsum crystallization ................. 159
5.4.3. Effect of silica colloids on the surface energy and nucleation rate of gypsum
crystallization .............................................................................................................. 163
5.4.4. Effect of silica colloids on the gypsum crystallization rate .............................. 165
5.4.5. Effect of silica colloids on crystal size distribution .......................................... 166
5.4.6. Effect of silica colloids on the morphology of gypsum crystals ....................... 166

5.4.6.1. Length-to-width ratio ................................................................................. 168
5.4.6.2. Implications of epitaxial growth of gypsum for crystal morphology ........ 169

5.5. Conclusions .............................................................................................................. 171

Acknowledgments ........................................................................................................... 172

Nomenclature .................................................................................................................. 173

Reference ........................................................................................................................ 174

SS. SUPPLEMENTARY DATA .................................................................................... 179

CHAPTER SIX

CRYSTALLIZATION OF GYPSUM IN THE PRESENCE OF COLLOIDAL SILICA.
11: IN A REVERSE OSMOSIS MEMBRANE SYSTEM

Abstract ........................................................................................................................... 183
6.1 . Introduction .............................................................................................................. 184
6.2. Experimental ............................................................................................................ 187
6.2.1. Feed suspensions ............................................................................................... 187
6.2.2. Experimental apparatus ..................................................................................... 189
6.2.3. Reagents, silica particles and R0 membranes .................................................. 190
6.2.4. Microscopical analysis ...................................................................................... 190

6.2.5. Protocols for membrane fouling experiments with less saturated suspensions 191
6.2.6. Study the effect of silica particles on gypsum crystallization with supersaturated

suspensions ................................................................................................................. 1 92
6.3. Results and discussion ............................................................................................. 193
6.3.1. Combined colloidal fouling and gypsum scaling in less saturated suspensions 193
6.3.1.1. Effect on RO membrane performance ....................................................... 193
6.3.1.2. Effect of colloidal fouling on gypsum scaling ........................................... 198
6.3.1.3. Effect of gypsum scaling on colloidal fouling ........................................... 200
6.3.1.4. Microscopical analysis of the fouling layer on membrane surface ............ 200
6.3.2. Combined colloidal fouling and gypsum scaling in supersaturated suspensions
..................................................................................................................................... 203
6.3.2.1. Effect of silica particles on gypsum crystallization ................................... 203
6.3.2.2. Effect on RO membrane performance ....................................................... 207
6.3.2.3. Microscopical analysis of the fouling layer on membrane surface ............ 207
6.4. Conclusions .............................................................................................................. 210
Acknowledgments ........................................................................................................... 21 1
Reference ........................................................................................................................ 212

LIST OF TABLES

Table 1.1. Examples for combined fouling occurring in practice ....................................... 5
Table 1.2. Published studies of combined fouling in membrane systems .......................... 7
Table 2.1. R0 experimental protocol ................................................................................ 48
Table 2.2. Measurement of the membrane salt permeability constant ............................. 50
Table 82.1. Measurement of the membrane salt permeability constant ........................... 72
Table 3.1. Membranes used in this study .......................................................................... 83
Table 4.1. Rm in different experiments .......................................................................... 123
Table 5.1. Composition of model solutions .................................................................... 154

Table 5.2. Estimated critical size of nuclei in the absence and presence of silica colloids
................................................................................................................................. 162

xi

LIST OF FIGURES

Figure 2.1. Schematic of the experimental apparatus ....................................................... 39

Figure 2.2. Flux change during conditioning, at a transmembrane pressure difference of
300 psi (2.068 MPa) and an ionic strength of 10'2 M. .............................................. 49

Figure 2.3. (a) Normalized permeate ﬂux at three ionic strengths;, (b) observed salt
rejection; (c) salt concentration at membrane surface; ((1) mass of deposited particles;

(e) effective porosity, and (f) hydraulic resistance of deposited colloidal layer ....... 52

Figure 2.4. Effect of introduction of silica particles on initial observed salt rejection at

10'2 M ........................................................................................................................ 57
Figure 2.5 Energy consumption analysis for the system: ................................................. 62
Figure 2.6. Effect of the feed channel Spacers on membrane performance ...................... 64
Fig. 82.1. TEM imaging of the deposit layer formed at 0.1 M ......................................... 73

Figure 3.1. Schematic illustration of the four fouling mechanisms: (a) complete blocking,
(b) standard blocking, (c) intermediate blocking, and ((1) cake ﬁltration. ................ 77

Figure 3.2. Normalized instantaneous permeate flux (a) and revere cumulative ﬂux (b) as
a function of permeate volume for different membranes .......................................... 88

Figure 3.3. The distributions and contributions of the blocking mechanisms in terms of
permeate volume (a) and percentage of ﬂux decline (b). ......................................... 90

Figure 3.4. SEM images of the ﬁltration surfaces of membranes. (a) BW30-365; (b)
NF90; (c) UF-30kDa; (d) UF-lOOkDa; and (e) UF-300kDa. ................................... 93

Figure 3.5. Effective porosity of the layer of silica colloids deposited onto the surface of
BW30-365 and NF90 membranes. ........................................................................... 95

Fig. S3. 1. Application of blocking laws to the ﬁltration data of UFlOOkDa membrane: (a)
1/ Q versus t based on Eq. (3.7) for intermediate blocking; (b) t/ V versus I based

on Eq. (3.2) for standard blocking; and (c) t/V versus V based on Eq. (3.4) for
cake ﬁltration. ......................................................................................................... 101

Fig. 832. Application of blocking laws to the R0 and NF membrane ﬁltration data: t/ V
versus V based on Eq. (3.4) for cake ﬁltration. ..................................................... 102

xii

Fig. 83.3. Application of blocking laws to the R0 and NF membrane ﬁltration data: (a)
Q versus V based on Eq. (3.6) for complete blocking, and (b) 1/ Q versus I based

on Eq. (3.7) for intermediate blocking. ................................................................... 103
Figure 4.1. Idealized cross-section view of the porous matrix ....................................... 115
Figure 4.2. SEM images for the silica colloids (a) and carbon ﬁbers (b) ....................... 120
Figure 4.3. Permeate ﬂux (a) and hydraulic resistance of the deposit layer (b). ............ 121
Figure 4.4. SEM of the deposition layers (continued on next page) ............................... 125
Figure 4.4. SEM of the deposition layers. ...................................................................... 126
Figure 4.5. Porosity of the silica deposit layer as a ﬁinction of permeate volume ......... 130

Figure 4.6. Comparison between model prediction and experimental measurement for the
hydraulic resistance of the deposition layer ............................................................ 133

Figure 5.1. Schematic of the experimental batch system ................................................ 152
Figure 5.2. Comparison of different techniques for measuring gypsum induction time 157

Figure 5.3. The transmittance (a) and conductivity (b) of the feed solutions as a function
of time ..................................................................................................................... 160

Figure 5.4. Effect of silica colloids on (a) induction time, (b) nucleation rate, and (c)
crystallization rate ................................................................................................... 164

Figure 5.5. Size distribution of gypsum crystals formed in the absence and in the presence

of 50 mg/L silica colloids. ...................................................................................... 167
Figure 5.6. SEM image of gypsum crystals formed in the absence (a-d) and in the

presence (e-h) of 50 mg/L silica colloids ................................................................ 170
Figure SS. 1. Transmittance as a function of time at three wavelengths ......................... 179
Figure 85.2. Effect of particles on transmittance and conductivity ................................ 180
Figure 85.3. The plots for calculating the surface energy of the crystals ....................... 181

Figure 6.1. Comparison of membrane ﬂux decline in the RO membrane system during
gypsum scaling (indicated as “CaC12+ Na2804 only”), colloidal fouling (“CaC12+
NaCl + particles”), and combined fouling experiments (“CaC12+ NaZSO4 +
particles”). ............................................................................................................... 194

xiii

Figure 6.2. Permeate conductivity of the RO membrane during gypsum scaling, colloidal
fouling, and combined fouling experiments, respectively. ..................................... 196

Figure 6.3. Mass of silica particles deposited in the absence and presence of gypsum
scaling ..................................................................................................................... 199

Figure 6.4. Gypsum crystals in the absence of silica particles ....................................... 201

Figure 6.5. Gypsum crystals in the presence of silica particles at different magniﬁcations
................................................................................................................................. 201

Figure 6.6. X-ray energy dispersive spectroscopy (EDS) analysis of the fouling layer
consisting of gypsum crystals and silica particles .................................................. 202

Figure 6.7. Effect of silica particles on gypsum crystallization at different operating

conditions ................................................................................................................ 205
Figure 6.8. Normalized permeate ﬂux at three experimental conditions ........................ 206
Figure 6.9. The crystals formed in the absence of silica particles. ................................. 208
Figure 6.10. The gypsum crystals formed in the presence of silica particles. ................ 209

xiv

AF M
DAF F
DBPs
DLVO
EDS
EPS
MF
MWCO
NF
NOM
PA
PAN
PES
PTFE
PVDF
R0
SDI
TEM
TDS
TOC
TMP
SEM

W254

ABBREVIATIONS

Atomic force microscopy
Dissolved air ﬂotation and ﬁltration
Disinfection by-products
Deijaguin-Landau-Verwey-Overbeek
Energy dispersive spectroscopy
Extracellular polymeric substances
Microﬁltration

Molecular weight cutoff
Nanoﬁltration

Natural organic matter

Polyamide

Polyacrylonitrile

Polyethersulfone
Polytetraﬂuoroethylene
Polyvinylideneﬂuoride

Reverse osmosis

Silt density index

Transmission electron microscopy
Total Dissolved Solids

Total organic matter
Transmembrane pressure difference
Scanning electron microscopy
Ultraﬁltration

Ultraviolet absorbance measured at 254 nm

XV

CHAPTER ONE

COMBINED FOULING ON THE PERFORMANCE OF PRESSURE-
DRIVEN MEMBRANES TREATING WATERS OF COMPLEX
COMPOSITION: CURRENT UNDERSTANDING AND FUTURE
RESEARCH NEEDS

1.1. Introduction

Our growing demand for drinking water has led many water utilities to use alternative
water sources, such as brackish surface water and seawater, with elevated levels of
contaminants and salts [1-3]. Meanwhile, the increasingly stringent water standards, such
as Disinfectants/Disinfectants Byproducts Rule (D/DBP) and Long Term 2 Enhanced
Surface Water Treatment Rule (LT2ESWTR), require a higher removal of contaminates
[4]. Membrane ﬁltration processes are continuously gaining popularity because they can
serve as a reliable means for removing microorganisms, organics, precursors of

disinfection by-products (DBPs), colloids, and other suspended matter of concern [4-6].

1.1.1. Overview of membrane ﬁltration processes

Pressure-driven membrane ﬁltration processes, including microﬁltration (MF),
ultraﬁltration (U F), nanoﬁltration (NF), and reverse osmosis (R0), are being increasingly
used by water and wastewater treatment utilities [4, 6]. Low pressure membranes, MF
and UF are primarily used for removing turbidity, pathogens, and particles from fresh

waters [7]. They are used as stand-alone processes [8, 9] or as pretreatment for high

pressure salt—rejecting NP and R0 membranes [10, 11]. NF or R0 membranes are usually
used for softening, desalination, and removal of natural organic matter (NOM) and

precursors of DBPs [7].

The major obstacles to a wider application of membrane ﬁltration technology are
membrane fouling and disposal of concentrate. Membrane fouling occurs when the
rejected contaminants accumulate within the membrane or on the membrane surface,
resulting in a decrease in permeate ﬂux. Membrane fouling shortens membrane life,
decreases both permeate quantity and quality, and increases capital and operational costs

[6, 12].

Depending on the properties of the contaminants being rejected by a membrane,

membrane fouling can be roughly categorized into four types [7]:

(l) Particulate fouling by particles in colloidal range (colloidal fouling) or larger;
(2) Organic fouling by NOM and other organic molecules;
(3) Biological fouling from the accumulation or growth of microorganisms; and

(4) Scaling resulting from sparingly soluble salts such as carbonate and sulfate.

Understanding membrane fouling is crucial to developing low-fouling membranes,
controlling membrane fouling, and ultimately reducing operating costs. However, our
knowledge of membrane fouling is limited. For example, theoretical predictions of
permeate ﬂux underestimate the real membrane performance, the phenomenon is known

as the “ﬂux paradox” [13]. An accurate estimation of permeate ﬂux still remains elusive

F_' A—‘ D—

in part due to the failure of predicting the fouling layer structure [14]. This is mainly due
to the fact that, in practice, fouling is caused by several foulants and involving different

fouling mechanisms simultaneously.

1.1.2. Combined fouling in practice

It was reported that treated wastewater presently accounts for 5%, brackish water for 22%
and seawater for 58% of the water produced by desalination technologies, among which
the membrane ﬁltration process is predominating [15]. Since the ﬁrst high-ﬂux,
anisotropic reverse osmosis membranes were invented in the early 1960s [16], NF and

R0 membranes have played an increasingly important role for desalination.

Brackish water is deﬁned as water containing total dissolved solids (TDS) levels that are
higher than potable water, but lower than seawater (in the range of 1,000 to 25,000 mg/l
TDS) [l7]. Waters found in coastal areas, aquifers, surface waters, and groundwater can
be brackish [2, l7]. Brackish water usually has a complex composition. For example, the
Colorado River, the primary water source for the American Southwest and the feed water
for many membrane treatment plants, has a complex composition [18] and includes high

amounts of organic matter, colloids, and salts.

For seawater, in addition to a high concentration of salts, a substantial amount of
inorganic particles of a wide range of sizes is present [19, 20]. The particulate matter

includes elemental sulfur, silica, and ferric and aluminum hydroxides [19, 21-24].

Although feed waters for NF/RO membrane systems usually are pretreated and thus the
colloidal fraction is largely removed, the performance of the pretreatment systems varies
seasonally [25-27] and cannot guarantee a complete removal of components of small size
[19—21, 24, 28]. Boerlage and co-workers [26] evaluated colloidal fouling in two RO pilot
desalination plants. In these two plants, the feed waters were pretreated by a slow sand
ﬁltration and a UF membrane system, respectively, before entering the RO membrane
system. It was found that both of the pretreatment methods were effective in removing
particles larger than 0.45 pm, but not effective in removing the smaller particles. Based
on ca. 2 years of monitoring in a seawater R0 plant, Chua et al. [19] reported that
although both a conventional pretreatment system and an ultraﬁltration (UF) membrane
system were used to remove the suspended particles [19], the silt density index (SDI) of
the feed water for the RO membrane system remained above 2.5 and a substantial

quantity of suspended particles were not removed.

Examples of combined fouling observed in practice are summarized in Table 1.1.

Table 1.1. Examples for combined fouling occurring in practice

 

 

 

 

 

 

Feed water Location Membrane Pretreatment Foulants Ref.
Oise river, NF - Organics, [29]
Paris, France colloids, scale
United R0 Ozonation, Biological [30]
_ Kingdom dissolved air matter, iron,
Bmk‘Sh flotation, silica colloids
surface soﬁening, media
water ﬁltration, and
cartridge ﬁlter
Colorado RO MF Colloids and [31]
River, USA scale
Zarzis and R0 Disinfection, sand Organic matter, [32,
Djerba Island, ﬁltration, silica, iron, and 33]
Brackish Tunisia oxidation, and MF scale
groundwater Minjur aquifer, R0 Media ﬁlter, and Iron, silica [34]
Saudi Arabia cartridge ﬁlter colloids, scale
Seawater Taiwan Strait, R0 Media ﬁltration, Organic matter, [3 5]
mixed with Taiwan and cartridge ﬁlter scales, and EPS
groundwater
Umm Lujj, NF Media ﬁltration, Organic matter, [36]
Saudi Arabia and cartridge ﬁlter iron, scale,
chromium and
Seawater fungus
Eilat, Israel R0 Coagulation, UF Iron, chromium, [3 7]
(or sand and organic and
cartridge ﬁlter) biological matter
Sydney, MF Biological Colloids, organic [38]
Australia treatment matter, and scale
, _ Greensboro, RO/N F MF/UF Organic, [39]
MunIcrpal NC, USA inorganic, and
wastewater biological matter
California, R0 MF/UF NOM, colloidal [40]
USA silicates, and

scale

 

1.2. Current research on the effect of combined fouling on membrane performance

The studies published to date on studying the effect of combined fouling so far are listed
in Table 1.2. Surprisingly, although the fouling mechanisms have been extensively

examined individually, very few studies have been conducted on the combined fouling.

1.2.1. Colloidal fouling and salt concentration polarization in salt-rejecting

membrane systems

1.2.1.1. Colloidal fouling

During membrane ﬁltration, colloids can adhere to the inner pores of porous membranes
and/or form a cake on membrane surface. Colloidal fouling increases the hydraulic
resistance to the permeate ﬂow and thus decreases permeate ﬂux. The two dominating
factors determining the effect of colloidal fouling on membrane performance are 1) the
amount of colloids on the membrane surface, and 2) structure of the deposited layer

(cake).

Colloidal deposition is inﬂuenced by hydrodynamics (crossﬂow rate, transmembrane
pressure or initial permeate ﬂux), solution chemistry (ionic strength and composition, and
pH), membrane properties (surface charge, roughness, and
hydrophilicity/hydrophobicity), and colloid properties (particle size and charge). Chellam

and Wiesner [49] studied particle transport and deposition in laminar crossﬂow

 

Table 1.2. Published studies of combined fouling in membrane systems

 

 

Combined fouling Membrane References
Colloidal and organic fouling MF, UP, and NF [41, 42][43][44-46]
Colloidal fouling and scaling R0 [47]

Organic fouling and scaling NF [48]*

 

*Brackish water and synthetic water were used. In other studies, only synthetic water was used.

membrane ﬁltrations and found that the mechanism controlling lateral migration
depended on inertial liﬁ, gravity, and permeation drag forces. Belfort et al. [13]
extensively reviewed back-transport of particles away from the membrane surface due to
Brownian diffusion, shear-induced diffusion and inertial liﬁ. Higher crossﬂow rate and
lower transmembrane pressure (or initial permeate ﬂux) help to alleviate colloidal fouling
[50, 51]. A faster crossﬂow produces greater shear at the membrane surface, and thus
increases the back diffusion of particles into the bulk suspension [52]. A lower
transmembrane pressure results in a slower initial permeate ﬂow, which brings a small
amount of colloids to the membrane surface. This phenomenon leads to the concept of
“critical ﬂux”, at which no colloidal fouling occurs due to the fact that the back diffusion
and convection of the colloids balance their deposition onto the membrane surface ([53]

and the references therein).

Colloidal deposition is usually more severe when the membrane has a rougher surface
[54, 55]. Boussu et al. [56] has demonstrated that roughness played an important role
only when the colloids were relatively small. 0n the contrary, Van der Bruggen et al. [5 7]
have shown no correlation between surface roughness and colloidal deposition rate.
Surface charge and hydrophilicity/hydrophobicity have important effects on colloidal
deposition as well. Generally speaking, hydrophilic membranes with higher surface
charge are preferred to minimize the colloidal adhesion to the membrane surface [16, 56,
58, 59]. Faibish et al. [51] demonstrated that small particles resulted in a higher ﬂux
decline than large particles at the same particle loading because smaller size resulted in a

higher hydraulic resistance and thus a lower permeability. Van der Bruggen et a1. [46]

observed the same trend in MF membrane ﬁltration, which was due to the fact that the
smal 1 particles blocked the pores and reduced permeate ﬂux rate. However, in NF
membrane ﬁltration from the same study, large particles fouled membrane more severely,

which was attributed to the slower back diffusion of the large particles from the

membrane surface to the bulk solution.

1.2. l -2. The structure of colloidal deposit layers

The structure of colloidal deposition layer is known to be a complex ﬁrnction of
operational variables such as hydrodynamics, colloid size and charge, and solution
chemistry [60-65]. Solution chemistry-mediated interactions between particles within the
deposited layer determine the structure and, therefore, speciﬁc hydraulic resistance of the
deposit [66]. Ionic strength inﬂuences the surface charge and thus the structure of the
Colloidal deposit layer as well [60]. At a high ionic strength, the repulsive forces between
the particles are reduced, which is attributed to the compression of the ionic double layer
surrounding particles, as predicted by the classic Derjaguin-Landau-Verwey-Overbeek
(DLVQ) theory. As a result, the porosity of the deposit layer is usually lower and
COHOi dal fouling is more severe at higher ionic strengths. This has been conﬁrmed by
numerous studies (e.g. [50, 64, 67, 68]). Similar to the effects of ionic strength, pH
inﬂuences colloidal fouling by changing the surface charge of the particles and
membranes. The effect of pH is strongly dependent on the isoelectric point of the
paﬁlcles and membranes. The further deviation from the isoelectric point (resulted from

either increasing pH above it or decreasing pH below it) usually increases the absolute

value of the surface charge of the particles and membranes [50, 69]. Qin et a1. [70]
demonstrated that a lower pH led to less fouling in a RO membrane system for
reclamation of spent rinse water from metal plating. F aibish at el. [67] observed that pH
inﬂuenced the initial permeate ﬂux but not the stable permeate ﬂux. The higher initial
permeate ﬂux at lower pH was contributed to the apparent viscosity of the permeating

solution near the surface.

Theoretical and experimental studies indicate that colloidal cakes can be stratiﬁed and
can contain layers with dense and loose particle packing [61, 63, 64, 71]. It has been also
reported that the porosity of the colloidal fouling layer during the cross-ﬂow ﬁltration
decreased with ﬁltration time [72]. Although the cake structure was not measured
Continuously from the cross-ﬂow membrane unit, the results suggest that the structure of

the Fouling layer should be time-dependent.

It is important to note that, in practice, these effects are expected to act simultaneously.
For instance, change in pH or ionic strength affects both membrane and particle charges,
WhiCh Changes the interaction forces between membranes and colloids and also among

the colloids themselves [56].

1-2-1—3 - Coupled effect of colloidal deposition and salt concentration polarization

When a solute is rejected by a membrane, the concentration of the solute at the membrane

surface will be higher than solute concentration in the bulk of the feed solution. This

10

leads to the formation of a gradient of solute concentration toward the membrane surface

This phenomenon is known as concentration polarization.

When colloidal fouling is present, the fouling layer formed on the membrane surface
hinders the diffusion of solutes in the layer [73-75], which increases the solute
concentration at the membrane surface. The hindrance effect is strongly dependent on the

structure of the layer [74], which, as discussed above, is a complex function of ﬂow

hydrodynamics, colloid size and charge, and solution chemistry.

Rejection of NaCl by RO membranes was shown to decrease in the presence of
membrane fouling by iron hydroxide [76, 77] and by silica colloids [76, 78-80]. In a
Study of effects of nanoﬁltration membrane fouling on rejection, Shafer et al. [81]
1' eported that rejection of cations and negatively charged low molecular weight acids

depended on the charge of ferric chloride precipitate deposited on the membrane surface.

Hoek et al. [78, 79] have proposed a model that described how hindrance of the back-
diffusion of salt by a stagnant colloidal cake results in cake-enhanced concentration
POIariZation, an increase in osmotic pressure, and, consequently, lower ﬂuxes and salt
rejeCtiOns. Results of the model, which assumed the porosity of the colloidal silica layer
to be constant, agreed well with experimental data on permeate ﬂux and salt rejection. A
modified cake-enhanced concentration polarization model was later proposed by Ng and
Elimelech [82], wherein the colloidal layer porosity and cake-enhanced osmotic pressure

were assumed to be changing in time and were determined simultaneously using a

recursive algorithm.

11

Wang and Tarabara [68] proposed a different approach to study the interplay between
colloidal fouling and salt concentration polarization. The concentration of rejected salt at
the membrane surface when colloidal particles were deposited on the membrane was

deterrnined experimentally based on measured salt permeability constant. Then the

structure of the fouling layer was estimated in terms of effective porosity. This approach
allow ed for a clear identiﬁcation of individual contributions of concentration polarization
and colloidal fouling to the permeate ﬂux decline. It was found that, at a higher solution
ionic strength, the effective porosity of the deposit layer was smaller and the effect of

colloidal fouling on salt concentration polarization was more evident.

1.2.2- Combined colloidal and organic fouling

Organic matter is an important membrane foulant. The effects of hydrodynamics on
orgarri c fouling are similar to that on colloidal fouling. In addition, organic fouling is
often affected by the foulants’ physicochemical characteristics such as charge, chemical
ﬁlﬂCtiOnality, and molecular conformation ([42, 45, 83, 84], and the references therein).
Divalent cations, such as Ca2+ and Mg2+ form complexes with organics and greatly
increase organic fouling [85]. These factors lead to a selective removal of NOM by

membrane, as shown recently by a series of membrane ﬁltration experiments [86].

SCMfer et a1. [41] studied MF membrane ﬁltering surface water containing NOM and

hematite colloids. It was found that the solution chemistry (especially pH and calcium

12

{0115) played a very important role in the combined fouling. More speciﬁcally, at extreme
high or low pH the NOM was absorbed onto the surface of the hematite colloids, which
increased the stability of the colloids and thus reduced the membrane fouling. If the
colloids pre-aggregated in electrolyte solution prior to adsorption of NOM, the presence
of NOM increased the ﬂux decline. Jermann et al. [43] showed that NOM fouling and
colloidal fouling (alginate) interacted in a UF membrane ﬁltration system. The fouling
from alginate alone was mostly reversible. However, in the presence of NOM, alginate
resulted in more severe irreversible fouling. This work implied that substances with a

mirror individual inﬂuence might have a large impact in the presence of other foulants.

Chen et al. [42] used ultrasound to control the combined fouling from NOM and silica
Colloi ds in ME membrane ﬁltration. It was found that ultrasound was more effective in

enhancing permeate ﬂux under conditions of high pH, low ionic strength, and in the

absence of divalent cations.

Van der Bruggen at al. [46] demonstrated that the rejection of organic matter (benzyl
31901101 ) by the NF membrane was enhanced in the presence of the polystyrene particles,
WhiCh Was more pronounced for smaller particles than for larger particles. This effect was
attributed to the adsorption of the organic matter onto the particles. Lee et al. [44] studied
the combined fouling in NF membrane ﬁltration with Special attention given to the effect
of NC)1\v‘I—calcium complexation. It was found that the ﬂux decline in presence of the
combined fouling is less than the sum of the decline due to the individual colloidal and

NOM fouling. This effect was presumably attributed to the increased stability of the

13

colloids in the presence of NOM and the competition between the colloids and NOM for
calcium. Li and Elimelech [45] demonstrated a synergistic effect between colloidal and
organic fouling in a NF ﬁltration system. It was found that the combined fouling resulted
in a higher ﬂux decline rate than the sum of the ﬂux decline due to colloidal and organic

fouling alone. Solution chemistry and colloidal particle size was found to inﬂuence the

synergistic effect.

1.2-3- Colloidal feuling and scaling

Seal in g from sparingly soluble salts, such as carbonate and sulfate, greatly impedes the
application of R0 membranes. Scaling occurs when the solubility of the salts is exceeded.
Carbonate scaling can be controlled by decreasing the pH of the feed solution. However

Calcium sulfate dihydrate (gypsum) is insensitive to pH, and can not be controlled by
adjus ti n g pH.

The eﬁ‘ect of crossﬂow rate and system conﬁguration on gypsum scaling in membrane
ﬁltration systems has been extensively examined. Lee and Lee [87] demonstrated that in
NF and RO membrane systems the gypsum scaling rate was decreased at higher
CfOSSﬂOw rate and lower transmembrane pressure. Lee and Lueptow [88] used a rotating
R0 meIrrbrane system to reduce the gypsum scaling. Generally speaking, these studies
achieVed reduction in gypsum crystallization by decreasing the effect of concentration

polarization which resulted in lower CaSO4 concentration at membrane surface.

14

It has been found that at high temperature the induction time of gypsum crystallization
was decreased [89, 90]. The activation energy of gypsum crystallization can be estimated
based on the dependence of induction time on temperature. Cohen’s research group [91,
92] demonstrated progressive axial development of gypsum crystals along the membrane
Surf ace, consistent with the increase in concentration polarization. Also, the membrane

Surf ace coverage by gypsum increased at higher gypsum saturation index [91, 92].

Sheiklroleslami and co-workers ([93] and references therein) demonstrated that the

presence of other salts (such as CaCO3) affected the structure and strength of gypsum
crystals. Amjad and Hooley [94] showed that polyelectrolytes affected gypsum
Crystallization and the effect depended on the concentration, molecular weight, and

Composition of the polyelectrolytes.

It is surprising that there is very limited research published on the combined effect of
scalin g and colloidal fouling on the membrane performance. Most of the available
knOWledge comes from studies on the performance of heat exchangers, where the effect
0f Particles on scaling has been demonstrated to be complex [95]. Bansal et al. [96] found
that the effect of suspended particles on scaling depended on the crystallizing ability of
paﬂiCIes. Crystallizing CaSO4 particles greatly enhanced gypsum formation because they

provided extra nucleation sites. On the contrary, noncrystallizing alumina particles

reduced the gypsum formation. They believed that this effect was due to 1) particles
reduced the strength of crystals deposited and thus resulted in a higher crystal removal

rate. and 2) particles on crystal surface acted as a distorting agent. Andritsos and

15

Karabelas [97] showed that CaC03 precipitation rate was greatly enhanced by ﬁne
aragonite (CaCO3 polymorph) particles, but was not affected by silica colloids.
McGarVey and Turner [98] examined the interplay between CaCO3 precipitation and
colloidal deposition. It was found although the rate of CaCO3 precipitation was not

affected by the silt and hematite (Fe203) particles, the deposition of the particles was

emanced by CaCO3 precipitation.

Wang and Tarabara [47] studied gypsum scaling and colloidal fouling in a crossﬂow
fr\trati on system. It was found that in the presence of silica colloids, gypsum scaling
resulted in a much faster permeate ﬂux decline of the RO membrane, which was not the
Simple sum of the ﬂux decline from colloidal deposition and that from gypsum scaling.

SEM/ EDS studies demonstrated that colloids attached onto the surface of precipitated

gypsum crystals, increasing resistance of the membrane deposit and making hydraulic

cleanin g inefﬁcient.

1.2-4- Other combined fouling

Studying gypsum crystallization in a nanoﬁltration (NF) membrane system ﬁltering

agfiCUItmal drainage water, Le Gouellec and Elimelech [48] demonstrated that the
presence of NOM resulted gypsum crystals of different morphology, which possibly
effected the performance of the membrane. It was also found that a single type of

commercially available aquatic humic acid could not model the effect of the complex

NOM in natural water on gypsum crystallization.

16

1.3. The future research need

As is evident from the information presented in Tables 1.1 and 1.2, although combined
fouling always occurs, very few studies have been focused on the effect of combined
fouling on membrane performance. Systematic studies on combined fouling are needed in

order to better predict and control membrane fouling, as which occurs in practice.

Future studies can be expected to:

(l) Reveal how the individual fouling mechanisms interact, and study the effect of
the interactions on membrane performance;

(2) Provide guidelines on the optimal strategies for membrane pretreatment and
cleaning in speciﬁc combined fouling scenarios; and,

(3) Develop a membrane fouling index that more accurately predicts fouling

propensity of waters of complex compositions.

l7

Reference

10.

11.

12.

13.

D. Bursill, Drinking water treatment - understanding the processes and meeting
the challenges, Wat. Sci. Technol.: Water Supply 1 (2001) 1-7.

State of California, Findings and Recommendations, (2003).

M. A. Shannon, P. W. Bohn, M. Elimelech, J. G. Georgiadis, B. J. Mariﬂas and
A. M Mayes, Science and technology for water puriﬁcation in the coming decades,
Nature 452 (2008) 301-310.

United States Environmental Protection Agency, Report: Membrane ﬁltration
guidance manual (815-R—06-009), (2005).

AWWA Membrane Technology Research Committee, Committee report:
Membrane processes, J. Am. Water Works Assoc. 90 (1998) 91-105.

B. S. Kamik, S. H. R. Davies, K. C. Chen, D. R. Jaglowski, M. J. Baumann and S.
J. Masten, Effects of ozonation on the permeate ﬂux of nanocrystalline ceramic
membranes, Water Res. 39 (2005) 728.

J. S. Taylor and MR. Wiesner, Membranes in Water Quality And Treatment.
Fifth ed, ed. R.D. Letterman. 1999, New York: McGraw-Hill. 11.1-11.37.

S. K. Kang and K. H. Choo, Use of MP and UP membranes for reclamation of
glass industry wastewater containing colloidal clay and glass particles, J. Membr.
Sci. 223 (2003) 89.

M. D. Afonso and R. Borquez, Review of the treatment of seafood processing
wastewaters and recovery of proteins therein by membrane separation processes --
prospects of the ultraﬁltration of wastewaters from the ﬁsh meal industry,
Desalination 142 (2002) 29.

V. Bonnelye, L. Guey and J. Del Castillo, UF/MF as RO pre-treatment: the real
beneﬁt, Desalination 222 (2008) 59.

G. Pearce, Introduction to membranes: Manufacturers' comparison: part 1, Filtr.
Sep. 44 (2007) 36.

M. Barger and RP. Carnahan, Fouling prediction in reverse osmosis processes,
Desalination 83 (1991) 3-33.

G. Belfort, R. H. Davis and A. L. Zydney, The behavior of suspensions and

macromolecular solutions in crossﬂow microﬁltration, J. Membr. Sci. 96 (1995)
1-58.

18

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

MR. Wiesner and S. Chellam, The promise of membrane technology, Environ.
Sci. Technol. 33 (17), 360A (1999).

S. Lattemann and T. Hopner, Environmental impact and impact assessment of
seawater desalination, Desalination 220 (2008) 1.

R. W. Baker, Membrane Technology and Applications. 2004, England: McGraw-
Hill.

Sandia National Laboratories, Desalination and Water Puriﬁcation Technology
Roadmap, Desalination & Water Puriﬁcation Research & Development Program
Report # 95 (2003).

C. Gabelich, T. I. Yun, J. F. Green, I. H. Suffet and W. R. Chen, Evaluation of
Precipitative Fouling for Colorado River Water: Desalination Using Reverse
Osmosis, Desalination and Water Puriﬁcation Research and Development
Program Report No. 85 (2002).

K. T. Chua, M. N. A. Hawlader and A. Malek, Pretreatment of seawater: Results
of pilot trials in Singapore, Desalination 159 (2003) 225-243.

Harvey Winters, Twenty years experience in seawater reverse osmosis and how
chemicals in pretreatment affect fouling of membranes, Desalination 110 (1997)
93.

Y. R. Ning, L. T. Troyer and S. R. Tominello, Chemical control of colloidal
fouling of reverse osmosis systems, Desalination 172 (2005) 1-6.

B. G. Darton and M. F azel. A statistical review of 150 membrane autopsies. in
International Water Conference. 2001.

J. W. McCutchan and J. S. Johnson, Reverse osmosis at Coalinga, California, J.
Am. Water Works Assoc. 62 (1970) 346.

M. Luo and Z. Wang, Complex fouling and cleaning in-place of a reverse osmosis
desalination system, Desalination 141 (2001) 15-22.

J. van Agtmaal, H. Huiting, P. A. de Boks and L. L. M. J. Paping, Four years of
practical experience with an Integrated Membrane System (IMS) treating estuary
water, Desalination 205 (2007) 26.

S. F. E. Boerlage, M. Kennedy, M. P. Aniye and J. C. Schippers, Applications of

the MFI-UF to measure and predict particulate fouling in R0 systems, J. Membr.
Sci. (2003) 97-116.

19

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

A. D. Khawaji, I. K. Kutubkhanah and J. Wie, A 13.3 MGD seawater R0
desalination plant for Yanbu Industrial City, Desalination 203 (2007) 176.

Christopher J. Gabelich, Tae I. Yun, Bradley M. Coffey and I. H. Mel Suffet,
Pilot-scale testing of reverse osmosis using conventional treatment and
microﬁltration, Desalination 154 (2003) 207.

A. Plottu, B. Houssais, C. Democrate, D. Gate] and J. Cavard, Autopsies of
membranes fouled on Mery-sur-Oise pilot units: Many lessons for the behaviour
of the water treatment plant, Desalination 157 (2003) 367.

L. Y. Dudley and E. G. Darton, Membrane autopsy -- a case study, Desalination
105 (1996) 135 -141.

C. J. Gabelich, M. D. Williams, A. Rahardianto, J. C. Franklin and Y. Cohen,
High-recovery reverse osmosis desalination using intermediate chemical
demineralization, J. Membr. Sci. 301 (2007) 131.

A. Boubakri and S. Bouguecha, Diagnostic and membrane autopsy of Djerba
Island desalination station, Desalination 220 (2008) 403-411.

M. Karime, S. Bouguecha and B. Hamrouni, RO membrane autopsy of Zarzis
brackish water desalination plant, Desalination 220 (2008) 258 - 266.

F. H. Butt, F. Rahman and U. Baduruthamal, Characterization of foulants by
autopsy of RO desalination membranes, Desalination 114 (1997) 51 - 64.

H. L. Yang, C. Huang and J. R. Pan, Characteristics of RO foulants in a brackish
water desalination plant, Desalination 220 (2008) 353-358.

A. S. Al-Amoudi and A. M. Farooque, Performance restoration and autopsy of
NF membranes used in seawater pretreatment, Desalination 178 (2005) 261 - 271.

S. Belfer, J. Gilron, Y. Purinson, R. Fainshtain, N. Daltrophe, M. Priel, B. Tenzer
and A. Toma, Effect of surface modiﬁcation in preventing fouling of commercial
SWRO membranes at the Eilat seawater desalination pilot plant, Desalination 139
(2001) 169 - 176.

L. D. Nghiem and A. I. Schafer, Fouling autopsy of hollow-ﬁbre MF membranes
in wastewater reclamation, Desalination 188 (2006) 113 - 121.

J. Kim, F. A. DiGiano and R. D. Reardon, Autopsy of high-pressure membranes

to compare effectiveness of MP and UP pretreatment in water reclamation, Wat.
Research 42 (2008) 697 - 706.

20

40.

41.

42.

43.

45.

46.

47.

48.

49.

50.

51.

R. Y. Ning and T. L. Troyer, Colloidal fouling of RO membranes following
MF/U F in the reclamation of municipal wastewater, Desalination 208 (2007) 232.

A. I. Schafer, U. Schwicker, M. M. Fischer, A. G. Fame and T. D. Waite,
Microﬁltration of colloids and natural organic matter, J. Membr. Sci. 171 (2000)
151-172.

D. Chen, L. K. Weavers, H. W. Walker and J. J. Lenhart, Ultrasonic control of
ceramic membrane fouling caused by natural organic matter and silica particles, J.
Membr. Sci. 276 (2006) 135-144.

D. Jermann, W. Pronk, S. Meylan and M. Boller, Interplay of different NOM
fouling mechanisms during ultraﬁltration for drinking water production, Water
Res. 41 (2007) 1713-1722.

S. Lee, J. Cho and M. Elimelech, Combined inﬂuence of natural organic matter
(NOM) and colloidal particles on nanoﬁltration membrane fouling, J. Membr. Sci.
262 (2005) 27-41.

Q. Li and M. Elimelech, Synergistic effects in combined fouling of a loose
nanoﬁltration membrane by colloidal materials and natural organic matter, J.
Membr. Sci. 278 (2006) 72.

B. Van der Bruggen, J. H. Kim, F. A. DiGiano, J. Geens and C. Vandecasteele,
Inﬂuence of MP pretreatment on NF performance for aqueous solutions
containing particles and an organic foulant, Sep. Purif. Technol. 36 (2004) 203.

P. Wang and V.V. Tarabara. Coupled effects of scaling and colloidal deposition
on the performance of reverse osmosis membranes. in The 233rd ACS National
Meeting. March 2007. Chicago, IL.

Y. A. Le Gouellec and M. Elimelech, Calcium sulfate (gypsum) scaling in
nanoﬁltration of agricultural drainage water, J. Membr. Sci. 205 (2002) 279.

S. Chellam and M. R. Wiesner, Particle transport in clean membrane ﬁlters in
laminar ﬂow, Environ. Sci. Technol. 26 (1992) 1611-1621.

W. Richard Bowen, Anne Mongruel and Paul Williams, Prediction of the rate of
cross-ﬂow membrane ultraﬁltration: A colloidal interaction approach, Chem. Eng.
Sci. 51 (1996) 4321-4333.

S. Hong, R. S. Faibish and M. Elimelech, Kinetics of permeate ﬂux decline in

crossﬂow membrane ﬁltration of colloidal suspensions, J. Colloid Interface Sci.
196 (1997) 267.

21

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

G. Belfort, R.H. Davis and AL. Zydney, The behavior of suspensions and
macromolecular solutions in crossﬂow microﬁltration, J. Membr. Sci. 96 (1994)
1-58.

P. Bacchin, P. Aimar and R. W. Field, Critical and sustainable ﬂuxes: Theory,
experiments and applications, J. Membr. Sci. 281 (2006) 42.

E. M. Hock, S. Hong and M. Elimelech, Inﬂuence of membrane surface
properties on initial rate of colloidal fouling of reverse osmosis and nanoﬁltration
membranes, J. Membr. Sci. 188 (2001) 115.

X. Zhu and M. Elimelech, Colloidal fouling of reverse osmosis membranes:
measurements and fouling mechanisms, Environ. Sci. Technol. 31 (1997) 3654-
3662.

K. Boussu, A. Belpaire, A. Volodin, C. Van Haesendonck, P. Van der Meeren, C.
Vandecasteele and B. Van der Bruggen, Inﬂuence of membrane and colloid

characteristics on fouling of nanoﬁltration membranes, J. Membr. Sci. 289 (2007)
220.

B. Van der Bruggen, D. Segers, C. Vandecasteele, L. Bracken, A. Volodin and

C. Van Haesendonck, How a microﬁltration pretreatment affects the performance
in nanoﬁltration, Sep. Sci. Technol. 39 (2005) 1443-1459.

K. Boussu, C. Vandecasteele and B. Van der Bruggen, Relation between
membrane characteristics and performance in nanoﬁltration, J. Membr. Sci. 310
(2008)51.

A. V. R. Reddy and H. R. Patel, Chemically treated
polyethersulfone/polyacrylonitrile blend ultraﬁltration membranes for better
fouling resistance, Desalination 221 (2008) 318.

R. M. McDonogh, C. J. D. Fell and A. G F ane, Surface charge and permeability
in the ultraﬁltration of non-ﬂocculating colloids, J. Membr. Sci. 21 (1984) 285-
294.

K.-J. Hwang, H.-C. Liu and W.-M. Lu, Local properties of cake in cross-ﬂow
microﬁltration of submicron particles, J. Membr. Sci. 138 (1998) 181-192.

Y. Lee and M. M. Clark, Modeling of ﬂux decline during crossﬂow ultraﬁltration
of colloidal suspensions, J. Membr. Sci. 149 (1998) 181-202.

P. Harmant and P. Aimar, Coagulation of colloids in a boundary layer during
cross-ﬂow ﬁltration, Colloids Surf, A 138 (1998) 217-230.

22

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

V. V. Tarabara, I. Koyuncu and M. R. Wiesner, Effect of hydrodynamics and
solution chemistry on permeate ﬂux in crossﬂow ﬁltration: Direct experimental
observation of ﬁlter cake cross sections, J. Membr. Sci. 241 (2004) 65-78.

J. C. Chen, M. Elimelech and A. S. Kim, Monte Carlo simulation of colloidal
membrane ﬁltration: Model development with application to characterization of
colloid phase transition, J. Membr. Sci. 255 (2005) 291-305.

M. R. Wiesner, Morphology of particle deposits, J. Environ. Eng. ASCE 125
(1999)1124-1132.

R. S. Faibish, M. Elimelech and Y. Cohen, Effect of interparticle electrostatic
double layer interactions on permeate ﬂux decline in crossﬂow membrane

ﬁltration of colloidal suspensions: An experimental investigation, J. Colloid
Interface Sci. 204 (1998) 77-86.

F. Wang and V.V. Tarabara, Coupled effects of colloidal deposition and salt
concentration polarization on reverse osmosis membrane performance, J. Membr.
Sci. 293 (2007) 111-123.

A. E. Childress and M. Elimelech, Effect of solution chemistry on the surface

charge of polymeric reverse osmosis and nanoﬁltration membranes, J. Membr.
Sci. 119 (1996) 253.

J. J. Qin, M. H. 00 and F. S. Wong, Effects of pH and antiscalant on fouling of

RO membrane for reclamation of spent rinse water from metal plating, Sep. Purif.
Technol. 46 (2005) 46.

K. J. Hwang and C. L. Hsueh, Cross-ﬂow microﬁltration of aggregated
submicron particles, J. Membr. Sci. 201 (2002) 137-148.

H. Vyas, A. J. Mawson, R. .1. Bennett and A. D. Marshall, A new method for
estimating cake height and porosity during crossﬂow ﬁltration of particulate
suspensions, J. Membr. Sci. 176 (2000) 113-119.

L. J. Thibodeaux, Chemodynamics: Environmental Movement of Chemicals in
Air, Water, and Soil. 1979: John Wiley & Sons.

B. P. Boudreau, The diffusive tortuosity and porosity of ﬁne-grained sediments,
Geochim. Cosmochim. Acta 60 (1996) 3139-3142.

P. C. Carman, Fluid ﬂow through a granular bed, Trans. Inst. Chem. Eng. London
15(1937)150-166.

P. Lipp, R. Gimbel and F. H. Frimmel, Parameters inﬂuencing the rejection
properties of FT30 membranes, J. Membr. Sci. 95 (1994) 185-197.

23

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

89.

J. M. Jackson and D. Landolt, About the mechanism of formation of iron
hydroxide layers on reverse osmosis membranes, Desalination 15 (1973) 361-378.

E. M. Hoek, A. S. Kim and M. Elimelech, Inﬂuence of crossﬂow membrane ﬁlter

geometry on colloidal fouling of reverse osmosis membranes, Environ. Eng. Sci.
19 (2002) 357-372.

E. M. Hoek and M. Elimelech, Cake-enhanced concentration polarization: 3 new
fouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 17 (2003)
5581-5588.

S. Lee, J. Cho and M. Elimelech, Inﬂuence of colloidal fouling and feed water
recovery on salt rejection of R0 and NF membranes, Desalination 160 (2004) 1-
12.

A. I. Schafer, A. G. F ane and T. D. Waite, Fouling effects on rejection in the
membrane ﬁltration of natural waters, Desalination 131 (2000) 215-224.

H. Y. Ng and M. Elimelech, Inﬂuence of colloidal fouling on rejection of trace
organics contaminants by reverse osmosis, J. Membr. Sci. 244 (2004) 215-226.

M. Taniguchi, J. E. Kilduff and G. Belfort, Modes of natural organic matter
fouling during ultraﬁltration, Environ. Sci. Technol. 37 (2003) 1676 - 1683.

A. W. Zularisam, A. F. Ismail and R. Salim, Behaviours of natural organic matter
in membrane ﬁltration for surface water treatment - a review, Desalination 194-
231 (2006) 211.

A. I. Schafer, A. G. F ane and T. D. Waite, Nanoﬁltration of natural organic matter:
Removal, fouling and the inﬂuence of multivalent ions, Desalination 118 (1998)
109.

J. Kim, W. Shi, Y. Yuan and M. M. Benjamin, A serial ﬁltration investigation of
membrane fouling by natural organic matter, J. Membr. Sci. 294 (2007) 115.

S. Lee and CH. Lee, Effect of operating conditions on CaSO4 scale formation
mechanism in nanoﬁltration for water soﬁening, Water Res. 34 (2000) 3854.

S. Lee and RM. Lueptow, Control of scale formation in reverse osmosis by
membrane rotation, Desalination 155 (2003) 131.

Fathi Ahmi and Abdellatif Gadri, Kinetics and morphology of formed gypsum,
Desalination 166 (2004) 427.

24

90.

91.

92.

93.

94.

95.

96.

97.

98.

A. Lancia, D. Musmarra and M. Prisciandaro, Measuring induction period for
calcium sulfate dihydrate precipitation, AIChE J. 45 (1999) 390 - 397.

W.Y. Shih, A. Rahardianto, R.W. Lee and Y. Cohen, Morphometric
characterization of calcium sulfate dihydrate (gypsum) scale on reverse osmosis
membranes, J. Membr. Sci. 252 (2005) 253-263.

A. Rahardianto, W.Y. Shih, R.W. Lee and Y. Cohen, Diagnostic characterization
of gypsum scale formation and control in RO membrane desalination of brackish
water, J. Membr. Sci. 279 (2006) 655.

R. Sheikholeslami, Mixed salts -- scaling limits and propensity, Desalination 154
(2003) 117.

Z. Amjad and J. Hooley, Inﬂuence of polyelectrolytes on the crystal growth of
calcium sulfate dihydrate, J. Colloid Interface Sci. 111 (1986) 496.

D. Hasson. Progress in precipitation fouling: a review, in:Understanding Heat
Exchanger Fouling and its Mitigation, T.R. Bott et al. (Eds). 1999: Begell House,
New York.

B. Bansal, H. Muller-Steinhagen and X. D. Chen, Effect of suspended particles on
crystallization fouling in plate heat exchangers, J. of Heat Transfer, Transactions
ASME 119 (1997) 568.

N. Andritsos and A.J. Karabelas, The Inﬂuence of particulates on CaCO3 scale
formation, J. Heat Transfer 121 (1999) 225-227.

G.B. McGarvey and CW. Turner, Synergism models for heat exchanger

fouling mitigation. In: T.R. Bott et al.Understanding Heat Exchanger Fouling and
its Mitigation, Begell House, New York, (1999) 155-161.

25

CHAPTER TWO

Wang, Fulin; Tarabara, Volodymyr V. (2007). Coupled effects of colloidal deposition
and salt concentration polarization on reverse osmosis membrane performance. J. Membr.
Sci. 293 111—123

26

CHAPTER TWO

COUPLED EFFECTS OF COLLOIDAL DEPOSITION AND SALT
CONCENTRATION POLARIZATION ON REVERSE OSMOSIS
MEMBRANE PERFORMANCE

ABSTRACT

Concentration of rejected salt at the membrane surface when colloidal particles were
deposited on the membrane was determined experimentally based on measured salt
permeability constant. This approach allowed for a clear identiﬁcation of individual
contributions of concentration polarization and colloidal fouling to the permeate ﬂux
decline. The observed transient behavior of permeate ﬂux, salt rejection, mass of
deposited colloids, and effective porosity of the deposited layer unequivocally pointed to
the importance of the two-way coupling between salt concentration polarization and
colloidal deposition. After the initial increase in the effective porosity of the fractal
colloidal deposit, the porosity reached a maximum, and gradually decreased to a steady
state value that depended strongly on the solution ionic strength. Both the resistance of
the colloidal deposit and osmotic pressure, which could be enhanced in the presence of
colloids, contributed signiﬁcantly to the permeate ﬂux decline. A short term increase in
salt rejection was observed upon introduction of colloidal particles into the feed in the
absence of membrane channel spacer. This increase could be sustained over the long term
in the presence of spacers and was attributed to the depolarization of the salt boundary

layer due to the local mixing by the concentrated ﬂowing layer of colloidal particles.

27

2.1. INTRODUCTION

The shrinking water supply together with our continuously increasing demand for this
resource has stimulated a renewed interest in reverse osmosis (RO) as a water treatment
unit process. Desalination, softening, disinfection by-product control, and removal of
speciﬁc inorganic contaminants such as arsenic, barium, nitrate, nitrite and others
exemplify possible applications of R0 in water treatment. A wider use of RO
membranes, however, is impeded by a group of phenomena that lead to permeate ﬂux
decline, deterioration of permeate water quality, shortening of membrane life, and thus to
a substantial increase in operational costs [1, 2]. The relevant physicochemical
phenomena include concentration polarization, scaling, adsorptive and biological fouling,

and colloidal fouling.

Although source waters usually undergo pretreatment that is designed to remove the
colloidal fraction, in practice both dissolved and residual suspended phases are present in
R0 feed [3-5]. The origins, surface chemistry, concentration, and morphology of the
smaller colloids that persist through the pretreatment stage are likely to depend on the
water source and the type of pretreatment. Examples of colloidal matter in R0 feed
waters include elemental sulfur, silica, and ferric and aluminum hydroxides [3-7].
Because of their small size, these residual colloids can form membrane deposits with a
high hydraulic resistance contributing to the permeate ﬂux decline. Also, colloids
deposited on the membrane surface can inhibit back-diffusion of dissolved species

resulting in precipitation of sparingly soluble salts in the pores of the deposited layer and

28

in the enhanced osmotic pressure [8, 9]. If the source water pretreatment is inadequate or
fails, colloidal fouling can be especially egregious. Given that practical applications of
R0 usually involve multicomponent feeds and because RO membranes provide an
absolute barrier to most components of the feed water, potential interactions of colloids
and other rejected materials have to be taken into account in predicting membrane

performance and deciding on the optimal strategy for membrane cleaning [3, 8].

Fouling of RO membranes by iron hydroxide [IO-I3], polystyrene [l4], and silica [9, 15-
17] colloids has also been investigated in controlled laboratory studies. In all cases,
permeate ﬂux was reported to decrease with the introduction of colloids into the feed.
Another important metric of membrane performance, solute rejection, has been shown to
be inﬂuenced by the presence of colloidal layer on the membrane surface. Rejection of
NaCl by RC membranes was shown to decrease in the presence of membrane fouling by
iron hydroxide [10, 18] and by silica colloids [9, 16, 18, 19]. Recently, Hoek et al. have
proposed a model that described how hindrance of the back-diffusion of salt by a stagnant
colloidal cake results in cake-enhanced concentration polarization, an increase in osmotic
pressure, and, consequently, lower ﬂuxes and salt rejections [9, 16]. Results of the model,
which assumed the porosity of the colloidal silica layer to be constant, agreed well with
experimental data on permeate ﬂux and salt rejection. A modiﬁed cake-enhanced
concentration polarization model was later proposed by Ng and Elimelech [17], wherein
the colloidal layer porosity and cake-enhanced osmotic pressure were assumed to be

changing in time and were determined simultaneously using a recursive algorithm.

29

In experiments with other colloids or colloidal mixtures, the rejection of salt has been
reported to exhibit more complex behavior. RO membrane fouling by humic substances
alone has been shown to result in a slight increase [20] or a marginal decrease in NaCl
rejection [21]. However, when silica colloids were present in the feed, the rejection of
NaCl was shown to remain unchanged with respect to that when no fouling occurred [21].
Similarly, for suspensions containing iron oxide colloids and humic substances, the
rejection of NaCl was shown to improve with respect to NaCl rejection observed when
humic substances were the sole foulant [20]. In a study of effects of nanoﬁltration
membrane fouling on rejection, Shafer et al. [22] reported that rejection of cations and
negatively charged low molecular weight acids depended on the charge of ferric chloride
precipitate deposited on the membrane surface. In case of micro- and ultraﬁltration, the
accumulation of larger colloids on the membrane surface is also known to modify
rejection of smaller colloids and solutes. For example, rejection of dextran was shown to
decrease in the presence of bentonite deposit on the surface of rrricroﬁltration membrane
and was attributed to retention of dextran molecules in the cake layer and in membrane
pores [23]. In another study on the microﬁltration of colloidal mixtures, the retention of
50 nm gold colloids was signiﬁcantly reduced in the presence of a membrane deposit of 1
pm latex colloids [24]. Improvement of solute rejection in the presence of dynamically
formed layers (secondary or dynamic membranes) of different types of particles on
membrane surfaces has also been reported ([25] and references therein). These
observations point to a strong dependence of membrane rejection on the type of the

foulants involved and, more generally, on the properties of the fouling layer.

30

The structure of the colloidal fouling layer is known to be a complex function of
operational variables such as ﬂow hydrodynamics, colloid size and charge, and solution
chemistry [26-31]. The propensity of colloids to deposit affects the layer’s hydraulic
resistance both directly via the overall deposit thickness and indirectly via the
accumulative drag force on deposited colloids that plays a role in determining the
equilibrium distance between colloidal particles. Therefore, accounting for various
mechanisms of particle transport to membrane surface in crossﬂow channelsiis essential
for the accurate prediction of the permeate ﬂux [32]. It should be noted that the
hydrodynamics of colloids in membrane channels can be drastically modiﬁed in the
presence of spacers. It is surprising that despite the wide application of spacers in practice

[33], no information is available on their effect on the colloidal fouling of RO membranes.

Solution chemistry-mediated interactions between particles within the deposited layer is
another signiﬁcant factor that determines the structure and, therefore, speciﬁc hydraulic
resistance of the deposit [34]. Theoretical and experimental studies indicate that colloidal
cakes can be stratiﬁed and can contain layers with dense and loose particle packing [27,
29, 30, 35]. There is also evidence for the existence of the concentrated ﬂowing layer of

colloids between the stagnant cake of deposited colloids and the bulk suspension [36-39].
One objective of this study is the direct experimental determination of (I) the effect of

rejected salt on the structure of colloidal layer and (2) how the two-way coupling between

salt concentration polarization and colloidal deposition inﬂuences permeate ﬂux and salt

31

rejection. Another objective is the evaluation of the effect of spacers on RO ﬂux and

rejection in the presence of colloidal particles in the feed.

2.2. BACKGROUND

Reverse osmosis membrane performance is characterized in terms of permeate ﬂux and
solute rejection. When the feed water is free of solutes and potential foulants, the pure

water permeate ﬂux J 0 is linked to the transmembrane pressure differential AP by the

Darcy’s equation:

 

J0 = 9 (2-1)

where Rmo is the hydraulic resistance of the membrane to pure water, and ,u is the

dynamic viscosity of water. When a solute is present in the feed and there is a gradient of

solute concentration across the membrane, osmotic pressure (Arrm) must be overcome.

In the simple case of dilute 1-1 electrolyte, van’t Hoff equation can be used to calculate

Air", :

Mm = 2RT(Cm—Cp), (2.2)

where R is the universal gas constant, T is the absolute temperature, Cm is the solute

concentration at the membrane surface, and C p is the solute concentration in the

32

permeate. Due to the effect of concentration polarization, solute concentration at the
surface of solute-rejecting membranes is higher than solute concentration in the bulk of

the feed solution. The boundary layer ﬁlm model can be used to estimate C m [40]:

92.—fez“ (£]_ex (1] t 23
Cf-Cp p D P k , (-)

where C f is the solute concentration in the bulk of the feed solution, J is the volumetric

permeate ﬂux, D is the solute diffusion coefﬁcient in water, 6 is the concentration
. . . D . .
polarization boundary layer thickness, and k = 5 IS the mass transfer coeffrcrent. For

laminar ﬂow in a ﬂat rectangular channel, k can be estimated from Sherwood correlation

[41]:
1/3
Sh=-k%h—=(3.663 +1.613 °Re-Sc--quZ-] , (2.4)

where Sh is the Sherwood number, L is the channel length , d}, is the hydraulic
diameter (d), z2h) of the membrane channel and h is the channel height, Re is the

Reynolds number, and Sc is Schmidt number. Equation (2.4) holds when the condition

0.1<(Re-Sc-i:—)<104 is satisﬁed. Equations (2.2) and (2.3) can be combined to

produce an expression for Arm:

33

J
Alrm = ZRTCfRObs emf/2), (25)

where Robs is the observed solute rejection:

C
Rob, =1- C?" (2.6)

When colloidal foulants are present in the feed and are rejected by the membrane, they
may accumulate on the membrane surface and form a deposit that exerts hydraulic

resistance (Rd) to the permeate ﬂux and contributes to its decline. Also, the membrane
resistance (Rm) tends to increase measurably during the conditioning stage as the

membrane is exposed to the electrolyte solution free of foulants. As will be shown in

section 4.2, it becomes important to account for this difference between Rm and Rmo

when estimating osmotic pressures at lower ionic strengths.

When these three factors - Azrm , Rd , and (Rm —R,,,0) - are accounted for and the

membrane reﬂection coefﬁcient is assumed to be equal to unity, the expression for water

ﬂux across the membrane becomes:

__ AP— An,"
”(Rm + Rd)

 

(2.7)

34

The interaction between the deposited layer and rejected salts makes it difﬁcult to factor

out individual contributions of Arr,n and Rd to the ﬂux decline. On one hand, in the

presence of the fouling layer, correlations such as Eq. (2.4) for the mass transfer

coefﬁcient k are not applicable and, therefore, Eq. (2.5) can not be used to compute Cm
and Arm. On the other hand, Rd depends on the deposit’s mass and porosity that are

functions of the ionic strength. Thus, without experimentally measuring one of the two

unknowns (Azz'm or Rd ), it is not possible to experimentally determine separate effects

of osmotic pressure and colloidal deposition on the membrane performance.

This dilemma can be resolved by measuring the membrane salt permeability constant, B .

According to the solution-diffusion model [42], solute ﬂux J s can be related to Cm and

Cas

p

J, = B(C,,, — C p). (2.8)

The mass solute ﬂux J 5 (Eq. (2.8)) and the volumetric water ﬂux J (Eq. (2.7)) are

related via the concentration of salt transported across the membrane:

JC

Ks
ca
II

p. (29)

Equations (2.8) and (2.9) can be combined to determine B when colloidal foulants are

not present in the feed. As will be shown in the section 2.4.3, B was found to be

35

approximately constant in the experiments reported in this paper. This allowed us to use

the measured value of B to determine Air", when colloids were deposited on the

membrane surface. By combining Eqs. (2.2), (2.8), and (2.9) we get:
JCp
Air," = 2RT(Cm—Cp) = 2RT -—B— , (2.10)

where C p and J are determined experimentally. After Art", is determined from Eq.

(2.10), Eq. (2.7) can be used to compute Rd. Kozeny-Carman equation [43] for Rd can

be used then to evaluate the effective porosity e of the deposited layer:

 

1801-5 M
Rd= (23) P, (2.11)
ppdpe

where A is the membrane ﬁltration area, M p is the mass of particles deposited on
membrane surface, and dp and ,0p are the diameter and the density of the silica

particles, respectively.

36

2.3. EXPERIMENTAL METHODS

2.3.1. Experimental apparatus

The schematic of the bench-scale crossﬂow RO unit is shown in Fig. 2.1. Two identical
Sepa CF 11 high pressure membrane modules in medium/high foulant conﬁguration (GE
Osmonics, Minnetonka, MN) were connected in parallel. Positive displacement pump
(M03 Hydracell, Wanner Engineering, Minneapolis, MN) equipped with a ﬂowrate
control unit (KBMD-240D, KB Electronics Inc., Coral Springs, FL) was used to deliver
feed water to the ﬁltration cell. Pulsation dampener (Model H1020V, Blacoh Fluid
Control Inc., Riverside, CA) was installed immediately downstream from the pump outlet.
Back pressure regulator (BP-3, Circle Seal Controls Inc., Corona, CA) was used to
maintain the transmembrane pressure differential at a constant predetermined value in the
range of 150 psi (1.034 MPa) to 500 psi (3.477 MPa); the pressure was continuously

monitored by an analog pressure transducer (PX303-500G5V, Omega, Stamford, CT).

In-line digital ﬂow meters (Models 101-8 and 101-3, McMillan Co., Georgetown, TX)
were used to continuously monitor the retentate rate and permeate ﬂux. The meters were
calibrated using a digital weighing balance (ARC120, Ohaus Corp., Pine Brook, NJ)
interfaced with a computer via a built-in RS-232 port. The permeate was collected to a
vessel positioned on the balance, as indicated by thick dashed lines in Fig. 2.1, only when
determining water and salt permeability constants. In all other experiments, the permeate
was recirculated back into the feed tank. In all experiments, the retentate ﬂow rate was
kept at 4.8 L/min, which resulted in a crossﬂow velocity of 0.24 m/s and a Reynolds

number of 816.

37

The membrane channel mesh spacers (Model YCAFSOS705, GE Osmonics, Minnetonka,
MN) were used in several experiments to study the effect of spacers on permeate ﬂux and
rejection. Cross section of the mesh spacer elements was approximately rectangular with

the height of ca. 1.12 mm and the width of ca. 0.56 mm.

Flow-through conductivity probes interfaced with a conductivity meter (Orion 550,
Thermo Electron Corp., Beverly, MA) were installed in the permeate line. One
conductivity probe (Orion, 013060A, Thermo Electron Corp.) with a ﬂow-through cell
was used to measure conductivity in the range of 0.001 uS/cm to 300 uS/cm; another
conductivity probe (Model 013005A, Thermo Electron Corp.) was equipped with a
house-made glass ﬂow-through cell and was used to measure conductivity in the range of
1 uS/cm to 200 mS/cm. The conductivity cells had a hydraulic detention time that ranged
from 0.5 min to 1 min, depending on the value of the permeate ﬂux. During membrane
fouling experiments, the permeate conductivity was recorded every 2 min during the ﬁrst

4 h of the experiment and every 6 minutes afterwards.

The data from the pressure transducers, the weighing balance (when used), the

conductivity meter, and the ﬂow meters were logged to the computer via the data

acquisition module (PCI-6023E/SC-2345, National Instruments Corp., Austin, TX) using

38

_/\/_

 

 

system

LabVlew
data acquisition

digital balance

'C--.‘-----.-.---.---.-DOD-COCOODODOO-OC--------.------.0------------

J

 

l

 

..
.m .H.
IIIIIIIIIIIIIII J rm mm
....... H mm mm mm
nnnnn . . . . ..
.. o . u ..
"m mm w" n"
mumm" ﬂ. mm
"ms-mu ﬂ ~H~r
FEE mm
”.r rrrrrrrrrrr Lstr rrrrrrr . mm
m m m n"
" um " "m
m mm - Mm I
m Pu vmm "u
N am. w..." "m
u .- emu" "u
.::i.b®l man "u
mr c. m"
mm m .m
mm m m"

 

 

 

 

   

flowmeter

 

modules

 

 

I—I

 

 

 

 

ID: RO membrane

pulse
dampener

lIllI]

     

external tempeéure sensor

 

 

 

    

um

r-—--coco-no-Q-ouocoucoo-o-u-o-o-oo-uu‘-------o-u-o-o-u-u-u-u-uu—uuu

 

 

II cooling liquid line

  

 

        

Feed tank

1. Schematic of the experimental apparatus

Figure 2

39

a program written in LabView (version 7.1, National Instruments). These data were

acquired at a 30 s time interval and were smoothed using the moving average procedure.

The temperature of the feed water in the 28.4 L (7.5 US gal) high density polyethylene
feed tank (Nalgene Labware, Rochester, NY) was maintained at (20.0 i 0.5) °C using a
programmable circulating chiller (Model 9512, PolyScience, Niles, IL) with an external
temperature probe. The chiller circulated the cooling liquid through a custom-made
stainless steel (SS316) chilling element immersed in the feed tank. The cooling liquid
was a 1:1 (volumezvolume) mixture of distilled water and ethylene glycol. The
temperature of the feed suspension was recorded and permeate ﬂux was adjusted to

account for the change in viscosity with temperature.

The concentration of Silica particles in the feed tank was monitored
spectrophotometrically (Multi-Spec 1501, Shimadzu, Kyoto, Japan) by diverting a small
portion of the retentate into a ﬂow-through UV-vis sample cell (Model 178.710 - QS,
Hellma GmbH & Co KG, Milllheim) for absorption measurement before returning this

sample into the feed tank.

After each experiment, the crossﬂow apparatus was cleaned with a detergent solution and

then ﬂushed with deionized water four times.

2.3.2. Reagents and chemical solutions

40

All reagents were of ACS analytical grade or higher purity (Fisher Scientiﬁc, Pittsburgh,
PA) and were used without further puriﬁcation. The ultrapure water used in the
experiments was supplied by a commercial ultrapure water system (Lab Five, USFilter
Corp., Hazel Park, MI) equipped with a terminal 0.2 pm capsule microﬁlter (PolyCap,
Whatman P1c., Sanford, ME). The resistivity of water was greater than 16 MQ-cm. The
pH of water was monitored (pH meter D12, Horiba, Kyoto, Japan) and was in the range
of 6.4 to 6.8 when in equilibrium with atmosphere. No buffers were added to the feed

water.

2.3.3. Silica colloids and R0 membranes

Silica particles (SnowTex-XL, Nissan Chemical America Corp., Houston, TX) were
chosen as the model colloidal foulant. Zeta potential measurements for the silica colloids
at several ionic strengths were performed using the zeta potential analyzer (ZetaPALS,
Brookhaven Instrument Corp., Holtsville, NY). Before zeta potential measurements, the

pH and the ionic strength of the solution were adjusted using KOH, HCl, and KCl.

Dynarrric light scattering (Bl-MAS particle sizing module, ZetaPALS, Brookhaven
Instrument Corp., Holtsville, NY) was used to measure particle size distribution. Samples
were diluted in a 10 mM/L KCI solution to reach the recommended count rate range of 50
kcps to 500 kcps. Transmission electron microscopy (JEOL 100 CX II TEM operated at

100 kV) was used to verify light scattering data.

41

Polyamide thin-ﬁlm composite RO membrane (BW30-365, Dow-FilmTec, Minneapolis,
MN) was used in all experiments. The membrane was supplied as a ﬂat sheet and was
stored in a sealed plastic container at room temperature. Steaming potential analyzer (BI-
EKA, Brookhaven Instrument Corp., Holtsville, NY) was used to measure streaming
potential of membranes. Before each experiment, two rectangular membrane coupons
were cut out of the membrane sheet and soaked in deionized water for 24 h at 4 °C with
water exchanged after the ﬁrst 12 h of storage, as recommended by the membrane

manufacturer.

2.3.4. Experimental protocol

Table 2.1 describes the chronological sequence of experimental tasks. In what follows,

the tasks are discussed in detail.

2.3.4.1. Membrane compaction

It is well known that most RO membranes exhibit limited and largely irreversible
compaction behavior [40]. At the beginning of each experiment, deionized water was
ﬁltered through the membranes to ensure that the irreversible compaction would not
contribute to the ﬂux decline observed in ﬁltration experiments. Optimization of the
compaction procedure involved choosing both compaction transmembrane pressure
differentials and compaction time. Pressure at the compaction stage was chosen to exceed
pressure to be used in the following membrane fouling experiment. Due to some scatter
in the hydraulic resistance of clean membrane coupons and different osmotic pressures in

different experiments, the transmembrane pressure differential in ﬁltration tests was

42

adjusted to ensure the same initial permeate ﬂux in all experiments. Pressures used in
ﬁltration experiments with deionized water and 10‘2 M, and 10‘1 M NaCl aqueous
solutions as feeds were in the range of 250 psi (1.724 MP3) to 400 psi (2.758 MPa). The
deionized water had ionic strength of 610'6 M as determined from the standard curve for
the conductivity of NaCl solutions. Prior to experiments with these feed waters, the
membranes were compacted at transmembrane pressure differentials of 300 psi (2.068
MP3), 350 psi (2.413 MPa), and 450 psi (3.103 MPa), correspondingly. A ﬂux change of
less than 1% over 12 h of the continuous compaction was chosen as the quantitative
criterion of the complete compaction. It was found that 48 h were sufﬁcient to meet the
criterion in all experiments. The compaction was performed with the deionized water that

was exchanged once during the compaction.

2.3.4.2. Pure water permeate ﬂux and measurement of membrane hydraulic

resistance

After compaction, permeate ﬂux was recorded for a series of transmembrane pressure
differentials applied: 250 psi (1.723 MPa), 300 psi (2.068 MPa), 350 psi (2.413 MPa),

400 psi (2.758 MPa) and 450 psi (3.103 MPa). The hydraulic resistance of the clean

membrane (Rmo) was calculated from Eq. (2.1) using linear least squares ﬁtting.

2.3.4.3. Membrane conditioning

After membrane hydraulic resistance was determined, sodium chloride was added into

the feed tank to condition the membrane with the solution of the same ionic strength as

43

was to be used in the following ﬁltration experiment. After 12 h of conditioning, stable
salt rejection and permeate ﬂux were achieved. The quantitative criterion of the complete
conditioning was that the change in ﬂux was less than 0.5% over 2 h of the continuous
conditioning. At the end of conditioning, the transmembrane pressure differential was
adjusted to achieve the desired initial ﬂux. In our experiments, the initial ﬂux was chosen
to be in the 54 L/(mz-h) to 56 L/(mz-h) (1.50-10'5 m/s to 1.55-10'5 m/s) range for all
. ﬁltration experiments. This ensured similar initial hydrodynamic conditions for fouling
experiments with different membranes. It should also be noted that this value of the
initial permeate ﬂux is higher than values of permeate ﬂux commonly encountered in
practice (up to 30 L/(mz-h)). This relatively high value was reported in other recent
studies on colloidal fouling of RO membranes [9, 16, 17, 19, 21], where similar permeate
ﬂuxes were employed. The consequence of this choice of initial permeate ﬂux was that
during the initial stage of ﬁltration (a) larger mass of colloids was deposited onto the
membrane and b) more signiﬁcant concentration polarization occurred than one expects
for ﬂuxes employed in R0 practice. After the steady state was achieved, however, the

value of permeate ﬂux in our experiments was comparable to values observed in practice.

2.3.4.4. Membrane fouling experiments

At the end of the membrane conditioning stage, a measured amount of silica particle
stock suspension was added into the 25 liters of the feed suspension to achieve the
concentration of silica particles of 200 mg/L. The pH of the Silica suspension during the

fouling experiments was between 7.1 and 7.9 and was not adjusted. The fouling

44

experiments were stopped after 24 h of ﬁltration, which was sufﬁcient for the steady state
permeate ﬂux to be established. The membrane coupons were removed from the

membrane cell and stored at 4 °C for future microscopy analyses.

To determine the mass of silica particles deposited on the membrane, absorbance of the
suspension in the retentate line was continuously recorded at a 2 nrin time interval. The

concentration of silica particles were calculated based on the absorption of the suspension

at 220 nm using a standard curve recorded beforehand ( ,02 = 1). This wavelength was
chosen because of the increasing absorption by NaCl at wavelengths lower than ca. 210
nm and decreasing absorption of silica particles with increasing wavelength. A simple
mass-balance equation was applied to calculate the mass of particles deposited on

membrane surface (M p ).

2.4. RESULTS AND DISCUSSION

2.4.1. Characteristics of colloidal particles and the membrane

The density of colloidal silica particles was calculated gravimetrically to be 2.36 g/cm3.
Manufacturer-supplied data indicated that the silica particle diameter ranged from 45 nm
to 60 nm. Light scattering data showed that the particle size distribution was relatively
narrow (polydispersity factor of 0.074) with a mean hydrodynamic particle diameter of
(55.6 d: 1.3) nm. This was close to the value of particle diameter of (51.8 i: 2.4) nm

determined by TEM imaging. Because TEM data were less representative of the entire

45

particle population, the particle diameter of 55.6 nm was used in all calculations. Zeta
potential values were measured to be (-35.1 i 0.7) mV, (-28.7 i: 1.9) mV, (-25.4 :t 1.5)
mV, and (-20.9 :1: 2.1) mV for solution ionic strengths of 6106 M, 1-10'3 M, 1-10'2 M,
and 1:101 M, respectively. The general trend for zeta potential of silica particles to
become less negative with increasing ionic strength was attributed to the compression of

the ionic double layer surrounding particles.

The contact angle for the BW30-365 membrane with deionized water was measured to be

(72.4 :1: 2.7) °C. The average hydraulic resistance of the membrane was (11.4 :1: 0.3)-10l3

ml (95% conﬁdence interval and p2 > 0.998 for all measurements). Streaming potential

of the membrane was measured to be (-16.6 i 0.4) mV at pH 7.2 in the 1 mM KCl
electrolyte. The minimal rejection reported by the manufacturer was 99% when measured
for 2 g/L NaCl electrolyte at 225 psi (15.3 bar) transmembrane pressure differential, 25°C,

pH 8 and 15% recovery.

2.4.2. Changes in membrane resistance during conditioning. Correction factor

During the compaction stage, membrane resistance gradually increased by ca. 20%
before stabilizing after approximately 50 h of compaction (Fig. 2.2). During the
conditioning stage (also known as pretreatment stage [44]), the permeate ﬂux dropped
initially with the introduction of NaCl electrolyte into the feed. The magnitude of this
decrease could be accurately predicted based on the calculated value of the osmotic

pressure due to the introduced electrolyte. Following this initial step-wise decrease,

46

another gradual decline (ca. 3%) in permeate ﬂux was observed that lasted approximately
2 h before the ﬂux stabilized; this decline was observed consistently in all conditioning
runs. To our knowledge, the only report of such behavior of permeate ﬂux during
membrane conditioning is the communication by Marinas [45] We attribute the observed

ﬂux decrease to the gradual change in Rm as the salt sorbs to the membrane material,

which binds more water and swells as a result. It should be noted that pure water ﬂux test
can not be used to measure the increased resistance of the membrane to pure water
because exposing the membrane to pure water would reverse the electrolyte-induced
changes in the membrane structure. More experiments were conducted under strictly
laminar ﬂow and Eqs. (2.4) and (2.5) were used to determine the osmotic pressure to

factor out the ﬂux decline due to the change in Rm alone. It was found that the correction

factor a deﬁned as:

a =Rm/R0 (2.12)

was in the (1.02 to 1.06) range. The value of a =l.04 was chosen in calculations of the
true osmotic pressure. Taking the correction factor into account is especially important
for low osmotic pressures. For instance, an osmotic pressure of 20 psi (0.138 MPa) would
be overestimated to be 30 psi (0.206 MPa) at a transmembrane pressure differential of

300 psi (2.068 MPa) if the correction factor was not taken into the account.

47

Table 2.1. R0 experimental protocol

 

.UQF'JT".C,

 

 

 

 

 

 

 

 

 

Task a Measured
' values
1 Membrane compaction J 0
(duration 60 h)
2 Measurement of the hydraulic J 0 Rmo (2.1)
resistance of the membrane to pure
water

.1. 7‘. ,‘. __..‘.‘ . 7 . . .1
'.' ’1 r-v f» 1.1 Yrs-why ‘ .. .
Urqﬁ.‘ .~o,.l"1114' 1w. ’.
- ‘4 v ,1 U‘Hf' . ‘1'“ .- L k! T 9.1' . ‘v -

 

coditioning
(duration 12 h)

(2.6)

 

4 Measurement of the hydraulic J , C
resistance of the membrane to pure

Air," (2.4), (2.5)

 

water after membrane conditioning b Rm (2'7),
Rd = 0
5 Measurement of the osmotic pressure J , C p Ayrm (2.7),
and salt permeability constant R d =
Cm (2.2)

 

 

 

 

 

 

  

_ [R u_ 4303‘ -. in - ‘.',‘ r v . ,
‘1 :‘r 11}-‘R- we"! [*1 L:.§‘J$f.’l.gi‘ .:‘F'- r ‘3 . ,3 ,. e“
-' ‘ w" ibut‘i'jieﬂ - {diff-‘5“; -. ,s.;g,u.. .
' ii ‘. T 1 .T'. ‘ ‘L‘u'..*..{n"f..'f‘.‘y..'_; . . ._ ,

6 Mebran

   

 

 

 

 

 

fouling experiment I I J C p d . Robs (2.6)
(duration ca. 24 h) Aﬂm (2.10), B
measured at
step 5
Cm (2.2)
Rd (2.7)
e (2.11)

 

   
 

a . . . .
( Values maintained constants during the experiments are AP , T , and retentate ﬂowrate.
Unless mentioned otherwise, the ﬂow regime was transitional with

Re-Sc-dh/L=ll,640

this step was performed for several membranes only, for which the average correction
factor a (Eq. (2.12)) was computed. This average value was used in all other ﬁltration
experiments.

The ﬂow regime was laminar with Re = 307 and Re- Sc - d h /L = 4,380 .)

48

2.4.3. Measurement of the membrane salt permeability constant

 

I 90 __
I
I
I 85 __
.‘ Conditioning
A ‘ 80 «a
.C 0
N O.
E '0 75 ‘~
El .'e
3 Compaction ll 5 |*-—- Decline due to Arrm
‘L 65 . , . ,
60 Jw I
Decline due to change in Rm
ll- - - - 4’ Tl * t t i i

 

-60-24-2 0 2 4 6 81012
Time(h)

Figure 2.2. Flux change during conditioning, at a transmembrane pressure difference of
300 psi (2.068 MPa) and an ionic strength of 10‘2 M.

49

Table 2.2. Measurement of the membrane salt permeability constant

 

 

a __
AP J0 Cp C... A”... PF ” B 3°
(psi) (1.761.211) (mmoVL) (moi/L) (psi) - (m/s) (m/s)
251 44.70 0.2904 0.0236 16.23 2.36 1.54107
-7
299 51.50 0.2886 0.0269 18.54 2.69 1.5510 -
-7 (1.53i:0.01) ~10 7
351 60.34 0.2877 0.0319 22.04 3.19 1.52-10
-7
400 66.69 0.2967 0.0361 24.93 3.61 1.53-10

 

(Measurements were conducted at the feed ionic strength of 10'2 M. After complete
compaction and conditioning, ﬁltrations at different pressures were conducted for at least

2 h to achieve steady state. Re = 307, Sc = 621, k = 1.4210"5 m/S, d}, = 0.0068 m.

u The apparent ﬂuxes (without adjustment based on feed solution temperature), which
were used to calculate B , using Eqs. (2.3), (2.5), and (2.10).

b Polarization factor deﬁned as PF = Cm /Cf.

c The average value. The error range corresponds to 95% conﬁdence interval.)

50

A series of experiments were conducted in a laminar ﬂow to determine the membrane salt

permeability constant. Crossﬂow retentate ﬂow rate set at 1.8 L/min translated into the

Reynolds number of 307 and Re- Sc - 6% = 4,3 80 < 104 making Eq. (2.4) applicable.

After membrane compaction and conditioning were completed, the transmembrane

pressure differential was adjusted to achieve a range of different values of C m. Table 2.2

shows the membrane salt permeability constant (B) calculated using Eq. (2.10). The
results indicated that B was approximately constant at different pressures that

corresponded to different values of permeate ﬂux and C m. This allowed us to use B
values to calculate Air," from Eq. (2.10) during ﬁltration experiments when colloids

were present in the feed.

2.4.4. Permeate flux, N aCl rejection, and structure of the deposited layer at different

ionic strengths

The permeate was recirculated back into the feed tank to keep the feed salt concentration
constant and to run longer experiments so that the steady state values for permeate ﬂux,
rejection and the mass of deposited colloids could be determined. More importantly,
maintaining the constant feed composition enabled modeling of the temporal evolution of
ﬂux across the membrane in a particular location (e. g. entrance) along the membrane
module. This is in contrast to the experimental design when the permeate is discarded; in

such situation, the feed concentration increases with time and the permeate ﬂux observed

51

(a) (b)

deionized water

 
    
      
 

 

      

 

 

x 0.9 , 98 Cf: 0.1"
a 08
8 ' g as
.5 07
"i 3: “1
0.6 i 3
O 5 924 cf 0.01M
z 0.5] ;
90 ~—————4—-—r —-4——a—ﬁ
0.4 o 4 a 1216 2024
Time (h)
(C) (d)
0.04., «E12
Amos 1° cit-0.01 u
d V I
.5 3 Cf 0.1"
E ‘l'.’
0 0.01 a ‘ deionized water
3 2
oi _ .- .-—- —.~ --o.1 go . . . . 1
o 4 s 12 1e 20 24 a o 4 a 12 1e 20 24
Timeihi Time (h)
(e) (t)
0.6 '1 8 l
1 090.1141
0.5 .
E ’ A 6 1
EM :5
0.3 “c, 41
2 s
S 0'2 E 2]
E 0.1 I
o Lava—4--tamer—“4-- - o 4
o 4 8 12 1e 20 24 0 4 8 12 16 20 24
Time (h) Timth)

Figure 2.3. (a) Normalized permeate ﬂux at three ionic strengths;, (b) observed salt
rejection, Robs ; (c) salt concentration at membrane surface, Cm; ((1) mass of

deposited particles, M d ; (e) effective porosity, e , and (f) hydraulic resistance of
deposited colloidal layer, Rd.

52

(No spacers were used in these experiments.
Transmembrane pressure differentials:
- for deionized water (no salt added): AP = 276 psi (1.903 MPa)

— for 10‘2 M salt solution: AP = 290 psi (1.999 MPa)

- for 10'1 mol/L salt solution: AP = 400 psi (2.758 MPa);
Initial conditions

- for deionized water (no salt added): J = 54.70 L/(mzh);
- for 10'2 M salt solution: J = 55.78 L/(mzh), C,,, = 0.028 mol/L, C1, = 4.38-10

mol/L, Azrm = 19.4 psi (0.134 MPa), B = 2.41107 m/S;
- for 10’1 mol/L salt solution: J = 54.45 L/(mzh), Cm = 0.17 mol/L, C p =
2.07103 mol/L, Arr," = 120 psi (0.827 MPa), B = 1.94107 m/s.)

.4

53

at different times of the ﬁltration experiment corresponds to permeate ﬂux at different
times and at different locations along the membrane module, which makes ﬂux and
rejection data difﬁcult to interpret.

Figure 2.3 illustrates transient behavior of J , Robs , Cm , M 5 , and Rd during

p ,
colloidal fouling experiments with 200 mg/L silica suspension at three different ionic
strengths. There was no appreciable permeate ﬂux decline for the silica suspension in

deionized water (Fig. 2.3a). While it is clear from M p data that particles were depositing

on the membrane surface (Fig. 2.3d ), the resistance of the deposit was much smaller
than that of the membrane itself and therefore the deposition did not lead to an observable
increase in the overall resistance to the permeate flow. Because no ﬂux decline was
observed, it was not possible to derive the effective porosity of the particle deposit from
the permeate ﬂux data. To estimate the porosity of the deposited layer, another
experiment was carried out using polyamide thin-ﬁlm composite ultraﬁltration membrane
(Model YMGKSP1905, GE Osmonics). The hydraulic resistance of this ultraﬁltration
membrane was determined to be 2.84 '1013 m-l. After the full membrane compaction (3
h duration), ﬁltration of 200 mg/L silica particle suspension in deionized water was
conducted with the same initial flux of 52.6 L/(m2 h) (1.46 -10'5 m/s) as that employed in
R0 experiments. Using Eq. (2.11), the average porosity of the deposit layer was

calculated to be 0.63.

For non-zero electrolyte concentrations, the permeate ﬂux decreased signiﬁcantly before

a steady state value was reached. The steady state value of the permeate ﬂux was smaller,

54

while the time to reach the steady state was higher, for the higher ionic strength (Fig.
2.3a); the same trend for the steady state ﬂux and the time to reach steady state ﬂux as a
function of ionic strength have been reported previously [15, 30]. Based on the permeate

ﬂux (Fig. 2.3a) and salt rejection (Fig. 2.3b) data, Cm as a function of ﬁltration time was

calculated (Fig. 2.3e, Table 2.1). In the course of the experiment, C m increased to reach a

maximum (after 6 h for 10'2 M and after 8 h for 10'1 M ionic strength) and then decreased
slightly resulting in an overall increase with respect to the initial value. The initial
increase and the resulting overall increase were attributed to the hindrance effect of the

fouling layer on salt back diffusion [9]. Possible reasons for the decrease in Cm are

discussed in section 2.4.5. Effective porosity a of the particle deposit was calculated to
be 0.41 and 0.27 for ionic strengths of 0.01 M and 0.1 M, respectively. This was much
lower than that for the case of particles suspended in deionized water. The decrease in
porosity was attributed to the lower surface charge and smaller repulsive forces between
the particles with increasing ionic strength (section 2.4.1). It should be noted that
although more particles were deposited on the membrane surface at 10'2 M than at 10" M
(Fig. 2.3d) ionic strength, the deposited layer at 10'1 M had a considerably larger
hydraulic resistance (Fig. 2.31) because of the smaller effective porosity (Fig. 2.3e). Both

Aﬂ'm and Rd were larger at 10'1 M, which resulted in a larger ﬂux decline at 10'1 M (Fig.

2.3a).

Figure 2.3e provides details on the evolution of the effective porosity of the deposit. For
both electrolyte concentrations, the effective porosity increased in the ﬁrst hour and then

gradually decreased for the remainder of the experiment. To verify whether the initial

55

increase in porosity corresponded to fractal growth, the following scaling expression [46]

was applied: 2 oc N f , where I; is the mean surface height of the deposit, N is the

number of particles deposited, and f is the scaling exponent that relates to the fractal

dimension D via D =1+l/ f . N was calculated from the mass of the deposit, M p ,
while il- was estimated from M p and the porosity, 8 using Eq. (2.11). With a very good

scaling observed ( p2 > 0.996 ), the fractal dimension was calculated to be 2.1 — the value

that is characteristic of reaction-limited conditions of deposit growth. The decrease in
deposit porosity in the later stages of ﬁltration was attributed to the processes of breakup

and restructuring.

Analysis of the permeate ﬂux and rejection data at the very early stages of colloidal
fouling can help to elucidate how newly deposited particles interact with the
concentration polarization layer of salt. To record such data, a series of R0 experiments
with silica suspension of two solid fractions (200 mg/L and 500 mg/L) was conducted in
laminar ﬂow. The experimental sequence for each test was: (1) complete membrane
compaction (60 h); (2) membrane conditioning (10'2 M, 12 h); (3) introduction of silica
stock solution to achieve the desired silica particle loads, (4) colloidal fouling experiment.
Transmembrane pressure differential was maintained at 300 psi (2.068 MPa) without any
adjustments of system pressure. This ensured that the salt concentration proﬁle was well
stabilized before the introduction of particles, and that any changes in permeate ﬂux and
rejection were due to the presence of silica particles. As Fig. 2.4 demonstrates, the

introduction of colloids resulted in a short term increase in Robs . For 200 mg/L silica

56

 

     

 

 

""""

 

  
 

 

 

(a)
100 100 I’
98 98 W
S 96 :5 90 ;‘ I
3 ,i 3 Conditioning ; ,' Filtration
O ,' I .. o " l, q
a 9‘ : I m 94 '1 'l'
92 Conditioning"; ‘ Flltratlon 92 j; ' E
m i l I "l," J , . . 90 |____ _ __T_ 'l'jr_. _.
-4 .3 .2 4.11 o 1 2 3 4 .4 .3 .2 £4 0 1 2 3 4
:‘Tlmo (h) .17 Time (h)
(c) (d) :'
0
10° ' 100 '
99 " 99 ‘
3 1: Q:
a; 98 E. ooooooOoOoOOii 9°
.0 ﬂ
0 000000000000 .0 T %
n: o
97 n: ‘bqb
97 -
c°"dm°"'"9 ”hawk Conditioning Flltratlon 0%
96 r T :7. l l ‘1 96 h 7 I r I $91
450 40 -20 o 20 4o 60 .50 .40 40 o 20 4o 60
Time (Will) Time (Will)

Fi are 2.4. Effect of introduction of silica particles on initial observed salt rejection at
10' M: (a) 200 mg/L silica particles; (b) 500 mg/L silica particles. (0) and (d) show the
initial ﬁltration in detail for (a) and ((1), respectively.

57

concentration, the initial rejection increased from 97.5% to 97.9%; for 500 mg/L silica
concentration, the increase was from 98.3% to 99.2%. Similar increase in the initial
rejection during nanoﬁltration (NF) of 200 mg/L colloidal loading, 10'2 M NaCl feed
suspension was reported earlier [9]. Because the conductivity measurement interval (2
min for ﬁltration stage) was larger than the conductivity cell hydraulic detention time (0.6

min), the real initial increase in Robs was even larger than what the data suggested.

It should be noted that upon addition of the stock silica suspension (pH 9.7 i 0.1) the
particles-free feed water (pH 6.1 to 6.80, the pH of the resulting suspension was in the
range of 7.1 to 7.9. Because of the known dependence of membrane rejection on pH,
rejection of the membrane was recorded as a function of pH. It was found that an increase
in feed solution pH from 6.1 to 7.9 resulted in an increase in salt rejection of less than
0.1%. Another control experiment was conducted to ascertain that the observed increase
of rejection was due to the added colloids and not due to changes in solution chemistry.
First, the stock silica suspension was ﬁltered through a 10 kDa ultraﬁltration membrane
to get 9.7 mL solution free of particles. The 10 KDa membrane was chosen to ensure 1)
the complete removal of silica particles and b) zero rejection of all solutes preset in the
silica suspension. Second, membrane compaction and conditioning was conducted using
the same approach as that used in all R0 experiments conducted in this work. Permeate
conductivity was continuously recorded. Third, after membrane conditioning, the 9.7 mL
of the particle-free solution was added into the feed tank. During the experiment no

pressure adjustment was made to ensure that there was no measurable change in permeate

58

conductivity upon the addition of silica suspension, which conﬁrmed that the increase in

salt rejection was indeed due to the effect of added silica colloids.

2.4.5. Hypothesis of the concentrated flowing layer of silica particles and its

depolarizing effect on NaCl transport

The observed transient behavior of salt rejection and permeate ﬂux can be explained by
taking into the account the complex structure of colloidal deposits on membrane surfaces;
such deposits consist of a stagnant layer and a concentrated ﬂowing layer. At the earlier
stages of colloidal fouling, the mass of colloidal particles in the ﬂowing layer is
signiﬁcant in comparison with the mass of particles in the growing stagnant layer. Under
these conditions, mixing of the salt concentration polarization layer by the concentrated
ﬂowing layer of colloidal particles is relatively important with respect to the
enhancement of the concentration polarization due to the formation of the stagnant
colloidal layer. As the stagnant part of the deposit grows, the relative importance of

mixing becomes increasing smaller and the rejection decreases.

The analysis of transient behavior of J, Robs , Cm, and M p for 10'2 M ionic strength

case shows that after the mass of deposited particle continues to grow while the permeate
ﬂux remains stable at the steady state value. It can be concluded from these observations
that the depositing particles do not get incorporated into the stagnant deposit. Instead they
become a part of the concentrated ﬂowing layer, which results in an increase in the
mixing of the salt concentration polarization layer. This proposed scenario is

corroborated by the observed gradual decrease in Cm and the corresponding increase in

59

observed rejection aﬁer the steady state permeate ﬂux is achieved. Ng and Elimelech [17]

also reported an increased Robs after ﬂux rate had stabilized.

At 0.1 M ionic strength, the depolarizing effect of the concentrated ﬂowing layer is less

pronounced. With Cm (Fig. 2.3e) and M p (Fig. 2.3d) remaining almost constant aﬁer 8

h of the experiment, the gradual decrease in permeate ﬂux (Fig. 2.3a) appears to be due

entirely to the gradual changes in cake porosity (Fig. 2.36)

2.4.6. Evolution of relative contributions of osmotic pressure and resistance of the

colloidal deposit to the permeate ﬂux decline

To illustrate the relative importance of osmotic pressure and colloidal deposit, we adapt
the graphical approach used by Elimelech and co-workers [9, 17] wherein expression for

permeate ﬂux (Eq. (2.7)) is re-written as AP = AP," (t) + APd (t) + Azrm (t) , or:

 

1=(Apm(t)+APd(t)+Aﬂm(t)], (2.13)

APAP AP

where AP is the transmembrane pressure differential, AP," = ,ulRm is the pressure drop
due to the hydraulic resistance of the membrane, and APd = ,ule is the pressure drop

due to the hydraulic resistance of the deposited layer of colloidal particles. Equation
(2.13) can be interpreted as an energy balance that describes how the input energy is

expended in overcoming the three resistances to the ﬂow (Fig. 2.5). Dashed line in Fig.

60

2.5 corresponds to the osmotic pressure Aﬂ'm in the absence of colloidal fouling. At
steady state, the contribution of the colloidal layer resistance, APd , to the ﬂux decline was

about the same for both ionic strength and was ca. 18 %. The relative enhancement of the
osmotic pressure was also approximately the same in both cases but due to the relatively
higher importance of the osmotic pressure at the higher ionic strength, cake-enhanced
osmotic pressure contributed signiﬁcantly to the ﬂux decline only at 10'1 M. For example,
while the steady state contributions of membrane resistance and additional resistances
related as 70% to 30% in the absence of colloids, the relationship became 42% to 58%
with colloidal foulants present in the feed. At 10'2 M, except for the ﬁrst 3 h of ﬁltration,
pressure drop due to colloidal layer was the more signiﬁcant “energy sink”. Thus, in
contrast to the case of ﬁltration of colloids suspended in deionized water, colloidal
fouling in the presence of even small amount of electrolyte (such as 10'2 M) becomes
signiﬁcant. The changes in colloid layer porosity and resistance in the presence of
electrolyte have to be accounted for to predict permeate ﬂux accurately. For example,
when the cake-enhanced concentration polarization model [9, 16] was applied to our

experimental data and a constant porosity of 0.4 was assumed, the Cm value calculated

from the model was generally more than 3 times higher than those shown in Fig. 2.3c. At
the end of the experiment with 10'l M ionic strength, this would produce the osmotic
pressure higher than the operating transmembrane pressure differential of 400 psi (2.758

MPa).

2.4.7. Effect of the feed channel spacers on N aCl rejection and permeate ﬂux

61

(a) (b)

 

 

 

 

V o

5 80“ “g- 60“ Apmk

'g ‘5.

g 604 §45iAﬂmo

3 ,0, gum.

>| APd% A'I'l'mok >. APd%

U: 20- 9151

'5 pal .

C C

"4 oY T—._—_f-—A-, 11.1 of r , T A A
0 4 81216 2024 0 4 312162024

Time ('1) Time "ll

Figure 2.5. Energy consumption analysis for the system: (a) at 10'2 M; (b) at 10'1 M.
(operating conditions are listed in the caption for Fig. 2.3.)

62

To shift the balance from the formation of the (undesirable) stagnant layer of deposited
particles to the formation of the (desirable) concentrated ﬂowing layer, additional mixing
in the vicinity of membrane surface is needed. An accepted method of providing such
mixing in membrane systems is the use of membrane spacers. This goal as well as the
need to test our hypothesis in the context of industrial applications motivated the part of

this study where feed channel spacers were used.

Figure 2.6 shows the normalized ﬂux and observed salt rejection as a function of
ﬁltration time when a spacer was placed in the feed channel. Filtration of silica
suspension in deionized water was not conducted because there was no ﬂux decline
observed even when spacers were not used (Fig. 2.3a). When the spacers were used, no
ﬂux decline was observed at 10'2 M, and the ﬂux decline was much smaller at 10'1 M
compared with runs without the spacers (Fig. 2.3a). lmportantly, after the initial increase

caused by the introduction of particles into the feed, Robs remained stable at this

elevated level. This was despite a decrease of particle concentration in the feed as
detected by UV absorbance. It appears that the effect of spacers is that of “trapping” of
particles in the interﬁlament domain of feed channel spacers where particles remain
without depositing on the membrane surface. In terms of the concentrated ﬂowing layer
model, introducing a spacer inhibits the formation of the stagnant deposit so that mixing
by colloidal particles in the ﬂowing layer becomes the dominating effect of particles on

the salt concentration polarization.

63

(a)

 

 

 

 

 

1 °-
5 0'9“ c 001ME ""
f= .
E 0.8- /
g 01‘ Cf= 0.1M
N
E 0.6-
0
2 0.5-
0.4 '“_“‘—“”T—"‘“——i———~ —r— —-—~—T —~T—-——— 1
o 4 8 12 16 20 24
Time (h)
(b)
i Cf=0.01M Cf=0.1M
g 96
0
m 95 Conditioning Filtration
°‘ 1.
93 r—— “—4 . . . A
.4 o 4 812162024

Time (it)

Figure 2.6. Effect of the feed channel spacers on membrane performance

((a) Normalized permeate ﬂux; (b) Observed salt rejection. Silica particle loading: 200
mg/L; Operating transmembrane pressure differential: for 10'2 M, AP = 250 psi (1.723
MPa), and for 10'1 M, AP = 330 psi (2.274 MPa); Initial ﬂux: for 10'2 M, J = 55.7
L/(mzh) (1 5410'5 m/s) and for 10‘1 M J = 60.0 L/(mzh) (1.55.10'5 m/s).)

64

Large “ﬂuidized particles” have been successﬁllly applied in ultraﬁltration with tubular
membranes: mass transfer was shown to improve when 0.7 mm and smaller glass
particles were used as turbulence promoters in ultraﬁltration of polyethylene glycol [47]
and when 3 mm stainless steel particles were used in the separation of gelatin solutions
[48]. As our results indicate, colloid-sized particles can be used to achieve improvements
in ﬂux and rejection for salt rejecting membranes as well. This opens new, perhaps
counter-intuitive, possibilities for controlling the salt rejection and permeate ﬂux by using
spacers in conjunction with introducing particles with low deposition propensity into the

feed.

2.5. CONCLUSIONS

Individual contributions of osmotic pressure and colloidal fouling to the RO permeate
ﬂux decline can be identiﬁed by determining salt permeability constant and measuring
salt transport across the membrane. By using this approach, we demonstrate that the two-
way coupling between salt concentration polarization and colloidal deposition is essential
in determining both permeate ﬂux and rejection. On one hand, porosity of the deposit and
its resistance to the permeate ﬂux are measured to be strong functions of the solution
ionic strength. The initial growth of the colloidal deposit on the surface of the salt-
rejecting membrane is found to be fractal with the fractal dimension of 2.1 that
corresponds to reaction-limited aggregation. After the breakup and restructuring of the
deposit, its resistance increases dramatically and contributes signiﬁcantly to the overall

ﬂux decline. On the other hand, deposition of colloidal particles on the membrane surface

65

inﬂuences salt transport to and across the membrane in a complex way. Formation of the
stagnant colloidal deposit results in the hindrance of the back diffusion of salt away from
the membrane surface. Under certain conditions, however, the presence of colloidal
particles improves membrane performance. Speciﬁcally, the introduction of colloidal
particles into the feed results in a short term increase in salt rejection; this increase can be
sustained over the long term when feed channel spacers are used. Based on the above
observations, we hypothesize that the concentrated ﬂowing layer of colloidal particles is
responsible for the local mixing of the concentration polarization layer of salt and that the
overall effect of colloidal deposition on permeate ﬂux and rejection is determined by the
balance between the stagnant and ﬂowing parts of the colloidal deposit. These ﬁndings
point to the potential of using particles with low deposition propensity as “mobile
mixers” to complement feed channel spacers as means of improving performance of salt-

rejecting membranes.

ACKNOWLEDGEMENTS

Financial support of this work by the National Water Research Institute (project no. 05-
TM-007) is gratefully acknowledged. We also thank Dow-FilmTec for providing us with
membrane samples and Nissan Chemical America Corp. for supplying the silica

suspension.

66

NOMENCLATURE

(WOW:
5

h

membrane ﬁltration area (m2)

solute permeability constant (ms-1)

solute concentration at membrane surface (molL'l)

solute concentration in permeate (molL'l)

feed solute concentration (molL'l)

solute diffusion coefﬁcient in water (mzs'l)

hydraulic diameter (m)

particle diameter (m)

channel height (m)

volumetric permeate ﬂux (ms'l)

volumetric pure water permeate ﬂux (ms'l)

salt ﬂux (molm'zh'l)

channel length (m)

mass of the particles deposited on membrane surface (kg)
transmembrane pressure differential (MPa)

pressure drop due to resistance of the membrane (MPa)
pressure drop due to resistance of the deposited colloidal layer (MPa)
universal gas constant (J K'lmol'l)

resistance of the deposited colloidal layer (m'l)

initial hydraulic resistance of the membrane to pure water (m'l)
hydraulic resistance of the membrane to pure water after membrane
conditioning (m'l)

observed solute rejection

absolute temperature (K)

correction factor

osmotic pressure across the membrane (MPa)

dynamic viscosity of water (kgm'ls'l)

concentration polarization boundary layer thickness (m)

effective porosity of the deposited colloidal layer

particle density (kgm'3)

correlation coefﬁcient

67

REFERENCE

10.

11.

12.

13.

14.

M. Barger and RP. Carnahan, Fouling prediction in reverse osmosis processes,
Desalination 83 (1991) 3-33.

Committee Report: Recent advances and research needs in membrane fouling, J.
Am. Water Works Assoc. 97 (2005) 79-89.

M. Luo and Z. Wang, Complex fouling and cleaning in-place of a reverse osmosis
desalination system, Desalination 141 (2001) 15-22.

K. T. Chua, M. N. A. Hawlader and A. Malek, Pretreatment of seawater: Results
of pilot trials in Singapore, Desalination 159 (2003) 225-243.

Y. R. Ning, L. T. Troyer and S. R. Tominello, Chemical control of colloidal
fouling of reverse osmosis systems, Desalination 172 (2005) 1-6.

B. G. Darton and M. Fazel. A statistical review of 150 membrane autopsies. in
International Water Conference. 2001.

J. W. McCutchan and J. S. Johnson, Reverse osmosis at Coalinga, California, J.
Am. Water Works Assoc. 62 (1970) 346.

S. 1. Graham, R. L. Reitz and C. E. Hickman, Improving reverse osmosis
performance through periodic cleaning, Desalination 74 (1989) 113-124.

E. M. Hoek and M. Elimelech, Cake-enhanced concentration polarization: a new
fouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 17 (2003)
5581-5588.

J. M. Jackson and D. Landolt, About the mechanism of formation of iron
hydroxide layers on reverse osmosis membranes, Desalination 15 (1973) 361-378.
E. Staude and W. Assenmacher, A contribution to fouling in hyperﬁltration
induced by chemical reaction, Desalination 49 (1984) 215-228.

R. D. Cohen and R. F. Probstein, Colloidal fouling of reverse osmosis membranes,

J. Colloid Interface Sci. 114 (1986) 194-207.

S. G. Yiantsios and A. J. Karabelas, The effect of colloid stability on membrane
fouling, Desalination 118 (1998) 143-152.

J. C. Schippers, J. H. Hanemaayer, C. A. Smolders and A. Kostense, Predicting
ﬂux decline of reverse osmosis membranes, Desalination 38 (1981) 339-348.

68

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

X. Zhu and M. Elimelech, Colloidal fouling of reverse osmosis membranes:
measurements and fouling mechanisms, Environ. Sci. Technol. 31 (1997) 3654-
3662.

E. M. Hoek, A. S. Kim and M. Elimelech, Inﬂuence of crossﬂow membrane ﬁlter

geometry on colloidal fouling of reverse osmosis membranes, Environ. Eng. Sci.
19 (2002) 357-372.

H. Y. Ng and M. Elimelech, Inﬂuence of colloidal fouling on rejection of trace
organics contaminants by reverse osmosis, J. Membr. Sci. 244 (2004) 215-226.

P. Lipp, R. Gimbel and F. H. Frimmel, Parameters inﬂuencing the rejection
properties of FT30 membranes, J. Membr. Sci. 95 (1994) 185-197.

S. Lee, J. Cho and M. Elimelech, Inﬂuence of colloidal fouling and feed water
recovery on salt rejection of R0 and NF membranes, Desalination 160 (2004) 1-
12.

C. W. van Oers, M. A. G. Vorstman and P. J. A. M. Kerkhof, Solute rejection in
the presence of a deposited layer during ultraﬁltration, J. Membr. Sci. 107 (1995)
173-192.

S. Lee, J. Cho and M. Elimelech, Combined inﬂuence of natural organic matter
(N OM) and colloidal particles on nanoﬁltration membrane fouling, J. Membr. Sci.
262 (2005) 27-41.

A. I. Schafer, A. G. Fane and T. D. Waite, Fouling effects on rejection in the
membrane ﬁltration of natural waters, Desalination 131 (2000) 215-224.

R. Jiraratananon, D. Uttapap and P. Sampranpiboon, Crossﬂow microﬁltration of
a colloidal suspension with the presence of macromolecules, J. Membr. Sci. 140
(1998) 57-66.

S. S. Madaeni and A. G. F ane, Microﬁltration of very dilute colloidal mixtures, J.
Membr. Sci. 113 (1996) 301-312.

V. T. Kuberkar and R. H. Davis, Modeling of fouling reduction by secondary
membranes, J. Membr. Sci. 168 (2000) 243-258.

R. M. McDonogh, C. J. D. Fell and A. G Fane, Surface charge and permeability
in the ultraﬁltration of non-ﬂocculating colloids, J. Membr. Sci. 21 (1984) 285-
294.

K.-J. Hwang, H.-C. Liu and W.-M. Lu, Local properties of cake in cross-ﬂow
microﬁltration of submicron particles, J. Membr. Sci. 138 (1998) 181-192.

69

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

Y. Lee and M. M. Clark, Modeling of ﬂux decline during crossﬂow ultraﬁltration
of colloidal suspensions, J. Membr. Sci. 149 (1998) 181-202.

P. Harmant and P. Aimar, Coagulation of colloids in a boundary layer during
cross-ﬂow ﬁltration, Colloids Surf, A 138 (1998) 217-230.

V. V. Tarabara, I. Koyuncu and M. R. Wiesner, Effect of hydrodynamics and
solution chemistry on permeate ﬂux in crossﬂow ﬁltration: Direct experimental
observation of ﬁlter cake cross sections, J. Membr. Sci. 241 (2004) 65-78.

J. C. Chen, M. Elimelech and A. S. Kim, Monte Carlo simulation of colloidal
membrane ﬁltration: Model development with application to characterization of
colloid phase transition, J. Membr. Sci. 255 (2005) 291-305.

S. Sethi and M. R. Wiesner, Modeling of transient permeate ﬂux in cross-ﬂow
membrane ﬁltration incorporating multiple particle transport mechanisms, J.
Membr. Sci. 136 (1997) 191-205.

J. Schwinge, P. R. Neal, D. E. Wiley, D. F. Fletcher and A. G. Fane, Spiral wound
modules and spacers. Review and analysis, J. Membr. Sci. 242 (2004) 129-153.

M. R. Wiesner, Morphology of particle deposits, J. Environ. Eng. ASCE 125
(1999)1124-1132.

K. J. Hwang and C. L. Hsueh, Cross-ﬂow microﬁltration of aggregated
submicron particles, J. Membr. Sci. 201 (2002) 137-148.

E. F. Leonard and C. S. Vassilief, The deposition of rejected matter in membrane
separation processes, Chem. Eng. Comm. 30 (1984) 209-217.

R. H. Davis and S. A. Birdsell, Hydrodynamic model and experiments for cross-
ﬂow microﬁltration, Chem. Eng. Sci. 49 (1987) 3203-3209.

D. Airey, S. Yao, J. Wu, V. Chen, A. G. Fane and J. M. Pope, An investigation of
concentration polarization phenomena in membrane ﬁltration of colloidal silica
suspensions by NMR micro-imaging, J. Membr. Sci. 145 (1998) 145-158.

V. Chen, A. G. Fane, S. Madaeni and I. G. Wenten, Particle deposition during
membrane ﬁltration of colloids: transition between concentration polarization and
cake formation, J. Membr. Sci. 125 (1997) 109-122.

M. Mulder, Basic Principles of Membrane Technology. 2nd ed. 1996, Dordrecht:
Kluwer Academic Publishers.

R. Rautenbach and R. Albrecht, Membrane Processes. 1989, New York: John
Wiley & Sons Ltd.

70

42.

43.

45.

46.

47.

48.

H. K. Lonsdale, U. Merten and R. L. Riley, Transport properties of cellulose
acetate osmotic membranes, J. Appl. Polym. Sci. 9 (1965) 1341-1362.

P. C. Carman, Fluid ﬂow through a granular bed, Trans. Inst. Chem. Eng. London
15 (1937) 150-166.

W. J. Koros, Y. H. Ma and T. Shimidzu, Terminology for membranes and
membrane processes, Pure & Appl. Chem. 68 (1996) 1479-1489.

B. J. Marinas, Effects of solute concentration and membrane compaction on the
performance of a fully aromatic polyamide reverse osmosis membrane, in
National meeting - American Chemical Society, Division of Environmental
Chemistry. 1996. p. 125-126.

S. Veerapaneni and M. R. Wiesner, Deposit morphology and headloss
development in packed bed ﬁlters, Environ. Sci. Technol. 31 (1997) 2738-2744.

M. J. van der Waal, P. M. van der Velden, J. Koning, C. A. Smolders and W. P.
M. van Swaay, Use of ﬂuidized beds as turbulence promoters in tubular
membrane systems, Desalination 22 (1977) 465-483.

G. M. Rios, H. Rakatotoarisoa and B. Tarodo de la Fuente, Basic transport

mechanics of ultraﬁltration in the presence of ﬂuidized particles, J. Membr. Sci.
34 (1987) 331-343.

71

S.2. SUPPLEMENTARY DATA

8.2.1. Determination of the salt permeability

As demonstrated in Section 2.4.3, A series of experiments were conducted in a laminar
ﬂow to determine the membrane salt permeability constant with varied feed
concentration. Table 82.1 shows the membrane salt permeability constant (B) calculated
using Eq. (2.10). The results indicated that B was approximately constant at different

pressures that corresponded to different values of permeate ﬂux and C m .

Table 82.1. Measurement of the membrane salt permeability constant

 

 

 

 

 

 

 

Fee
concentrgtion AP J0 Cp Cm A” m PF B
(Cf ) (psi) (1&1211» (mmol/L) (mol/L) (psi) - (m/s)

251 44.70 0.2904 0.0236 16.23 2.36 1.54107

0.001 M 299 51.50 0.2886 0.0269 18.54 2.69 155-10':
351 60.34 0.2877 0.0319 22.04 3.19 1.5210
400 66.69 0.2967 0.0361 24.93 3.61 1.53107
250 38.76 0.5721 0.9809 38.97 2.11 0978-107

0003 M 302 45.66 0.5514 0.9816 51.50 2.42 0969-10“:
362 53.96 0.5730 0.9809 53.64 2.84 1013-10
400 58.95 0.5505 0.9817 67.78 3.13 0963-107
250 34.10 0.8871 0.0966 66.71 1.93 0.876107

0005 M 304 41.99 0.8295 0.1126 77.93 2.25 0863-10“:
353 47.43 0.8223 0.1252 86.69 2.50 0869-10
402 53.69 0.7908 0.1415 98.05 2.83 0836-107
250 31.46 1.0689 0.1286 88.9 1.84 0.73010“7

0007 M 310 38.81 0.9789 0.1484 102.8 2.12 0714-10“:
352 45.22 0.9807 0.1681 116.5 2.40 0735-10
399 49.57 0.9987 0.1830 126.8 2.61 0754-107
247 27.15 1.3929 0.1521 105.1 1.69 0.695-10’C

0.009 M 302 32.82 1.3227 0.1698 117.5 1.89 0714-10:
350 38.45 1.3659 0.1894 131.1 2.10 0774-10
400 46.60 1.2966 0.2220 153.9 2.47 0758-107
250 26.82 1.9509 0.1845 127.3 1.68 0794-107

0.0“ M 300 32.81 1.7709 0.2074 143.3 1.89 0783-10:
354 40.08 1.7799 0.2388 165.2 2.17 0.834-10
401 45.65 1.7259 0.2662 184.3 2.42 0.825107

 

72

8.2.2. Imaging the fouling layer using transmission electron microscopy (TEM)

Fig. 82.1 shows a cross-section image of the deposit of the silica particles formed at 0.1
M using transmission electron microscopy (TEM). The cross-section was imaged using
Joel 6400V scanning electron microscope (Japan Electron Optics Laboratories, Japan)
with a ﬁeld emission (Oxford EDS, Oxford Instruments, UK) in backseattered electron
(BSE) imaging mode with an accelerating voltage of 20 kV. We used following
procedures to prepare the cross-section samples of fouled membranes for imaging.
Samples were inﬁltrated with resin of Poly/Bed Mixture (Polysciences, Warrington, PA)
in gelatine capsules in agitation for 24 h. Samples were polymerized at 60 °C for 24 h and
were cross-sectioned with a thickness of 80-100 nm using an ultramicrotome (Model:

MTX, RMC, Boekeler Instruments, Tucson, AZ) equipped with a diamond knife.

 

Fig. 82.1. TEM imaging of the deposit layer formed at 0.1 M

73

CHAPTER THREE

Wang, Fulin; Tarabara, Volodymyr V. (2008). On the identiﬁcation of pore blocking
mechanisms during early stages of membrane fouling by colloids. (Submitted for
publication in Journal of Colloid and Interface Science)

74

CHAPTER THREE

ON THE IDENTIFICATION OF PORE BLOCKING MECHANISMS
DURING EARLY STAGES OF MEMBRANE FOULING BY
COLLOIDS

ABSTRACT

A method based on a simple linear regression ﬁtting was proposed and used to determine
the type, the chronological sequence, and the relative importance of individual fouling
mechanisms in the experiments on the dead-end ﬁltration of colloidal suspensions with
membranes ranging from loose ultraﬁltration (U F) to non-porous reverse osmosis (RO).
For all membranes, ﬂux decline was consistent with one or more pore blocking
mechanisms during the earlier stages and with the cake ﬁltration mechanism during the
later stages of ﬁltration. For ultraﬁltration membranes, pore blocking was identiﬁed as
the largest contributor to the observed ﬂux decline. The chronological sequence of
blocking mechanisms was interpreted to depend on the size distribution and surface
density of membrane pores. For salt-rejecting membranes, the ﬂux decline during the
earlier stages of ﬁltration was attributed to either intermediate blocking of relatively more
permeable areas of the membrane skin, or to the cake ﬁltration in its early transient stages,
or a combination of these two mechanisms. The ﬁndings emphasize the practical
importance of the clear identiﬁcation of the mechanisms of colloid-membrane
interactions as determining the potential for the irreversible fouling and the efficiency of

membrane cleaning.

75

3.1. INTRODUCTION

Blocking ﬁltration laws describe the following four mechanisms of membrane fouling by
colloidal particles (Fig. 3.1): complete pore blocking (also called pore sealing), standard
pore blocking (also called pore constriction), intermediate pore blocking, and cake
ﬁltration. For the complete blocking, it is assumed that each particle reaching the
membrane blocks a pore without superimposing over other particles. For the standard
blocking, it is assumed that particles deposit within pores and the pore volume decreases
proportionally to the volume of deposited particles. For the cake ﬁltration, it is assumed
that depositing particles do not block pores either because membrane is dense and there
are no pores to block or because the pores are already covered by other particles and
therefore are not available to block. For the intermediate blocking, it is assumed that
some particles deposit on other particles (as in cake ﬁltration) while other particles block

membrane pores (as in complete blocking).

Blocking ﬁltration laws were ﬁrst proposed by Hermans and Bredée [1] and subsequently
developed by Gonsalves [2]. In an experimental study with several membranes, Grace [3]
determined that standard blocking occurred with each microﬁlter examined. Hermia
offered a uniﬁed analytical description of all four blocking mechanisms and extended

their application to power-law non-Newtonian ﬂuids [4].

Although initially developed for a membrane with parallel cylindrical pores, blocking

laws have been shown to have predictive power also in applications involving polymeric

76

 

 

 

 

ill .

Figure 3.1. Schematic illustration of the four fouling mechanisms: (a) complete blocking,
(b) standard blocking, (c) intermediate blocking, and ((1) cake ﬁltration.

) (C) (d)

77

membranes with irregular pore morphology [5-7]. The blocking laws have been used for
studying fouling of porous membranes ﬁltering suspensions of microorganisms [8],
proteins [9-11], natural organic matter (NOM) [12], as well as natural water [6] and
wastewater [13]. Blocking laws have been used to interpret permeate ﬂux behavior for
non-porous membranes as well [14, 15]. However, to our knowledge there were no
studies with salt-rejecting (NF, RO) membranes where the effect of the change in osmotic
pressure during ﬁltration was duly considered and the separation was carried out in the
dead-end geometry. It should be noted in this regard that the theory of blocking laws
assumes no back-transport of the colloids from the vicinity of the membrane surface to
the bulk of the feed suspension; this assumption narrows the applicability of blocking

laws to unstirred dead-end ﬁltration [4], [16].

For membrane ﬁltration carried out in a constant transmembrane pressure mode with the
feed ﬂow normal to the membrane surface and with foulants of spherical shape that are
completely retained, equations describing the relationship between the total ﬁltered
volume, V , and ﬁltration time, t, for each of the four fouling mechanisms are as follows

[1,4]:

 

K bV = Q0 (1 _ e-be) (complete blocking), (3.1)
K St _ t 1 (standard blocking), (3,2)
2 H V Q0

78

Ki V = ln(l + K tht) (intermediate blocking), (3-3)

KCV = 3t _ _2_ (cake ﬁltratlon), (3.4)

VQ0

where Qo is the initial ﬂow rate, and K is the constant with the subscript indicating the

blocking mechanism. The ﬁrst three mechanisms are categorized as pore blocking.
Hermia [4] demonstrated that a common characteristic equation for different fouling

mechanisms can be derived by differentiating Eqs. (3.1-3.4):

2 n
_d___t_ = k .‘1’. , (3,5)
dyz dV

where k is a constant and n is the blocking index equal to 2, 1.5, 1, or 0 for complete
blocking, standard blocking, intermediate blocking, and cake ﬁltration, respectively.

2
Equation (3.5) has been used to identify blocking mechanisms by plotting —d-—;— against
dV

it— and determining the value of the blocking index n [9, ll, 12, 17-19]. The method is

d V

advantageous in that based on a single plot one can, in principle, i) identify all blocking
mechanisms and ii) determine whether the data can be interpreted in terms of blocking
laws; for example, in several studies the value of the blocking index n for later stages of
ﬁltration was reported to be negative [9, l 1, 12, 17, 18] indicating the failure of blocking

laws to explain the ﬂux data. A signiﬁcant disadvantage of the direct application of Eq.

79

(3.5) is the high sensitivity of the value of n with respect to the noise in the ﬂux data and,

relatedly, to the choice of the time interval used in computing derivatives in Eq. (3.5).

An alternative approach to identifying mechanisms of colloidal fouling is the application
of the blocking laws in their integrated forms, Eqs. (3.1-3.4). Grace [3] identiﬁed

standard blocking (Eq. (2)) by observing a linear dependence when the experimental data
were plotted as t versus ~19; for the early stages of ﬁltration, the data did not fall on a

straight line implying that fouling mechanism(s) other than standard blocking led to the

initial ﬂux decline. Similarly, cake ﬁltration (Eq. (3.4)) was identiﬁed as the dominant
fouling mechanism by observing the linear relationship between V and Tt; [19-21]; in

these studies, the data corresponding to the earlier stages of ﬁltration did not follow the
linear dependence, revealing that some form of pore blocking occurred prior to the start
of the cake ﬁltration regime. The linearity of the integrated expressions for standard
blocking (Eq. (3.2)) and cake ﬁltration (Eq. (3.4)) makes their use straightforward and
facilitates interpretation of the results. Expressions for complete blocking and
intermediate blocking given by Eqs. (3.1) and (3.3), respectively, are not linear but linear

expressions can be derived by differentiating experimental V(t) data to obtain permeate

ﬂux, Q:
Q = Q0 - KbV (complete blocking) (3.6)
1 1 . . .
— = — + K1" (1ntermedlate blocking) (3.7)
Q Q0

80

The main advantage of the use of linear expressions (Eqs. (3.2), (3.4), (3.6), and (3.7)) is
that it does not require additional data processing; instead, it involves only a
straightforward linear least squares ﬁt, with an easily determinable correlation coefﬁcient,

and allows for a facile identiﬁcation of individual fouling mechanisms.

A clear identiﬁcation of and differentiation between the blocking mechanisms has three
important practical implications in that the knowledge of fouling mechanisms can guide
decisions on the optimal choice of the membrane and membrane cleaning strategy [22]. It
should be noted that although colloidal deposition during the early stages of ﬁltration is
the most likely cause of irreversible fouling [22], little is known about the early stages of
colloidal fouling of porous (with irregular pore morphology) and, especially, non-porous
membranes. The need in such information motivated this study. In this work, we
conducted unstirred dead-end ﬁltration experiments with membranes spanning a range of
molecular weight cut-offs. We then ﬁt Eqs. (3.2), (3.4), (3.6), and (3.7) to the
experimental ﬂux data using the linear least square ﬁtting to identify i) the type, ii) the

chronological sequence, and iii) the relative importance of individual fouling mechanisms.

3.2. Materials and methods

3.2.1. Membranes and colloids

Table 3.1 lists the membranes used in this study. Prior to ﬁltration experiments, each
membrane was soaked in deionized water for 24 h at 4 °C with water exchanged after the

ﬁrst 12 h of storage. Silica colloids (SnowTex-XL, Nissan Chemical America Corp.,

81

Houston, TX) were used as the model colloids. Using dynamic light scattering
(ZetaPALS, Brookhaven Instrument Corp., Holtsville, NY) the mean hydrodynamic

diameter of colloids was determined to be (55.6 i 1.3) nm.

3.2.2. Filtration experiments

Dead-end ﬁltration experiments were conducted at a constant transmembrane pressure
using a stainless steel ﬁltration cell (HP4750, Sterlitech Corp., Kent, WA) without
stirring. The pressure was delivered from a tank of compressed ultrapure zero grade air.
The permeate ﬂux was recorded continuously using an electronic mass balance (AV8100,
Ohaus Corp., Pine Brook, NJ) interfaced with a computer. For each experiment, 1 L of
deionized water was ﬁltered to compact the membrane ﬁrst; then, clean water permeate
ﬂux was recorded for a series of transmembrane pressure values to determine the
hydraulic resistance of the compacted membrane; ﬁnally, a 200 mL of 200 mg(SiOz)/L
aqueous suspension of silica colloids was poured into the cell and the membrane fouling
experiment was started. Permeate ﬂux was calculated based on the mass of permeate
collected on the balance. The transmembrane pressure was adjusted to bring the initial
ﬂuxes in experiments with different membranes as close one to another as possible;
however, because of the very wide difference in permeability of different membranes, it
was not possible to set the initial ﬂux at the same value. The initial ﬂuxes were 23.4 urn/s,
26.9 urn/s, 108.9 urn/s, 133.7 urn/s, and 189.3 um/s for BW30-365, NF90, UF-30kDa, UF-

lOOkDa, and UF-300kDa membranes, respectively.

82

Table 3.1. Membranes used in this study

 

 

Membrane Membrane Nominal Hydraulic Manufacturer Model number
type material MWCO resistarlce ,
RO PA‘ 100 to 200 Da 1.1-10l4 Dow-FilmTech BW30-365
NF PA 200 to 400 Da 3.8-10l3 Dow-FilmTech NF90
UF PVDFb 30 kDa 1.8-1012 GE Osmonics YMJWSP3001
UF PANc 100 kDa 121012 GE Osmonics YMMWSP1905
UF PESd 300 kDa 4.8-10” Millipore Corp. PBMK

 

a Polyamide; b Polyvinylidene ﬂuoride; c Polyacrylonitrile; d Polyethersulfone

83

3.2.3. Scanning electron microscopy (SEM) analysis

Small coupons were cut from the membranes and were soaked in deionized water at 4°C
overnight. After the coupons were dried in air, they were ﬁrst coated for 20 s using a pure
osmium coater (N eoc-AN, Meiwa Shoji Co. Ltd., Kyoto, Japan), and then imaged using
Hitachi S-47OOII Field Emission scanning electron microscope (Hitachi High
Technologies America, Inc. Pleasanton, CA) with a secondary electron detector. The
microscope was operated in ultra-high resolution mode with an accelerating voltage of 15

kV and an emission current of 10 11A.

3.2.4. Data ﬁtting procedure

The following procedure was adapted to assign portions of the permeate ﬂux dataset to a
particular fouling mechanism: i) a portion (“core”) of the entire ﬁltration dataset that ﬁt
one of the linear Eqs. (3.2), (3.4), (3.6), and (3.7) with a coefﬁcient of determination
higher than 0.99 was identiﬁed; ii) the “core” was expanded data point by data point by
evaluating if a data point immediately adjacent to the “core” belonged to the 95%
conﬁdence interval computed for this data point based on the linear ﬁt for the “core”
using a t-test [23]. If the data point was within the interval, the data point was included
into the “core” and a new ﬁtting line was determined. Then, the t-test was repeated for
the next adjacent data point. The process was stopped when the next data point was found
to be located outside of the interval. Each ﬁltration experiment was repeated once and

very reproducible ﬁtting results were obtained.

84

3.2.5. Determination of the effective porosity of the colloidal deposit

The hydraulic resistance, Rd , of a layer of spherical incompressible uniform particles

can be estimated using Kozeny-Carman equation:

1801— M
Rd: (2‘? A”, (3.8)
ppdpe

 

where A is the membrane ﬁltration area, a is the effective porosity of the particle

deposit, M p is the mass of particles deposited on membrane surface , and tip and p p
are the diameter and the density of the particles, respectively. M p is estimated based on

the feed concentration of silica colloids and the permeate volume. This approach
presumes that the back transport of silica colloids from the membrane surface to the bulk
of the feed is negligible, which is consistent with the assumptions of the blocking laws
theory. When ﬁltration experiments are conducted at a constant transmembrane pressure,

the instantaneous value of the hydraulic resistance of the cake can be determined by:

Rd = [J°J_ J]Rma (3.9)

 

where Rm is the hydraulic resistance of clean membrane, J 0 and J are the initial and

instantaneous values of the permeate ﬂux, respectively. Combining Eqs. (3.8) and (3.9),

the effective porosity, a , can be determined.

85

3.3. Results and discussion

3.3.1. Flux performance of different membranes

Figure 3.2a shows the normalized permeate ﬂux as a function of permeate volume in
experiments with different membranes. The larger decline in the normalized permeate
ﬂux observed for the ultraﬁltration membranes was partly due to their lower hydraulic
resistance (measured in a clean water ﬂux test prior to the fouling experiments); for lower
resistance membranes the additional resistance to the permeate ﬂow due to the fouling
layer has a relatively higher impact on the ﬂux. Another factor responsible for the
membrane-to-membrane difference in the transient behavior of permeate ﬂux was the
difference in the type and timing of fouling mechanisms. In Fig. 3.2b the results of
ﬁltration experiments for all membranes are presented on the plot of t/V versus V ,
which should be linear when the fouling mechanism is cake ﬁltration (Eq. (3.4)). While
the cake ﬁltration appears to be the primary fouling mechanism for all membranes during
the later stage of ﬁltration, the non-linearity of the curves during the early stages of
experiments implies that pore blocking of one or more types preceded the cake ﬁltration
stage. Similar trends were reported by others [19-21] but fouling mechanisms acting
during the early stage of ﬁltration were not clearly identiﬁed. The following discussion

focuses on this early stage of ﬁltration experiments.

Figure 3.3 illustrates the chronological sequence of fouling mechanisms occurring in

ﬁltration experiments with different membranes. The error bars correspond to the 95%

86

conﬁdence interval for the beginning and the end of the portion of each experiment where

the given blocking mechanism dominates.

Although only one mechanism is assigned for one speciﬁc section of ﬁltration at the
initial stage, it is reasonable to expect that more than one of these blocking processes
occurs simultaneously. The blocking mechanism identiﬁed in this study as acting during
a particular stage of ﬁltration is the one which dominates the fouling process during this
stage. During certain periods of ﬁltration, the data could satisfactorily ﬁt more than one
linear equation. Such “overlap” has a straightforward and intuitive interpretation as
corresponding to the transitional stage wherein two fouling mechanisms act

simultaneously.

87

Normalized permeate ﬂux

0.2

0.4

0.3

W (minlmL)

0.2

0.1

0.8
0.6

0.4

 

 

  
     

 
  

....r\r.(¢...r.n\.;pzrr 1'
\ .111 gtﬁméfg
.kw‘ik

El BW30-365
O NF90

<> UF-30kDa
D U F-100kDa
UF-300kDa

   
 

     
   
 

 

 
   
   

 

 

 

 

 

 

     
 

 

  
 
 
 

 

 

 

l A 1 l

l l I r
i
M :1 BW30-365 i
. o NF90
F1 o UF-30kDa
_ o UF-100kDa
; i :1 UF-300kDa
’ II
.190 A,,,.'.'.'1'.'2.
E

1 1 1 1 1 1 l 1 1 1 1

o 50 100 150 200

Permeate volume (mL)

Figure 3.2. Normalized instantaneous permeate ﬂux (a) and revere cumulative ﬂux (b) as
a function of permeate volume for different membranes. The transmembrane pressures
for the BW30-365, NF 90, UF30kDa, UFlOOkDa, UF 300kDa membranes are 300 psi
(2.07 MPa), 100 psi (0.69 MPa), 40 psi (0.28 MPa), 30 psi (0.21 MPa), 20 psi (0.14
MPa), respectively. Note: the higher scatter in the ﬂux data at the later stages (Fig. 3a) of
experiments with R0 and NF membranes was due to the dissolution of air in the feed
resulting in bubbles on the low pressure side of the membrane.

88

3.3.2. Blocking mechanisms for UF membranes

In ﬁltration with the UP membranes, Eq. (3.7), but not Eq. (3.6), ﬁtted the ﬂux data well,
indicating that intermediate blocking, and not complete blocking, occurred. An example
of the ﬁtting approach is shown in Fig. 81 in the Supplementary Data (SD). With UF-
300kDa membrane, standard blocking occurred ﬁrst, followed by intermediate blocking
and then cake ﬁltration (Fig. 3.3). This was due to the large pore size and surface porosity
of the UF-300kDa membrane (Fig. 3.4). Particles ﬁrst deposited in the membrane pores
and reduced the area for water to pass through. As expected, it resulted in the fastest ﬂux
decline (Fig. 3.2). In contrast, intermediate blocking, standard blocking, and then cake
ﬁltration were found to occur successively in ﬁltration experiments with UF-30kDa and
UF-lOOkDa membranes (Fig. 3.3). Given the presence of large (up to 500 nm) surface
pores in these membranes, standard blocking was clearly occurring at the beginning of
the ﬁltration; however, the analysis of the ﬂux data (Fig. 3.3) indicates that intermediate
blocking was dominating the overall fouling process. This can be rationalized by
accounting for the relatively lower (with respect to UF-300kDa) apparent surface
porosity and wider pore size distribution of UF-30kDa and UF-lOOkDa (Fig. 3.4): for
such pore morphologies it can be expected that more particles deposit ﬁrst onto smaller
pores and other particles, while relatively fewer particles deposit into the inner channels
of larger pores (Fig. 3.4). An explanation based on differences in membrane pore
morphology was also offered by Bowen et al. [17] to explain their ﬁnding that complete

blocking was followed by a standard blocking, and then again by complete blocking:

89

 

      

 
                 

 

 

 

 

 

 

 

 

 

200' '1 ’2 . 100%
2’3 I’3 x/
1601 a a ,3 111.
5 "l Cake ¢ 2
g $ :5 liitntion é -
'120I 5. 5 7 ~/ 60963
i :4 v: 4 a g
- .5 V5? Standard ¢ '=
2 I ﬁt blocking ¢ ‘6
so a /_a . / 40%
i ,4 at a
4 2' a a
E -- -/ intermediate é
a. ‘0 $4 é blocking - '/ 20% g
I; I é
o u a o - o o o 1 o
awso- NF- UF- ur- UF- awao- NF- ur- UF- us-
365 90 30110. 10010: 30011011 365 90 soitoa 1ooitoa 300kDa

Figure 3.3. The distributions and contributions of the blocking mechanisms in terms of
permeate volume (a) and percentage of ﬂux decline (b).

90

such sequence of fouling mechanisms was interpreted to be due to the wide size

distribution of membrane pores and foulants.

The relative importance of different fouling mechanisms can be evaluated by
comparatively analyzing the volume of ﬁltrate collected (Fig. 3.3a) and the percent of
ﬂux decline observed (Fig. 3.3b) during the ﬁltration intervals corresponding to
individual fouling mechanisms. For UF membranes, the contribution of pore blocking to
the overall ﬂux decline was signiﬁcantly larger than that of cake formation and growth.
For all three UF membranes tested, cake ﬁltration corresponded to more than 50% of
permeate collected but to less than 25% of the decrease in ﬂux (Fig. 3.3). It was pore
blocking that was the main contributor to the ﬂux decline. In experiments with UF-
300KDa membrane for example, standard blocking was the dominant fouling mechanism
during the ﬁltration interval wherein only ca. 40% of permeate was collected, but resulted
in more than 85% of ﬂux decline. The above comparison is conservative: had the
experiments lasted longer, the disparity of the contributions of cake ﬁltration and pore

blocking to the overall ﬂux decline would have been even more pronounced.

3.3.3. Blocking mechanisms for salt-rejecting membranes

As in the case with UF membranes, cake ﬁltration was clearly identiﬁed as the dominant
fouling mechanisms responsible for the ﬂux decline during later stages of separation with
BW30-365 and NF90 membranes (Fig. 3.3b) as indicated by the very good ﬁt of the
experimental data to Eq. (3.4) (Fig. 83.2 in the SD). The part of R0 and NF ﬂux data that

corresponds to the collection of the ﬁrst 45 mL of the permeate could be ﬁt by Eqs. (3.6)

91

and (3.7) equally well; this is consistent with the fact that Eqs. (3.1) and (3.3) have the
same ﬁlnctional form when linearized to describe the early stages of ﬁltration (Fig. 83.3
in the SD). Thus, it was not possible to differentiate between complete and intermediate
blocking mechanisms for R0 and NF membranes based on the ﬁtting results alone. A
simple calculation shows, however, that at least 70 layers of silica colloids were
deposited on the membrane surface after 45 mL of the feed suspension passed the
membrane. Based on this estimation, it is reasonable to conclude that complete blocking,
which assumes that every particle seals a pore, could not be responsible for the ﬂux
decline throughout this entire period of the experiment. This conclusion leaves
intermediate blocking as the mechanism possibly responsible for the initial ﬂux decline
raising the question of the availability of membrane surface pores for the particles to
block. Given that the skin layer of the RO membrane is non-porous and the pores of NF
membranes are, by the deﬁnition of nanoﬁltration, smaller than 5 nm (Fig. 3.4), the fact
that intermediate blocking and not cake ﬁltration described the experimental ﬂux data the
best is surprising and hints at the fortuitous nature of the ﬁt. One possible physical
explanation of the good ﬁt by the intermediate blocking model is that some variance
exists in the permeability of the membrane skin. In this case, more permeable areas
(surface defects) [24] of the membrane skin would function as “pores” during initial
stages of the separation experiment resulting in higher colloidal deposition on these areas.
After the membrane surface is completely covered by the colloids, no “pores” on the
membrane surface would remain to be blocked and only cake ﬁltration would occur

afterwards. This explanation is consistent with and corroborated by the reported short-

92

 

a BW30-365 . b NF90

 

(e) UF—300kDa

Figure 3.4. SEM images of the ﬁltration surfaces of membranes. (a) BW30-365; (b)
NF90; (c) UF-30kDa; (d) UF-lOOkDa; and (e) UF-300kDa.

93

term increases in the rejection of salt by nanoﬁltration membranes immediately upon the

introduction of colloids into the electrolyte-only feed [25, 26].

An additional, and not necessarily alternative, explanation of the deviation from the cake
ﬁltration law is the transient behavior of the cake porosity during the initial stages of the
cake growth. As the estimation of the effective porosity using Carrnan-Kozeny equation
demonstrates (Fig. 3.5), the effective porosity increased initially for a short period and
then decreased until a relatively steady value was achieved; this implies that with the
deposition of more particles an initially loose deposit grew into a more compact structure
until a stable and uniform cake was formed. The surface roughness of BW30-365 and

NF 90 membranes (Fig. 3.4) may be contributing to this transient behavior of cake

porosity.

While it is commonly assumed that for NF and R0 membranes cake ﬁltration is the only
mechanism of colloidal fouling, our results indicate that as much as 18% (for BW30-365)
and 53% (for NF90) of ﬂux decline for salt—rejecting membranes can not be
unequivocally assigned to cake formation. Given that colloidal deposition contributing to
the ﬂux decline at the early stage of ﬁltration is the likely cause of irreversible portion of
the fouling, the reported ﬁndings point to the importance of more detailed understanding
of the colloid-membrane interactions that manifest themselves and can be interpreted as

pore blocking.

94

0.4 I I I I I I I I I l I I I I I I f I

 

.0
(a)
1
1

      

     

   

   

 

 

 

 

 

 

g ”Elli ......... --'- [with . .
e .0 ; ﬂ . ~
é. 02 _ Q NF90 . .- ._
z - El BW30-365 '
E 4 , .1
a:
I“ t -
0.1 r _
,
i
1 1
" T
0 1 1 1 1 1 1 1 1 1 l 1 1 1 1 1 1 1 1 1
0 50 100 150 200

Permeate volume (mL)

Figure 3.5. Effective porosity of the layer of silica colloids deposited onto the surface of
BW30-365 and NF90 membranes.

95

3.4. Conclusions

Experiments on the dead-end ﬁltration of colloidal suspensions using membranes ranging

from loose ultraﬁltration to reverse osmosis were conducted to determine the type, the

chronological sequence, and the relative importance of individual fouling mechanisms.

To identify fouling mechanisms, a method was proposed that involves a straightforward

linear least squares ﬁtting with an easily determinable goodness of ﬁt. It was

demonstrated that

1)

2)

3)

4)

For all membranes, ﬂux decline during the later stages of ﬁltration is consistent with
the cake ﬁltration mechanism, but deviates from this behavior during earlier stages of
ﬁltration.

For all membranes, ﬂux decline during the earlier stages of ﬁltration could be
interpreted in terms of one or more pore blocking mechanisms. The observed
temporal overlap of different fouling mechanisms was interpreted as the transitional
stage wherein contributions of the two fouling mechanisms to the ﬂux decline were
commensurate.

For ultraﬁltration membranes, pore blocking and not cake formation was the largest
contributor to the observed ﬂux decline. The chronological sequence of blocking
mechanisms was interpreted to depend on the size distribution and surface density of
membrane pores.

For salt-rejecting membranes, ﬂux decline during the earlier stages of ﬁltration was

attributed to either intermediate blocking of relatively more permeable areas of the

96

membrane skin, or to the cake ﬁltration in its early transient stages, or a combination
of these two mechanisms.
Clear identiﬁcation and differentiation between fouling mechanisms acting during the
early stages of membrane separation provides useful insights into details of colloid-
membrane interactions that are implicated in controlling permeate ﬂux performance and

the efﬁciency of membrane cleaning.

Acknowledgements

This work was supported by the National Science Foundation (research grant OISE-
0530174) and the National Water Research Institute (project no. 05-TM-007). We thank
Dow-FilmTec for providing membrane samples and Nissan Chemical America Corp. for

supplying the silica suspension.

97

Reference

10.

ll.

12.

13.

P. H. Hermans and H. L. Bredée, Principles of the mathematical treatment of
constant-pressure ﬁltration, J. Soc. Chem. Ind. 55T (1936) 1-4.

V. E. Gonsalves, A critical investigation on the viscose ﬁltration process, Rec.
Trav. Chim. des Pays-Bas 69 (1950) 873.

H. P. Grace, Structure and performance of ﬁlter media. II. Performance of ﬁlter
media in liquid service, AIChE. J. 2 (1956) 316-336.

J. Hermia, Constant pressure blocking ﬁltration laws - application to power-law
non-Newtonian ﬂuids, Trans. Inst. Chem. Eng. 60 (1982) 183 -187.

K. J. Hwang and T. T. Lin, Effect of morphology of polymeric membrane on the
performance of cross-ﬂow microﬁltration, J. Membr. Sci. 199 (2002) 41-52.

E. Iritani, N. Katagiri, Y. Sugiyarna and K. Yagishita, Analysis of ﬂux decline
behaviors in ﬁltration of very dilute suspensions, AIChE J. 53 (2007) 2275-2283.

H. Huang, T. A. Young and J. G. Jacangelo, Uniﬁed membrane fouling index for
low pressure membrane ﬁltration of natural waters: principles and methodology,
Environ. Sci. Technol. 42 (2008) 714-720.

W. Xu and S. Chellam, Initial stages of bacterial fouling during dead-end
microﬁltration, Environ. Sci. Technol. 39 (2005) 6470 -6476.

E. Iritani, Y. Mukai, Y. Tanaka and T. Murase, Flux decline behavior in dead-end
microﬁltration of protein solutions, J. Membr. Sci. 103 (1995) 181-191.

L. Palacio, C. C. Ho, P. Pradanos, A. Hernandez and A. L. Zydney, Fouling with
protein mixtures in microﬁltration: BSA-lysozyme and BSA-pepsin, J. Membr.
Sci. 222 (2003) 41-51.

C. C. Ho and A. L. Zydney, A combined pore blockage and cake ﬁltration model
for protein fouling during microﬁltration, J. Colloid Interface Sci. 232 (2000) 389.

W. Yuan, A. Kocic and A. L. Zydney, Analysis of humic acid fouling during
microﬁltration using a pore blockage-cake ﬁltration model, J. Membr. Sci. 198
(2002) 51-62.

J. A. Suarez and J. M. Veza, Dead-end microﬁltration as advanced treatment for
wastewater, Desalination 127 (2000) 47-5 8.

98

14.

15.

l6.

17.

18.

19.

20.

21.

22.

23.

24.

25.

M. K. Purkait, P. K. Bhattacharya and S. De, Membrane ﬁltration of leather plant
efﬂuent: Flux decline mechanism, J. Membr. Sci. 258 (2005) 85—96.

T. Mohammadi, M. Kazemimoghadarn and M. Saadabadi, Modeling of
membrane fouling and ﬂux decline in reverse osmosis during separation of oil in
water emulsions, Desalination 157 (2003) 369-375.

P. Aimar. Recent progress in understanding particle fouling of ﬁltration
membranes. in The Fifth International Membrane Science & Technology
Conference. 2003. Sydney, Australia.

W. R. Bowen, J. I. Calvo and A. Hernandez, Steps of membrane blocking in ﬂux
decline during protein microﬁltration, J. Membr. Sci. 101 (1995) 153-165.

C. Duclos-Orsello, W. Li and C. C. Ho, A three mechanism model to describe
fouling of microﬁltration membranes, J. Membr. Sci. 280 (2006) 856.

E. M. Tracey and R. H. Davis, Protein fouling of track-etched polycarbonate
microﬁltration membranes, J. Colloid Interface Sci. 167 (1994) 104-116.

K. J. Kim, V. Chen and A. G. Fane, Ultraﬁltration of colloidal silver particles:
Flux, rejection, and fouling, J. Colloid Interface Sci. 155 (1993) 347 - 359.

S. F. E. Boerlage, M. D. Kennedy, M. R. Dickson, D. E. Y. El-Hodali and J. C.
Schippers, The modiﬁed fouling index using ultraﬁltration membranes (MFI-UF):

characterisation, ﬁltration mechanisms and proposed reference membrane, J.
Membr. Sci. 197 (2002) 1-21.

H. Choi, K. Zhang, D. D. Dionysiou, D. B. Oerther and G. A. Sorial, Effect of
permeate ﬂux and tangential ﬂow on membrane fouling for wastewater treatment,
Sep. Purif. Technol. 45 (2005) 68-78.

D. C. Montgomery, A. E. Peck and G. G. Vining, Introduction to Linear
Regression Analysis. Fourth ed. 2006: Wiley-lnterscience.

B. Mi, C. L. Eaton, J H. Kim, C. K. Colvin, J. C. Lozier and B. J. Marinas,
Removal of biological and non-biological viral surrogates by spiral-wound
reverse osmosis membrane elements with intact and compromised integrity,
Water Res. 38 (2004) 3821-3832.

B. M. Hoek and M. Elimelech, Cake-enhanced concentration polarization: a new

fouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 17 (2003)
5581-5588.

99

26. F. Wang and V.V. Tarabara, Coupled effects of colloidal deposition and salt
concentration polarization on reverse osmosis membrane performance, J. Membr.
Sci. 293 (2007) 111-123.

100

SS. SUPPLEMENTARY DATA

 

 

   

 

 

 

 

 

 

 

 

025 .ﬁ.,...,-..,..-,...,.-.‘ 0,26...-,-...,--..,s...,...MWA
1 ' i : 5 . L —' IO.138+0.00436X .R- 0.998 1
; —y=o.~1o4+o.o111x; as 0.997 1‘): . 1' ; s s , 1b).
0.225 2. ., , . .1 _. 024 _._ . 1 "gt a
- ~ . 1 ‘ t - . g ' --" : ‘
9 0.2 _- 30.22 - ~7
E : i; l 1
20.175 - °- 02 - -
.. E .
e .. .
Q 0.15 — 3 0.18 - i
1- ‘ <
0.125 1 0.16 — -
1 1 *
0.1[1L_1_l111i1_1111_t1_1111111114 0.14”111114441111L1141111111111111
0 2 4 6 8 10 12 0 5 10 15 20 25 30
Time (min) Time (min)
(13 ."* . , . . . , . - . , . . a , . . .+
- 3 1 (C) :
—y80.0s746+0.000926x th1 1
0.27 s ; .. .. 1_ .. 1. ..... ,2
:1‘
E 0.24 7
E 1
5. :
§;(121 -
0.18 :-
0'15 ’ 1 1 1 1 1 1 1 1 1 . . 1 1 1 1 1 1 1 ‘

 

 

 

100 120 140 160 180 200
Permeate volume (mL)

Fig. S3.]. Application of blocking laws to the ﬁltration data of UFlOOkDa membrane: (a)
1/ Q versus t based on Eq. (3.7) for intermediate blocking; (b) t/ V versus t based

on Eq. (3.2) for standard blocking; and (c) t/V versus V based on Eq. (3.4) for
cake ﬁltration.

101

0.6

 

 

    

 

 

 

 

 

 

. y= 0.473 + 0.000402x R= 0.999 1
. ""'""Y=0.423+0.00(i569x R=1§ 1
0.55 9 = =
LT
g .
.5 0.5
g 1
g 1-
0.45 1
i +ewao-365
’ "0—NF90
0.4 b 1 1 1 1 l 1 1 1 1 l 1 1 1 l 1 1 1 a
0 50 100 150 200

Permeate volume (mL)

Fig. 83.2. Application of blocking laws to the R0 and NF membrane ﬁltration data: t/V
versus V based on Eq. (3.4) for cake ﬁltration.

102

 

I ﬁi I r I

' l

2,3 4 1 1 , . 1 1 1
y = 1.98 - 0.001865)! R= 0.9901 :

 

 
 
  

 

—El— BW30-365 Z

..1

 

 

Q (cm‘3lmin)
N

.x
(D
T I I I

 

 

 

..s
m

80

 

 

 

 

 

 

 

 

 

0 20 40 60 100
Permeate volume (mL)
0.6.........,..-2,ﬁ,.,,..,,
f (b)-—-y=0.5053+0.0009971x R= 0.9901 i
‘ y=o.4404+0.00219x Rs 0.9987
0.55 — 1 . ; _ _
6‘
<
5 .
a 0 5 ~
E 1
E I NF90 +BW30-365
=0 45 1 —e—Ni=9o
. ‘ i
1 T
04 11111111111111111111
0 10 20 30 40 50
Time (min)

Fig. 83.3. Application of blocking laws to the R0 and NF membrane ﬁltration data: (a)
Q versus V based on Eq. (3.6) for complete blocking, and (b) 1/ Q versus t based

on Eq. (3.7) for intermediate blocking.

103

As Fig. S3 shows, both Eq. (3.6) and Eq. (3.7) ﬁt the initial part of ﬂux data for both R0

and NF membranes well. The underlying reason is explained as follows.

K b in Eq. (1) and K ,- in Eq. (3.3) were calculated numerically based on the V(t) data
obtained during initial stages of R0 and NF experiments. It is found that for both the R0

and NF membranes, the value of K b is on the order of 10'16 or 10'17 min", and the value
of K ,- is on the order of 10'17 cm'3. This means that Kbt << 1 and K iQOt << 1.

—K b1

For Eq. (1) which corresponds to complete blocking, K b V = Q0 (1 — e ), Taylor series

K312 K {:73 . . . . .
2 + 6 ).D1fferent1at1ngth1s equatlon leads

 

expansion gives us, K bV z Q0 (K bt —

IO:

2 2
dV Kbt
=—~ l—K 1+ . (3.1)
Q dt Q0( b 2 )

 

For Eq. (3) which corresponds to intermediate blocking, K ,-V = ln(1+ K iQOt) , a similar

approach gives us,

Q = % ~ Q0(1—K1Q61+K12Q3t2)- ($32)

The similarity between Eq. (83.1) and Eq. ($3.2) demonstrates that, when K bt << 1 and
K iQot << 1, Eqs. (3.1) and (3.3) predict the same relationship between ﬂux and time.

This is the reason why both Eqs. (3.6) and (3.7) ﬁt the initial part of ﬂux data equally

104

well. This implies that, when K bt << 1 and K iQot << 1, Eqs. (3.6) and (3.7) can not be

used to discriminate between complete blocking and intermediate blocking.

105

CHAPTER FOUR

Wang, Fulin; Tarabara, Volodymyr V. (2008). Modeling the permeability of ﬁber-ﬁlled
porous media using Kozeny-Carman-Ethier equation. (In preparation for publication in a
journal)

106

CHAPTER FOUR

MODELING THE PERMEABILITY OF FIBER-FILLED POROUS
MEDIA USING KOZENY-CARMAN-ETHIER EQUATION

ABSTRACT

In this study, a model is proposed for predicting the permeability of mixed porous media
of particles and ﬁbers. The model combines Kozeny-Carman equation and Ethier’s
theoretical model to predict the permeability of the mixed media based on the
permeability of the porous bed of particles and on the volume fraction of the ﬁbers.
Membrane ﬁltration experiments were conducted using silica colloids and carbon ﬁbers
with varied ﬁber mass fraction. The model prediction is in great accordance with the

experimental results for the ﬁber mass fraction of less than 50%.

107

4.1. INTRODUCTION

Flow through porous media is a very important process for a variety of natural
phenomena and engineering applications, such as, ﬂow of groundwater through soil,
water transport through bodily tissues, ﬂow through deposit layers on membrane surfaces,
ﬂow through fuel cells and permeation through polymeric networks. Given the
kaleidoscopic shapes of natural and engineering matter, the components of the media can
be spherical particles, elongated particles, rods, cubes, ﬁbers (cylinders), or a
combination of them. Permeability is the key parameter for designing and controlling this
process. The permeability depends on the structure of the media, such as, the shape of the

components, the porosity, the packing densities, etc [1, 2].

Based on Darcy’s law, theoretical equations have been developed for predicting the
permeability of porous media. For media of uniform spherical particles, the well-known
Kozeny-Carman equation [3] is widely used, which predicts the permeability (K) based

on porosity (e) and particle diameter (dp ) as shown in Eq. (4.1),

d283
___ P 2 . (4.1)
180(1 — e)
Darcy’s law has also been extended to model the permeability of ﬁbrous media. The ﬂow
through media of ﬁbers with uniform radius that are placed randomly or in arrays has
been modeled by several researchers [4-8]. Jackson and James [2] summarized the

historical development of this subject. A general equation can be presented as [1, 2],

108

K
—2 = f (it). (4.2)

a

where a is the ﬁber radius, f ((25) is a model-dependent function, and ¢ is the ﬁber

volume fraction. However, it is only applicable for monodisperse ﬁbrous media, but not

polydisperse ﬁbrous media.

Ethier [1] theoretically modeled the ﬂow in a composite porous media composed of two
types of ﬁbers of highly dissimilar sizes using the Debye-Brinkman equation. The overall
permeability was determined based on the permeability of the ﬁne ﬁbers and the volume

fraction of the coarse ﬁbers (more detail in Section 4.2.3).

Fibrous ﬁlters have been widely used in industries such as production of chemicals,
environmental safety, food industry, pharmaceutical, and semi-conductor industries ([9,
10]) to remove gas impurities. When solid particles accumulate inside the ﬁlters, the
ﬁlters turn into a matrix of particles and ﬁbers. Thomas et al. [10] modeled the
performance of ﬁbrous ﬁlters of varied characteristics removing solid aerosol particles in
air under different operating conditions. Pressure drop and ﬁlter efﬁciency were modeled
based on cake packing density. In this study, the trapped particles formed dendrites and
functioned as new collecting ﬁbers. Recently, new nanoﬁbrous membranes have been
developed using various materials, such as TiOz nanowire [ll], polyurethane [12],
polyvinyl alcohol [13, 14], polysulfone [15], and nylon [16]. As demonstrated in [15, 16],

when the nanoﬁbrous membranes are fouled by particles, they become ﬁltration media of

109

ﬁbers and particles. In media of ﬁbers and particles, ﬁbers increase tortuosity and
obstruct some permeable area, which greatly reduces the total permeability [1].
Knowledge of the permeability of such media is crucial for designing membrane systems

and predicting their performance.

However, to our best knowledge, so far there have been no theoretical studies on
predicting the hydraulic permeability of porous media consisting of particles and ﬁbers.
To predict the performance of these media is in pressing need. In this study, we
developed a theoretical model for predicting the hydraulic permeability of media
consisting of particles and ﬁbers, based on Kozeny-Carman equation and Ethier’s
theoretical model. The model prediction is in great accordance with the experimental

results.

4.2. Theoretical background

4.2.1. The hydraulic resistance of a deposit layer in membrane systems

When the feed water is free of solutes and potential foulants, the pure water permeate

ﬂux J 0 is linked to the transmembrane pressure AP by the Darcy’s equation:

JO :——, (4.3)

110

where Rm is the hydraulic resistance of the membrane, and ,u is the dynamic viscosity

of water.

When there are foulants deposited on membrane surface, the permeate ﬂux will decrease

due to the hydraulic resistance of the deposit layer (Rd ). The equation for permeate ﬂow

becomes:

: AP
#(Rm '1' Rd (1)) ,

 

1 (4.4)

where J, is the permeate ﬂux at ﬁltration time t, and Rd (t) is the hydraulic resistance

of the deposit layer as a function of t. J , /J0 is deﬁned as the normalized permeate ﬂux.

When ﬁltration experiments are conducted at a constant transmembrane pressure, Eqs.

(4.3) and (4.4) lead to a simple equation for determining Rd (t) as:

 

Rd (1) = [JOJ’ J’ :lRm. (4.5)
1

Equation (4.5) can also be used for determining the Rd as a function of permeate volume

(Rd (V) ) as:

 

Rd (V) = [JOJ‘VJV JR," , (4.6)

111

where J V is the permeate ﬂux when the cumulative permeate volume is V .

4.2.2. The hydraulic resistance of porous media of particles

For deposit layers consisting of spherical incompressible particles of uniform size,

Kozeny-Carman equation [3] is widely used to calculate Rd :

180(1—s) Mp
,apdﬁr;3 A

 

Rd = 1 (47)

where A is the membrane ﬁltration area, a is the porosity, M p is the mass of particles
deposited on membrane surface, and tip and p p are the diameter and the density of the

particles, respectively.

The speciﬁc hydraulic resistance lid is deﬁned as:

.1 R _ 2
R __d_=180(1£)

— 2 4.8
5d die3 ( )

where 6d is the thickness of the deposit layer. The permeability K is the reciprocal of

the hydraulic speciﬁc resistance, as K = 1/ Rd.
4.2.3. The permeability of highly polydisperse ﬁbrous media

Ethier [1] theoretically modeled the ﬂow in a composite porous media composed of two

types of ﬁbers of highly different sizes using the Debye-Brinkman equation. The ﬁne

112

matrix appeared to be homogeneous on the length scale of the coarse ﬁbers. In his model,

the coarse ﬁbers were embedded into the ﬁne ﬁbers (with permeability K f) to form a

porous matrix, as shown in Figure 4.1.

Ethier has shown that, for a medium of randomly oriented ﬁbers, when K f / ac2 -—> 0

where ac is the radius of the coarse ﬁbers, the total permeability (K m) can be predicted
based on the permeability of the ﬁne ﬁbers (K f) and the volume fraction of the coarse

ﬁbers as:

Ktotz 1—¢c (49)
Kf 1+2/3¢c’ '

 

where 05, is the volume fraction of the fraction of the coarse ﬁbers, which is deﬁned as:

_ VC
Vtot ,

 

¢c (4.10)

where VC is the volume occupied by the coarse ﬁbers, and Vm, is the total volume.

4.3. Experimental method

4.3.]. Experimental apparatus

Dead-end membrane ﬁltrations were conducted using a high-pressure stainless steel

ﬁltration cell (HP 4750, Sterlitech Corp., Kent, WA) at a transmembrane pressure of 100

113

psi (0.689 MPa) without stirring. The permeate ﬂux was recorded using a digital
weighing balance (ARC120, Ohaus Corp., Pine Brook, NJ) interfaced with a computer
via a built-in RS-232 port. The permeate was delivered to a beaker positioned on the
balance. The data from the balance were logged to the computer via a program written in
LabView (version 7.1, National Instruments, National Instruments Corp., Austin, TX)
with a time interval of 10 s. The volume of the feed suspensions was 100 mL. The
experiment was conducted at the room temperature (21 °C). Because there was no stirring

in the ﬁltration cell, a constant temperature of the feed suspension was assumed.

114

 

Figure 4.1. [1] Idealized cross-section view of the porous matrix

(The coarse ﬁbers are shown to be parallel to one another in this ﬁgure, but they can be
randomly oriented as well)

115

4.3.2. Reagents and chemicals

All reagents were of ACS analytical grade or higher purity (Fisher Scientiﬁc, Pittsburgh,
PA) and were used without further puriﬁcation. The deionized water used in the
experiments was supplied by a commercial ultrapure water system (Lab Five, USFilter
Corp., Hazel Park, MI) equipped with a terminal 0.2 pm capsule microﬁlter (PolyCap,
Whatman Plc., Sanford, ME). The resistivity of the water was greater than 16MQ-cm.
The pH of water was measured to be in the range of 6.1 to 6.4 (pH meter D12, Horiba,

Kyoto, Japan). No buffers were added to the water.

4.3.3. Ultraﬁltration (U F) membrane, silica colloids, and carbon ﬁbers

A polyethersulfone (PES) UF membrane that has a nominal molecular weight cut-off
(MWCO) of 10 kDa according to the manufacturer was used (DiaryUF, Hydranautics
Corp., Oceanside, CA). This UF membrane was chosen because it did not reject salts and
thus no possible osmotic pressure was expected to occur across the membrane. The
membrane was supplied as a ﬂat sheet and was stored in a sealed plastic container at
room temperature. Before the experiment, two rectangular membrane coupons were cut
out of the membrane sheet and soaked in deionized water for 24 h at 4 °C with water

exchanged after the ﬁrst 12 h.

Silica colloids (SnowTex-XL, Nissan Chemical America Corp., Houston, TX) were used
as the model colloids. Carbon ﬁbers were provided by Applied Sciences, Inc. (Pyrograf®
PR-19-PS-LD, Cedarville, Ohio) with “PS-LD” indicating that the ﬁbers were low

density (LD) pyrolytically stripped (PS) carbon ﬁbers.

116

4.3.4. Scanning electron microscopy (SEM) analysis

The colloids, the carbon ﬁbers, and the deposit layers on membrane surface were ﬁrst
coated for 20 s using a pure osmium coater (Neoc-AN, Meiwa Shoji Co. Ltd., Kyoto,
Japan), and then imaged using Hitachi S-47OOII Field Emission scanning electron
microscope (Hitachi High Technologies America, Inc. Pleasanton, CA) with a secondary
electron detector. The microscope was operated in ultra-high resolution mode with an

accelerating voltage of 15 kV and an emission current of 10 11A.

4.3.5. Experimental protocol

4.3.5.1. Preparation of the feed suspensions

Carbon ﬁbers of were weighed at 200 mg and added to a 100 mL 2 mM sodium dodecyl
sulfate (SDS) solution. The suspension was continuously sonicated with a 100 W
ultrasonic cell disrupter (VirSonic 100, SP Industries Inc., Gardiner, NY) for 40 min.
After that the ﬁbers were evenly and stably dispersed in the suspension. After 24 hours,
no visible aggregates were observed. The suspension was used as the stock suspension of

the carbon ﬁbers.

4.3.5.2. Membrane compaction

To eliminate the effect of membrane compaction on our determination of hydraulic
resistance of deposit layers based on permeate ﬂux, two UF membrane coupons were
compacted at a transmembrane pressure of 120 psi (0.827 MPa) using deionized water in

a crossﬂow ﬁltration system for around 48 h until a stable permeate ﬂux was achieved.

117

The crossﬂow ﬁltration system was described in our previous study [17]. After the
compaction, the membrane coupons were taken out and several round pieces (with a
diameter of 47.6 mm) were gently cut out of the coupons. The pieces were suitable for
the dead-end ﬁltration cell and were stocked in deionized water at 4 °C for subsequent

use.

4.3.5.3. Measurement of membrane hydraulic resistance using pure water

The permeate ﬂux was recorded for a series of transmembrane pressures: 60 psi (0.414
MPa), 80 psi (0.552 MPa), 100 psi (0.689 MPa) and 120 psi (0.827 MPa). The hydraulic

resistance of the clean membrane (Rm) was calculated from Eq. (4.1) using linear least

squares ﬁtting.

4.3.5.4. Membrane filtration experiments

One hundred milliliters of feed suspensions of varied concentration of carbon ﬁbers and
200 mg/L silica colloids were prepared using the stock suspensions of ﬁbers and silica
colloids. The feed suspensions were sonicated for 5 minutes using a sonicator (Model
SOT, VWR International Inc., West Chester, PA) to ensure the ﬁne dispersion of silica
colloids and carbon ﬁbers. After that the feed suspensions were added into the ﬁltration
cell and membrane ﬁltration experiments were conducted at a constant transmembrane
pressure of 100 psi (0.689 MPa). The permeate ﬂux was recorded. The resistance of the
deposit layer was calculated based on Eqs. (4.5) and (4.6) as a function of time and

permeate volume as well.

118

4.4. Results and discussion

4.4.1. Characterization of the silica colloids and carbon ﬁbers

The SEM image of the silica colloids is shown in Fig. 4.2(a). The colloids have a density
of 2.36 g/cm3 and a diameter of (55 .6 i 1.3) nm, as characterized in our previous study

[17].
Fig. 4.2(b) shows the SEM image of the carbon ﬁbers. According to the manufacturer,

the ﬁbers have a density of 1.95 g/cm3 and the diameter and the length range from 0.1 to

0.2 pm and from 30 to 100 pm, respectively.

119

5
.3) \‘ r 1 'V (,
Y3" ‘3 (I ...,
r J k 1 k ’1
%“P~f3ooﬂ~1v:s
050 r-C 36 1‘0:

 

(a) (b)
Figure 4.2. SEM images for the silica colloids (a) and carbon ﬁbers (b)

120

ﬁﬁﬁﬂliﬂi‘iﬂ
. r

      

9
co

.0
a:

      

  
 

3’40.

- '- 4.0.:
mums-we, .

'::.1 ﬁ-‘(o , '

' x1 , "nu/.1

“r ’0’,” i '92.}?

(((‘p‘b/l “

    
 

    

  

 

Normalized permeate ﬂux
.0 .O
O) \l

 

 

 

 

 

 

 
 
 
 
   

 

 

9 100 mgIL carbon fibers
0 50 mglL carbon ﬁbers
0 5 A 25 mg/L carbon fibers _______ .............. _:
O 200 mglL carbon ﬁbers 3 .
D Silica particles only g 1
0 4 £21 200 mgIL carbon fiber only i
0 20 40 60 80 100
Permeate volume (mL)
1 , . . . r .
0 100 mgIL carbon fibers § § .%
O 50 mglL carbon fibers 5 g.-.-:-:.
... A 25 mgIL carbon fibers 3 ._ .
5 0-8 0 200 mglL carbon fibers """"""" """""" - . k
M D Silica particles only 5 . -- 3
P2 ﬂ. .
8 05 ..............
c
8
.2
In
8 0.4
.2
3
E 'a ..
3.
I 02 .................................................
O

 

 

 

O 20 40 60 80 100
Permeate volume (mL)

Figure 4.3. Permeate flux (a) and hydraulic resistance of the deposit layer (b).
Transmembrane pressure was 100 psi (0.689 MPa) for each experiment. Rm in each

experiment was shown in Table 1. All of the experiments were conducted with 200 mg/L
silica colloids, except the data line for “200 mg/L carbon ﬁbers only” in (a) when no
colloids were added. The hydraulic resistance of the deposit layer from “200 mg/L carbon
ﬁbers only” was not shown in (b) because it was too low to be measurable in our
experiments.

121

4.4.2. Permeate flux and hydraulic resistance of the deposit layer

Table 4.1 shows the membrane hydraulic resistance (Rm) in the experiments. Rm was
used to calculate the hydraulic resistance of the deposit layer (Rd) based on Eqs. (4.5)

and (4.6).

The normalized permeate ﬂux of the UP membrane ﬁltrating different suspensions is
shown in Fig. 4.3(a). All of the experiments were conducted with 200 mg/L silica
colloids, except the experiment with carbon ﬁbers only when no silica colloids were used.
From Fig. 4.3(a) it can be seen that no evident permeate ﬂux was observed in the
experiment with 200 mg/L nanoﬁbers only, which indicates the ﬁbers formed a very
porous structure that had a relatively low hydraulic resistance to the membrane hydraulic
resistance. SEM images (Fig. 4.4(e)) conﬁrmed that the deposit layer from this

experiment was quite porous with visible large open spaces among the ﬁbers.

With 200 mg/L silica colloids, the permeate ﬂux decline rate increased when the
concentration of the carbon ﬁbers increased from 25 mg/L to 100 mg/L (Fig. 4.3(a)),
indicating the hydraulic resistance increased with increased carbon ﬁber concentration
(Fig. 4.3(b)). However, when it reached 200 mg/L, permeate ﬂux decline rate greatly
decreased, indicating a lower hydraulic resistance. This is due to the fact that in this
experiment the deposition layer was more porous with open spaces among the ﬁbers, as

shown in Fig. 4.4(d), which dramatically reduced the hydraulic resistance.

122

Table 4.1. Rm in different experiments

 

 

Experiment Name R m (x1olym'1)
200 mg/L carbon ﬁbers only 1.17
200 mg/L silica colloids only 1.16
25 mg/L carbon ﬁbers + 200 mg/L silica colloids 1.22
50 mg/L carbon ﬁbers + 200 mg/L silica colloids 1.22
100 mg/L carbon ﬁbers + 200 mg/L silica colloids 1.16
200 mg/L carbon ﬁbers + 200 mg/L silica colloids 1.43

 

123

It is worthy noticing that in Fig. 4.3(a), the permeate ﬂux from the experiment with 200
mg/L silica colloids only decreased slightly faster than that with 200 mg/L silica colloids
and 200 mg/L carbon ﬁbers. But the hydraulic resistance of the deposit layer for the
former was slightly lower than that for the latter (Fig. 4.3(b)). This is due to the lower

hydraulic resistance of the membrane used in the former, as shown in Table 4.1.

4.4.3. SEM images of the deposit layers

4.4.4. Modeling the hydraulic resistance of the deposit layers consisting of silica

colloids and fibers

4.4.4.1. Theoretical development

In our experiments, the ﬁbers (with a radius a f = 0.05 ~ 0.1 pm) and silica colloids can
be regarded as the coarse and ﬁne ﬁbers in Ethier’s theoretical model, respectively.
Because in our experiments K p M} < 0.001 where K p is the permeability of the
porous bed formed by the silica colloids, Ethier’s model can be used to predict the
permeability of the deposit layer based on the permeability of the silica colloid
aggregates. Replacing the permeability K by hydraulic speciﬁc resistance Ii , Eq. (4.9)

can be re-written as:

Riot =1+2/3¢f (411)
RP 1—¢f '

 

124

 

Figure 4.4. SEM of the deposition layers (continued on next page).

125

 

(0)

Figure 4.4. SEM of the deposition layers. (a) 25 mg/L carbon ﬁbers and 200 mg/L silica
colloids; (b) 50 mg/L carbon ﬁbers and 200 mg/L silica colloids; (c) 100 mg/L carbon
ﬁbers and 200 mg/L silica colloids; (d) 200 mg/L carbon ﬁbers and 200 mg/L silica
colloids; (e) 200 mg/L carbon ﬁbers only (without silica colloids)

126

where 13,0, is the hydraulic speciﬁc resistance of the total deposit layer, 11,, is the

speciﬁc hydraulic resistance of porous bed of the silica colloids, and eff is the volume

fraction of the carbon ﬁbers.

Combining Eqs. (4.7), (4.8) and (4.11), an equation for predicting the hydraulic resistance

of a deposit layer consisting of ﬁbers and particles can be written as:

. lsoaf-s )2 r+213¢
Rd = Rdé‘d = 2 5 [ f ]§d, (4.12)
dpep 1—¢f

 

where 6d is the thickness of the deposit layer and 8,, is the effective porosity of the

porous bed of silica colloids. We name this equation as Kozeny-Carman—Ethier equation.

The total volume of the deposit layer V can be calculated as:

(0!

mp mf
+ ’
pp(1_8p) pf

 

V10: = Vp + Vf = (4.13)

where Vc and Vp are the volume of carbon ﬁbers and the porous bed of silica colloids
on membrane surface, respectively, p f and p p are the density of the carbon ﬁbers (1.95

g/cm3) and the silica colloids (2.36 g/cm3), respectively.

The thickness of the deposit layer (5d ) can be calculated as:

127

mp ’"f
+
5d = V10. = ”PO-Sp) pf : ”'pr +mfpp0‘8p) (4.14)
A A Apfppﬂ-ep) ’

 

 

 

 

where A is the area of the deposit layer (i.e. the effective ﬁltration area of the membrane).

The volume fraction of the carbon fibers ¢f is:

 

 

m
_L
V p m p (1-6 )
¢f=Vf = f = f P P . (4.15)
to, mf+ mp mfpp(1-£p)+mppf
pf pp(l-6p)

Combining Eqs. (4.12), (4.14) and (4.15), the equation for predicting the hydraulic
resistance of the deposit layer consisting of carbon ﬁbers and silica colloids in our

experiments is:

 

(4.16)

R 180(1—6p) 3mppf+5mfpp(1-£p):|[mppf+mfpp(l-£p)
d= -

dg£p3 3’"pr Apfpp

4.4.4.2. The effective porosity of the porous bed of silica colloids

We assumed that the back transport of silica colloids from membrane surface to the feed
was negligible, which means that the mass of the silica colloids deposited can be
calculated based on permeate volume and the feed concentration of the silica colloids.

This assumption is reasonable because the ﬁltration was performed using a dead-end

128

mode without stirring. This assumption has been widely used in the models for the
performance of membrane operated at a dead-end non-stirring mode. For examples, it is
used in the well-know blocking ﬁltration laws [18] which study different blocking

mechanisms in membrane fouling processes.

Thus, the mass of silica colloids deposited on the membrane surface can be predicted

based on the permeate volume (V ) and the feed concentration of the colloids (C p ), as:

p. (4.17)

Based on Eqs. (4.6), (4.7) and (4.17), the eﬁ‘ective porosity of the deposit layer of silica

colloids only can be determined as a function of permeate volume, as shown in Fig. 4.5.

It can be seen that the effective porosity gradually increased until it reached a stable value,
which is believed to be attributed to the processes of breakup and restructuring of the
deposit layer. The averaged effective porosity was 0.353 based on over 130 data points.

This average porosity was used for the following modeling analysis.

129

 

0.8 V l I I r r I I l v I I r r r l

 

 

 

 

 

 

 

0.6
E
3
8
$0.4
§
3:
Ill
0.2_ ....................................................................................... __
J
+ _ , . ‘ 4
041.111L11111...i
0 20 40 60 80 100

Permeate volume (mL)

Figure 4.5. Porosity of the silica deposit layer as a function of permeate volume

130

4.4.4.3. Prediction of the hydraulic resistance of the deposit layer consisting of silica

colloids and carbon ﬁbers

We made four assumptions for applying the Kozeny-Carman-Ethier equation on our

experimental data:

(1) The back transport of silica colloids from membrane surface to the feed was
negligible, as discussed above. This assumption enabled us to calculate the mass
of carbon ﬁber and silica colloids on membrane surface based on permeate
volume, as indicted in Eq. (4.17).

(2) The porosity of the silica aggregates did not change in the presence of carbon
ﬁbers. This is a valid assumption because all of the experiments were operated
at dead-end non-stirring mode with the same solution chemistry. As suggested
in the literature [17, 19-21], solution chemistry (i.e. ionic strength and pH) is the
main factor controlling the structure of deposit layers on membrane surfaces
because it affects the surface charge of the colloids and thus inﬂuences the
interactions among the colloids. As mentioned above, the averaged porosity
(0.353) was used for the bed of silica colloids in the mixed media.

(3) The carbon ﬁbers and colloids are uniformly dispersed in the deposit layers.
This is ensured in our experimental approaches. As shown in Section 4.3.5.1,
the carbon ﬁbers were stably and evenly dispersed after the intense sonication.
In addition, immediately before each ﬁltration experiment the feed suspension

was sonicated again for 5 minutes to achieve a well-mixed suspension.

131

(4) In the mixed media, all of the open spaces between the ﬁbers are occupied by
the colloids. This is the assumption made in Ethier’s approach [1], which
ensures that the permeate ﬂow occurs among particles. As the SEM images in
Fig. 4.4 suggests, this assumption was valid for the deposit layers formed with
25 to 100 mg/L ﬁbers with 200mg/L silica colloids (Fig. 4.4(a-c)), but was not
valid for the 200 mg/L carbon ﬁbers with the silica colloids Fig. 4.4(d) where

there were open spaces among the ﬁbers.

Based on these assumptions, Rd , which was theoretically predicted using Eq. (4.16), was

compared with experimental measurement from Eq. (4.6). The results are shown in Fig.

4.6.

From Fig. 4.6, we can see that the model prediction was very good for the deposit layers
from 25 mg/L, 50 mg/L and 100 mg/L carbon ﬁbers with 200 mg/L silica colloids.
However, it became not applicable when the carbon ﬁber concentration was increased to
200 mg/L. This is because there were open spaces among the ﬁbers (Fig. 4.4(d)), which
violated the fourth of the aforementioned assumptions. The open spaces greatly decreased

the hydraulic resistance of the deposit layer.

132

 

 

1 . .1 ..s,. 14 .4., 4 4.

.............

 

P
a

 

E] EXPO” “I Musunmm ..... ............... .1 0.8

E]
0 Model prediction «114-7 Experimental measurement

0 Model prediction

   

   
   

 

 

 

.0
a:
T

I 0.6

  
   

 

 

 

 

 

 

 

 

  

 

: LLJ ' t
0.4 .......... ...... . ...... 0.4
0.2 0.2 '
.2 22’
t z j , (c) . ' (d) :
1 Cl Experimental measurement “ I j 2 i ‘ ‘ ------------ .. ‘P
0 Model prediction ‘ D Experimental measurement
. : 3)“; 1 0 Model prediction
0.3 _ ........................................................ in, ............... a 1.6 3 :

  

 

 

 

 

 

 

Deposition layer resistance (10“ m") Deposition layer resistance no” 111")
. O

0.6 ~ 1.2 eeeee
0.4 J 0.8» ,
. . j ‘ ”1,!
-' .' .7 '. 1 :
0 I... . r . r . . A r A 1 r n A 0 . . . r . A . . r r L
0 20 40 60 80 100 0 20 40 60 80 100

Permeate volume (mL) Permeate volume (mL)

Figure 4.6. Comparison between model prediction and experimental measurement for the
hydraulic resistance of the deposition layer from the experiments with: (a) 25 mg/L

carbon ﬁbers and 200 mg/L silica colloids ( p2 = 0.92); (b) 50 mg/L carbon ﬁbers and
200 mg/L silica colloids ( ,02 = 0.98); (c) 100 mg/L carbon ﬁbers and 200 mg/L silica
colloids (,02 = 0.97); (d) 200 mg/L carbon ﬁbers and 200 mg/L silica colloids.

133

4.5. Conclusions

In this study, we presented a Kozeny-Carrnan—Ethier equation to predict the permeability
of a deposit layer consisting of ﬁbers and particles. The equation was based on the
permeability of the porous bed of silica colloids and the volume fraction of ﬁbers. Dead-
end membrane ﬁltration experiments were conducted to experimentally determine the
permeability of porous media with varied mass fraction of carbon ﬁbers and silica
colloids. The model prediction is in great accordance with the experimental results when
the mass fraction of carbon ﬁbers was less than 50%. When the mass fraction of carbon
ﬁbers exceeded 50%, the model was not applicable because the ﬁbers created some open
spaces, which were not occupied by particles, and thus greatly increased the permeability

of the deposit layer.

134

Acknowledgments

Financial support of this work by the National Water Research Institute (project no. 05-
TM-007) is gratefully acknowledged. We thank Nissan Chemical America Corp. for
supplying the silica suspension, Hydranautics Corp. for the ultraﬁltraion (UF) membrane,
and Applied Sciences, Inc. for the carbon ﬁbers. We also thank Ms. Ewa Danielewicz
from the Center for Advanced Microscopy and Dr. Jue Lu from the Composite Materials
and Structures Center at Michigan State University for their assistance with osmium
coating of samples for SEM imaging and preparation of carbon ﬁber suspensions,

respectively.

135

Nomenclature

ﬁber radius (m'l)

radius of coarse ﬁbers (m'l)

radius of ﬁbers in deposit layers consisting of particles and
ﬁbers (m'l)

membrane ﬁltration area (m2)

feed concentration of silica colloids (kg/m3)

diameter of particles (m'])

a model-dependent function

volumetric permeate flux (ms'l)
volumetric pure water permeate ﬂux (ms'l)

volumetric pure water permeate ﬂux at ﬁltration time t (ms'l)
volumetric pure water permeate ﬂux at cumulative volume V

(ms")

permeability (m2)

permeability of ﬁne ﬁbers (m2)

permeability of porous bed formed by the silica colloids (m2)

permeability of total media (m2)
mass of particles in deposit layers consisting of only particles

(kg)

mass of ﬁbers in deposit layers consisting of particles and
ﬁbers (kg)

mass of particles in deposit layers consisting of particles and
ﬁbers (kg)

hydraulic resistance of membranes to pure water (m'l)
hydraulic resistance of deposit layers (m’l)

speciﬁc hydraulic resistance (m'z)

ﬁltration time (h)

cumulative permeate volume (m3)

volume occupied by coarse ﬁbers (m3)

volume occupied by ﬁbers in deposit layers consisting of
particles and ﬁbers (m3)

volume occupied by particles

total volume of deposit layers consisting of particles and ﬁbers
(m3)

transmembrane pressure (MPa)

136

Greek letters

1” dynamic viscosity of water (kgm'ls'l)

5d thickness of the deposit layer (m)

6 effective porosity of deposit layer

5 p effective porosity of porous bed formed by silica colloids

(15 volume fraction

¢c volume fraction of coarse ﬁbers

¢f volume fraction of ﬁbers in deposit layers consisting of
particles and ﬁbers

pp particle density (kgm'3)

p f ﬁber density (kgm'3)

p2 correlation coefﬁcient

137

Reference

10.

ll.

12.

13.

C. R. Ethier, Flow through mixed ﬁbrous porous materials, AIChE J. 37 (1991)
1227-1236.

G.W. Jackson and DP. james, The permeability of ﬁbrous porous media, Can. J.
Chem. Eng. 64 (1986) 364-374.

P. C. Carman, Fluid flow through a granular bed, Trans. Inst. Chem. Eng. London
15 (1937) 150-166.

J. Happel, Viscous ﬂow relative to arrays of cylinders, AIChE J. 5 (1959) 174-
177.

L. Spielman and Goren S.L., Model for predicting pressure drop and ﬁltration
efﬁciency in ﬁbrous media, Environ. Sci. Technol. 279-286 (1968).

GP Yu and T.T. Soong, A random cell model for pressure drop prediction in
ﬁbrous ﬁlters, Trans. ASME J. Appl. Mech. 301 (1975).

S. Kuwabara, The forces experienced by randomly distributed parallel circular
cylinder or spheres in viscous ﬂow at small Reynolds numbers, J. Phys. Soc. of
Japan 14 (1959) 527-532.

A. S. Sangani and A. Acrivos, Slow ﬂow past periodic arrays of cylinders with
application to heat transfer, Int. J. Multiphase Flow 1982 (1982) 193-206.

P.C. Raynor and D. Leith, Evaporation of accumulated multicomponent liquids
from ﬁbrous ﬁlters, Ann. Occup. Hyg. 43 (1999) 181 - 192.

D. Thomas, P. Penicota, P. Contala, D. Leclerca and J. Vendelb, Clogging of
ﬁbrous ﬁlters by solid aerosol particles Experimental and modelling study, Chem.
Eng. Sci. 56 (2001) 3549 - 3561.

X. Zhang, A. J. Du, P. Lee, D. D. Sun and J. O. Leckie, Ti02 nanowire membrane
for concurrent ﬁltration and photocatalytic oxidation of humic acid in water, J.
Membr. Sci. 313 (2008) 44-51.

C. Yao, X. Li, K. G. Neoh, Z. Shi and E. T. Kang, Surface modiﬁcation and
antibacterial activity of electrospun polyurethane ﬁbrous membranes with
quaternary ammonium moieties, J. Membr. Sci. 320 (2008) 259-267.

Y. Hong, T. Shang, Y. Li, L. Wang, C. Wang, X. Chen and X. Jing, Synthesis

using electrospinning and stabilization of single layer macroporous ﬁlms and
ﬁbrous networks of poly(vinyl alcohol), J. Membr. Sci. 276 (2006) 1-7.

138

14.

15.

16.

17.

18.

19.

20.

21.

Y. Wang and Y. L. Hsieh, Immobilization of lipase enzyme in polyvinyl alcohol
(PVA) nanoﬁbrous membranes, J. Membr. Sci. 309 (2008) 73-81.

R. Gopal, S. Kaur, C. Y. Feng, C. Chan, S. Ramakrishna, S. Tabe and T.
Matsuura, Electrospun nanoﬁbrous polysulfone membranes as pre-ﬁlters:
Particulate removal, J. Membr. Sci. 289 (2007) 210-219.

D. Aussawasathien, C. Teerawattananon and A. Vongachariya, Separation of
micron to sub-micron particles from water: Electrospun nylon-6 nanoﬁbrous
membranes as pre-ﬁlters, J. Membr. Sci. 315 (2008) 11-19.

F. Wang and V.V. Tarabara, Coupled effects of colloidal deposition and salt
concentration polarization on reverse osmosis membrane performance, J. Membr.
Sci. 293 (2007) 111-123.

J. Hermia, Constant pressure blocking ﬁltration laws - application to power-law
non-Newtonian ﬂuids, Trans. IChemE. 60 (1982) 183 -187.

W. R. Bowen, A. Mongruel and P. Williams, Prediction of the rate of cross-ﬂow
membrane ultraﬁltration: A colloidal interaction approach, Chem. Eng. Sci. 51
(1996) 4321-4333.

R. S. F aibish, M. Elimelech and Y. Cohen, Effect of interparticle electrostatic
double layer interactions on permeate flux decline in crossﬂow membrane
ﬁltration of colloidal suspensions: An experimental investigation, J. Colloid

Interface Sci. 204 (1998) 77-86.
V. V. Tarabara, I. Koyuncu and M. R. Wiesner, Effect of hydrodynamics and

solution chemistry on permeate ﬂux in crossﬂow ﬁltration: Direct experimental
observation of ﬁlter cake cross sections, J. Membr. Sci. 241 (2004) 65-78.

139

CHAPTER FIVE

Wang, Fulin; Tarabara, Volodymyr V. (2008). Crystallization of gypsum in the presence
of colloidal silica. I: in a stirred batch system. (In preparation for publication in a journal)

140

CHAPTER FIVE

CRYSTALLIZATION OF GYPSUM IN THE PRESENCE OF
COLLOIDAL SILICA. I: IN A STIRRED BATCH SYSTEM

ABSTRACT

Although both scaling and colloidal fouling in membrane ﬁltration processes have been
intensively studied independently in their respective ﬁelds, there is very limited research
in the context that these two fouling mechanisms occur simultaneously. In practice,
fouling often involves different mechanisms due to the fact that feed waters usually have
complex components. In this work, we studied the effect of silica colloids on gypsum
scaling in a batch system using supersaturated calcium sulfate suspensions of varied
saturation index in the absence and presence of silica colloids. The induction time was
determined based on the light transmittance at 400 nm. Measuring conductivity of the
feed suspensions was used as a convenient technique to determine the crystallization rate
of gypsum. The results showed that the presence of silica colloids retarded the induction
of gypsum crystallization and slightly decreased the nucleation rate. This was presumably
due to the slightly increased critical size for nuclei as a result of the collision between
silica colloids and newly formed nuclei. It was also found that gypsum crystallization rate
was increased in the presence of silica colloids. This was attributed to the fact the silica
colloids changed the morphology of gypsum crystals and resulted in larger crystals, as
revealed by scanning electron microscopy (SEM) images. In the absence of silica colloids,

there were mostly needle-shaped crystals along with a signiﬁcant amount of clusters of

141

smaller crystals. On the contrary, when silica colloids were present, there were some

needle-shaped crystals with much larger plate—shaped crystals.

142

5.1. Introduction

Fouling - the undesired deposition and accumulation of materials on surfaces - is one of
the most serious problems as it lowers the efficiency of many industrial processes. It was
estimated that fouling-related costs correspond to ca. 0.25% of USA’s gross domestic
product (GDP) [1]. In aqueous systems, fouling can be especially severe; membrane-
based separation [2] (reverse osmosis and nanoﬁltration [3, 4], ultra- and microﬁltration
[5, 6], membrane distillation [7, 8], membrane bioreactors [9, 10]) and heat transfer [11]
(such as tubular heat exchangers [12], and plate heat exchangers [13]) technologies suffer
signiﬁcant losses due to this phenomenon. Fouling reduces the quality and quantity of
permeate, increases energy consumption, shortens equipment lifetime, and thus raises

both operational and capital costs.

Because of the complexity of natural waters, fouling is usually caused by a suite of
distinct phenomena, each corresponding to a different fouling mechanism. For instance,
in reverse osmosis (RO) membrane systems, precipitation of sparingly soluble salts
(scaling) is accompanied by biological and particulate fouling. In heat exchange systems,
scaling and particulate fouling are known to occur simultaneously [14]. While
contribution of one type of fouling can be dramatic, there is growing evidence that
interactions between different fouling mechanisms may play an important [9, 14, 15] or

even dominant role [16] in determining the overall system performance.

143

Scaling, or colloidal fouling — the usually undesirable precipitation process that can
severely impact permeate ﬂux and rejection of membranes [l7] and performance of heat
exchangers ([18, 19]) - has been shown to be inﬂuenced by surfactants [20], ionic species
[21, 22], and natural organic matter (NOM) [23]. Surfactants have been shown to have an
effect on gypsum crystallization in terms of induction time, growth efﬁciency, surface
energy and crystal structure [20]. The effects varied with surfactants types. The cationic
surfactant cetyltrimethylammonium bromide (CTAB) decreased the induction time, and
increased the growth efﬁciency and surface energy, while the anionic surfactant had the
contrary effects. The presence of other precipitating salts affected the structure and
strength of gypsum crystals [21]. Polyelectrolytes have been shown to affect gypsum
crystallization and the effect depended on the concentration, molecular weight, and
composition of the polyelectrolytes [24]. Jiang et al. [25] have shown that the effect of
borax on gypsum scaling relied on its concentration: borax expedited gypsum scaling at
low concentration, but retarded it at high concentration. Studying gypsum crystallization
in a nanoﬁltration (NF) membrane system ﬁltrating agricultural drainage water, Le
Gouellec and Elimelech [23] demonstrated that the presence of NOM resulted in shorter
gypsum crystals. It was also found that a single aquatic humic acid could not model the

effect of the complex NOM on gypsum crystallization.

Although both scaling and colloidal fouling have been intensively studied independently
in their respective ﬁelds, there is very limited research in the context that these two
fouling mechanisms occur simultaneously [26]. All studies published to date can be

roughly grouped into two categories: a) fundamental studies on the nucleation and growth

144

of crystals in the presence of colloids and b) practical studies concerning performance of

a speciﬁc technology.

With no research published on the effect of particles on colloidal fouling of membranes;
most of the available knowledge comes from studies on the performance of heat
exchangers. In heat transfer systems, the effect of particles on scaling has been
demonstrated to be complex [27]. Bansal et a1. [19] found that the effect of suspended
particles on scaling depended on the crystallizing ability of particles. Crystallizing CaSO4
particles greatly enhanced gypsum formation because they provided extra nucleation sites.
On the contrary, noncrystallizing alumina particles reduced the gypsum formation. They
believed that this effect was due to 1) particles reduced the strength of crystals deposited
and thus resulted in a higher crystal removal rate, and 2) particles on crystal surface acted
as a distorting agent. Andritsos and Karabelas [14] showed that CaCO3 precipitation rate
was greatly enhanced by ﬁne aragonite (CaCO; polymorph) particles, but was not
affected by silica colloids. McGarvey and Turner [28] examined the interplay between
CaCO3 precipitation and colloidal deposition. It was found although the rate of CaCO3
precipitation was not affected by the silt and hematite (Fe203) particles, the deposition of

the particles was enhanced by CaCO3 precipitation.
The morphology of gypsum crystals is very important in industry. As pointed out in [20],

it is important for gypsum precipitation processes involving subsequent solid-liquid

separation, such as removal of by-product gypsum in phosphoric acid manufacturing

145

process [29]. During the reuse of calcium sulfate dihydrate, it is preferable to form large

plate-shaped crystals [30].

The objective of this work was to study the effect of colloids on the scaling of calcium
sulfate dihydrate (gypsum) in a batch system. Model feed waters were prepared by
suspending silica colloids in calcium chloride and sodium sulfate solutions of different
degrees of supersaturation for gypsum. The effect of colloids on gypsum crystallization
was examined in terms of induction time, rates of nucleation and crystallization, and

morphology of gypsum crystals.

5.2. Background

There are three steps in a crystallization process: achievement of supersaturation,

formation of nuclei (i.e., nucleation), and their growth into larger crystals [31].

The degree of supersaturation in solution is represented by the gypsum saturation index

(Sg) as:

2+ 2-
7 2+ [Ca 17 2- [SO I

g Ksp ’

146

where K Sp is the solubility product for CaSO4-2H20; [Ca2+] and [803-] are molar

concentrations of calcium and sulfate ions in solution, respectively; 7C 2+ and 7802‘
a

are the activity coefﬁcients for calcium and sulfate ions, respectively; K for

5P

CaSO4-2H20 in water at 20 °C has been shown to be 2.623E-4 [32] .

Davis approximation can be used to estimate the activity coefﬁcient as a function of ionic

strength:

I

+x/7

 

log y = —(Az2 1 — 0.21), (5.2)

where z is the ion charge; I is the ionic strength (M); A z 0.5 for aqueous solutions.
Induction time is deﬁned as the time that elapses between the moment the supersaturation
condition is achieved and the moment crystals are detected in the solution. Induction time

is known to be a function of saturation index, agitation intensity, temperature, presence of

impurity, etc. [31, 33] and can be viewed as consisting of three contributions [31, 34]:
find =tr +tn +tg2 (5.3)

where t, is the relaxation time required for the system to achieve a quasi-steady-state

distribution of molecular clusters; tn is the time required for formation of a stable

147

nucleus, which will grow into crystals instead of dissolving into the feed solution [35].

t g is the time for the nucleus to grow to a detectable size.

As the deﬁnition of tg implies, the value of tind depends on the sensitivity of the

technique used for detecting the change in the physical properties of the system due to the
formation of crystals [33, 34]. Two types of techniques - optical and electrochemical —
have been used to determine the induction time for gypsum crystallization. Optical
techniques include monitoring of scattered and transmitted light [35], measurement of
turbidity [20, 36-38], and direct naked-eye observation [39]. Electrochemical techniques
include measuring the conductivity of the feed suspensions [40] or the concentration of
calcium in the solution using calcium selective electrodes [36, 41, 42]. The technique-
speciﬁc value of the induction time is usually deﬁned operationally as the time for one of
the above parameters to change to a certain extent because of the formation of nuclei and

crystals in the solution [33, 34].

According to the classical homogeneous crystallization theory, the homogeneous

nucleation rate (J s) can be estimated as [31, 34, 43]:

AG ~
J S = F exp[— ﬁg], (5.4)

where F is a frequency constant known as the pre-exponential factor and has a

theoretical value of 1030 nuclei/(cm3-s), which has been used for gypsum crystallization

148

[20]; AGcrm-ca] is critical free energy for nucleation; k is the Boltzmann constant, which

has a value of R/N A where R is the gas constant (J/(mol-K)) and N A is the

Avogadro’s number (mol'l); T is the absolute temperature (K).

AGm-n-w, is related to surface energy ( 73) and the critical size( Ram-cal), the minimum

radius required for a stable nucleus, as:

2
47:73 Rcrilical (5'5)

AGcritical = 3 ,

According to the classical homogeneous crystallization theory, the relationship between

the critical size (Rcritical) and the surface energy of the gypsum nuclei is governed by

Gibbs-Thompson equation as [31, 43]:

273Vm

Rcritical = mi (5.6)
g

where Vm is the molecular volume for gypsum (m3/mol). This equation implies that

surface energy tend to increase with increasing size of nuclei.

Combing Eqs. (5.4-5.6), we can get an expression for nucleation rate which has been

used for gypsum crystallization [20, 35] as:

149

 

 

-16 3V2N
J3 = Fexp ”:5 m ’42 . (5.7)
3(RT) (In S g)
The surface energy (ys) can be calculated from the equation below [20, 34]:
1 ﬁriViNAfte)
logtmd = 8+ 3 2 3 , (5.8)
T (logSg) (2.3R)

where B is a dimensionless constant.

It is worthy noticing that although Eqs. (5.4) and (5.5) are based on the assumption that
the crystallization is ideally homogeneous (spontaneous) which is not common in most
laboratory systems, they have been widely used for estimating the properties of gypsum
crystals formed in stirred batch systems [20, 35, 37, 44, 45]. To make our results
comparable to those studies, the two equations were adopted in this study to estimate the

nucleation rate and surface energy of gypsum crystals.

For bulk crystallization, the rate of gypsum crystallization can be expressed as [46]:

dm
—=k C -C , 5.9
dt b( b s) ( )

150

where m is the amount of gypsum formed; kb is the gypsum bulk crystallization rate
(also called diffusion-controlled crystallization constant [46]); Cb is the CaSO4

concentration in bulk solution; C S is the saturation concentration of CaSO4.

5.3. Material and methods

5.3.1. Experimental apparatus

The experimental batch system is shown in Fig. 5.1. A 1,500 mL Pyrex® beaker with a
1,000 mL of feed suspension was placed in a water bath on the top of a multi-stirrer
(model 1286, Lab-Line Instruments Inc., Melrose Park, IL). The temperature of the water
in the bath was maintained at (20 :1: 0.2) °C using a programmable circulating chiller
(model 9512, PolyScience, Niles, IL) equipped with an external resistance temperature
detector. Keeping the feed temperature constant is very important because temperature
inﬂuences not only gypsum crystallization but also measurement of conductivity. The
water in the bath was mixed by two 3 inch magnetic stirrer bars at a rate larger than 200
rpm. The feed suspension in the beaker was mixed by an identical stirrer bar at a rate of
120 rpm. The suspension was continuously circulated by a gear pump (model EW—74013-
20, Cole-Partner, Vernon Hills, Illinois) at a ﬂow rate of 1.6 L/min. A small portion of
the ﬂow was diverted into a ﬂow-through UV-vis sample cell (model 178.710 — QS,
Hellma GmbH & Co KG, Miillheim) for the on-line transmittance measurement using a

spectrophotometer (Multi-Spec 1501, Shimadzu, Kyoto, Japan). The transmittance in the

151

Conductivity meter

 

    

[1 Spectrophotometer _ii-‘l

 

 

  
 

 

 

 

 

 

 

 

 

 

 

A“ . 9 Data acquisition system
i- '(-' " ' E
i
l
O : External temperature sensor
Gear purrp ' ri> ..................
Cooling liquid line I l E
..z
: 20" . .
It}? ‘ . OO— -
Water bath and the multl-stirrer Chiller

Figure 5.1. Schematic of the experimental batch system

152

range of 400 nm to 800 nm was recorded in real-time. The probe (model 013005A,
Thermo Electron Corp. Beverly, MA) of a conductivity meter (model Orion 550, Thermo
Electron Corp.) was placed in the suspension for on-line conductivity measurement. The
data from the conductivity meter and the spectrophotometer were logged to a PC every

30 sec or 2 min.

5.3.2. Regents, silica colloids, and feed suspensions

All reagents were of ACS analytical grade or higher purity (Fisher Scientiﬁc, Pittsburgh,
PA) and were used without further puriﬁcation. The deionized water used in the
experiments was supplied by a commercial ultrapure water system (Lab Five, USFilter
Corp., Hazel Park, MI) equipped with a terminal 0.2 pm capsule microﬁlter (PolyCap,
Whatman Plc., Sanford, ME). The resistivity of the water was greater than 16 MQ-cm.

No buffers were added to the feed water.

Silica colloids (SnowTex-XL, Nissan Chemical America Corp., Houston, TX) were
chosen as the model colloids. Light scattering data (ZetaPALS, Brookhaven Instrument
Corp., Holtsville, NY) showed that the particle size distribution was relatively narrow
(polydispersity factor of 0.074) with a mean hydrodynamic particle diameter of
(55.6 $1.3)nm [47]. Silica colloids were chosen because of the abundance of the
particulate silica in water. The experiments were conducted by adding CaC12-2H20 and

Na2S04 separately. The amount of CaClz°2H20 and Nast4 added is shown in Table 5.1.

153

Table 5.1. Composition of model solutions

 

Saturation index (S g) CaClz-2H20, g/L Na2804, g/L Ionic strength,

 

mM
2.5 3.80 3.67 155.1
3.0 4.35 4.21 177.8
4.0 5.38 5.20 219.6
5.0 6.24 6.02 254.5
6.0 7.09 6.85 289.4

 

154

5.3.3. Characterization of the gypsum crystals

At the point in the crystallization experiment that corresponded to the 70% decrease in
transmittance, 10 mL of the feed suspension was taken from the feed vessel using a
syringe and was ﬁltered through a polycarbonate track-etch membrane (Nuclepore®,
Whatman Plc.) with a 0.1 pm nominal pore size. Immediately after the ﬁltration, 5 mL of
methanol was passed through the membrane to remove the residual suspension on
membrane surface to prevent formation of new gypsum crystals. The membrane was then

dried in air at room temperature.

A ca. 1 cm2 piece of the dried membrane was cut out and coated with osmium for 30 s
using an osmium coater (Neoc-AN, Meiwa Shoji Co. Ltd., Kyoto, Japan). Scanning
electron microscopy (SEM) images of the membrane surface were recorded using Hitachi
S-47OOII Field Emission scanning electron microscope (Hitachi High Technologies
America, Inc. Pleasanton, CA). The microscope was operated in the ultra-high resolution

mode with an accelerating voltage of 15 kV and an emission current of 10 11A.

The size distribution of gypsum crystals in the feed tank was measured using a particle
Sizer (Malvem Mastersizer 2000, Malvem Instruments Ltd, Worcestershire, UK). When
the transmittance of the feed suspension decreased by 70%, a 150 mL sample was taken
from the feed vessel and placed into the Hydro SM® small volume sample dispersion
module of the particle sizer. The initial feed suspensions were used as the background to
remove the effect of ions and silica colloids. The sample in the dispersion module was

stirred at a speed of 2,000 rpm. Data reported in this paper represent the average size

155

distribution computed based on the three data sets automatically recorded with a time

interval of 5 s between them.

5.4. Results and discussion

5.4.1. Determination of induction time

In this work, transmittance, calcium ion concentration, and conductivity were used and

compared for determining the induction time.

In our experiments, transmittance of the feed suspensions in the range of 400 nm to 800
nm was monitored in real-time. We used the 5% decline in transmittance as the starting
time of gypsum crystallization as proposed by Lancia and co-workers’ [35]. We found
that, a lower wavelength led to a relatively lower induction time (Fig.SS.I in
Supplementary Data), which indicated that transmittance at a lower wavelength was more
sensitive for detecting the change in the properties of feed suspension due to nucleation.

For example, in the experiment with a supersaturated suspension (S g = 5) in the absence

of silica colloids, the results showed that the induction time was 23 min, 27 min, and 29
min based on the transmittance at 400 nm, 600 nm, and 800 nm, respectively. The

induction time reported in this work was based on the transmittance at 400 nm.

A calcium selective electrode (Orion 97-20 Ionplus®, Thermo Electron Corp.) was also
used to determine the induction time. The electrode measures the calcium potential

difference across the sensing membrane between the target solution and the reference

156

 

 

 

 

 

 

 

 

 

 

 

 

 

    

    
  

 

 

 

 

 

 

 

 

 

 

 

 

 

-37 $ a 1 4
1.35
-38
s 1.3
E. 9
_ 1.25 3
e -39 e
c 8
0
out 1.2 <
8 e
5 '40 1.15%:
0 -
T. O
o 1.1 3
-41
1.05
_42 c. ................................................................................... 1
Time (min)
100, .................. j ............ i ____________ . . .......... 13.8
.- . , (b)
'§?«(((((“?«( 2 13 7
80 '
:3 13.6 g
g 60 _ . E
,‘3 El Transmittance 13.5 <
E - - 5
g 40 _ ........................................... K4 .................................. g i a
3 93513.4 :4
.3.
20 _ ............................................... ifs ................................. 13‘s
0 1 1 l 1 ...-T.mil”iililiiiliiiiiliiiilililiiiiiil’.‘ 132
0 10 20 30 40 50 60 70
Time (min)

Figure 5.2. Comparison of different techniques for measuring gypsum induction time
(S g = 5; no silica colloids): (a) calcium potential and conductivity measurement for

determining the induction time; (measurements were for the diluted permeate of the UP
membrane ﬁltration); (b) transmittance and conductivity measurement for determining
the induction time (measurements were for the feed suspension).

157

solution. The calcium potential is inﬂuenced by the ionic strength and temperature of the
target solution. The measurement required an equilibration time of ca. 25 min, which was
too long for in-situ measurement in our experiments. Thus, an experiment described
below was conducted to determine the change of calcium ion concentration in the feed
suspension. A gypsum crystallization experiment was carried out at a saturation index of
5 in the batch system without silica colloids. Every 5 minutes, a 5 mL suspension was
taken from the feed suspension and ﬁltered through an ultraﬁltraion (UF) membrane
using a stirred stainless steel dead-end ﬁltration cell (HP 4750, Sterlitech Corp., Kent,
WA) at a transmembrane pressure of 150 psi (1.03 MPa) to remove the possible nuclei
formed in the feed suspension. The UF membrane (Omegam, Pall Corp., East Hills, NY)
has a nominal molecular weight cut-off (MWCO) of 1,000 Da and a CaSO4 rejection of
ca. 4.6% at this transmembrane pressure. A 2 ml permeate was collected and diluted 15
times with deionized water. The dilution ensured that 1) the volume was enough for the
measurement, and 2) the diluted permeate solution was much less saturated, and no new
nuclei formed during the measurement. Both the calcium potential and conductivity of
the diluted permeate solution were measured. The results are shown in Fig. 5.2(a), from
which it can be seen that neither calcium potential nor conductivity of the diluted

permeate could clearly indicate the induction time.

The above experiment was repeated with continuous measurement of the conductivity
and transmittance of the feed suspension, as reported in Fig. 5.2(b). After 23 min,
transmittance decreased by 5%. However, conductivity did not decrease until 38 min.

This comparison indicates that, in our study, transmittance was more sensitive than

158

conductivity for determining the induction time. One possible reason for this is that, at
the beginning of nucleation, light transmittance could detect the extremely small nuclei,

which do not cause conductivity decline because they are still dissolved in the suspension.

5.4.2. Effect of silica colloids on the induction of gypsum crystallization

The transmittance and conductivity of the feed suspensions as a function of time at

different saturation indices were shown in Fig. 5.3.

It was found that the presence of silica colloids resulted in a longer induction time. This
effect was more evident at a low saturation index (F ig. $52 in Supplementary Data). The
induction time with or without silica colloids at different saturation indices was shown in

Fig. 5.4 (a).

There is possibly a certain amount of chemicals, such as surfactants, in the stock
suspension of the silica colloids. An eXperiment was conducted to test the effect of these

chemicals on gypsum nucleation. First, 0.095 mL (the same amount as used in our
experiments) of the stock silica suspension was taken and added into to 50 mL deionized
water. Second, the diluted suspension was ﬁltered through a 10 kDa ultraﬁltration (UF)
membrane to obtain a solution free of particles. The 10 kDa membrane was chosen to
ensure a) the complete removal of silica particles and b) zero rejection of all chemicals

present

159

 

 

1:] No colloids

 

80 _,_
0 With colloids Ea
é » . ., ‘_“.'
3 60 ..
i
E
r: 40
i!
'—

N
O

 

 

 

 

 

 

 

 

0
16
15
15 o
"’ o
35 14 a
= c
3 1:1 No Colloids 14 3
° 13 . . 5.
2I 0 With coIIOIds .2
E 13 ’3‘
2 12 91
in O
5, 12 '3
a
E 11 11 g
‘5
1E 10 g
2.
8 10 a.
9
9
1 1000

100
Time (min)

Figure 5.3. The transmittance (a) and conductivity (b) of the feed solutions as a function
of time (transmittance was reported at a reduced frequency)

160

in the silica suspension. Third, a gypsum crystallization (S g = 4) experiment was

conducted with the solution added into the feed suspension. The recorded induction time
was compared to that in the experiment without adding the solution, and no evident
difference was observed. This experiment indicates that the observed effect on gypsum
crystallization was due to the colloids and not due to the chemicals in the stock

suspension.

It was also found that after conductivity started to decline, it followed a straight line 9Fig.
5.3). This was used for determining the gypsum crystallization rate, as discussed in

Section 5.4.4.

Based on the comparison of Fig. 5.3(a) and Fig. 5.3(b), it can be seen that transmittance
measurement was more sensitive for measuring induction time than conductivity

measurement, as also indicated in Fig. 5.2(b).

Based on Eq. (5.6) and the surface energy, Ram-ca, for different saturation indices is

estimated and shown in Table 5.2.

We suspect that the increase in critical size in the presence of silica colloids in our study
was due to the collision between the silica colloids and newly formed nuclei. The
collision made newly formed nuclei more unstable. As a result, nuclei had to grow into a

larger size to stay stable in the suspension instead of dissolving back into the suspension.

161

Table 5.2. Estimated critical size of nuclei in the absence and presence of silica colloids

 

 

Sg Critical size of gypsum nuclei (Rcrmcal , nm)
In the absence of colloids In the presence of colloids

3 0.609 0.628

4 0.482 0.496

5 0.416 0.427

6 0.373 0.383

 

162

This increase in Rem-”ca, retarded the induction, resulting in a larger induction time and a

higher surface energy.

5.4.3. Effect of silica colloids on the surface energy and nucleation rate of gypsum

crystallization

Based on Eq. (5.8), log(t) was plotted against 1/(log S g )2 at different saturation indices
(Fig. $53 in the Supplementary Data). From the slope, the surface energy was calculated
to be 10.9 :t 1.2 J/(mol-K) ( p2 = 0.99) for the crystals formed in the absence of silica

colloids and 11.2 i 1.1 J/(mol-K) ( p2 = 0.99) for those in the presence of 50 mg/L silica

colloids. This indicates that the presence of silica colloids slightly increased the surface

energy of crystals.
Based on Eq. (5.7), the theoretical nucleation rates of gypsum crystallization at different

saturation indices were calculated and presented in Fig. 5.4(b). The presence of silica

colloids slightly decreased the nucleation rates.

163

 

\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

    

    

 

 

,\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

 

 

 

 

 

 

   

  

 

 

 

  

 

 

       

 

 

E] N il Id
With ll id
4
au 0

olloids on (a) induction time, (b) nu

 

22222

   

 

 

 

 

 

 

 

 

 

 

 

0

mo. 000000
o. 0.

mosaic—... “$588 :23 22.2 r... .538. 2:3..mnbo

rate, and (c)

cleation

te

5.4. Effect of silica c

Figure
crystallization ra

5.4.4. Effect of silica colloids on the gypsum crystallization rate

In our experiments, the feed suspension volume is always 1 L, which enables us to

change the Eq. (5.9) to:

dC .
gt-=kb(Cb-Cs), (5.10)

where k'b is the crystallization rate (min'l), which has been shown to be dependent on the

saturation index [48, 49].

Transmittance can not be used to determine the change in the concentrations because it
depends on both the number and size of the nuclei and crystals [50]. In our experiments,
the decline in conductivity after induction time was used to calculate the crystallization

rate. The measurement was easy and straightforward. The decline of conductivity is
proportional to the decrease of [Ca 2+] and [802'] , due to the formation of gypsum

crystals. Thus the decline rate of conductivity (calculated from the slopes of the lines for
conductivity versus time (Fig. 5.3)) can be used for determining dC/ dt based on a

previously recorded standard curve of conductivity and CaSO4 concentration.

Gypsum crystallization rates at different saturation indices in the absence and presence of
silica colloids were shown in Fig. 5.4(c). At all of the saturation indices, the presence of
silica colloids resulted in a higher gypsum crystallization rate, with the difference being

larger for higher degrees of saturation. We suspect that this effect was due to that fact that

165

silica colloids inﬂuenced the epitaxial growth and morphology of gypsum crystals, which

is discussed in Section 5.4.6.

5.4.5. Effect of silica colloids on crystal size distribution

Laser diffraction-based analysis of gypsum crystal size distribution has been used by
Mahmoud [20]. However, it should be noticed that the laser analyzer’s measurement is
usually performed assuming the particles are spherical. Because the gypsum crystals are
needle-shaped or plate-shaped, extreme care should be taken when one reports the size
data based on this type of measurement. Thus, in this study, it was used only for

qualitative analysis of gypsum crystal size distribution.

Our measurement for crystal size distribution is shown in Fig. 5.5, which indicates there
was a difference in the crystal size distribution in the absence and in the presence of the

silica colloids.

5.4.6. Effect of silica colloids on the morphology of gypsum crystals

Scanning electron microscopy (SEM) imaging was used for a direct observation of the
morphology of the gypsum crystals. As shown in Section 5.3.3, when transmittance
decreased by 70% at a saturation index of 5, 10 mL of feed suspensions was taken from

the feed suspension. The samples were taken at 40 min and at 44 min for the experiments

166

  
       
   

No Colloids

l

l

      
 
     

With Colloids

M (.0 4:8 0'1
._ _W

a 1 l . l . ..
o c _ a i 1__; TT4:_.__
> i ' i ' MHz? -
lELL-11 U 1 ..--..1 10 103 1000 $00
Particle size (pm)

Figure 5.5. Size distribution of gypsum crystals formed in the absence and in the presence
of 50 mg/L silica colloids. Conditions: S g = 5, transmittance = 50% (Laser obscuration

for the measurement without and with colloids is 9.7% and 11.3%, respectively.)

167

without and with silica colloids, respectively. This choice of sampling at the same
transmittance ensured that the crystals had the same time periods of growth. After the
samples were ﬁltered through the 0.1 pm polycarbonate (PC) track-etch membrane, 5 mL
of methanol was used to remove the residual water on the membrane surface, which
prevented further growth of the crystals on membrane surface. This approach has
increased the comparability of the morphology of crystals for identifying the effects of

silica colloids.

The SEM images in the absence and presence of silica colloids are shown in Fig. 5.6.

5.4.6.1. Length-to-width ratio

From Fig. 5.6, it can be seen that, in the absence of silica colloids, most of the crystals
were needle-shaped with a typical length-to-width ratio of ca. 20. In contrast, when the
silica colloids were present, most of crystals were plate-shaped, with a typical length-to-
width ratio of ca. 2. There were still some needle-shaped crystals, the structure of which

was not uniform in terms of length-to-width ratio.

There were a signiﬁcant amount of clusters of smaller crystals (Fig. 5.5 (a-d)) observed in
the silica colloid-free feeds. Having observed the needle-shaped crystals only at later
stages of growth, we consider smaller crystals to correspond to earlier stages in the
development of needle-shaped crystals. The clusters of smaller crystals were not found in

feeds containing silica colloids.

168

5.4.6.2. Implications of epitaxial growth of gypsum for crystal morphology

It was reported that the presence of cetyltrimethylamonium bromide (CTAB) slightly
increased the surface energy of gypsum crystals, whose length-to-width ratio was smaller
compared to those formed with no additives [20]. A similar result was observed
elsewhere [51]: 0.1 g/L nitrilotrimethylenephosphonic acid (NTMP) increased the surface
energy of gypsum crystals by ca. 50% and resulted in the formation of much larger plate-
shaped crystals. Based on these studies, a tentative conclusion can be made that gypsum
crystals with a higher surface energy tend to form larger and plate-shaped crystals,
compared to those with a lower surface energy at the same conditions. In our
experiments, however, the presence of silica colloids slightly increased the surface

energy of crystals, as shown in Section 5.4.3.

We suspect that the formation of large plate-shaped crystals in the presence of silica
colloids was due to the effect of the colloids on the epitaxial growth of the gypsum
crystals. As discussed in the references (e.g. [46] and [43]), the three key growth sites,
kinks, steps, and terraces, dominate the crystal growth process. The size of the crystals
depends on the growth rates of the three sites while the structure (shape) depends on the
relative importance between the three growth rates. The presence of the silica colloids
was expected to inﬂuence the mixing condition near the crystal surface [52]. We suspect

that the resultant change in local mixing enhanced the surface ﬂux of the crystals and

169

 

Figure 5.6. SEM image of gypsum crystals formed in the absence (a-d) and in the
presence (e-h) of 50 mg/L silica colloids. Conditions: S g = 5, transmittance = 50%)

170

thus created more kinks and steps. This effect had two immediate consequences on
crystal growth: 1) it increased the overall crystal growth rate, as we observed in
conductivity measurement, and 2) it changed the epitaxial growth of the crystals and led

to large, plate-shaped crystals.

5.5. Conclusions

The effect of silica colloids on gypsum scaling was studied in a stirred batch system.
Three techniques, light transmittance, conductivity and calcium ion concentration
measurement, were compared for determining induction time. It was found that light
transmittance was the most sensitive. The presence of silica colloids retarded the
induction of gypsum crystallization and slightly decreased the nucleation rate. This was
presumably due to the slightly increased critical size for nuclei, as a result of the collision
between silica colloids and newly formed nuclei. The silica colloids increased the
gypsum. crystallization rate. This was attributed to the fact that the silica colloids changed
the morphology of gypsum crystals. In the absence of silica colloids, there were mostly
needle-shaped crystals, along with a signiﬁcant amount of clusters of smaller crystals. On
the contrary, when silica colloids were present, there were fewer needle-shaped crystals

with much larger plate-shaped crystals.

171

Acknowledgments

Financial support of this work by the National Water Research Institute (project no. 05-
TM-OO7) is gratefully acknowledged. We thank Pall Corp. for providing the UP
membrane samples and Nissan Chemical America Corp. for supplying the silica
suspension. We thank Ms. Ewa Danielewicz from the Center for Advanced Microscopy

at Michigan State University for her assistance with osmium coating of SEM samples.

172

Nomenclature

A dimensionless coefﬁcient

B dimensionless constant

C gypsum concentration (mol/L)

C b gypsum concentration in bulk solution (mol/L)
C s saturation concentration of gypsum (mol/L)
F frequency constant

f (9) correction factor

I ionic strength (M)

J3 nucleation rate (nucleus/(cm3-s))

k Boltzmann constant (1 .38>< 10'23 UK)

kb gypsum bulk crystallization rate (g'L/(min-mol»
k}, gypsum crystallization rate (min'l)

Ksp solubility product

m mass of gypsum crystals (g)

N A Avogadro’s number (mol'l)

R gas constant (J/(mol-K))

Ram-ca] critical size for nuclei (m)

S g saturation index

T absolute temperature (K)

tind induction time (min)

V m molecular volume (m3/mol)

2 ion charge

AGcrm-cal critical free energy for nucleation (J)

Greek letters

7 activity coefﬁcient
75 activity coefﬁcient
p2 correlation coefﬁcient

173

Reference

10.

ll.

12.

S. J. Pugh, G. F. Hewitt and H. Muller-Steinhagen, Fouling during the use of
seawater as coolant - the development of a user guide, Heat Transfer Engineering
26 (2005) 35 - 43.

Committee Report: Recent advances and research needs in membrane fouling, J.
Am. Water Works Assoc. 97 (2005) 79-89.

R. Y. Ning, T. L. Troyer and R. S. Tominello, Chemical control of colloidal
fouling of reverse osmosis systems, Desalination 172 (2005) 1.

W. Song, V. Ravindran, B. E. Koel and M. Pirbazari, Nanoﬁltration of natural
organic matter with H202/UV pretreatment: fouling mitigation and membrane
surface characterization, J. Membr. Sci. 241 (2004) 143.

S. Sethi and MR. Wiesner, Modeling of transient permeate flux in cross—ﬂow
membrane ﬁltration incorporating multiple particle transport mechanisms, J.
Membr. Sci. 136 (1997) 191-205.

T. Carroll, S. King, S. R. Gray, B. A. Bolto and N. A. Booker, The fouling of
microﬁltration membranes by NOM after coagulation treatment, Wat. Res. 34
(2000) 2861.

M. Gryta, M. Tomaszewska, J. Grzechulska and A. W. Morawski, Membrane
distillation of NaCl solution containing natural organic matter, J. Membr. Sci. 181
(2001) 279.

Surapit Srisurichan, Ratana Jiraratananon and A. G. Fane, Humic acid fouling in
the membrane distillation process, Desalination 174 (2005) 63.

R. S. Trussell, R. P. Merlo, S. W. Herrnanowicz and D. Jenkins, The effect of
organic loading on process performance and membrane fouling in a submerged
membrane bioreactor treating municipal wastewater, Wat. Res. 40 (2006) 2675.

S. Rosenberger, C. Laabs, B. Lesjean, R. Gnirss, G. Amy, M. Jekel and J. C.
Schrotter, Impact of colloidal and soluble organic material on membrane

performance in membrane bioreactors for municipal wastewater treatment, Wat.
Res. 40 (2006) 710.

J. Taborek, T. Aoku, R. B. Ritter, Paeln J.W. and JG. Knudsen, Fouling - the
Major Unresolved Problem in Heat Transfer, Chem. Eng. Prog. 68 (1972) 69 - 78.

D. Butterworth, Design of shell-and-tube heat exchangers when the fouling
depends on local temperature and velocity, Appl. Therm. Eng. 22 (2002) 789.

174

13.

14.

15.

l6.

17.

18.

19.

20.

21.

22.

23.

24.

25.

S. Grandgeorge, C. Jallut and B. Thonon, Particulate fouling of corrugated plate
heat exchangers. Global kinetic and equilibrium studies, Chem. Eng. Sci. 53
(1998) 3050.

N. Andritsos and A.J. Karabelas, The Inﬂuence of particulates on CaCO3 scale
formation, J. Heat Transfer 121 (1999) 225-227.

S. Lee, J. Cho and M. Elimelech, Combined inﬂuence of natural organic matter
(NOM) and colloidal particles on nanoﬁltration membrane fouling, J. Membr. Sci.
262 (2005) 27—41.

F. Wang and V.V. Tarabara. Coupled effects of scaling and colloidal deposition
on the performance of reverse osmosis membranes. in The 233rd ACS National
Meeting. March 2007. Chicago, IL.

M. Uchymiak, A. Rahardianto, E. Lyster, J. Glater and Y. Cohen, A novel R0 ex
situ scale observation detector (EXSOD) for mineral scale characterization and
early detection, J. Membr. Sci. 291 (2007) 86.

A. Karabelas, Scale formation in tubular heat exchangers--research priorities,
International Journal of Thermal Sciences 41 (2002) 682.

B. Bansal, H. Muller-Steinhagen and X. D. Chen, Effect of suspended particles on
crystallization fouling in plate heat exchangers, J. of Heat Transfer, Transactions
ASME 119 (1997) 568.

M. H. H. Mahmoud, M. M. Rashad, I. A. Ibrahim and E. A. Abdel-Aal, Crystal
modiﬁcation of calcium sulfate dihydrate in the presence of some surface-active
agents, J. Colloid Interface Sci. 270 (2004) 99-105.

R. Sheikholeslami, Mixed salts -- scaling limits and propensity, Desalination 154
(2003) 117.

Z. Amjad and J. Hooley, Inﬂuence of polyelectrolytes on the crystal growth of
calcium sulfate dihydrate, J. Colloid Interface Sci. 111 (1986) 496.

Yann A. Le Gouellec and Menachem Elimelech, Calcium sulfate (gypsum)
scaling in nanoﬁltration of agricultural drainage water, J. Membr. Sci. 205 (2002)
279.

M. Okazaki and S. Kimura, Scale formation on reverse-osmosis membranes, J.
Chem. Eng. Jpn. 17 (1984) 145.

W. Jiang, H. Pan, J. Tao, X. Xu and R. Tang, Dual roles of borax in kinetics of
calcium sulfate dihydrate formation, Langmuir 23 (2007) 5070-5076.

175

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

R. Sheikholeslami, Composite fouling of heat transfer equipment in aqueous
media - a review, Heat Transfer Engineering 21 (2000) 34 - 42.

D. Hasson. Progress in precipitation fouling: a review, in:Understanding Heat
Exchanger Fouling and its Mitigation, T.R. Bott et al. (Eds). 1999: Begell House,
New York.

GB. McGarvey and CW. Turner, Synergism models for heat exchanger
fouling mitigation. In: T.R. Bott et al.Understanding Heat Exchanger Fouling and
its Mitigation, Begell House, New York, (1999) 155-161.

M. M. Rashad, H. M. Baioumy and E. A. Abdel-Aal, Structural and spectral
studies on gypsum crystals under simulated conditions of phosphoric acid

production with and without organic and inorganic additives, Cryst. Res. Technol.
38 (2003) 433 - 439.

Y. Kojima and T. Yasue, Synthesis of large plate-like gypsum dihydrate from
waste gypsum board, J. Euro. Ceram. Soc. 26 (2006) 777-783.

J. W. Mullin, Crystallization. 4th ed. 2001, Oxford: Butterworth-Heinemann.

D.K. Nordstrom, L.N. Plummer, D. Langmuir, E. Busenberg, H.M. May, B.F.
Jones and D.L. Parkhurst, Revised chemical equilibrium data for major water-
mineral reactions and their limitations, in Bassett, R.L. and Melchior, D. eds.,

Chemical modeling in aqueous systems 11: Washington DC, American Chemical
Society Symposium Series 416, Chapt. 31, (1990) 398-413.

0. S6hnel and J. W. Mullin, A method for the determination of precipitation
induction periods, J. Crystal Growth 44 (1978) 377 - 382.

O. Schnel and J. W. Mullin, Interpretation of crystallization induction periods, J.
Colloid Interface Sci. 123 (1988) 43-50.

A. Lancia, D. Musmarra and M. Prisciandaro, Measuring induction period for
calcium sulfate dihydrate precipitation, AIChE J. 45 (1999) 390 - 397.

W.Y. Shih, K. Albrecht, J. Glater and Y. Cohen, A dual-probe approach for
evaluation of gypsum crystallization in response to antiscalant treatment,
Desalination 169 (2004) 213.

S. He, J. E. Oddo and M. B. Tomson, The inhibition of gypsum and barite

nucleation in NaCl brines at temperatures from 25 to 90[degree sign]C, Appl.
Geochem. 9 (1994) 561.

176

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

D. Hasson, A. Drak and R. Semiat, Inception of CaSO4 scaling on RO membranes
at various water recovery levels, Desalination 139 (2001) 73 - 81.

W. H. Lam and K. M. Ng, Diastereomeric salt crystallization synthesis for chiral
resolution of ibuprofen, AIChE J. 53 (2007) 429 - 437.

P.G. Klepetsanis and PG. Koutsoukos, Kinetics of calcium sulfate formation in
aqueous media: effect of organophosphorous compounds, J. Crystal Growth 193
(1998) 156 -163.

Y.A. Le Gouellec and Elimelech. M., Control of calcium sulfate (gypsum) scale
in nanoﬁltration of saline agricultural drainage water, Environ. Eng. Sci. (2002)
387-397.

A. Hina, G. H. Nancollas and M. Grynpas, Surface induced constant
composition crystal growth kinetics studies. The brushite—Gypsum system, J.
Crystal Growth 223 (2001)213 -224.

AS. Myerson, Handbook of Industrial Crystallization. 2nd ed. 2002, Boston:
Butterworth-heinemann. 33-63.

F. Alimi, H. Elﬁl and A. Gadri, Kinetics of the precipitation of calcium sulfate
dihydrate in a desalination unit, Desalination 158 (2003) 9.

T. A. Hoang, H. M. Ang and A. L. Rohl, Effects of temperature on the scaling of
calcium sulphate in pipes, Powder Technol. 179 (2007) 31.

G.H. Nancollas, The growth of crystals in solution, Adv. Colloid Interface Sci. 10
(1979) 215-252.

P. Wang and V.V. Tarabara, Coupled effects of colloidal deposition and salt
concentration polarization on reverse osmosis membrane performance, J. Membr.
Sci. 293 (2007) 111-123.

B. R. Smith and F. Sweett, The crystallization of calcium sulfate dihydrate, J.
Colloid Interface Sci. 37 (1971) 612-618.

S. T. Liu and G. H. Nancollas, Linear crystallization and induction-period studies
of the grth of calcium sulphate dihydrate crystals, Talanta 20 (1973).

T. Naganurna and Y. Kagawa, Effect of particle size on light transmittance of

glass particle dispersed epoxy matrix optical composites, Acta Mater. 47 (1999)
4321 - 4327.

177

51.

52.

M. Prisciandaro, E. Olivieri, A. Lancia and D. Musmarra, Gypsum precipitation
from an aqueous solution in the presence of nitrilotrimethylenephosphonic acid,
Ind. Eng. Chem. Res. 45 (2006) 2070 - 2076.

E. Meiburg, E. Wallner, A. Pagella, A. Riaz, C. Haertel and F. Necker,

Vorticity dynamics of dilute two-way-coupled particle-laden mixing layers, J.
Fluid Mech. 421 (2000) 185—227.

178

SS. SUPPLEMENTARY DATA

    
 
  
  

100

 

((((((((((((((((

     

 

 

 

 

 

 

 

‘ :«o; :«(r r; , ,
ﬁr, g
80 ._ ............................................... ........................... _
. Z
A .............................................. _<>._: ............................ _
.\° * <2):
8 60 ' . '0'?) ........................... 3
g D Wavelength = 400 nm %% (2) g
E O Wavelength=600 nm ----- [j-Cgs‘gw-W-f ------------ ~
«E, O Wavelength = 800 nm Cl 0 1
a: 40 , . . -.
e 5 . -
'- _ ............. a ...........................................
20 L .............................................
r ............................................
0 L l A
0 10 20 30
Time (min)
Figure 55.1. Transmittance as a function of time at three wavelengths (S g = 5; no

particles)
Fig. S5.1 shows that the induction time was 23 min, 27 min, and 29 min based on the

transmittance at 400 nm, 600 nm, and 800 nm, respectively. The induction time reported

in this work was based on the transmittance at 400 nm.

179

 

 

 

 

 

 

 

 

10.2

9
Asuaé 3.28

9

9.00

9.

 

....................

 

.........................

S

 

With colloids

D No colloid
O

 

 

_ u

........................................

 

.............................

 

 

 

0 0
6 4

3% 355538...

b
0
8

2

300 400
Tlme (min)

200

100

Figure $5.2. Effect of particles on transmittance and conductivity (gypsum
crystallization rate) at S g = 3, (a) Conductivity, (b) Transmittance

180

 

3 I I I I I I l
E—y= 0. 49068 + o. 41353x R= 0.99292

 

 

 

 

 

 

 

 

----- y= 0.48498 + o. 43729x R= 0.99655
2.5 I ........... . ......... _.
-B— No colloids .0
2,0 --O--With colloids ‘ i ' '
c : T :
:‘I 2 _ ..................................................... , .............................. J
U?
2
1.5 —————— . -- ......... ........... .......... ......... _
1 L 1 l 1 l 1 I
1 1.5 2 2.5 3 3.5 4 4.5 5
1l(logSg)2

Figure 85.3. The plots for calculating the surface energy of the crystals

181

CHAPTER SIX

Wang, Fulin; Tarabara, Volodymyr V. (2008). Crystallization of gypsum in the presence
of colloidal silica. II: in a reverse osmosis membrane system. (In preparation for
publication in a journal)

182

CHAPTER SIX

CRYSTALLIZATION OF GYPSUM IN THE PRESENCE OF
COLLOIDAL SILICA. II: IN A REVERSE OSMOSIS MEMBRANE
SYSTEM

Abstract

Gypsum scaling usually occurs along with colloidal fouling in reverse osmosis (R0)
membrane systems treating brackish surface water and seawater, because pretreatment
processes vary seasonally and can not guarantee a complete removal of colloidal particles
in the R0 feed waters. This work presents the ﬁrst study on the combined colloidal
fouling and gypsum scaling in R0 membrane systems. The combined fouling was
observed to have a signiﬁcant synergistic effect on the performance of R0 membranes. In
experiments with less saturated suspensions, gypsum scaling enhanced colloidal
deposition and the combined fouling resulted in a much faster permeate ﬂux decline,
which was larger than the sum of the ﬂux decline from colloidal fouling and that from
gypsum scaling alone. In experiments with supersaturated suspensions, the presence of
silica particles retarded induction time but increased the crystallization rate. When
transmembrane pressure was applied, crystals that formed on the membrane surface
through surface crystallization in the early stage greatly shortened the induction time and

increased the rate of gypsum crystallization.

183

6.1. Introduction

The decreasing water supply and the growing demand for water have led many water
utilities to use alternative water sources, such as brackish surface water and seawater,
with elevated levels of contaminants and salts [1, 2]. Meanwhile, the increasingly
stringent water standards, such as Disinfectants/Disinfectants Byproducts Rule and Long
Term 2 Enhanced Surface Water Treatment Rule, require a higher removal of
contaminates to ensure human health [3]. These two factors have raised the popularity of
reverse osmosis (R0) membrane ﬁltration, which can serve as a reliable means for
removing microorganisms, organics, precursors of disinfection by-products (DBPs),
colloids, and other suspended matter of concern [3, 4]. However, membrane fouling, such
as gypsum scaling and colloidal fouling, decreases the quality of product water, increases

operating cost, and thus impedes a wider application of R0 membrane ﬁltration.

Scaling ﬁ'om sparingly soluble salts, such as carbonate and sulfate, is a serious type of
fouling for R0 membranes. Scaling occurs when the solubility of the salts is exceeded at
a high recovery rate. Scales formed on a membrane surface reduce permeate ﬂux, shorten
membrane life and are difﬁcult to be removed. Carbonate scaling can be controlled by
decreasing the pH of the feed water. Calcium sulfate dihydrate (gypsum), however, is

insensitive to pH, and can not be controlled by adjusting it.

The effect of crossﬂow rate and system conﬁguration on gypsum scaling in membrane

ﬁltration systems has been extensively examined. Lee and Lee [5] demonstrated that in

184

nanoﬁltration (NF) and R0 membrane systems, the gypsum scaling rate was decreased at
a higher crossﬂow rate and a lower transmembrane pressure. Lee and Lueptow [6] used a
rotating R0 membrane system to reduce the gypsum scaling. These studies achieved a
reduction in gypsum scaling by decreasing the effect of concentration polarization, which
resulted in a lower CaSO4 concentration at the membrane surface. Cohen’s research
group [7] demonstrated a progressive axial development of gypsum crystals along the
membrane surface, consistent with the increase in concentration polarization. Also, the

membrane surface coverage by gypsum increased at a higher gypsum saturation index [7].

A substantial amount of suspended particles are present in seawater [8, 9], brackish water
and wastewater [10, 11]. Pretreatment for R0 membrane systems varies seasonally [12-
14], and can not guarantee a complete removal of colloidal particles in the R0 feed
waters [8, 9, 13, 15-17]. The colloidal particles can adhere to the inner pores of porous
membranes and/or form a deposit layer on the membrane surface. The so-called colloidal
fouling increases the hydraulic resistance to the permeate ﬂow and thus decreases
permeate ﬂux. Numerous studies have been conducted on the effects of colloidal fouling

on R0 membrane performance (e. g. [4] and the references therein).

In practice, gypsum scaling usually occurs along with colloidal fouling in R0 systems
treating brackish surface water or seawater. Although the pilot R0 membrane ﬁltration
system treating brackish the Colorado River water had a microﬁltration (MF) membrane
system as pretreatment, the R0 membrane was simultaneously fouled by colloidal

fouling and scaling [18]. In another study, an R0 membrane ﬁltration system in a

185

brackish water treatment plant used a pre-treatment process including coagulation,
ﬂocculation, dissolved air ﬂotation and ﬁltration (DAFF), pH correction, and cartridge
ﬁlters [19]. However, an autopsy of the fouled R0 membrane clearly identiﬁed several
kinds of foulants on the membrane surface, including colloids, extracellular polymeric

substances (BPS), scales and natural organic matter (N 0M) [19].

Based on ca. 2 years of monitoring in a seawater R0 (SWRO) plant, Chua et al. [8]
reported that although both a conventional pretreatment system and an ultraﬁltration (UF)
membrane system were used to remove the suspended particles [8], the silt density index
(SDI) of the R0 feed water remained above 2.5 and a substantial amount of suspended
particles remained. The SDI peaks indicated that the leakage of particles through
membrane pretreatment was possible. Boerlage and co-workers [13] evaluated colloidal
fouling in two R0 pilot desalination plants. In these two plants, the feed waters were
pretreated by a slow sand ﬁltration and a UF membrane, respectively, before entering the
R0 membrane systems. It was found that both of the ﬁltration methods were effective in

removing particles larger than 0.45 pm, but not the smaller particles.

Although there are numerous studies for the combined colloidal fouling and gypsum
scaling in heat transfer systems (e.g. [20] and references therein), so far there has been no
study available for the effects of combined colloidal fouling and gypsum scaling on the
performance of R0 membranes. This study aimed at examining the interplay between the
colloidal fouling and gypsum scaling in R0 membrane systems. A signiﬁcant synergistic

effect was observed during the combined gypsum scaling and colloidal fouling.

186

6.2. Experimental

6.2.1. Feed suspensions

There are two pathways for gypsum scaling in membrane systems: surface crystallization
and bulk crystallization [21-23]. In surface crystallization, gypsum crystals grow on the
membrane surface. On the contrary, in bulk crystallization, gypsum nucleation occurs in

the supersaturated feed solution and gypsum crystals deposit onto membrane surface.
The rate of gypsum crystallization in membrane system can be expressed as [5, 24, 25]:

dm
E=k3(Cm—Cs)2, (6.1)

for surface crystallization, and

dm
—=k C —C , 6.2
dt b( b s) ( )

for bulk crystallization, where m is the amount of gypsum formed, ks and kb are the

gypsum surface and bulk crystallization rates (also called surface and diffusion controlled

crystallization constant, [24], respectively), respectively, Cb and Cm are the CaSO4
concentrations in bulk solution and at the membrane surface, respectively, C s is the

CaSO4 saturation concentration. Surface crystallization occurs at a low saturation index

187

 

 

 

on the membrane surface while bulk crystallization occurs at a high saturation index in

the bulk feed water.

The experiments were conducted by adding soluble CaC12-2H20 and Na2S04 separately

instead of directly adding CaSO4°2H20. The advantages of this method included:

(1) It was convenient to prepare supersaturated solutions with respect to calcium
sulfate [26];

(2) The amount of undissolved salts was minimized [26];

(3) It was close to the situation in practice when the concentrations of ions are
increased due to increase in recovery, the saturation index for CaSO4-2H20
exceeds 1, and gypsum formation begins; and,

(4) Filtration of CaClz solution for 12 h conditioned membrane and reduced the effect
of the change in the properties of membrane when in contact with salty

suspensions.

For experiments with less saturated feed suspensions, 35.9 g CaC12-2H20 and 34.7 g
Na2S04 were dissolved separately and then were added into the deionized water. This
resulted in an ionic strength of 0.073 M. Using the equations in our previous study [27],

the saturation index (S g) of the feed suspensions was 0.91. In the control experiment

with only soluble salts and silica particles, Na2804 was replaced by 28.5 g NaCl to

achieve a close ionic strength and ion composition.

188

 

For experiments with supersaturated feed suspensions, 107.9 g CaC12-2H20 and 104.0
Na2804 were added, resulting in an ionic strength of 0.219 M and a saturation index of
4.04. In the control experiments, 85.7 g NaCl was used to replace N32S04. Feed channel

spacers were used in these experiments.

6.2.2. Experimental apparatus

The experimental apparatus has been described in our previous work [28]. The retentate
ﬂow rate was kept at 3.8 L/min. The 20 L feed suspension was mixed in a 30 L feed tank
by the retentate ﬂow and a 3 inches long magnetic stirrer bar. The verge of the membrane
coupons were covered by impermeable transparent tape for 2 cm and 1 cm in width at the
entrance and exit, respectively, to minimize the effect of inﬂow and outﬂow on gypsum
scaling. The effective ﬁltration area was reduced to 113.3 cm2 for each membrane

coupon.

After the experiments, one of the two membrane coupons was kept for microscopic
analysis. Another one was put into stirred deionized water of 200 mL to dissolve the
gypsum on the coupon surface. The concentration of calcium ion in the solution was
measured using an atomic absorption spectrometer (AAnalysis 800, PerkinElmer Inc.,
Wellesley, MA) in the ﬂame operation mode at a wavelength of 422.7 nm with a slit of
0.7 nm. Standard calcium solution was purchased from ULTRA Scientiﬁc Inc. (ECK-020,

North Kingstown, RI). The standard curve had a correlation coefﬁcient higher than

189

99.3%. The mass of gypsum on membrane coupon surface was calculated based on the

measured concentration.

6.2.3. Reagents, silica particles and R0 membranes

All reagents and chemicals were as reported previously [29]. Recently received silica
particles (SnowTex-XL, Nissan Chemical America Corp., Houston, TX) were chosen as
the model colloidal foulants. Polyamide thin-ﬁlm composite R0 membrane (BW30-365,
Dow-FilmTec, Minneapolis, MN) was used in all experiments. The characterization of
the colloids and membrane, and the membrane storage procedure were as reported

previously [29] as well.

6.2.4. Microscopical analysis

The surface of fouled membrane samples was imaged using Hitachi S-47OOII F icld
Emission scanning electron microscope (Hitachi High Technologies America, Inc.
Pleasanton, CA) with a secondary electron detector. The microscope was operated in
ultra-high resolution mode with an accelerating voltage of 15 kV and an emission current
of 10 ILA. Samples were coated with gold for around 4 min at a current of 20 mA using a

sputter coater (Emscope SC 500, Polaron Equipment Ltd. UK) before SEM imaging.

The cross-section of the fouled membrane samples was imaged using Joel 6400V

scanning electron microscope (Japan Electron Optics Laboratories, Japan) with a ﬁeld

190

emission (Oxford EDS, Oxford Instruments, UK) in backscattered electron (BSE)
imaging mode with an accelerating voltage of 20 kV. BSE imaging and X-ray energy
dispersive spectroscopy (EDS) were used to examine structure of the fouling layer
consisting of gypsum crystals and silica particles. We used following procedures to
prepare the cross-section samples of fouled membranes for imaging. Samples were
inﬁltrated with LR White resin (Electron Microscopy Sciences, Hatﬁeld, PA) in gelatine
capsules in agitation for 24 h. LR White resin is an organic chemical (80% Polyhydroxy
substituted bisphenol A dimethacrylate resin and 19.6% C12 methacrylate ester) that does
not dissolve gypsum crystals. Samples were polymerized at 60 °C for 24 h and were hand
free cross-sectioned. Carbon was coated onto the cross-section with a carbon string

evaporator (Ernest F. Fullam, Inc. Latham, New York.).

6.2.5. Protocols for membrane fouling experiments with less saturated suspensions

Experimental protocol is similar to what we used in previous study [29]. After
compacting membrane for 36 h and determining its hydraulic resistance, CaC12°2H20 was
dissolved in the 20 L feed suspension to condition the membrane for 12 h. Na2S04 or
NaCl was dissolved and added into the feed suspension. The initial permeate ﬂux was
controlled at 53 L/(mZ-h) to 54 L/(m2°h)(1.47'10‘5 m/s to 1.50405 m/s). In experiments
with colloidal fouling, 1.90 mL ST-XL stock solution was added into the feed suspension
to achieve a 50 mg/L silica particle loading. The pH of the feed suspension during the
fouling experiments was between 5.8 and 6.2 and was not adjusted. The fouling

experiment was stopped after 12 h. Membrane coupons were taken out and handled as

191

mentioned above. Aﬁer each experiment, the crossﬂow system was cleaned thoroughly
with a detergent solution. The pH of detergent solution was increased to 10 using NaOH
solution. Then it was ﬂushed with deionized water four times. The Teﬂon tubes in the

retentate line were replaced by new ones to remove the possible silica particles deposited.

To determine the mass of silica particles deposited on the membrane, a 25 mL sample
was take from the suspension every 2 h and was diluted 2 times to dissolve the possible
gypsum crystals. The absorbance at 220 nm of the resulted suspension was measured.

The concentration of silica particles in the suspension was calculated based on the

absorption using a standard curve recorded previously ( p2 = 1). A simple mass-balance

equation was applied to calculate the mass of particles deposited on membrane surface

(Mp).

6.2.6. Study the effect of silica particles on gypsum crystallization with

supersaturated suspensions

Four experiments were conducted to determine the effect of silica particles on gypsum
crystallization with supersaturated suspension: (1) with silica particles, with
transmembrane pressure; (2) without silica particles, with transmembrane pressure; (3)
with silica particles, without transmembrane pressure; and (4) without silica particles,
without transmembrane pressure. The transmittance at 400 nm of the suspension in the
feed tank was continuously recorded using a spectrophotometer (Multi-Spec 1501,

Shimadzu, Kyoto, Japan) by diverting a small portion of the retentate into a ﬂow-through

192

UV—vis sample cell (Model 178.710 — QS, Helhna GmbH & Co KG, Miillheim) before
returning this sample into the feed tank. A 5% drop in the transmittance was used to
identify the beginning of induction. The reason for this choice was discussed in our

previous study [27].

6.3. Results and discussion

6.3.]. Combined colloidal fouling and gypsum scaling in less saturated suspensions

6.3.1.1. Effect on RO membrane performance

Fig. 6.1 compares the permeate ﬂux decline of the RO membrane during gypsum scaling,
colloidal fouling, and combined gypsum scaling and colloidal fouling with less saturated
suspensions, respectively. It can be seen that ﬂux decline due to gypsum scaling after

12 h ﬁltration was ca. 12%, close to that when the RO membrane was fouled by colloidal
fouling alone. When gypsum scaling and colloidal particles were both present, the ﬂux
decline was ca. 42%, much larger than the sum of the decline from gypsum scaling and

colloidal fouling alone.

We suspect that three factors contributed to this observation. First, the silica particles
occupied the open space between the crystals (as shown in Figs. 6.4-6.6). The resultant
fouling layer consisted of large gypsum crystals (>20 pm in length and 2~3 pm in

diameter) and tiny particles (55 nm in diameter). Because the gypsum crystals were not

193

 

   

  
   
  

  
  

 
   

“I I l . '
o I l I I ... I"

k(t' ( ( -
: ( (““W'R'It'm‘ “(£32.

    

  

  

 

 
  

.................

 

 

"( ””‘lm .,
0_9 _ ............................... (flu
X
3 _
'0: .
+3 ~ s
.. 0.8 — ........
E .
8 . .
1: ~ 3 3 .
g 07 _. ........... .
To “ 1 5 3
E 2 =. 1
z 0.6 D CaCl2 + Na2804 only
0 CaCl2 + NaCI + Particles 5
A CaCl2 + Na2$04+ Particles 3
0_5m1.l...l...l...

 

 

0 2 4 6

Filtration time (h)

Figure 6.1. Comparison of membrane ﬂux decline in the RO membrane system during
gypsum scaling (indicated as “CaC12+ Na2SO4 only”), colloidal fouling (“CaC12+ NaCl
+ particles”), and combined fouling experiments (“CaC12+ Na2S04 + particles”). No
spacers were used; ST-XL particles loading (if added): 50 mg/L; Sg (for gypsum

scaling): 0.91; Initial hydraulic resistance of the membrane to pure water Rmo: (10.1 :1:

0.1)><10l3 m'l; Transmembrane pressure AP: (310 :i: 4)psi (2.37 MPa); Initial permeate
ﬂux Jo: (53.6 a: 0.5) L/(mz'h) (1.49-10's m/s); Retentate ﬂow rate: 3.8 L/min; Feed

suspension temperature: (20.0 :1: 0.5) °C.

194

permeable to water, the water passed the fouling layer through the open spaces between
the particles or those between the particles and crystals. The effect of the crystals on the
ﬂow included larger tortuosity and “area obstruction” [30], which greatly increased the
hydraulic resistance of the fouling layer. A theoretical model was developed in another
study to predict the permeability of this type of porous media [31]. Second, gypsum
scaling enhanced the deposition of silica particles, as shown in Fig. 6.3 and discussed
later. Third, the retentate ﬂow was less effective in removing the fouling layer resulted
from the combined fouling than that from either gypsum scaling or colloidal fouling
alone. After each experiment, a small portion of the fouled membrane was cut out and
gently washed by deionized water from a wash bottle. SEM imaging of these washed
samples demonstrated that after the cleaning, the fouling layer from gypsum scaling or
colloidal fouling alone was effectively removed. On the contrary, on the surface of the
membrane fouled by the combined fouling, a signiﬁcant amount of crystals and particles
remained after washing. The above three factors resulted in a much faster ﬂux decline

when the combined fouling occurred.

Fig. 6.2 demonstrates the permeate conductivity of the RO membrane during gypsum
scaling, colloidal fouling, and combined gypsum scaling and colloidal fouling,
respectively. In these experiments, there were three or four ions in the permeate, all of
which contributed to permeate conductivity. The RO membrane used in our study
(BW30-365, Dow-FilmTec) had a rejection of higher than 99.5% for CaSO4 or CaClz and
a NaCI rejection of 97% to 98%, which indicated that NaCl was the major contributor to

conductivity. Thus, although it was not feasible to determine the concentration of CaSO4

195

 

 

 

 

Permeate conductivity (uslcm)

b I l -1
Is . ( . .
100 6,1440%)", . . .' 33:5, ......... ‘. ....... LL11!” .............. 3 .............. ;. ............ _n
e ’1”), [’5’]; I! Fﬂb’L-L _ ' '
. "IIIIHllllumm I"
" "'lllliiII-I- E
80 -"- "'""!".'I_'Inin; -------
O CaCl2 + NaCl + Particles 9‘1"“? 'mmmnn

A CaCl2 + Na2$04 + Particles
Cl CaCl2 + Na2$04 only

60 ____________________ i.

 

 

 

Filtration time (h)

Figure 6.2. Permeate conductivity of the RO membrane during gypsum scaling, colloidal
fouling, and combined fouling experiments, respectively. (Experimental conditions were
shown in the caption of Fig. 6.1.)

196

and NaCl in permeate, respectively, based on permeate conductivity, it was acceptable to

use the permeate conductivity as an “indictor” of salt rejection.

As shown in Fig. 6.2, in the experiment with gypsum scaling only, the permeate
conductivity increased for a short period and then decreased afterwards. This was in
accordance with the observations by other researchers using CaClz and Na2SO4 for
gypsum scaling experiments [32]. One possible reason for the short increase in permeate
conductivity in the beginning is that initially some part of the ions on the membrane
surface formed nuclei or crystals instead of passing through the membrane, and thus the

concentration of salt at the membrane surface (Cm) decreased because of the formation

of gypsum crystals on the membrane surface.

In the experiment with colloidal fouling alone, the permeate conductivity increased
because the fouling layer enhanced the salt concentration polarization, as discussed in [28,

33].

In the experiment with the combined gypsum scaling and colloidal fouling, the permeate
conductivity decreased for the initial 3 h and increased aﬁerwards. This indicated that
after 3 h, the effect of enhanced salt concentration polarization by colloidal fouling was

more important.

197

6.3.1.2. Effect of colloidal fouling on gypsum scaling

As shown in Section 6.2, the mass of gypsum crystals on the membrane at the end of the
experiments was determined by dissolving the crystals on the membrane surface in
deionized water and measuring calcium ion concentration. The results showed that (246.1
:1: 4.7) mg gypsum formed in the experiment with gypsum scaling alone, and (242.9 :E 4.5)
mg gypsum formed with the combined gypsum scaling and colloidal fouling. This
indicated that the presence of the silica particles did not have evident effect on the
amount of gypsum crystals on membrane surface. It is worthy noticing that, compared
with the ﬂux with gypsum scaling alone, the ﬂux in the experiments with the combined
fouling was signiﬁcantly lower (Fig. 6.1), which was expected to result in a lower
concentration of the ions on the membrane surface, according to the theory of
concentration polarization. Thus less gypsum crystals were expected to form on
membrane surface based on the kinetics of gypsum crystallization. However, as we
indicated above, in the presence of silica particles the retentate ﬂow became less effective
in removing the fouling layer consisting of silica particles and gypsum crystals. As a
result, gypsum crystals in the combined fouling experiment were more likely to stay on
the membrane than those in the experiment with gypsum scaling alone. We suspect that
this is the reason why we did not observe difference on the amount of gypsum crystals on

the membrane surface in absence and presence of silica particles.

It should be noted that the amount of calcium ion for the gypsum formed on the

198

 

350”T”?""!""I""I~~I~HI..
300? _________ _:
250:— ----------- ........ E ____________ 3 ____________________________________ In“:
200_ j ...... M, _______ _:

l- : I
o r
I— n ... u
. o I
150 _ ................................... o ............................................ .2
l I I I I
l- I e . n u .4
I I I I I
I. I l I e e
I I l 0 0

   
 

 

100 1 ---------------------- ' ' ' ‘ '
—-a— CaCl2 + Na2304 + Particles
.-' , --O-- CaCl2 + NaCl + Particles

)-
.
n
j— ------------- . ......... 'u—.--n-a-c-I'ﬁ'I'CCIFUIC‘III'I. IIIIIIIIIIIII 1f‘T IIIIIIIIII "v‘.
n a u n u
I.
I I i c n
r e v o I
i l o o I
I I

 

 

Mass of particles deposited (mg)

 

 

 

0 2 4 6 8 10 12
Filtration time (h)

Figure 6.3. Mass of silica particles deposited in the absence and presence of gypsum
scaling (Experimental conditions were shown in the caption of Fig. 6.1.)

199

membrane surface accounted for less than 0.6% of calcium in the feed suspensions,
which indicated that the amount of gypsum formed on membrane surface did not

signiﬁcantly affect the CaSO4 concentration in feed suspensions.

6.3.1.3. Effect of gypsum scaling on colloidal fouling

Fig. 6.3 shows the amount of silica particles deposited on the membrane surface in the
absence and presence of gypsum scaling. Although the permeate ﬂux was lower (shown
in Fig. 6.1), in presence of gypsum scaling more silica particles deposited on the
membrane surface than in the experiment without gypsum scaling. We suspect that this
was due to two factors. First, the gypsum crystals resulted in a rougher membrane surface.
Colloidal particles were more likely to attach on a rougher membrane surface [34].
Second, as discussed above, the fouling layer consisting of both particles and gypsum
crystals was more difﬁcult to be removed by the retentate ﬂow. As a result, in the

presence of gypsum scaling, more silica particles deposited on the membrane surface.

6.3.1.4. Microscopical analysis of the fouling layer on membrane surface

Figs. 6.4 and 6.5 show the SEM images of the crystals in the absence and presence of the
silica particles, respectively. Without the silica particles, gypsum crystals grew from
different spots and formed large and dendritic clusters of crystals. This is the typical

structure of crystals formed through surface crystallization. In the presence of silica

200

 

Figure 6.4. Gypsum crystals in the absence of silica particles (Experimental conditions
were shown in the caption of Fig. 6.1.)

   

Figure 6.5. Gypsum crystals in the presence of silica particles at different magniﬁcations
(Experimental conditions were shown in the caption of Fig. 6.1.)

201

/ Fouling layer \

‘ ,.;¢ '

Membrane /'

 

Figure 6.6. X-ray energy dispersive spectroscopy (EDS) analysis of the fouling layer
consisting of gypsum crystals and silica particles (Experimental conditions were shown
in the caption of Fig. 6.1.)

202

particles, the structure of the crystals did not change. The open space among crystals was
occupied by the silica particles. It was also observed that in the absence of the silica
particles, there were no crystals in the entry area. More crystals grew along the flow
direction with most crystals appearing in the exit area. This to was due to increase in
concentration polarization along the ﬂow direction and the ﬂush by the retentate ﬂow,
which was in accordance with that reported by other researchers [7, 35]. On the contrary,
in the presence of silica particles, SEM imaging (pictures not shown) revealed that there
were signiﬁcant gypsum crystals in both the entry and exit areas. We suspect that this is
also due to the fact the fouling layer consisting of silica particles and crystals was more

difﬁcult to be removed than that of crystals only.

Fig. 6.6 reveals the structure of the fouling layer consisting of gypsum crystals and silica
particles based on the X-ray energy dispersive spectroscopy (EDS) analysis of the cross-
section of the fouling layer. It also demonstrates that the silica particles were embedded
into the inclusion provided by gypsum crystals. The distribution of silica particles in the
fouling layer was not uniform. Relatively more silica particles deposited closer to the

membrane surface.

6.3.2. Combined colloidal fouling and gypsum scaling in supersaturated suspensions

6.3.2.1. Effect of silica particles on gypsum crystallization

Fig. 6.7 shows the inﬂuence of silica particles on gypsum crystallization at different

operating conditions with supersaturated suspensions in terms of transmittance and

203

conductivity. From the transmittance (Fig. 6.7(a)), it can be clearly seen that the presence
of silica particles resulted in a slower transmittance decline and increased the induction
time. This is in accordance with our previous study in a batch crystallization system [27].
It is also shown that when transmembrane pressure was applied, the induction time was
greatly shortened. It is due to the fact that when transmembrane pressure was applied, the
permeate ﬂow increased the concentration on the membrane surface although the spacers
were used in the feed channel. This led to some crystals forming on the membrane
surface through surface crystallization (as shown in Figs. 6.9 and 6.10). These crystals

acted as “seeds” and greatly decreased the induction time.

As proposed in our previous study [27], the crystallization rate can be estimated based on
the decline of conductivity. Eq. (6.2) can be written as,

dC .
E— = kb(Cb — C, ), (6.3)

where k'b is the crystallization rate (min'l). Based on the conductivity data in Fig. 6.7(a),

the crystallization rate under different operating conductions was estimated as:

(1) Without silica particles, without transmembrane pressure: k1 = 0.00053 mini;
(2) With silica particles, without transmembrane pressure: k2 = 0.00072 min";

(3) Without silica particles, with transmembrane pressure: k3 = 0.0011 min"; and

(4) With silica particles, with transmembrane pressure: k4 = 0.0015 min".

204

   
    

 

 

 

 

 

 

 

 

 

 

0 With A P, No Part.
"" A With A P, With Part.
) 3 g D No A P, No Part.
80 — <> NoAP,Wlth Part.
‘3 ~ (a) I ' ‘ '
o\
v P
8 60 _ ...........................
5 .
5 .
g 40 _ ...................................
h .—
.—
20 _ .........................................................
’- in}:
L -
o 1 l l I i l l l 1 l 1 l
12 ‘7 Y I I ! 1' I I Y F I l I I F r Y I T ! 1' T l V I V l T r I ! V T Y r
’1 II.- 4”!”,’” /: : I g
;;:-2.:z1;29;992w;*29ﬂ“-" 1 i \’ b"i.'jj"h’\‘.rs.,”/\\%\ i
/— - — 9-9/9. 9; :
118 _W' ........... ; ......... /// _./_I,,/. "an. "MM/9,, ........ .- ......... _
' . . ///’ t ‘ . "" ,1: r-_\ .’ -
. ' I . ?( - ' in”, / I
A (b) I ///l Q?" Int "Q? . .1
g z - ..., 9,, s
. a- . .. I J
0 1 ’////, a; " f)
B ' 5 ’9. 3 V9,,
._ .............................................. ’ ....a.‘ ....... .......... WEI-”I,
s11 6 . . . g .
’ 2 ff 1 I"‘ 5|
5‘ 0 With A P, No Part. A, 5 i9, 3 ’35:”
3? A With A P, With Part. "a; 9
3114 D NOAP,NO Part. _______ ’IA ________ ___________________
'g o No A P, With Part. ; - j
o . . . i 9
0 ' ' s ’4, ‘
- ' A
A
11.2 _ .......................................................... A ......................... _
1 1 1 l L L A 1 l l A l l A l l i l l 1 i 1 1 l l I: J l l 1 j l L L 1

 

 

 

0 50 100 150 200 250 300 350 400
Time (min)

Figure 6.7. Effect of silica particles on gypsum crystallization at different operating
conditions: (a) Transmittance of the feed suspensions; (b) Conductivity of the feed
suspensions. (Silica particles (if added) loading: 50 mg/L; Transmembrane pressure (if
applied): 300 psi; Experiments were conducted with spacers using supersaturated
suspensions with S g = 4.04 .)

205

 

7’] 79",“ (:14! “— r r r r l r r ‘—
1 -””’/II//,),')“ .,.-—'».(.(««((‘<««««(«((«t(«t««««(«(«««««««««(««(««(((««««««t(«(««r-

 

 

 

 

 

 

”II,;"'='a‘-.=.- .
- i
0'8 i : "‘ """ """""" ; """""" ’- """""" """"" --------- ﬂ
X : : : : : : ; .
3

a: _
2 s s 2 s s s «
3 0.6 ....... .......... ............ ........... ......... _
g 2 I 1 i I ‘
g. ; : : : 5 i
3 o 4 5 5 5 ‘ .........EEWKQEQPPFPP ..... _
.5 ' 3 3 3 = meastrred in-i-llne .
.-. z 2 £ , / a i J

g i 5:515:- : : :
g 0.2 ; 4 , ”Ix/Illl/l/l/I/I/I/I/I/l/l/l/H/l/Ill/[Illl/I/l/l/I'A‘

o CaCl2 + NaCl + Particles E '

D CaCl2 + Na2$04 only

0 A CaCl2 + Na2$04 + [Particles ’ I i

0 0.5 1 1.5 2 2.5 3 3.5 4
Filtration time (it).

Figure 6.8. Normalized permeate ﬂux at three experimental conditions. Experimental
conditions were shown in the caption of Fig. 6.7.)

206

As shown in Fig. 6.7(b), the presence of silica particles increased the crystallization rate.
This is in accordance with what we observed in the batch crystallization system [27].
Similarly, when the transmembrane pressure was applied, the crystals formed on the
membrane surface greatly increased the crystallization rate. Bansal et al. [26] found
adding gypsum crystals greatly enhanced gypsum formation because they provided extra

nucleation sites.

6.3.2.2. Effect on RO membrane performance

Fig. 6.8 compares the permeate ﬂux decline of the RO membrane during gypsum scaling,
colloidal fouling, and combined gypsum scaling and colloidal fouling with supersaturated
suspensions, respectively. It can be seen that when scaling occurred in supersaturated
suspensions, the permeate ﬂux decreased dramatically after only ca. 2 h ﬁltration, and the

presence of silica particles led to a slightly faster permeate ﬂux decline.

6.3.2.3. Microscopical analysis of the fouling layer on membrane surface

Fig. 6.9 shows the crystals formed in the initial stage and after a 12 h ﬁltration,
respectively. It is found that in the initial stage the crystals were large and dendritic,
which conﬁrmed our statement that at the early stages crystals formed ﬁrst on membrane
surface through surface crystallization. Different to those shown in Fig. 6.4, the clusters
in Fig. 6.9(a) had some large plate-shaped crystals. As shown in Fig. 6.9(b), after 12 h,

the crystals were

207

 

Figure 6.9. The crystals formed in the absence of silica particles. (a) 20 min; (b) 12 h
(Experimental conditions were shown in the caption of Fig. 6.7.)

208

'—s_
.‘iliiiiiii

I i
12 3mm x1 m» SEiMi ill/5;“ )5 15 02 El'.‘ Mm

 

Figure 6.10. The gypsum crystals formed in the presence of silica particles. (a) 20 min;
(b)12 h. (Experimental conditions were shown in the caption of Fig. 6.7.)

209

mainly needle-shaped, which is the typical structure of gypsum crystals formed through

bulk crystallization.

Fig. 6.10 shows that the crystals formed in the presence of silica particles in the initial
stage were also through surface crystallization. After a 12 h ﬁltration, the crystals formed
through bulk crystallization were mainly needle-like, similar to those formed in the

absence of the particles.

6.4. Conclusions

The interactions between colloidal fouling and gypsum scaling in RO membrane systems
were investigated with both less saturated and supersaturated suspensions. A signiﬁcant
synergistic effect was observed during the combined colloidal fouling and gypsum
scaling. When gypsum scaling was dominated by surface crystallization, gypsum scaling
enhanced colloidal deposition and the combined fouling resulted in a much faster
permeate ﬂux decline, which was larger than the sum of the ﬂux decline from colloidal
deposition and that from gypsum scaling alone. When bulk crystallization dominated, the
presence of silica particles retarded induction time, but increased the crystallization rate.
When transmembrane pressure was applied, crystals formed on the membrane surface
through surface crystallization in the early stage greatly shortened the induction time and
increased the gypsum crystallization rate. This study implied that it was important to take
into consideration the effects of combined fouling when designing and Operating RO

membrane systems.

210

Acknowledgments

Financial support of this work by the National Water Research Institute (project no. 05-
TM-007) is gratefully acknowledged. We thank Dow-FilmTec for providing us with
membrane samples and Nissan Chemical America Corp. for supplying the silica
suspension. We also thank Dr. Ewa Danielewicz and Dr. Alicia Pastor from the Center
for Advanced Microscopy at Michigan State University for their assistance with the

cross-section sample preparation and EDS analysis.

211

 

 

Reference

10.

11.

12.

13.

D. Bursill, Drinking water treatment - understanding the processes and meeting
the challenges, Wat. Sci. Technol.: Water Supply 1 (2001) 1-7.

State of California, Findings and Recommendations, (2003).

United States Environmental Protection Agency, Report: Membrane ﬁltration
guidance manual (815-R-06-009), (2005).

AWWA Membrane Technology Research Committee, Committee report:
Membrane processes, J. Am. Water Works Assoc. 90 (1998) 91-105.

S. Lee and CH. Lee, Effect of operating conditions on CaSO4 scale formation
mechanism in nanoﬁltration for water softening, Water Res. 34 (2000) 3854.

S. Lee and RM. Lueptow, Control of scale formation in reverse osmosis by
membrane rotation, Desalination 155 (2003) 131.

A. Rahardianto, W.Y. Shih, R.W. Lee and Y. Cohen, Diagnostic characterization
of gypsum scale formation and control in RO membrane desalination of brackish
water, J. Membr. Sci. 279 (2006) 655.

K. T. Chua, M. N. A. Hawlader and A. Malek, Pretreatment of seawater: Results
of pilot trials in Singapore, Desalination 159 (2003) 225-243.

Harvey Winters, Twenty years experience in seawater reverse osmosis and how
chemicals in pretreatment affect fouling of membranes, Desalination 110 (1997)
93.

B. A. Winﬁeld, A study of the factors affecting the rate of fouling of reverse
osmosis membranes treating secondary sewage efﬂuents, Water Res. 13 (1979)
565.

Seung-Hyun Kim, Seong-Yong Moon, Cho-Hee Yoon, Seong-Keun Yim and Jae-
Weon Cho, Role of coagulation in membrane ﬁltration of wastewater for reuse,
Desalination 173 (2005) 301.

J. van Agtmaal, H. Huiting, P. A. de Boks and L. L. M. J. Paping, Four years of

practical experience with an Integrated Membrane System (IMS) treating estuary
water, Desalination 205 (2007) 26.

S. F. E. Boerlage, M. Kennedy, M. P. Aniye and J. C. Schippers, Applications of

the MFI-UF to measure and predict particulate fouling in R0 systems, J. Membr.
Sci. (2003) 97-116.

212

 

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

A. D. Khawaji, I. K. Kutubkhanah and J. Wie, A 13.3 MGD seawater RO
desalination plant for Yanbu Industrial City, Desalination 203 (2007) 176.

M. Luo and Z. Wang, Complex fouling and cleaning in-place of a reverse osmosis
desalination system, Desalination 141 (2001) 15-22.

Y. R. Ning, L. T. Troyer and S. R. Tominello, Chemical control of colloidal
fouling of reverse osmosis systems, Desalination 172 (2005) 1-6.

Christopher J. Gabelich, Tae I. Yun, Bradley M. Coffey and I. H. Mel Suffet,
Pilot-scale testing of reverse osmosis using conventional treatment and
microﬁltration, Desalination 154 (2003) 207.

C. J. Gabelich, M. D. Williams, A. Rahardianto, J. C. Franklin and Y. Cohen,
High-recovery reverse osmosis desalination using intermediate chemical
demineralization, J. Membr. Sci. 301 (2007)l31.

T. Tran, B. Bolto, S. Gray, M. Hoang and E. Ostarcevic, An autopsy study of a
fouled reverse osmosis membrane element used in a brackish water treatment
plant, Water Res. 41 (2007) 3915-3923.

D. Hasson. Progress in precipitation fouling: a review, in:Understanding Heat
Exchanger Fouling and its Mitigation, T.R. Bott et al. (Eds). 1999: Begell House,
New York.

A.G. Pervov, Scale formation prognosis and cleaning procedure schedules in
reverse osmosis systems operation, Desalination 83 (1991) 77.

M. Okazaki and S. Kimura, Scale formation on reverse-osmosis membranes, J.
Chem. Eng. Jpn. 17 (1984) 145.

S. Lee, J. Kim and CH. Lee, Analysis of CaSO4 scale formation mechanism in
various nanoﬁltration modules, J. Membr. Sci. 163 (1999) 63.

G.H. Nancollas, The growth of crystals in solution, Adv. Colloid Interface Sci. 10
(1979) 215-252.

W.Y. Shih, A. Rahardianto, R.W. Lee and Y. Cohen, Morphometric
characterization of calcium sulfate dihydrate (gypsum) scale on reverse osmosis
membranes, J. Membr. Sci. 252 (2005) 253-263.

B. Bansal, H. Muller-Steinhagen and X. D. Chen, Effect of suspended particles on

crystallization fouling in plate heat exchangers, J. of Heat Transfer, Transactions
ASME 119 (1997) 568.

213

27.

28.

29.

30.

31.

32.

33.

34.

35.

F. Wang and V. V. Tarabara, Crystallization of gypsum in the presence of
colloidal silica. I: in a stirred batch system, In preparation (2008).

F. Wang and V.V. Tarabara, Coupled effects of colloidal deposition and salt
concentration polarization on reverse osmosis membrane performance, J. Membr.
Sci. 293 (2007) 111-123.

F. Wang and V.V. Tarabara, Coupled effects of colloidal deposition and salt
concentration polarization on reverse osmosis membrane performance, J. Membr.
Sci. (submitted) (2006).

C. R. Ethier, Flow through mixed ﬁbrous porous materials, AIChE J. 37 (1991)
1227-1236.

F. Wang and V. V. Tarabara, Modeling the permeability of ﬁber-ﬁlled porous
media using Kozeny-Carman-Ethier equation, In preparation (2008).

M. Uchymiak, A. Rahardianto, E. Lyster, J. Glater and Y. Cohen, A novel R0 ex
situ scale observation detector (EXSOD) for mineral scale characterization and
early detection, J. Membr. Sci. 291 (2007) 86.

E. M. Hoek and M. Elimelech, Cake-enhanced concentration polarization: a new
fouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 17 (2003)
5581-5588.

X. Zhu and M. Elimelech, Colloidal fouling of reverse osmosis membranes:
measurements and fouling mechanisms, Environ. Sci. Technol. 31 (1997) 3654-
3662.

A. P. Mairal, A. R. Greenberg, W. B. Krantz and L. J. Bond, Real-time

measurement of inorganic fouling of RO desalination membranes using ultrasonic
time-domain reﬂectometry, J. Membr. Sci. 159 (1999) 185-196.

214

 

..
m

 

   

IIiﬂil'ijiﬂiiilﬂiziﬂijieilIiﬂilieilijliil