 

 

 

 

 

 

 

a. ad...
553.... .

3...“..033 ‘1 1 a.

.. éawaghwwuz

3. 4
u 1
a...
.z ...
H

M...

2. :3

.n: ...u

. . 5..
\ . $9.15,.

v:

. a ..
.1 v .
'::z..\a..¢ .
3...“...1. .13.)». :1
$11.31....“ FALL—wt .
.Ix...§..9.32.|.. nix!
iiiuﬂco a V‘I‘r .- t\
, .ﬁ .
. . ﬁlial-{IurSQ-ai
:.s.§v§..\aa;.
.\-I 1..
\‘i (.11..
7 Ii}
.9 .
iii...
.1131? k:

I. I HUN...“ 15.3.1413... . .1...
s x 5.! . II:

{$1.33}.

 

m Michigan State
.1 O - University

 

This is to certify that the
dissertation entitled

NANOPARTICLE-POLYMER COMPOSITE MEMBRANES:
SYNTHESIS, CHARACTERIZATION AND ENVIRONMENTAL
APPLICATIONS

presented by

JULIAN S. TAUROZZI

has been accepted towards fulﬁllment
of the requirements for the

Ph.D. degree in Environmental Engineering

 

 

m

Majaf’ Profeséor's Signature

Wave/a /{l @297

V

 

Date

MSU is an Afﬁrmative Action/Equal Opportunity Employer

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 KzlProleccaPreleIRClDateDmindd

 

NANOPARTICLE-POLYMER COMPOSITE MEMBRANES: SYNTHESIS,
CHARACTERIZATION, AND ENVIRONMENTAL APPLICATIONS

By

Julian S. Taurozzi

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Environmental Engineering

2009

ABSTRACT
NANOPARTICLE-POLYMER COMPOSITE MEMBRANES: SYNTHESIS,
CHARACTERIZATION, AND ENVIRONMENTAL APPLICATIONS
by

Julian S. Taurozzi

Advances in nanotechnology and materials science offer new possibilities for the
development of novel water treatment technologies. A salient example is the emergence
of a new generation of nanostructured membranes with improved separation properties
and, in some cases, multiﬁmctional capabilities. The present work explores different
aspects of the formation of nanoparticle-enabled membranes, focusing on the dependence
between synthetic methods, nanoparticle properties, and the performance of the obtained

membranes.

The effects of nanoparticle loading, size and morphology, and methods of nanoparticle
incorporation on the functional properties of the resulting nanocomposite membranes are
studied. Additionally, empirical predictors of membrane performance for the rational
design of nanocomposite membranes and novel strategies for the manufacture of

membrane-based sensors and biofouling resistant membrane spacers are presented.

The ﬁrst chapter of the dissertation provides an overview of the theory of phase inversion
in polymer blends - one of the main methods for the fabrication of ultraﬁltration

membranes. The second chapter describes the study of the effect of shape and loading of

carbon nanoparticle ﬁllers on the hydraulic properties of nanocomposite membranes. In
the third chapter, the effects of casting mixture composition and nanoparticle
incorporation route on the morphological structure and separation performance of
nanocomposite membranes are described and the potential use of these nanocomposites
for the mitigation of membrane biofouling is discussed. Finally, in the last chapter, a
method for the synthesis of a nanoparticle-enabled, membrane-based sensor for water

quality control is presented.

Copyright by
JULIAN S. TAUROZZI
2009

ACKNOWLEDGEMENTS

This work would not have been possible without the unconditional support and guidance
of my PhD advisor and committee chairman, Dr. Volodymyr V. Tarabara. I would like to
specially thank him for his invaluable advice, insightful discussions and permanent
encouragement and instruction towards the execution of every task with the highest level

of excellence.

I would like as well to express my deep gratitude to my committee members: Dr. Merlin
L. Bruening, Dr. Thomas C. Voice, Dr. Alison M. Cupples and Dr. Jerome Rose, for their

positive feedback, supportive disposition and constructive criticism.

I wish to thank former and current members of my research group at the Environmental
Nanotechnology Research Lab: Dr. Fulin Wang, Dr. Jeonghwan Kim, Dr. Seokjong
Byun, Wenqian Shan, Alla Alpatova, Elodie Pasco and Adam R. Rogensues, for their
wholehearted help and unreserved assistance in conducting experiments, analyzing data,

discussing results and enabling a warm and amiable work environment.

My special thanks to Ms. Lori Hasse and Mr. Yang-lyang Pan for their constant

administrative and technical support.

My gratitude to Dr. Baokang Bi, Dr. Alicia Pastor-Lecha, Dr. Melinda Frame, Ms. Ewa
Danielewicz and Ms. Carol S. Flegler for the helpful discussions and inestimable aid in

the conduction of microscopy experiments.

I would like to stress my endless appreciation to the Department of Civil and
Environmental Engineering at Michigan State University and the National Science

Foundation for their ﬁnancial support throughout my PhD.

Finally, I wish to express my eternal gratitude to my family, and specially my mother and

sister, for their relentless support, unconditional love and endless patience.

vi

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................. x
LIST OF FIGURES ........................................................................................................... xi
INTRODUCTION .............................................................................................................. 1
Signiﬁcance and application potential ............................................................................ 3
Multifunctional membranes ........................................................................................ 4
Hydraulic properties .................................................................................................... 5
Mechanical properties ................................................................................................. 7
Tunable hydrophilicity ................................................................................................ 9
Adaptive membranes ................................................................................................ 11
Dissertation Overview .................................................................................................. 14
References ..................................................................................................................... 16
CHAPTER ONE ............................................................................................................... 21
Theoretical backgorund of the phase inversion process ............................................... 21
1.1. Phase Inversion ...................................................................................................... 21
1.2.Liquid-Liquid Demixing ......................................................................................... 23
1.3.Ternary system demixing ........................................................................................ 35
1.4.Wet phase inversion ................................................................................................ 44
1.4.1.Solvent/non solvent afﬁnity ............................................................................. 49
1.4.2.Polymer concentration in the casting mixture ................................................. 50
1.4.3.Addition of non solvent to the casting mixture ................................................ 52
1.4.4.1mmersion bath temperature ............................................................................ 54
1.4.5.Addition of solvent to the immersion bath ...................................................... 56
References ..................................................................................................................... 58
CHAPTER TWO .............................................................................................................. 59
Fullerene-polysulfone nanocomposite ultraﬁltration membranes: Effect of ﬁller shape
and loading on membrane properties ............................................................................ 60
Abstract ......................................................................................................................... 60
2.1 Introduction ............................................................................................................. 61
2.2 Experimental ........................................................................................................... 64
2.2.1 Materials .............................................................................................................. 64
2.2.2 Preparation of polysulfone membranes ........................................................... 64
2.2.2.1 Preparation of nanoparticle-ﬁlled membranes .......................................... 66
2.2.3 Characterization of casting mixtures and cast membranes .............................. 67
2.3. Results and Discussion .......................................................................................... 69
2.3.1 Casting mixture properties ............................................................................... 69
2.3.1.1 Nanoparticle dispersions in N-methyl-pyrrolidone .................................. 69
2.3.1.2 Viscosity of nanoparticle (C60) — polymer (PSt) - solvent (NMP) —
porogen (PVP) blends: Effect of nanoparticle loading ......................................... 69

vii

2.3.1.3 Demixing behavior of polymer (PSI) — nanoparticle (nC6o) — solvent
(NMP ) — porogen (PVP) blends exposed to nonsolvent (water): Effect of

nanoparticle loading .............................................................................................. 72
2.3.2 Membrane hydraulic properties ....................................................................... 76
2.3.3 Nanocomposite membranes doped with nanoparticles of different shapes ..... 78

2.4. Conclusions ............................................................................................................ 81
Supporting Documentation ........................................................................................... 84
References ..................................................................................................................... 92
CHAPTER THREE .......................................................................................................... 96
Effect of ﬁller incorporation route on the properties of polysulfone-silver
nanocomposite membranes of different porosities ....................................................... 97
Abstract ......................................................................................................................... 97
3.1 Introduction ............................................................................................................. 98
3.2 Experimental ......................................................................................................... 102
3.2.1 Materials ........................................................................................................ 102
3.2.2 Preparation of polysulfone membranes ......................................................... 102
3.2.3 Preparation of polysulfone-silver nanocomposite membranes ...................... 105
3.2.4 Characterization of membranes: SEM, TEM, EDS, FTIR, XPS ................... 106
3.2.5 Permeability and rejection tests ..................................................................... 107
3.2.6 Measuring silver losses .................................................................................. 108
3.2.7 Membrane biofouling characterization .......................................................... 109

3.2.7.1 Bioﬁlm growth inhibition under diffusive conditions ............................ 109

3.2.7.2 Bioﬁlm growth inhibition under convective conditions ......................... 110

3.2.7.3 Assessment of silver leaching ................................................................. 111

3.3. Results and Discussion ........................................................................................ 113
3.3.1 Silver-free polysulfone membranes ............................................................... 113

3.3.1.1 Surface morphology ................................................................................ 114

3.3.1.2 Cross-sectional morphology ................................................................... 114
3.3.2 Polysulfone-silver nanocomposite membranes .............................................. 116

3.3.2.1 Cross-sectional morphology ................................................................... 116

3.3.2.2 Surface morphology ................................................................................ 118

3.3.2.3 F onnation-structure links for nanocomposite membranes ..................... 121

3.3.2.4 Permeability and rejection of nanocomposite membranes ..................... 123
3.3.3 Antibacterial properties of nanocomposites ................................................... 125
3.3.4 Practical aspects of the application of silver nanocomposites as membrane and
membrane spacer materials ..................................................................................... 130

3.3.4.1 Effect of antibacterial activity on the efﬁciency of hydraulic cleaning. 130
3.3.4.2 Antibacterial efficiency of nanocomposite membranes under convective

transport conditions ............................................................................................. 131

3 .4. Conclusions .......................................................................................................... 135
Acknowledgements ..................................................................................................... 137
Supporting Documentation ......................................................................................... 138
References ................................................................................................................... 151
CHAPTER FOUR ........................................................................................................... 155

viii

Silver nanoparticle arrays on track etch membrane support as ﬂow-through optical

sensors for water quality control ................................................................................. 156
Abstract ....................................................................................................................... 156
4.1. Introduction .......................................................................................................... 157
4.2. Experimental ........................................................................................................ 161
4.2.1 Materials ........................................................................................................ 161
4.2.2 Synthesis of silver nanoparticles .................................................................... 161
4.2.3 Membranes and membrane surface modiﬁcation .......................................... 162
4.2.3.2 APTMS used as a linker ......................................................................... 163
4.2.3.3 CysA used as a linker .............................................................................. 164
4.2.4 XPS, UV-vis, Raman and SERS measurements ............................................ 164
4.2.5 Particle size and zeta potential measurements ............................................... 167
4.2.6 SEM and TEM imaging ................................................................................. 167
4.2.7 SERS characterization of APTMS modiﬁed membranes .............................. 168
4.2.8 Membranes ﬁltration performance ................................................................ 168
4.2.9 Preconcentration studies ................................................................................ 169
4.3. Results and Discussion ........................................................................................ 170
4.3.] Silver nanoparticle hydrosol .......................................................................... 170
4.3.2 Modiﬁed membranes characterization ........................................................... 173
4.3.2.1 Surface chemistry .................................................................................... 173
4.3.2.2 Silver nanoparticle array morphology .................................................... 174
4.3.2.3 Optical properties of modiﬁed membranes ............................................. 176
4.3.2.4 Filtration performance of modiﬁed membranes ..................................... 184
4.3.2.5 Preconcentration ..................................................................................... 187

4.4. Conclusions .......................................................................................................... 192
Acknowledgements ..................................................................................................... 193
References ................................................................................................................... 194

ix

LIST OF TABLES

Table 2.1.Components of membrane casting mixtures (% w/w). Weight % of ﬁller is
expressed with respect to the mass of PSf. ............................................................... 65

Table SD.2. 1 .Cluster sizes for different silica powders sonicated at a variety of intensities
and times. 100 % sonication intensity corresponds to a maximum power output of
600 W. PEG denotes Polyethyleneglycol (400 Mw). ............................................... 89

Table 3.1 .Components of membrane casting mixture (% mass) .................................... 104

Table 3.2.Resistance and rejection values for all membrane types. Mass fractions of silver
were the same for PSf/Agcx and PSf/Agm casting mixtures .................................... 124

Table SD.3. 1 .Statistical data analysis for released protein on diffusive biofouling
experiments without hydraulic rinsing (Figure 3.5 of main document). Mean values
are normalized to that of PSf/Type I membranes, errors are calculated with a 90%
conﬁdence level, the “t values” columns indicate the comparative basis for each
pair. ......................................................................................................................... 149

Table SD.3.2.Statistical data analysis for released protein on convective biofouling
experiments (Figure 3.6 of main document). Mean values are normalized to that of
PSf/Type I membranes, errors are calculated with a 90% conﬁdence level, the “t
values” columns indicate the comparative basis for each pair ................................ 150

LIST OF FIGURES

Figure 1.1.Plot of AGm vs. (1)2 for the polymer/solvent system at temperature T4, above
the critical point. ....................................................................................................... 26

Figure 1.2.Polymer/solvent system’s AGm vs. (1)2 plot at a temperature T2 below the
critical point .............................................................................................................. 29

Figure 1.3.AGm vs. (1)2 curves for the polymer/solvent system at different temperatures,
from the completely mixed system at T4, above the critical temperature (Tc), to T1,

ﬁnal temperature of the demixed system. ................................................................. 34
Figure 1.4.Temary phase diagram of a three component system (solvent, non-solvent and

polymer). ................................................................................................................... 37
Figure 1.5.Solubility diagram for the three component system, showing the binodal and

spinodal curves, tie lines, and the critical point. ....................................................... 39
Figure 1.6.Demixing and gelation regions for the three component system. ................... 42

Figure 1.7.Schematic of solvent (J s) and non-solvent (JNs) transport for the wet phase
inversion process. ...................................................................................................... 45

Figure 1.8.Demixing paths for a three component system during wet phase inversion 47

Figure 1.9.Eﬁect of polymer concentration (point a*) on the demixing path of ternary
system ....................................................................................................................... 51

Figure 1.10.Effect of non solvent addition to the casting mixture (point a*) on the
demixing path of ternary system ............................................................................... 53

Figure 1.1 1.Effect of temperature increase on the demixing path of the ternary system. 55

Figure 1.12.Effect of solvent addition to the coagulation bath on the demixing path of
ternary system. .......................................................................................................... 57

Figure 2.1. Viscosity and demixing front propagation rate for PSf-PVP-C6o blends (a)
and hydraulic resistance and 150 kDa rejection for Crag-ﬁlled PSf membranes (b) at

0% (control), 0.7 % and 10 % w/w C60/PSf loadings. .............................................. 70

Figure 2.2. Schematic (A) of an isothermal phase diagram for a three component
(solvent-polymer-nonsolvent) system and laser-scanned brightﬁeld images (B) of

demixing front propagation in PSf-Cw-PVP-NMP blends as a function of time (in
seconds) of exposure to nonsolvent (water) .............................................................. 75

xi

Figure 2.3.Hydraulic resistance (circles) and rejection (triangles) of 150 kDa dextran for
PSff membranes ﬁlled with C60 and SWCN Ts at a ﬁller/polymer weight loading of

0.7 %. ........................................................................................................................ 79
Figure SD.2.].SEM images of 1 mg/mL SWCNT/NMP (A) and C6o/NMP (B)

organosols. ................................................................................................................ 86
Figure SD.2.2.Cloud points for PSf-NMP-water system .................................................. 88

Figure SD.2.3.Front thickness vs. the square root of time plots for 0, 0.7 and 10 % C60
loading casting mixtures. .......................................................................................... 91

Figure 3.1. SEM surface (a, b, c) and cross-section (d, e, f) images of PSf Type I (a, d),
Type II (b, e) and Type III (c, f) membranes. Inset A shows the permeate side
surface of PSf Type II membranes .......................................................................... 115

Figure 3.2. Schematic illustration of the distribution of silver nanoparticles in the silver-
polysulfone nanocomposites prepared using different methods of silver
incorporation. On micrographs, the black dots with arrows pointing to them
correspond to silver nanoparticles. TEM micrographs (A, C and D) show membrane
cross-sections, while SEM micrograph (B) shows the top cross section layer, with
large silver clusters embedded into the polysulfone matrix .................................... 117

Figure 3.3.Representative SEM images of the surfaces of PSf Type I (a), PSf/Agin Type I
(b) and PSf/Ag“ Type I (c) membranes. ................................................................ 120

Figure 3.4.Typical SEM images of the surface of biofouled PSf Type I silver-free (a),
PSf/Agm (b) and PSf/Ag':x (c) membranes. Evident is the stark contrast between the
thick bioﬁlm developed on the surface of a silver-free membrane (a), and the
absence of bioﬁlm growth on the nanocomposite surface. The sample preparation
procedure is detailed in section 3.2.4 ...................................................................... 127

Figure 3.5.Bradford assay results describing the bioﬁlm growth on the surface of PSf
membranes and PSf/Ag nanocomposites of different porosities under diffusive
transport conditions ................................................................................................. 129

Figure 3.6.Bioﬁlm growth on PSf/Ag Type I nanocomposites after 48 hr of continuous
ﬁltration of a) LB medium (ﬁrst two columns) and b) leachate obtained by
continuously ﬁltering DI water through PSf/Ag Type I nanocomposite for 2 days
(third column). The leachate was Ag-loaded water obtained from the ﬁltration of DI
water through a bacteria-free PSf/Ag Type I nanocomposite; the leachate was then
mixed with LB medium to be ﬁltered through an inoculated Ag-free membrane. 133

Figure SD.3.].TEM image (top) of silver nanoparticles (Spectrum 1) embedded in the

PSf matrix (Spectrum 2) of a PSf/Agm Type I membrane, and EDS spectra (bottom)
of both the PSf matrix (yellow ﬁlled spectrum) and the Ag nanoparticles (line

xii

spectrum) showing the silver peaks for the nanoparticle. The x-axis for the EDS
spectra is in KeV units. ........................................................................................... 141

Figure SD.3.2.FTIR spectra of PSf Type I (Top) and PSf/Agin Type I (bottom)
membranes. ............................................................................................................. 143

Figure SD.3.3.XPS survey spectrum and atomic concentration (relative) table for a

PSf/Agm Type I membrane. The silver signal is not visible in the survey spectrum.
................................................................................................................................. 145

Figure SD.3.4.Bradford assay results describing the bioﬁlm growth on the surface of PSf
membranes and PSf/Ag nanocomposites of different porosities under diffusive
transport conditions, both with and without hydraulic rinsing. .............................. 147

Figure 4.1 .Proposed mechanism for the APTMS reaction with polycarbonate and further
silver chemisorption to the incorporated nitrogen. ................................................. 165

Figure 4.2.Micrographs of Ag nanoparticles: SEM image of ﬁltered hydrosol (A), and
TEM image of Ag nanoparticles deposited onto TEM grids by drop-coating (B). 171

Figure 4.3.Size distribution proﬁle of silver nanoparticle suspension. Also shown are the
predominant spherical diameters for the two major distribution peaks, and the
estimated rod lengths for the two high end size peaks (Schematics not to scale) .. 172

Figure 4.4.SEM micrographs of PCTE membrane ﬁlter that is non-modiﬁed (A),
modiﬁed using APTMS linker (B, C, D), and modiﬁed using CysA linker (E, F). 175

Figure 4.5.UV-vis spectra of a modiﬁed PCTE membrane (bare membrane used as
baseline), and of silver hydrosol (250 [IL in 2 mL ultrapure water, ultrapure water
used as baseline). .................................................................................................... 177

Figure 4.6.Absorption (A) and Raman spectra (B) of methylene blue. .......................... 179

Figure 4.7 .Raman spectra of an untreated membrane (bottom) and modiﬁed membrane
(top). ........................................................................................................................ 180

Figure 4.8.Rarnan spectrum of a 1 mmol/L aqueous solution of MB (bottom), a 10-5

mol/L aqueous solution of MB in pre-coagulated silver sol (middle), and a 10'5
mol/L aqueous solution of methylene blue after “drop and dry” on a modiﬁed

membrane (top). ...................................................................................................... 183
Figure 4.9.Box plots for enhancement factors of MB on precoagulated sol and modiﬁed

membranes (A) and after preconcentration of modiﬁed membranes (B). .............. 185
Figure 4.10.Results of clean water ﬂux test for bare and modiﬁed membranes. ........... 186

xiii

Figure 4.11.Raman spectrum of a 10'7 mol/L aqueous solution of MB after 2 hours of
pre-concentration on a modiﬁed membrane. .......................................................... 188

Images in this dissertation are presented in color

xiv

INTRODUCTION

Nanoparticle-polymer composite membranes: Synthesis, characterization, and

environmental applications

Membrane-based separation is a well established water treatment process and an integral
component in a growing number of drinking water and wastewater treatment facilities. A
membrane is a semi-permeable barrier designed for the separation of targeted compounds
from source waters. Membranes perform a variety of operations involving the removal of
pollutants from contaminated waters, the recovery of valuable or recyclable components
from treated streams, or the identiﬁcation and quantiﬁcation of targeted compounds. The
versatility of membranes in treating source waters is further enhanced by their minimal

environmental footprint.

An emerging ﬁeld of materials engineering with promising applications for membrane
manufacture is that focused on the development of Nanocomposites. Nanocomposites are
materials that combine two or more components, wherein at least one of the components
has nanoscale dimensions. Materials at the nanoscale - or nanoparticles - with at least one
dimension of less than 100 nm, exhibit unique properties that are fundamentally different
from those observed for the same materials at the macroscopic scale. Nanocomposite
structures thus incorporate distinctive properties to yield exceptional mechanical,

thermal, or electrical features, among others.

The increasing choice of available nanoparticles offers new possibilities for the
manufacture of a new generation of nanoenabled membranes. Examples of such potential
include the bottom-up design Of nanoparticle-enabled multifunctional membranes capable
Of performing multiple tasks or the incorporation of nanoparticles within the membrane
structure to obtain membranes with improved mechanical and separation properties.
Research at the interface of nanomaterials science and membrane manufacture holds
promise for more efﬁciently safeguarding the quality and ensuring the availability of

drinking water.

Signiﬁcance and application potential

The expanding availability and diversity of nanoparticles stresses the potential Of their
application towards the synthesis of a new generation of nanoparticle-enabled
membranes. Exceptional features such as supramagnetism or enhanced catalytic and
optical properties have been reported for nanoparticles of different materials, while the
possibility of chemically functionalizing the nanoparticle surface further broadens the

spectrum of potential nanoenabled features [1, 2].

A conventional approach for the synthesis of nanocomposite membranes relies on the
incorporation of nanoparticles to a polymer blend or a pre-formed membrane. The
development of such porous nanocomposites however, should not be undertaken without
focusing on issues that will affect the operational viability Of the ﬁnal product. The
structure of the resulting membranes can be expected to be a function of the physical and
chemical properties of both the polymer phase and nanoparticles, as well as the method
of nanoparticle incorporation. However, only a few of the studies available to date have
focused on the understanding of links between the operational properties of the obtained

membranes, and the characteristics and incorporation method of the nanoparticles.

In what follows, different examples of the application potential of the use of
nanoparticles towards the development of novel or improved membrane-based water
treatment strategies, and the importance of understanding the underlying mechanisms and

fundamental aspects of nanocomposite membrane manufacture are discussed.

Multifunctional membranes

Multifunctional membranes couple separation with one or more additional functions,
allowing for a decrease in operational redundancy and thus lowering overall costs and

minimizing the environmental footprint.

The increasing availability and diversity of nanoparticles with unique physicochemical
properties is opening up new possibilities for the fabrication of multifunctional
membranes. An example of multifunctional membranes is that of membrane-based
sensors [3, 4], which couple the separation and pre-concentration functions afforded by
the membrane with additional sensing features, allowing for the detection, identiﬁcation
and quantiﬁcation of targeted compounds as they are separated from treated streams.
This dissertation (Chapter 4) describes a novel approach for the synthesis of a

nanoparticle-enabled membrane sensor.

Another example Of multifunctional membranes are those which can controllably release
a speciﬁc compound to the treated stream [5-7]. Such membranes are of special interest
for the mitigation of biofouling, a major cause of membrane breakdown and high
replacement costs. In particular, silver nanoparticles show promise as nanoscale ﬁllers
that can controllably release bactericidal silver ions directly into the membrane feed
stream and thus prevent the formation of bioﬁlm on the membrane surface. Chapter 3
describes the synthesis and characterization of polysulfone-Si] ver nanocomposite

membranes and their potential towards biofouling prevention and control.

Hydraulic properties

Wet phase inversion is the main method of preparation of integrally skinned, polymeric
ultraﬁltration membranes [8, 9] . A comprehensive, systematic study of the connection
between the process of phase inversion in nanoparticle-polymer blends, and the
properties of the resulting membranes is thus required. The results of the studies
published to date strongly imply that the microstructural changes induced by
nanoparticle incorporation can potentially be controlled, to prepare membranes with

desirable hydraulic properties.

Recent works on the incorporation of nanoparticles using the wet phase inversion process
have resulted in the development of membranes with enhanced hydraulic performance

(signiﬁcant simultaneous improvement of ﬂuxes and selectivities).

Available knowledge on the processes governing the formation of porous nanocomposite
membranes is insufﬁcient and limited in scope. Most studies published so far involved
the incorporation of passive ﬁllers into polymeric matrices and indicated that the
presence of nanoparticles affected both the porosity (pore size and density) of the
membrane skin and the macrovoid morphology of the asymmetric support layer [6, 10-
23]. A number of such published studies described the synthesis of organic-inorganic
porous (10 kDa to 200 kDa) composite membranes by adding particles to the solution of

the membrane-forming polymer [6, 10-23]. Incorporation of increasing loadings of silica

nanoparticles into polysulfone membranes was reported to result in higher porosity and
interconnectedness of macrovoid walls [12], increased density and size of surface pores
and thicker skin layer [11]. For clay-polysulfone composite membranes, skin layer
thickness was found to decrease with increasing clay loadings [21]. Macrovoid
suppression and increased porosity and thickness of the membranes skin layer for higher
ﬁller concentrations was also reported for hybrid TiOz-polysulfone ultraﬁltration
membranes [19]. Similarly, incorporation of activated carbon was found to diminish
macrovoid formation in composite polysulfone membranes; while the cross sectional
asymmetry and the uniformity in the mean pore size of membranes increased with an
increase in the size of carbon ﬁller particles [24]. In a recent study, nanotube ﬁllers were
used to prepare a nanocomposite PSf membrane with acid-treated multiwall carbon
nanotubes [25]. The average porosity Of the composite membranes peaked at a nanotube
loading of 1.5%, which also corresponded to the highest permeate ﬂux. Interestingly, the
only pronounced change in membrane macroscopic morphology was an increase in

surface roughness-

These studies proved that skin layer properties (porosity, surface porosity, thickness) as
well as the macrovoid morphology of the support layer can be affected by nanoparticle
incorporation. Both of these structurally distinct membrane regions (skin and support) are
important; the skin layer determines the permeability, rejection, and selectivity of
membranes, while the structure of the support layer affects its compaction behavior [26,

27].

It is thus important to understand the links between nanoﬁller shape, size and inclusion
methodology, (namely loading concentration and synthetic incorporation route); and, the
eﬁ‘ect of these parameters on the morphological structure of the nanoﬁlled porous matrix.

This is the main focus of Chapters 2 and 3 of this dissertation.

Mechanical properties

Mechanical stability is of major importance for membranes, as it is required to prevent
structural breakage at high pressures and thus ensure the membrane’s Operational
viability. Improvement of the mechanical properties of polymeric membranes via
nanoparticle incorporation into polymer blends is a design strategy that is earning

considerable attention.

In recent studies on the development of polymer nanocomposites, signiﬁcant
improvements in tensile stiffness and strength, toughness, rigidity and impact resistance
were found to be directly dependent on nanoparticle size and loading [28—33]. Tensile
strength and elongation-at-break of UP polyvinylidene ﬂuoride (PVDF) membranes
prepared by phase inversion were improved by 50% upon addition of 10 nm alumina
particles at a 2 % wt loading [15]. A signiﬁcant increase in tensile modulus was also
reported for polysulfone-clay composite membranes [21]. Yang et a1. [19] reported a

signiﬁcant improvement in breaking and bursting strength values for porous UF

polysulfone membranes upon incorporation Of Ti02 nanoparticles (30 nm) via phase

inversion. Separately, Li et al. [22] showed that Ti02 nanoparticle (10-20 nm) ﬁlled,

phase inversion polyphtalazinone ether sulfone ketone (PPESK) membranes exhibited a

remarkable enhancement of tensile strength as compared to Ti02 free membranes.

A common undesirable feature of porous polymeric membranes is the decrease in
permeate ﬂow due to compaction. Given their intrinsically ﬂexible nature and porous,
sponge-like morphology, the matrices Of polymeric membranes can be compressed under
sufﬁciently high pressure, resulting in pore constriction and a consequent decline in ﬂow.
Compaction is thus detrimental to process efﬁciency, due to the reduction of permeate
ﬂUx and increased energy costs, since higher pressures are needed to compensate for the
decline in ﬂux upon compaction. Mitigation of membrane compaction is yet another
beneﬁt of improving the mechanical robustness of membranes. The introduction of
nanoparticles to the bulk membrane material and could reduce the membrane

compaction and thus reduce operating costs.

Nunes et a1. [34] studied the in situ formation of silica nanoparticles within porous
polyetherimide (PEI) phase inversion membranes, used as supports for dense composite
membranes. The nanopatricle-ﬁlled membranes experienced a permeate ﬂow loss of only
10% upon compaction, as compared to a higher than 60% permeate loss for silica-free

membranes. In a later work, Ebert et al. [16] Observed a signiﬁcant reduction (more than
50 %) in ﬂux loss upon compaction for porous, phase inversion, TiOz nanoparticle ﬁlled

PVDF membranes, with respect to nanoparticle-free controls.

In all of the studies cited above, the dependence of mechanical strength on nanoparticle
loading showed a peak at a specific nanoparticle concentration. This observed nonlinear
dependence of mechanical strength on nanoparticle loading points to the importance of
understanding the eﬁ‘ects of nanoparticle incorporation on the mechanisms of membrane
formation and membrane performance, in order to allow for an optimized design of

mechanically robust nanocomposites.

Tunable hydrophilicity

The most common component of natural waters is highly hydrophobic organic matter. A
straightforward approach to reduce membrane fouling due to deposition of organic
matter on the membrane surface is therefore to reduce the aﬁ‘inity between the membrane
material and the foulants. This can be attained by increasing the hydrophilicity of the
membrane material, thus decreasing its propensity to adsorb hydrophobic organics from

the feed stream.

The integration ofnanoparticles and polymeric materials is an emerging approach to
modify the physico-chemical properties of the host polymer matrix. Incorporating
nanoparticles into the bulk polymer, as opposed the traditional strategy of applying
chemical pre-treatments or chemically active surface coatings offers the advantage of
minimizing the environmental footprint of the hybrid composite by reducing the risk of
release of toxic compounds derived from chemical treatments. An additional advantage

of nanoparticle inclusions vs. chemical coatings is the increased durability of the

incorporated feature, as it becomes part of the bulk material, rather than simply a surface

feature, more prone to be removed under ﬂow through conditions.

Few studies addressed the effect of incorporated particles on the hydrophilicity of.
resulting nanocomposite membranes. Polysulfone-clay phase inversion composite
membranes showed a 25% higher hydrophilicity when compared to clay free membranes
[21]. Increasing loadings of alumina (A1203) nanoparticles in phase inversion PVDF
membranes were correlated with an increase in hydrophilicity of the modiﬁed
membranes [15], while incorporation of titanium (TiOz) nanoparticles to phase inversion
PSf [19] and PPESK [22] membranes showed hydrophilicity enhancements of above

40% in both cases.

Given that the surface area Of nanoparticles is related to the nanoparticle size, one can
expect signiﬁcant hydrophilic enhancements as the size of hydrophilic nano inclusions is
reduced, at equal loadings. Furthermore, similar hydrophilic enhancements to those
observed for higher nanoparticle loadings or micro-scaled particles could be attained with
lower loadings of smaller sized nanoparticles; thus allowing for further optimization and

cost reduction strategies.

In light of the potential use of hydrophilic nanoparticles as membrane ﬁlelrs for the
mitigation of membrane fouling, it is important to study the eﬁect of nanoparticle loading

and size on the performance of nanoparticle-ﬁlled membranes.

10

Adaptive membranes

A relevant example of membranes with a dynamic function is stimuli-responsive (or
adaptive) membranes that can reversibly adjust their permeability or rejection in

response to external stimuli [3, 35-52].

In recent years, numerous studies have focused on the development a new generation of
magnetically responsive dense polymer nanocomposites, or magneto-elastomers,
obtained when magnetically active nanoscale inclusions are embedded into an elastic
polymeric matrix. The ability of such composites to reversibly change their structural and
mechanical properties as they are subject to an external magnetic ﬁled, is raising
considerable attention towards various applications including “smart” actuators, dampers

and artiﬁcial muscles, among others [53-55].

An appealing aspect of the phase inversion approach relies in the additional possibilities
for the tailored design of nanocomposite membranes and the coupling between their
hydraulic and mechanical properties and the nanoparticles reactivity to external stimuli.
Magnetically susceptible nanoparticles are of particular interest for the development of
such reactive, or adaptive, porous nanocomposites. The introduction of magnetically
active nanoparticles to the matrix of a polymeric membrane could allow for the
possibility of externally controlling, via application of a magnetic ﬁeld, the structural

morphology and consequently the hydraulic properties of the nanostructured membranes.

11

Nanoparticle loading, size and morphology should all be expected to affect the degree of
susceptibility of the nano-structured composite to the external stimuli. Supra-magnetic
properties of nanoparticles are enabled and dictated by their size, as they are intrinsically
a surface-based effect. As reported by Thapa et al. [56], the enhancement of saturation
magnetization in magnetite nanoparticles was strongly size dependent, increasing at
lower particle diameters, but reaching a size threshold beyond which a sudden drop in

magnetization was observed.

Additionally, nanoparticle aspect ratio (e. g. nano-rods vs. nano-spheres) is expected to be
of critical relevancefor the development Of mechanically adaptive nanocomposites,
allowing for the anisotropic alignment of nano-inclusions within the polymeric matrix, to
achieve direction dependant enhancements of mechanical properties. Thostenson et al.
[57] compared the tensile behavior of polystyrene ﬁlms doped with aligned (via
extrusion) carbon nanotubes, with respect to non—aligned nanotube ﬁlled polystyrene
ﬁlms. Nanotube alignment resulted in a signiﬁcant improvement in the elastic modulus of
the composites, as compared to randomly oriented nanotube-polystyrene ﬁlms. In a later
work, Camponeschia et al. [58] showed that the mechanical properties (glass transition
temperature, Young’s modulus and elastic and viscous moduli) of carbon nanotube doped
epoxy composites could be altered upon exposure to an external magnetic ﬁeld.
Furthermore, upon magnetic alignment, the composites showed enhancements in their

mechanical properties that were not Observed for non-aligned composites.

12

The above ﬁndings point to promising research opportunities involving the development
of magnetically adaptive porous nanocomposites, and stress the importance of studying
the interplay between nanoparticle loading, size and morphology and the performance of

nanoparticle-ﬁlled composites.

l3

Dissertation Overview

This work explores various aspects of the fabrication of nanoparticle-enabled membranes
and its potential environmental applications. As follows, a brief summary of each chapter

is offered.

Chapter 1 is introductory. It offers a theoretical background covering the thermodynamic
and kinetic principles that govern the wet phase inversion process, which is one of the

synthetic methods used to fabricate polymeric membranes (see Chapters 2 and 3.)

Chapter 2 focuses on the effect of nanoparticle shape and loading on the hydraulic
resistance and rejection of polysulfone membranes ﬁlled with two types of carbon
nanoparticles: Fullerenes (nanospheres) and Single-walled carbon nanotubes (nanorods).
The rate of precipitation front advancement and viscosity of the casting mixtures are
found to correlate with the permeability and rejection of cast nanocomposite membranes.
These correlations are rationalized in terms of the effect of nanoparticle ﬁllers on the
demixing and gelation processes occurring during phase inversion. The propagation rate
of the precipitation front and the viscosity of the casting mixture are proposed as simple-
to—measure predictors of membrane performance with potential application for the

rational design of porous nanocomposites.

Chapter 3 describes the synthesis of ﬂat sheet porous polysulfone-silver nanocomposite

membranes by the wet phase inversion process. In particular, the effects of casting

14

mixture composition and nanoparticle incorporation route on the morphological and
separation properties of the prepared membranes. Furthermore, the built-in antibacterial
capacity due to the gradual release of ionic silver by the prepared nanocomposites was
found to be effective in reducing intrapore biofouling. Such nanocomposites could in
principle be used as materials for macroporous membrane spacers to inhibit the bioﬁlm

growth on downstream membrane surfaces.

Finally, Chapter 4 presents a method for the synthesis of a nanoparticle-enabled,
membrane-based sensor. Silver nanoparticle arrays are deposited on the surface of
polycarbonate track etch membranes. The nanoparticle-modiﬁed membranes are
characterized in terms of their hydraulic and optical properties and evaluated as ﬂow-
through surface—enhanced Raman scattering (SERS) sensors for water quality monitoring.
The membrane-supported nanoparticle arrays were found to be SERS-active with slightly
lower but signiﬁcantly more reproducible enhancements in comparison with
enhancements afforded by source nanoparticle suspensions. The ﬁndings point to the
promise of combining high speciﬁcation SERS-active systems and the ﬂow-through
design for the development of membrane based analytical sensors for the trace detection

of aqueous pollutants.

15

References

10.

11.

12.

W. Liu, Nanoparticles and their biological and environmental applications, J.
Biosci. Bioeng. 102 (2006) l—7.

F. E. Kruis, H. Fissan and A. Peled, Synthesis of nanoparticles in the gas phase
for electronic, Optical and magnetic applications--a review, J. Aerosol Sci. 29
(1998) 511-535.

B. Adhikari and S. Majumdar, Polymers in sensor applications, Prog. Polym. Sci.
29 (2004) 699-766.

P. Vadgama, Membrane based sensors: A review, J. Membr. Sci. 50 (1990) 141-
152.

D. G. Yu, M. Y. Teng, W. L. Chou and M. C. Yang, Characterization and
inhibitory effect of antibacterial PAN -based hollow ﬁber loaded with silver
nitrate, J. Membr. Sci. 225 (2003) 115.

W. L. Chou, D. G. Yu and M. C. Yang, The preparation and characterization of
silver-loading cellulose acetate hollow ﬁber membrane for water treatment,
Polym. Adv. Technol. 16 (2005) 600 - 607.

R. Bhattacharya, T. N. Phaniraj and D. Shailaja, Polysulfone and polyvinyl
pyrrolidone blend membranes with reverse phase morphology as controlled
release systems: experimental and theoretical studies, J. Membr. Sci. 227 (2003)
23.

R. E Kesting, Synthetic Polymeric Membranes, A Structural Perspective. Second
ed. 1985, New York: John Wiley & Sons.

M. Mulder, Basic Principles of Membrane Technology. Second ed. 2003,
Dordrecht, Netherlands: Kluwer Academic.

A. Bottino, G. Capannelli and A. Comite, Preparation and characterization of
novel porous PVDF-Zr02 composite membranes, Desalination 146 (2002) 35-40.

P. Aerts, I. Genne, S. Kuypers, R. Leysen, I. F. J. Vankelecom and P. A. Jacobs,
Polysulfone-aerosil composite membranes: Part 2. The inﬂuence of the addition
of aerosil on the skin characteristics and membrane properties, J. Membr. Sci. 178
(2000) 1-1 1.

P. Aerts, E. Van Hoof, R. Leysen, I. F. J. Vankelecom and P. A. Jacobs,
Polysulfone-aerosil composite membranes: Part 1. The inﬂuence of the addition
of Aerosil on the formation process and membrane morphology, J. Membr. Sci.

176 (2000) 63—73.

16

l3.

14.

15.

l6.

17.

18.

19.

20.

21.

22.

23.

24.

N. M. Wara, L. F. Francis and B. V. Velamakanni, Addition of alumina to
cellulose acetate membranes, J. Membr. Sci. 104 (1995) 43-49.

I. Genne, S. Kuypers and R. Leysen, Effect of the addition of ZrO2 to polysulfone
based UF membranes, J. Membr. Sci. 113 (1996) 343-350.

L. Yan, Y. S. Li, C. B. Xiang and S. Xianda, Effect of nano-sized A1203-particle
addition on PVDF ultraﬁltration membrane performance, J. Membr. Sci. 276
(2006) 162-167.

K. Ebert, D. Fritsch, J. K011 and C. Tjahjawiguna, Inﬂuence of inorganic ﬁllers on
the compaction behaviour of porous polymer based membranes, J. Membr. Sci.
233 (2004) 71-78.

D. G. Yu, M. Y. Teng, W. L. Chou and M. C. Yang, Characterization and
inhibitory effect of antibacterial PAN -based hollow ﬁber loaded with silver
nitrate, J. Membr. Sci. 225 (2003) 115-123.

X. He, Q. Shi, X. Zhou, C. Wan and C. Jiang, In situ composite of nano SiO2-
P(VDF-HFP) porous polymer electrolytes for Li-ion batteries, Electrochim. Acta
51 (2005) 1069-1075.

Y. Yang, H. Zhang, P. Wang, Q. Zheng and J. Li, The inﬂuence of nano-sized
Ti02 ﬁllers on the morphologies and properties of PSF UF membrane, J. Membr.
Sci. 288 (2007) 231-238.

W. K. Son, J. H. Youk, T. S. Lee and W. H. Park, Preparation of antimicrobial
ultraﬁne cellulose acetate ﬁbers with silver nanoparticles, Macromol. Rapid
Commun. 25 (2004) 1632-1637.

0. Monticelli, A. Bottino, I. Scandale, G. Capannelli and S. Russo, Preparation
and properties of polysulfone-clay composite membranes, J. Appl. Polym. Sci.
103 (2007) 3637-3644.

J. B. Li, J. W. Zhu and M. S. Zheng, Morphologies and properties of
poly(phthalazinone ether sulfone ketone) matrix ultraﬁltration membranes with
entrapped T102 nanoparticles, J. Appl. Polym. Sci. 103 (2006) 3623-3629.

A. Bottino, G. Capannelli, V. D'Asti and P. Piaggio, Preparation and properties Of
novel organic-inorganic porous membranes, Sep. Purif. Technol. 22-23 (2001)
269-275.

L. Ballinas, C. Torras, V. Fierro and R. Garcia-Valls, Factors inﬂuencing
activated carbon-polymeric composite membrane structure and performance, J.
Phys. Chem. Solids 65 (2004) 633-637.

17

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

J .-H. Choi, J. Jegal and W.-N. Kim, Fabrication and characterization of multi-
walled carbon nanotubes/polymer blend membranes, J. Membr. Sci. 284 (2006)
406-415.

K. M. Persson, V. Gekas and G. Tragardh, Study of membrane compaction and its
inﬂuence on ultraﬁltration water permeability, J. Membr. Sci. 100 (1995) 155-
162.

G. Jonsson, Methods for determining the selectivity of reverse osmosis
membranes, Desalination 24 (1977) 19-37.

R. Kotsilkova, D. Fragiadakis and P. Pissis, Reinforcement effect of carbon
nanoﬁllers in an epoxy resin system: Rheology, molecular dynamics, and
mechanical studies, J. Polym. Sci. Pol. Phys. 43 (2005) 522-533.

A. Zhu, A. Cai, J. Zhang, H. J ia and J. Wang, PMMA-grafted-silica/PVC
nanocomposites: Mechanical performance and barrier properties, J. Appl. Polym.
Sci. 108 (2008) 2189-2196.

M. Q. Zhang, M. Z. Rong, H. B. Zhang and K. Friedrich, Mechanical properties
of low nano-silica ﬁlled high density polyethylene composites, Polym. Eng. Sci.
43 (2003) 490-500.

J. N. Coleman, U. Khan and Y. K. Gun’ko, Mechanical reinforcement of
polymers using carbon nanotubes, Adv. Mater. 18 (2006) 689-706.

M. I. Sarwar, S. Zulﬁqar and Z. Ahmad, Polyamide-silica nanocomposites:
mechanical, morphological and therrnomechanical investigations, Polym. Int. 57

(2008) 292-296.

K. Yang, Q. Yang, G. Li, Y. Sun and D. Feng, Mechanical properties and
morphologies of polypropylene with different sizes of calcium carbonate
particles, Polym. Compos. 27 (2006) 443-450.

S. P. Nunes, K. V. Peinemann, K. Ohlrogge, A. Alpers, M. Keller and A. T. N.
Pires, Membranes of poly(ether imide) and nanodispersed silica, J. Membr. Sci.
157 (1999) 219-226.

Y. Wang, Z. Liu, B. Han, Z. Dong, J. Wang, D. Sun, Y. Huang and G. Chen, pH
Sensitive polypropylene porous membrane prepared by grafting acrylic acid in
supercritical carbon dioxide, Polymer 45 (2004) 855-860.

N. Minoura, K. Idei, A. Rachkov, Y. W. Choi, M. Ogiso and K. Matsuda,
Preparation of azobenzene-containing polymer membranes that function in
photoregulated molecular recognition, Macromolecules 37 (2004) 9571-9576.

18

37.

38.

39.

40.

41.

42.

43.

45.

46.

47.

48.

S. Petrov, T. Ivanova, D. Christova and S. Ivanova, Modiﬁcation of
polyacrylonitrile membranes with temperature sensitive poly(vinylalcohol-co-
vinylacetal), J. Membr. Sci. 261 (2005) 1-6.

W. Lequieu', N. I. Shtanko and F. E. Du Prez, Track etched membranes with
therrno-adjustable porosity and separation properties by surface immobilization of
poly(N-vinylcaprolactam), J. Membr. Sci. 256 (2005) 64-71.

N. I. Shtanko, V. Ya Kabanov, P. Yu Apel, M. Yoshida and A. I. Vilenskii,
Preparation of permeability-controlled track membranes on the basis of ‘smart'
polymers, J. Membr. Sci. 179 (2000) 155-161.

L. Liang, X. Feng, L. Peurrung and V. Viswanathan, Temperature-sensitive
membranes prepared by UV photopolymerization of N -isopropylacrylamide on a
surface of porous hydrophilic polypropylene membranes, J. Membr. Sci. 162
(1999) 235-246.

J. F. Hester, S. C. Olugebefola and A. M. Mayes, Preparation of pH-responsive
polymer membranes by self-organization, J. Membr. Sci. 208 (2002) 375-388.

A. Kumar, A. Srivastava, I. Y. Galaev and B. Mattiasson, Smart polymers:
Physical forms and bioengineering applications, Prog. Polym. Sci. 32 (2007)
1205-1237.

T. Peng and Y. L. Cheng, pH—responsive permeability of PE-g-PMAA
membranes, J. Appl. Polym. Sci. 76 (2000) 778-786.

L. Ying, P. Wang, E. T. Kang and K. G. Neoh, Synthesis and characterization of
poly(acrylic acid)-graft-poly(vinylidene ﬂuoride) copolymers and pH-sensitive
membranes, Macromolecules 35 (2002) 673 -679.

H. Iwata, I. Hirata and Y. Ikada, Atomic force microscopic analysis of a porous
membrane with pH-sensitive molecular valves, Macromolecules 31 (1998) 3671 -
3678.

H. Iwata, M. Oodate, Y. Uyama, H. Amemiya and Y. Ikada, Preparation of
temperature-sensitive membranes by graft polymerization onto a porous
membrane, J. Membr. Sci. 55 (1991) 119-130.

W. Y. Wang, L. Chen and X. Yu, Preparation of temperature sensitive
poly(vinylidene ﬂuoride) hollow ﬁber membranes grafted with N-
isopropylacrylamide by a novel approach, J. Appl. Polym. Sci. 101 (2006) 833-
837.

L. Y. Chu, T. Niitsuma, T. Yamaguchi and S. Nakao, Therrnoresponsive transport

through porous membranes with grafted PNIPAM gates, AICHE J. 49 (2003)
896-909.

19

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

X. Liu, K. G. Neoh and E. T. Kang, Redox-sensitive microporous membranes
prepared from poly(vinylidene ﬂuoride) grafted with viologen-containing
polymer side chains, Macromolecules 36 (2003) 8361 -8367.

Y. Ito, S. N ishi, Y. S. Park and Y. Irnanishi, Oxidoreduction-sensitive control of
water permeation through a polymer brushes-grafted porous membrane,
Macromolecules 30 (1997) 5856 -5859. ,

S.A. Piletsky, T.L. Panasyuk, E.V. Piletskaya, A.V. El'skaya, R. Levi, I.
Karube and G. Wulff, Imprinted membranes for sensor technology: Opposite

behavior of covalently and noncovalently imprinted membranes, Macromolecules
31 (1998) 2137-2140.

T. Yamaguchi, T. Ito, T. Sato, T. Shinbo and S. Nakao, Development of a fast
response molecular recognition ion gating membrane, J ACS 121 (1999) 4078-
4079.

G. Filipcsei, I. Csetneki, A. Szilagyi and M. Zrinyi, Magnetic ﬁeld-responsive
smart polymer composites, in Oligomers Polymer Composties Molecular
Imprinting. 2007. p. 137-189.

1. Park, K. J. Kim, J Nam, J. Lee and W. Yim, Mechanical, dielectric, and
magnetic properties of the silicone elastomer with multi-walled carbon nanotubes
as a nanoﬁller, Polym. Eng. Sci. 47 (2007) 1396-1405.

Z. Varga, G. Filipcsei, A. Szilagyi and M. Zrinyi, Electric and Magnetic Field-
Structured Smart Composites, Macromol. Symp. 227 (2005) 123-134.

D. Thapa, V. R. Palkar, M. B. Kurup and S. K. Malik, Properties of magnetite
nanoparticles synthesized through a novel chemical route, Materials Letters 58
(2004) 2692-2694.

B. T. Thostenson and T. W. Chou, Aligned multi-walled carbon nanotube-
reinforced composites: processing and mechanical characterization, Journal Of
Physics D-Applied Physics 35 (2002) L77-L80.

E. Camponeschi, R. Vance, M. Al-Haik, H. Garmestani and R. Tannenbaum,

Properties of carbon nanotube-polymer composites aligned in a magnetic ﬁeld,
Carbon 45 (2007) 2037-2046.

20

CHAPTER ONE
Theoretical backgorund of the phase inversion process

This chapter provides a comprehensive theoretical framework regarding the basic
principles that govern the wet phase inversion process; as this is the synthetic method
used to fabricate the porous polymeric matrices in the work detailed in Chapters Two and

Three.

1.1. Phase Inversion

The vast majority of both dense and porous commercially available polymeric
membranes, as well as the porous supports of composite membranes, are produced by the

phase-inversion process [1-3].

The phase inversion process was ﬁrst applied in the 1960s, when Loeb and Sourirajan [4]
developed the method to prepare asymmetric, integrally skinned, dense reverse osmosis
membranes. The development of polymeric asymmetric membranes constituted a turning
point towards overcoming a fundamental barrier for the extensive application of
membranes; the achievement of simultaneous high selectivities and permeation ﬂuxes.
Such breakthrough is now recognized as the milestone that gave rise to the expansion of

membrane technologies, and ultimately led to the widespread commercial application Of

21

membranes. Later, in 1971, Michaels [5] adapted the method to produce ultraﬁltration

membranes using a variety of commercial polymers.

In the phase inversion process, a viscous polymer solution is cast into a thin ﬁlm or a
hollow ﬁber, and immersed into a homogeneous phase (non solvent) which has a high
afﬁnity for the polymer solvent, but low or no afﬁnity for the polymer. As both phases
are brought into contact, a mass transfer process is initiated as solvent and non solvent are
exchanged, eventually resulting in the precipitation of a solid porous polymeric matrix

that constitutes the membrane.

The described “demixing” of the polymer solution can be brought about using a variety

of strategies that give rise to different phase inversion alternatives.

When the immersion phase is the atmosphere, or dry phase inversion, the polymer
precipitates as the solvent evaporates from the polymer solution and/or a non solvent

(dissolved in the gas phase) slowly diffuses into the viscous ﬁlm.

Phase inversion can also be induced by using a latent polymer solvent; which‘is a
solution component that behaves as a solvent at high temperatures, but becomes the non
solvent as temperature decreases; such is the underlying principle of the thermally

induced phase inversion process.

22

Finally, when the immersion medium is a non solvent liquid, the process is known as wet
phase inversion. This process can be as well preceded by a brief solvent evaporation
stage (dry-wet phase inversion). The wet phase inversion process is the preferred method
for the synthesis of porous membranes; for this reason, further analyses will be focused
on the basic principles of wet phase inversion and the effect of different process

parameters on the morphological features of the cast membranes.

The following sections offer a detailed qualitative description of the basic concepts and
fundamental thermodynamic and kinetic principles affecting membrane formation during
wet phase inversion. The understanding of such principles is of critical importance as
they dictate the membrane’s ﬁnal structural morphology, and thus have a critical impact

on its performance and viability.

1.2.Liquid-Liquid Demixing

In all phase inversion methods, the process starts with a thermodynamically stable
polymer solution, or casting mixture. For illustrative simplicity, the solution will be
composed solely of polymer and solvent. It must be noted though that in many cases
three or four solution components are used, incorporating additional pore forming agents

or a small fraction of non solvent into the casting mixture.

For a binary (polymer/solvent) system, the free energy of mixing (AGm), as derived from

the Flory-Huggins theory [6], is given by:

23

AG,"
RT

 

= ”11n¢1+nzln¢2 "”"1¢2;i.I

Where, n1 and 11; represent the total moles of solvent and polymer in the mixture,
respectively; (1)] and (p2 are the volume fractions Of solvent and polymer in the mixture

and x is the Flory-Huggins solvent-polymer interaction parameter, which encompasses
all excess enthalpic contributions due to the non ideal behavior of the solution

constituents.

The variation in chemical potential (Art) for each component upon mixing can be

obtained as:

Where [.10 represents the chemical potential Of the pure component.
For a given overall solvent/polymer ratio, at a given temperature, the two components

will tend to minimize their respective Au. Such tendency can be better visualized by

analyzing a polymer solution that is subject to progressive cooling.

24

At a sufﬁciently high temperature (T4), the two components will tend to mix at any

relative composition. This is illustrated in Figure 1.1.

25

the
>

 

 

 

 

 

 

Figure 1.1.Plot of AG", vs. (1)2 for the pol ymer/solvent system at temperature T4, above
the critical point.

26

At an arbitrary composition A, both components are completely mixed. The variation in

the chemical potential for each component upon mixing, with respect to each component

in its pure (non-mixed) state (Am and Aug) can be obtained from the intersections of the

tangent to the AGm curve (at composition A) with each Of the pure component axes.

As can be seen on Fig. 1.1, both All] and Au; are negative, indicating a decrease in the

chemical potentials for both components upon mixing, and thus a favorable mixing

tendency.

Let us now consider a hypothetical demixing process, by which the two components
separate from each other to form two separate phases, each of which is predominantly
concentrated in one of the components (points B and C). For such demixing to occur,
component 1 has to be transferred from the original mixture (A) to the new, component 1
rich, phase B. However, as can be seen from the graph, this transference process would
result in a tangent (at point B) line that would produce an increase (with respect to point
A) in the chemical potential for component 1. A similar analysis can be done for the
formation of phase C, in this case, this would result in an increase in the chemical
potential for component 2 (point C). The chemical potential increase for each component
is shown by the dotted arrows. Consequently, the demixing process is thermodynamically
unfavorable (the components tend to decrease their chemical potentials and thus the free
energy Of the system), and the two components will remain completely mixed with

composition A.

27

The situation, however, changes when temperature is decreased. Below a given

temperature - or Critical Temperature (TC) - which is speciﬁc to the studied component

pair; the AG," curve starts eliciting two inflection points (A’l and A’z), as shown in

Figure 1.2.

28

E
El

_.>
3?.

>..
N

>
N

 

Meta-stable
region

$2

——>

 

T2

Iél Unstable I ;:
Iéi region Iéi
r/ A I/r
in- ' :é:
~ ~‘ :/:
Polymer ““n. '/I
lean ~- \/,.
Polymef ‘ ~ - . ~
ﬁCh “ ‘~]

 

 

Figure 1.2.Polymer/solvent system’s AGm vs. O2 plot at a temperature T2 below the

critical point

29

The intersection of the tangent to the curve (at point A) with the pure component 1 axis
now yields a positive Au]. The original mixture of composition A (Fig. 1.1), stable at T4,

now lies in a thermodynamically unstable region and the mixture will thus tend to demix.
The system will demix by “pulling” the two components apart, forming two separate
phases, each of which will be relatively richer in one of the components. By doing so, the
system separates into two liquid phases, until a new equilibrium state of minimum free

energy is reached. This process is know as Liquid-Liquid (L-L) demixing.

From unstable composition A, the two components will tend to reduce their respective
chemical potentials, by moving along the AGm curve towards curve points where their
chemical potentials are reduced with respect to those in initial point A, in other words,

along points where 5Alli/5¢ < 0. This negative chemical potential variation region will

occur between points A and A’l for component 1 (solvent) and between points A and A’2

for component 2 (polymer). For this reason, this region is called the demixing or unstable

region, and any mixture inside this region will spontaneously experience L-L demixing,

compositions A’l and A’z are called the spinodal points. As demixing unfolds,
component 1 will move towards point A’1 and component 2 will do so towards A’2,
thereby progressively forming two separate phases enriched in either component 1 (A’l)

or component 2 (A’z). The demixing process will continue until the two phases reach

equilibrium with each other, or in thermodynamical terms, until the overall variation in

chemical potentials for each component is the same in both phases. This ﬁnal state is

shown in ﬁgure 1.2: Points A1 and A2; that share a common tangent, indicating they have

30

reached the same ﬁnal chemical potentials in the AGm curve. These two compositions
(A1 and A3) are knows as the binodal compositions. The common tangent line connecting

the binodal points is known as a tie line.

As the two components demix and separate, two distinct phases are formed. We will now

focus the analysis on the path followed by the polymer (component 2) molecules.

As demixing unfolds, the solvent molecules will progressively “dc-solve” the polymer
molecules. As solvent and polymer molecules separate, a ﬁnely interconnected network
of polymer lean and polymer rich regions are created. This process is known as spinodal

demixing, and will continue as long as there is a driving force for spontaneous demixing,
that is, SAM/5(1) < 0. Eventually, SAM/5(1) becomes cero at inﬂexion —spinodal- point A’g,

and there is thus no longer a driving force for the spontaneous separation of polymer

from the solvent within the system. However, the solvent has not yet reached its spinodal
point (A’ 1); this creates a thermodynamically meta-stable region (from A’2 to A2). Within
this region, the polymer rich phase will remain stable for small compositional
ﬂuctuations in the surrounding system (since SAM/5(1) 2 0). Simultaneously, component 1

(the solvent) is following a parallel trend as it moves towards the opposite end of the Gm

curve, forming the polymer lean (or solvent rich) phase Al, and thus continuously

“releasing” solvent to the system. These “free” solvent molecules will initially become
part of the polymer lean phase during spinodal demixing, however, when the polymer

rich phase has reached its meta-stable region, new solvent molecules will start forming

31

isolated nuclei within the continuous polymer rich phase. As the solvent molecules
separate from the polymer molecules to form the polymer lean phase, these freed polymer
molecules will create a thermodynamical pressure against the polymer rich phase, which
is now in its meta-stable state. The polymer molecules will be gradually incorporated to

the meta-stable polymer rich phase, “pushing” its composition towards stable binodal
point A2. What was originally an interconnected network of two coexisting phases during

spinodal demixing, now becomes a distinctly separated system of growing nuclei of
polymer lean phase within a “sea“ of polymer rich phase. This distinct demixing process
occurring within the meta-stable region, by which the system equilibrates itself via
formation and growth of polymer lean nuclei of composition Al within the polymer rich

phase, is called demixing by nucleation and growth. In parallel, and analogously, the

polymer lean phase evolves to its own stable composition A].

The preceding discussion focused on the formation of polymer lean phase nuclei within a
polymer rich phase. However, the inverse process can also occur, that is, the formation of
polymer rich blobs in a sea of solvent rich phase; what will determine the kind of
nucleation and growth mechanism will be which of the two phases reach the spinodal
point ﬁrst. If the spinodal point is ﬁrst reached by the polymer rich phase; then the
solvent will constitute the forming nuclei, as described above, if instead the polymer lean
phase reaches its spinodal point before the polymer rich phase, then polymer blobs will

grow within the solvent rich phase.

32

Figure 1.3 shows the AG“, curves for the polymer/solvent system at different

temperatures. Below a critical temperature (TC), the AGm curves start showing spinodal

and binodal points, the ﬁrst spinodal/binodal pair coinciding in the critical point. By
joining the binodal and spinodal points for each temperature; the binodal and spinodal
curves for the system, as a function of temperature, are Obtained. The miscibility gap is
delimitated by the binodal curve; any point lying inside the binodal curve will experience
demixing, and will separate into two distinct phases (polymer lean and polymer rich) of
equilibrium compositions given by the intersections between the tie lines and the binodal
curve, at a given temperature. A mixture, stable at a high temperature, will enter the
miscibility gap as it is subject to progressive cooling. If it enters the miscibility gap on
the left side of the critical point, that is closer to the polymer lean spinodal points, then
demixing will occur by nucleation and growth of the polymer rich phase (polymer blobs
within a “sea” of polymer lean phase); the opposite nucleation and growth process
(polymer lean nuclei growing in a polymer rich phase) will hold if the gap is entered on

the right side of the critical point. Figure 1.3 (adapted from [7]) shows the starting stable
point (A) of the described system at T4 and the endpoint binodal compositions of the
separated equilibrium phases (A; and A2) after cooling down the system to an arbitrary

temperature T1. In this case, demixing will occur by nucleation and growth of the

polymer lean phase.

33

 

   
    

Binodal curve Critical point

A Spinodal curvy
I

 

      

 

 

 

 

Figure 1.3.AGm vs. (I32 curves for the polymer/solvent system at different temperatures,

from the completely mixed system at T4, above the critical temperature (Tc), to T1, ﬁnal
temperature of the demixed system.

34

1.3.Ternary system demixing

The offered thermodynamic analysis applies to the demixing of a polymer/solvent
mixture as it is cooled down; wherein thermodynamic destabilization is induced by a
temperature decrease. However, temperature decrease is not the only way to destabilize a
polymer/solvent mixture. Demixing of the system can also be brought about by
incorporating a third component into the system; a non-solvent. At a constant

temperature, the progressive incorporation of non solvent to the system will have a
similar destabilizing effect (and AGm curves deformation patterns) than that of

progressive cooling for the polymer/solvent mixture. An equivalent analysis to the one
offered for cooling induced demixing can be performed for non-solvent induced
demixing. Analogously, binodal and spinodal curves can be obtained beyond a critical
point, this time not as a function of decreasing temperature, but as a function of an

increasing non-solvent concentration in the system.

In a three-component system (polymer, solvent and non solvent), the relative composition
of each component in the system can be graphically represented using a ternary phase
diagram (Fig. 1.4). Each vertex of the diagram represents the pure component (100 %
relative concentration), while the triangle side that is opposite to each component vertex
represents its respective zero concentration. For an arbitrary three-component mixture
(star point in Fig. 1 .4), the relative concentration of each component in the system can be
Obtained by drawing a line parallel to the zero axis for that component. The relative

position of the line between the pure component vertex and its respective zero axis will

35

represent its relative concentration within the system. This is exempliﬁed and illustrated
in Figure 1.4. For the example mixture the relative component concentrations are 15, 40

and 45% for non solvent, solvent and polymer, respectively.

36

Solvent

 

 

Non Solvent 45% Polymer

Figure 1.4.Ternary phase diagram of a three component system (solvent, non-solvent and
polymer).

37

As explained before, an increase in non solvent concentration will result in the demixing
of the system into two separate phases. The demixing region, or miscibility gap, can be
schematically represented by the critical point, and binodal and spinodal curves. In this
case, the critical point and demixing curves will appear as we move from the zero non

solvent axis (polymer/solvent axis) towards the pure non solvent vertex, as shown in Fig.

1.5.

38

Solvent

Binodal
curve

Spinodal
curve

   
 

Critical Point

Tie line

Non Solvent Polymer

Figure 1.5.Solubility diagram for the three component system, showing the binodal and
Spinodal curves, tie lines, and the critical point.

39

A ﬁnal consideration must be given to the solidiﬁcation of the polymer network.
Depending on the arrangement of the polymer molecules in the solid state, polymers can
be classiﬁed into three categories: Crystalline, with a regular and highly ordered
molecular arrangement in the solid state; amorphous, with randomly packed molecules in
no discernible order, and semi-crystalline, where both crystalline and amorphous regions
coexist in the solid state. Since the polymer of choice for the present work — Polysulfone
- is an amorphous polymer, the analysis will be focused on the solidiﬁcation of this type

of polymers.

The solid state for amorphous polymers can be visualized as a frozen (or “glassy”) state

of their randomly oriented molecules, which can be brought about by cooling the system
past a certain glass transition temperature (Tg). Beyond the Tg, the mobility of the

polymer chains is signiﬁcantly reduced and the solid state is achieved. This solidiﬁcation

process is known as vitriﬁcation.

An additional solidiﬁcation phenomenon of critical importance for membrane formation
is gelation. Amorphous gelation is a physical transition which results in a highly elastic,
solid three dimensional network, or gel. This transition can occur due to an increase in
the viscosity of a homogeneous polymer/solvent solution beyond a ﬂuidity threshold (at
increasingly higher polymer concentrations); or by enhancing the formation of polymer-
polymer hydrophobic bonds, reducing the mobility of the molecules to ﬁnally achieve the

gelled state. The latter can be brought about by incorporation of non solvent to a ternary

40

system, and can induce gelation at a lower polymer concentration threshold than that

needed for viscosity increase induced gelation in a binary mixture.

In a ternary system, gelation can occur at constant temperature, both due to breaching of
the viscosity threshold as the concentration of the polymer within the polymer rich phase
increases during demixing, and also due to non solvent induced enhancement of polymer
chain interactions. Gelation will thus occur beyond a polymer concentration threshold
within the studied ternary system, and is independent of the vitriﬁcation process. If the
molecules are sufﬁciently immobilized, the gel will not experience molecular structural
changes, while it may continue to loose liquid due to outward diffusion. Otherwise, and
depending upon the degree of immobilization of the polymer chains, the forming gel can
further shrink and contract its molecular structure as polymer molecules get entangled by
formation of hydrophobic bonds, resulting in an increased compaction of polymer

molecules and release of liquid within the gel matrix, in a process known as syneresis.

All Of the above discussed processes are illustrated in Fig. 1.6. Note that the ternary

diagram has been rotated for illustrative purposes.

41

Polymer

  
   

Gel
transition

G-L transition
tie line

 

 

 

Solvent Non Solvent

Figure 1.6.Demixing and gelation regions for the three component system.

42

The ﬁgure shows four distinct regions. A mixture point within each region will
experience different thermodynamic states and thus demixing or solidiﬁcation driving
forces. Within the thermodynamically stable (liquid state) region for the three component
mixture (L), delimitated by the gel transition line and the binodal curve, all mixture
components will be completely mixed forming a homogenous solution (A). As the
polymer concentration in solution increases at constant non solvent concentration (path
A-B); it will cross the gelation threshold (gel transition line) and enter the gelation region
(G), encompassed beyond the gel transition line. The solution will then experience
homogeneous gelation without a preceding liquid demixing step. Two other regions can
be deﬁned; both within the miscibility gap inside the binodal curve. These regions can be
entered by increasing the non solvent content in the mixture, moving towards the non
solvent/polymer axis (path A-C). The ﬁrst region (L-L) involves the formation of two
separate liquid phases due to liquid-liquid demixing. As explained in the previous
section, if a mixture in the stable region (L) enters the miscibility gap below the critical
point, and is able to reach the meta-stable region before gelling, demixing will occur by
nucleation and growth of the polymer rich phase, while if it enters the gap above the
critical point (path A-C), it will demix by nucleation and growth of the polymer lean
phase. If gelation occurs before reaching the meta-stable region, the demixing system will
be “frozen” in the spinodal state. The compositions of each demixed phase will be given
by the intersections of the tie lines with the opposite branches of the binodal curve. As
demixing progresses before gelation, the composition of the polymer rich phase will
follow the upper branch of the binodal curve, towards the polymer/non solvent axis.

Beyond a given non solvent concentration in the mixture (G-L transition tie line, point C)

43

the polymer rich phase will have a polymer concentration that falls within the gelation
region, the G-L transition tie line marks the beginning of the G-L region, that is, the
region where a gelled polymer rich phase coexists with the liquid polymer lean phase.
Once it has entered the G region, the gelled phase can either experience further syneresis,
contracting its polymer structure, or simply continue to loose liquid by diffusion, with a

ﬁxed polymer structure.

1.4.Wet phase inversion

We can now analyze the membrane formation process during wet phase inversion. As
explained in the previous sections, wet phase inversion occurs when a stable polymer
mixture (binary or ternary) is immersed in a non solvent bath. The progressive exchange
of solvent and non solvent within the system will eventually cause the precipitation of a

polymeric matrix that constitutes the membrane.

In order to better understand the evolution of the system as it undergoes phase inversion,
both diffusional and thermodynamic phenomena must be taken into consideration. As the
homogeneous solution is cast into a thin ﬁlm on a support (usually a glass plate) and
immediately immersed in a non solvent bath (Fig. 1.7), the solvent, which must have a

certain degree of afﬁnity for the non solvent, experiences a diffusional drive towards the

immersion bath (J 5), while the non solvent does so towards the cast mixture (J NS).

44

. Non solvent bath
JS JNS

    
 
 

Cast
mixture

Glass plate (support)

Figure 1.7 .Schematic of solvent (J s) and non-solvent (J NS) transport for the wet phase
inversion process.

45

The relative magnitudes of these two ﬂows will have a critical effect on both the degree
of asymmetry and the porosity of the ﬁnal membrane. As soon as the ﬁlm is immersed
into the non solvent bath, the solvent molecules at the ﬁlm/bath interface, which have a
high afﬁnity for the non solvent, will experience a high diffusive drive towards the non
solvent bath. The non solvent molecules will also experience a diffusive drive towards
the non solvent poor ﬁlm; however, they will also encounter a diffusive resistance from
the polymer molecules in the ﬁlm, which have no afﬁnity for the non solvent molecules.
Such a difference between the high solvent ouﬂow (from the ﬁlm interface) and low non
solvent inﬂux into the ﬁlm will result in a net solvent depletion at the ﬁlm/bath interface.
The ﬁlm top layer thus becomes concentrated in polymer content with respect to the
underlying bulk portion of the ﬁlm, which still retains the original mixture composition.
This is the basic phenomenon responsible for the formation of dense skin layers,
characteristic of asymmetric membranes. The process is illustrated in Figure 1.8,

represented by compositional path a—b.

46

Polymer

 

 

 

 

Solvent Non Solvent

Figure 1.8.Demixing paths for a three component system during wet phase inversion

47

The interface area concentrates in polymer (moves towards the polymer vertex), with a
less signiﬁcant incorporation of non solvent (gradual inclination towards the non solvent
vertex). Following this trend, most of the compositional path occurs within the stable
region (L), delaying its entrance to the demixing gap. For this reason, this process is
called delayed demixing. Eventually, the mixture enters the demixing zone; however, its
“residence time” within the L-L region is small, almost immediately crossing the gelation
threshold and solidifying in its polymer rich phase. During its short residence in the L-L
region, the polymer lean phase demixes by nucleation and growth (the miscibility gap
was crossed above the critical point), forming small solvent rich nuclei within the
polymer rich phase. Upon gelation of the polymer rich phase, these small nuclei will

constitute the ﬁnal pores of the skin layer.

As the top layer concentrates in polymer and as non solvent penetrates within the system
and mixes with the solvent, it gels creating a physical barrier to ﬂow, thus reducing the
diffusive drive for the solvent to leave the ﬁlm towards the immersion bath; solvent
outﬂow progressively slows down, causing the bulk ﬁlm layers to follow a different
compositional path (path a-c) than that of the toplayer. In this case, as solvent outﬂow is
restricted, the mixture experiences a less drastic polymer concentration step before
entering the miscibility gap; or, in other words, the compositional path does not markedly

moves towards the polymer vertex, shifting instead towards the derrrixing zone.

Moving along this path, the residence time of the sub skin layers of the ﬁlm in the

thermodynamically stable region (L) is minimal, entering the miscibility gap almost

48

immediately after immersion, and thus experiencing what is known as instantaneous

demixing. As sub skin layers enter the miscibility gap farther from the gelation threshold,
liquid-liquid demixing unfolds for a longer period of time, allowing for the formation and
growth of bigger nuclei before the onset of gelation and solidiﬁcation of the polymer rich

phase.

The above explained instantaneous demixing process becomes more pronounced towards
the bottom of the ﬁlm, where solvent ouﬂow is most restricted due to the increasing
resistance exercised by the coagulating layers above. Such decrease in solvent outﬂow
rates along the ﬁlm depth, from its fastest at the ﬁlm/bath interface, to the slowest rate at
the bottom of the cast ﬁlm, and its direct effect on delayed (toplayer, path a-b) vs.
instantaneous (towards the ﬁlm bottom, Fig. 1.8, path a-c) demixing kinetics will give
rise to the typical asymmetric structure observed in wet phase inversion membranes. If L-
L dernixing is allowed to unfold for a sufﬁcient period of time before gelation, the
solvent rich phase nuclei will grow and coalesce within the polymer rich phase, to form
distinct solvent rich macro-nuclei, which will grow until the polymer rich phase gels and
restricts further nuclei growth. These macro scale nuclei will reﬂect as macrovoids in the

membranes solidiﬁed structure.

As follows, an analysis of the main parameters affecting the phase inversion process and

thus directly inﬂuencing the membranes ﬁnal structure will be offered.

1.4.1.Solventlnon solvent afﬁnity

49

As the afﬁnity between solvent and non solvent increases, the higher will be the tendency
for these two components to mix. The increased propensity for mixing will result in an
increased non solvent inﬂow, thus pushing compositional paths of the ﬁlm’s top layer,
and sub layers, towards the binodal curve. Delayed demixing will shift towards
instantaneous demixing, producing a thinner, more porous skin, as well as enhancing

asymmetry and macrovoid formation.

1.4.2.Polymer concentration in the casting mixture

An increase in the polymer concentration of the mixture will shift the ﬁlm interfacial

polymer concentration towards the polymer vertex of the ternary system (Fig. 1.9, point
a*). This will result in delayed demixing, either causing the interface to enter the gelation

region before liquid demixing, or reducing the L-L demixing period. In any case, the end
result will be a denser and thicker skin, and in some cases further suppression of
macrovoid formation. An increase in polymer concentration before phase immersion can
be artiﬁcially induced by a pre-immersion evaporation step, during which a portion of the
solvent is lost to the atmosphere and the solution interface is therefore pre concentrated in
polymer before wet phase inversion proceeds. This combined approach is also known as

dry-wet phase inversion.

50

Polymer

 

 

 

 

Solvent Non Solvent

Figure 1.9.Effect of polymer concentration (point a*) on the demixing path of ternary
system

51

1.4.3.Addition of non solvent to the casting mixture

The addition of non solvent to the casting mixture will push the mixture composition
closer to the miscibility gap (Fig 1.10, point a*), therefore hastening the onset of L-L

demixing and producing a more porous structure. The addition of non solvent can thus

shift demixing kinetics from delayed to instantaneous.

52

Polymer

 

 

 

 

Solvent Non Solvent

Figure 1.10.Effect of non solvent addition to the casting mixture (point a*) on the
denrixing path of ternary system

53

1.4.4.1mmersion bath temperature

Increasing the temperature of the immersion bath can have different effects, speciﬁc to
the particular studied system and immersion temperature. Generally, an increase in
temperature results in an increase in the overall miscibility of the three components,
which translates into the shrinking of the miscibility gap (Fig 1.11, horizontal arrow). The
mixture thus remains in the homogeneous region (L) for a longer period of time during
solvent/non solvent exchange, until it eventually gels. If gelation occurs before demixing,
such delayed demixing effect results in a ﬁnal denser membrane. However, an increase
temperature can also delay the onset of gelation, as molecules have higher mobility at
higher temperatures. The latter effect results in a viscosity decrease, which can
simultaneously facilitate non solvent inﬂux and retard gel formation, thus promoting
instantaneous demixing (Fig 1.11, downwards arrow) and producing in turn a more
porous membrane. The interplay and relative weight of these two opposing responses to
an increase in temperature; namely delayed demixing due to binodal curve shrinkage
versus instantaneous demixing induced by a decrease in the viscosity of the mixture and
the gelation retardation, will determine the ﬁnal structure of the membrane. The resulting
trend will of course depend on the speciﬁc combination of solvent, polymer, non solvent

and bath temperature.

54

Polymer

Solvent Non Solvent

Figure 1.11.Effect of temperature increase on the demixing path of the ternary system.

55

1.4.5.Addition of solvent to the immersion bath

The effects of the addition of solvent to the coagulation bath can be evaluated in terms of
the overall tendency of solvent and non solvent to exchange between the ﬁlm interface
and the immersion bath. The addition of solvent to the immersion bath will essentially
reduce the driving forces (diffusive gradients) for both the non solvent to enter the ﬁlm
and the solvent to leave the ﬁlm. Which of these two tendencies is more signiﬁcantly
reduced will determine whether the demixing process is delayed (faster solvent outﬂow
than non solvent inﬂow, Fig 1.12 upwards arrow) or instantaneous (faster non solvent
inﬂow than solvent outﬂow, Fig 1.12 downwards arrow); and will thus result in either a
denser or more porous structure for the precipitated membrane, respectively. Again, this
effect will be strictly dependant on the polymer/solvent/non solvent combination, and the

concentration of solvent in the immersion bath.

56

Polymer

 

Solvent Non Solvent

Figure 1.12.Effect of solvent addition to the coagulation bath on the demixing path of
ternary system.

57

References

1. R. W. Baker, Membrane Technology and Applications. Second ed. 2004, West
Sussex, England: John Wiley & Sons.

2. R. E Kesting, Synthetic Polymeric Membranes, A Structural Perspective. Second
ed. 1985, New York: John Wiley & Sons.

3. M. Mulder, Basic Principles of Membrane Technology. Second ed. 2003,
Dordrecht, Netherlands: Kluwer Academic.

4. S. Loeb and S. Sourirajan, Sea water dernineralization by means of an osmotic
membrane, Adv. Chem. Ser. 38 (1962) 117-132.

5. A. S. Michaels, High Flow Membrane, US Patent 3.615.024. October 1971.

6. P. J. Flory, Principles of Polymer Chemistry. 1953, New York: Cornell Univ.
Press.

7. D. Nunn, J. Elliot and J. C. Tan, Thermodynamics of mixing and separating
polymers. 2008, University of Cambridge.

58

CHAPTER TWO

F ulIerene-polysulfone nanocomposite ultraﬁltration membranes: Effect
of ﬁller shape and loading on membrane properties

Original version submitted for a possible publication in

Journal of Membrane Science

by

Julian S. Taurozzi, Adam R. Rogensues, and Volodymyr V. Tarabara

Department of Civil and Environmental Engineering, Michigan State University,

East Lansing, MI 48824, USA

59

CHAPTER TWO

Fullerene-polysulfone nanocomposite ultraﬁltration membranes: Effect
of filler shape and loading on membrane properties

Abstract

Nanoparticle-ﬁlled polysulfone membranes were synthesized by the wet phase inversion
method. Two types of carbon nanoparticles of different shapes - C60 (nanospheres) and
SWCNTs (nanorods) were used as ﬁllers. In the study of the effect of nanoparticle loading,

the membranes with 0.7 % Cay/polymer loading had higher hydraulic resistance and rejection

as compared to polysulfone-only membranes and membranes with 10 % C60/polymer
loading. In the study of the effect of nanoﬁller shape, no signiﬁcant differences were found
in the hydraulic permeabilities and observed rejections for membranes ﬁlled with either C60

or SWCNTS. The demixing rate, indirectly evaluated as the rate of precipitation front
advancement, and the viscosity of the casting mixture were found to correlate with the
permeability and rejection of cast nanocomposite membranes. The correlations were
rationalized in terms of the effect of nanoparticle ﬁllers on the demixing and gelation
processes occurring during phase inversion. The observed dependencies on the nanoparticle
loading and shape were explained as resulting from the combined effect of 1) nanoparticle-
induced destabilization of the polymer blend including the hastening of solvent transfer to the
polymer-lean phase due to the hydrophobicity of the nanoﬁller and 2) nanoﬁller aggregation.
The propagation rate of the precipitation front and the viscosity of the casting mixture were
proposed as simple-to-measure predictors of membrane performance with potential

application for the rational design of porous nanocomposites.

60

2.1 Introduction

Porous nanocomposite membranes can be prepared by incorporating nanoscale ﬁller into
a casting mixture (polymer, solvent, porogen, other components) and exposing the blend
to a nonsolvent in order to initiate wet phase inversion [1-8]. The dependence of the wet
phase inversion process on the composition of polymer/solvent/nonsolvent system can be
understood within the framework of the Flory-Higgins model [9, 10]. The practical
beneﬁts of the model are limited because in practice, membranes are cast from blends
with four or more components and because the determination of relevant thermodynamic

parameters can be difﬁcult [10, 13].

Numerous published studies on the effects of pore-forming additives (i.e., porogens) on
the wet phase inversion process suggest that complex interactions arise from the porogen-
induced changes in rheological and thermodynamic properties of polymer blends [14-21].

Less is known about the effect of nanoparticulate additives. Research on the preparation

of phase inversion membranes from polymer blends doped with Ti02 [2, 3, 8], Ag0 [5,

22, 23], A1203 [6, 7], and Si02 [4, 20, 21] nanoparticles demonstrated that membrane

skin properties (thickness, overall porosity, surface porosity) as well as the macrovoid
morphology of the support layer can be affected by nanoparticle incorporation. The effect
of carbon nanoﬁllers on the properties of nanocomposite membranes prepared by wet
phase inversion was also studied recently [1,26]. Brunet et a1. [1] evaluated mechanical,
hydraulic, and antifouling properties of SWCNT-ﬁlled polysulfone (PSf) membranes (1.4

% w/w SWCNT/PSf loading) using polyvinylpyrrolidone as the porogen and found no

61

signiﬁcant changes in membrane properties, possibly due to the incomplete dispersion of
the nanotubes in the casting mixture. Choi et al. [26] incorporated different amounts of
multi-walled carbon nanotubes into PSf blends without porogen and correlated the
nanotube loading to the hydraulic properties of the resulting membranes. The authors
reported an increase in permeability and a decrease in rejection at intermediate loadings,

with a reversal of the trend at higher loadings.

To our knowledge, the only experimental study focused on the links between the phase
inversion process and properties of porous composite membranes formed from particle-

doped polymer blends was that by Aerts et al. [20, 21]. The authors used light
microscopy to investigate the effect of the addition of hydrophilic particles (SiOz, Aerosil

200) to polysulfone blends by measuring changes in viscosity and demixing behavior in
the porogen-free quaternary (polymer, ﬁller, solvent, nonsolvent) blends and related the
changes to morphological characteristics of the cast membranes. The presence of a small
amount of ﬁller was shown to lead to an increase in the demixing rate; a decrease in the
derrrixing rate with the further increase of the ﬁller was attributed to a higher viscosity,
which increased with an increase in ﬁller loading throughout the entire loading range
studied (up to 3 vol. %). Aerosil 200 primary particles are 12 nm spheres but the size of
the ﬁller dispersed in the solvent/polymer mixture or embedded in the cast membrane
was not reported. Aerts et al. used a low intensity dispersion technique (mixing rather
than sonication), which was likely not sufﬁcient to disperse silica powders into
aggregates of sizes comparable to the primary particle size (see Supporting

Documentation (SD)).

62

The present work is on the mechanisms of formation of phase inversion porous
membranes cast from fullerene-ﬁlled polymer blends. The study focused on the effects of
nanoparticle loading and shape and, indirectly, by contrasting our ﬁndings with those of
Aerts et al. [20, 21], on the effects of ﬁller hydrophilicity. Two types of carbon
nanoparticles - C60 (nanospheres) and single-walled carbon nanotubes (nanorods) - were
used as model ﬁllers representing idealized zero- and one-dimensional ﬁllers of very
similar surface chemistry. Polyvinylpyrrolidone, which is also a known dispersing aide
for C60 and carbon nanotubes [32, 33], was used as the porogen in all polymer blends
studied. The differences between nanoparticle-free and nanocomposite membranes were

interpreted in terms of nanoparticle-induced changes in the properties of the casting

mixture and resulting changes in demixing and gelation processes.

63

2.2 Experimental

2.2.1 Materials

N-Methyl-2-pyrrolidone (Aldrich, 99% ACS reagent), polyvinylpyrrolidone (Aldrich,
MW avg. 10 kDa), and polysulfone (Solvay, Udel P-3500 LCD pellets, MB8, MW avg.
79 kDa) were all used as received. C60 fullerenes (99+ % purity) were purchased from
MER Corp. Single-walled carbon nanotubes (purity > 90%) were purchased from
Cheaptubes, Inc. The ultrapure water used in all experiments was supplied by a
commercial ultrapure water system equipped with a terminal 0.2 pm cartridge microﬁlter.
The resistivity of water was greater than 16 MQ-cm. 150 kDa dextran (DIN certiﬁed

standard, Sigma-Aldrich) was used as rejection probe.

2.2.2 Preparation of polysulfone membranes

Polysulfone membranes were prepared by wet phase inversion. Polyvinylpyrrolidone

(PVP) was added to the casting mixtures as a pore forming agent (i.e., a porogen.) The

composition of each casting mixture is given in Table 2.1.

64

 

 

 

 

 

 

 

 

Membrane
Component
PSf PSf/C60-0.7 PSf/C6040 PSF/NT-0.7
PSf (% w/w) 12.0 12.0 11.9 12.0
PVP (% w/w) 5.0 5.0 4.9 5.0
NMP (% w/w) 83 82.9 82.0 82.9
Filler (% w/w) --- 0.7 10.0 0.7

 

 

 

 

 

Table 2.1.Components of membrane casting mixtures (% w/w). Weight % of ﬁller is
expressed with respect to the mass of PSf.

65

To prepare the membranes, PSf pellets and porogen were added to N-methyl-2-
pyrrolidone (NMP) and the mixture was stirred at 60 0C until PSf completely dissolved
and a homogeneous casting solution was formed. The resulting viscous solution was cast
as a thin ﬁlm on a 25 cm x 30 cm ﬂat glass plate using a rnicrometric ﬁlm applicator
(Model 3570, Elcometer, Inc). All of the cast ﬁlms had a drawdown thickness of (300 i
l) um. Immediately after casting, the ﬁlm was immersed in a nonsolvent (water) bath to
induce phase inversion. The water bath was maintained at the room temperature (ca. 22

0C). Free standing membrane sheets that formed were thoroughly rinsed with deionized

water to remove residual solvent.

2.2.2.1 Preparation of nanoparticle-ﬁlled membranes

Dry C60 powder was added to NMP to produce nC60 organosols with two different C60

loadings: 1 mg/mL and 15 mg/mL. Porogen (PVP) and polymer (PSf) were then added to
the organosols to obtain the casting mixtures used to prepare nanocomposite membranes
with two different ﬁller-to-polymer ratios (0.7 % and 10 % by weight, respectively).
Similarly, 0.7 % w/w SWCNT-ﬁlled membranes were prepared following the same
procedure and using 1 mg/mL SWCNT/NMP organosols. The 1 mg/mL organosols were
dispersed by sonication for 20 min at 25 W (horn probe, Sonicator 3000, Misonix, Inc.).
The obtained organosols were then transferred to 25 mL Erlenmeyer ﬂasks containing the
polymer (PSf) and the porogen (PVP). The resulting casting mixtures were then
homogenized by prolonged (ca. 1 day) stirring. The 15 mg/mL organosols were prepared

directly in 25 mL Erlenmeyer ﬂasks and dispersed by sonication for 24 h using a bath

66

sonicator (Aquasonic 50T, VWR Scientiﬁc). Probe sonication was not used in this case to
prevent nanoparticle losses, since at such high loadings a considerable amount of
organosol residue was observed to remain in the scintillation vials after transfer to the
Erlenmeyer ﬂasks when horn probe sonicator was used. Such residue was not observed
for the lower loadings; thus probe sonication was used for mixtures with lower NP

loadings to shorten dispersion time. PSf and PVP were added to the 15 mg/mL
ﬁller/NMP organosols and the mixtures were stirred at 60 0C until homogeneous

solutions were obtained. Nanoparticle-ﬁlled membranes were then cast from these

viscous mixtures following the procedure explained in section 2.1.2.2.

2.2.3 Characterization of casting mixtures and cast membranes

Particle size distribution in the organosols was measured using dynamic light scattering
(ZetaPALS, BI_MAS Option, Brookhaven Instrument Corp). Measurements were

recorded in 5 cycles of 2 min each using a 30% cutoff dust ﬁlter.

The steady zero-shear viscosity of casting mixtures with variable C60 loadings was

recorded at room temperature using a programmable rheometer (Model DV-III,

Brookﬁeld).

The inﬂux of the non-solvent (water) into the casting mixture was quantiﬁed using
confocal microscopy. For each casting mixture, a water droplet and a droplet of the

casting mixture were placed on a microscope slide; the two droplets were then brought

67

into contact by capillary action, induced by pressing a cover slip against the glass slide.
The advancement of the precipitated polymer front was monitored over time by laser-
scanned brightﬁeld imaging using an Olympus FlroiewTM FV1000 confocal
microscope with a 488 nm Ar laser line and a 4X PlanFLN objective (NA 0.13). For each
sample, 333 brightﬁeld images were taken with 1.13 s interval between images (Fig. 2.2

B).

The hydraulic resistance and rejection of the membranes were measured using a stirred
Amicon 8010 dead-end ultraﬁltration cell (Millipore Corp.) connected to a 5 L
pressurized feed tank (Alloy Products Corp) ﬁlled with ultrapure water. Prior to each test,
the membrane coupons were compacted at 40 psi until steady permeate ﬂux was attained.
The permeate ﬂux was measured continuously by collecting the ﬁltered water on a digital
balance (AV8101C, Ohaus Corp). The extent of membrane compaction was determined
by comparing the initial and steady state values of permeate ﬂux. To measure membrane
rejection, 20 mg/L aqueous solution of 150 kDa dextran was ﬁltered through the
compacted membrane. Dextran concentration in the permeate was measured using a total

organic carbon analyzer (Model 1010, 01 Analytical).

68

2.3. Results and Discussion

2.3.1 Casting mixture properties

2.3.1.1 Nanoparticle dispersions in N—methyl-pyrrolidone

Suspensions of SWCNTs and nanoscale C60 aggregates (nC60) in NMP were prepared by

mixing the nanoparticles with NMP and sonicating the resulting organosols for 20 min

(see section 2.2.2.1.) Particle size distribution was measured for 1 mg/L organosols of
SWCNTs and nC60 by dynamic light scattering. The measured sizes of nC60 aggregates

and SWCNT bundles in NMP (before other components - PSf, PVP - were added to the

casting mixture) were (181 i 20) nm and (183 i 17) nm, respectively. SEM images of
sonicated and dried 1 mg/mL C60/NMP and SWCNT/N MP organosols are given in SD

(Fig. SD.2.1)

2.3.1.2 Viscosity of nanoparticle (C60) — polymer (PSf) - solvent (NMP) — porogen (PVP)

blends: Effect of nanoparticle loading
Casting mixtures with 0.7% loading of nC60 had a higher viscosity than nanoparticle-free

controls; this trend was reversed when the nanoparticle loading was increased to 10%

(Fig. 2.1 a).

69

450

 

 

9'0

 

 

 

 

 

 

 

 

 

U
(D
420 - E
.5.
cm
"'15
H
A 390 — o
E a
E. E
3. 3607 B
o a.
.Ew.’ 8
> 330 — B
8
f
3004 g
_ 15 V
270 l 1 L l I 71/ H
l l T r l /l/ H— 100
2.4 ~ b
A 70
N
—2 >- 80 ﬁ’
x 2 _ J. ' ----* g.
-. :3
5 9..
8 ~ 60 _
g 1.6 - - E”:
:2. +--- g
3 - 40 ‘3’
‘-" o.
.9. g:
g 1'2 T 1: g
. :3
>» . _ A
0.8 e .
1 l r 1 1 /r/ _ 0
0 0.2 0.4 0.6 0.8 l 9.5 10
C60 loading (% weight)

Figure 2.1. Viscosity and demixing front propagation rate for PSf-PVP-C60 blends (a)
and hydraulic resistance and 150 kDa rejection for C60-ﬁlled PSf membranes (b) at 0%
(control), 0.7 % and 10 % w/w C60/PSf loadings.

70

Incorporation of different amounts of carbon ﬁllers (carbon black [29] and spherical
nanoparticles [30, 31]) into polymer blends were reported to increase blend viscosity with
higher increases observed for ﬁllers of smaller size. The observed effects of ﬁller loading
and size were attributed to a change in the relaxation modulus and the elastic recovery of
the blend, and were correlated to changes in the available polymer-ﬁller interfacial
surface area [29] that is affected by both loading and size of the ﬁller. The increase in
viscosity observed in our study for casting mixtures with 0.7 % loading of nanoparticles

is in agreement with the trends reported in these studies. Aerts et al. [20, 21] recorded an
increase of viscosity of PSf/NMP blends ﬁlled with hydrophilic (SiOz) particles with an

increase in ﬁller loading up to 3 vol. % and interpreted this effect as a consequence of
adsorption of polysulfone chains to the ﬁller surface. In our study, a reversal of the
viscosity trend at higher (10 %) ﬁller loading was observed. We hypothesize that the
reversal is due to ﬁller aggregation. This hypothesis is consistent with results of a recent
modeling study wherein the viscosity of polymer-nanoparticle (C60) melts was shown to
be highly sensitive to the degree of nanoparticle clustering and to be proportional to the
total surface area of nanoparticle aggregates [31]. The aggregation is also more likely for
blends with low nanoparticle—polymer afﬁnity, which is the case for the
fullerene/polysulfone pair. Adding the dispersing aid (PVP) improved the afﬁnity

between fullerenes and polysulfone.

It must be noted that even though the particle size distribution of C60 clusters suspended
in NMP at 1 mg/L (see section 2.3.1.1) was measured, the measured nC60 sizes may be

differ from the sizes of nC60 in the casting mixture that contains the polymer and the

71

porogen. Generally, the relative Chemical afﬁnity and interactions between the ﬁller and
other casting mixture components may alter the aggregation state of the ﬁller within the

melt throughout the membrane formation process. For example, the addition of PVP, a
known dispersing aide for C60 and carbon nanotubes [32, 33], can be expected to result

in a decrease in the size of ﬁller clusters especially at low ﬁller loadings where there is a
high excess of PVP with respect to the ﬁller. As the ﬁller loading is increased, the
porogen effect should decrease allowing for aggregation to occur and for larger clusters
with smaller speciﬁc surface area to form. This scenario is consistent with the increase of
viscosity that we observed to accompany the increase of ﬁller loading from 0.7% to 10%

(Fig. 2.1 a)

2.3.1.3 Demixing behavior of polymer (PSf) — nanoparticle (nC60) — solvent (NMP ) —

porogen (PVP) blends exposed to nonsol vent (water): Effect of nanoparticle loading

The results on the propagation rate of the polymer precipitation front (Fig. 2.1 a, Fig.

SD.2.3) show higher precipitation front propagation rates and indicate a faster gelation
for blends with 0.7 % C60 loading as compared to C6O-free controls and blends with 10 %
C60 loading. In what follows, an analysis of this behavior will be offered in terms of the

effect of the nanoparticle ﬁller on the demixing process.

Figure 2.2 A shows a schematic of a typical isothermal phase diagram for a three
component (polymer-solvent-nonsolvent) system. Point “a” represents a homogenous

polymer-solvent blend, while the “a” —>“c” transition illustrates the compositional path of

72

the blend as it incorporates nonsolvent and separates (i.e., demixes) into polymer-rich and
polymer-lean phases. This results in the formation of a solid polymer matrix that is the
membrane [10, 13]. Point “b” corresponds to the point of initiation of the phase
separation as the system crosses the binodal curve and undergoes liquid-liquid demixing
(region L-L) following compositional tie lines. At point “c”, the polymer concentration in
the polyrner-rich phase reaches critical value (gelation point) and the polymer solidiﬁes
forming the membrane. Pore size and porosity of the membrane are controlled by the
time available for solvent-rich (i.e. polymer-lean) nuclei to form and grow in the blend;
this is the time from the moment phase separation begins as the system enters the binodal
region (point b, Fig. 2.2 A) and the time when gelation of the polymer-rich phase occurs

(point c, Fig. 2.2 A). If gelation occurs earlier (i.e., faster a -—> c transition), the nuclei

will have less time to grow, thus resulting in the formation of smaller pores [9].

In this study, the rate of precipitation front propagation was used as an indirect measure
of the demixing rate (i.e., of how fast the gel is formed upon incorporation of
nonsolvent). As the nonsolvent diffuses into the droplet of polymer blend, ﬁrst the
outermost interfacial layer of the droplet demixes into polymer-rich and polymer-lean
phases (Region L-L, Fig. 2.2 A) and the polymer-rich phase within this layer reaches the
gelation point (“c”). The entire process (if not diffusion limited) takes less than 1 s and
manifests itself as an increase in the optical density of the precipitated portion of the
droplet. As the nonsolvent diffuses further into the droplet, successive layers of the blend
undergo transition from a completely mixed state to a gelled state. It is the reasoning that

the rate of the advancement of the precipitation front should be proportional to the rate of

73

derrrixing of every one of the successive differential layers that was at the base of the
postulate that the rate of precipitation front propagation can be used as a measure of the
demixing rate. In the past, precipitation front propagation data was used to discriminate
between instantaneous and delayed demixing [20]. By using this data only as a measure
of the demixing rate we take a more conservative approach that recognizes that the
difference between the instantaneous and the delayed demixing lies in the difference in
the amount of time it takes for the system to reach (not the gelation state but) the binodal

curve that can shift in response to the addition of a ﬁller.

It should also be noted that the hydraulic resistance that the precipitated gel exerts to the
nonsolvent is another factor that might control the propagation rate of precipitation front.
However, because there was a reverse correlation between the rate of precipitation front
propagation and the resistance of the resulting membrane (Fig. 2.1), we concluded that
the hydraulic resistance of the precipitated region to the inﬂowing nonsolvent was a

factor of secondary importance that played a mitigating role if any.

74

A)

Polymer

    
    

sun 91.]. "

 

Solvent N onsolvent
B)

    

    

artisans“: 7-961 irimsinstanuéé-e isnéisésranese-ir

Figure 2.2. Schematic (A) of an isothermal phase diagram for a three component
(solvent—polymer-nonsolvent) system and laser-scanned brightﬁeld images (B) of
demixing front propagation in PSf-C60-PVP-NMP blends as a function of time (in
seconds) of exposure to nonsolvent (water).

75

The data on the rate of precipitation front advancement can be analyzed in terms of the
thermodynamic destabilization effect of nanoparticulate ﬁllers. Previous studies on the
effect of porogen additives on the demixing of polymer blends showed that the addition
of a fourth component (porogen) increases the thermodynamic instability of the system,
hastening the onset of demixing [17, 19]. Addition of hydrophobic carbon nanoparticles
(a ﬁfth component) is likely to have an analogous effect on the demixing process; the
addition of nanoparticles would result in an overall reduction of the blend’s afﬁnity for
the polar nonsolvent (water), accelerating gelation as less nonsolvent is needed to initiate
phase separation. This effect however, should be expected to be strongly dependent on
the degree of dispersion of the ﬁller within the blend. As the nanoparticles aggregate, the

nanoparticle-polymer interfacial area will become smaller, decreasing the destabilizing

effect of nanoparticles. Thus, C60 aggregation (see section 2.3.1.2) would explain why the

propagation front trend with C60 loading is reversed at higher (10 %) C60 content.

2.3.2 Membrane hydraulic properties

Figure 2.1 b shows the permeability and observed rejection of 150 kDa dextran for C60-

ﬁlled membranes as a function of C60 loading. The membranes with 0.7 % C60 loading
show a simultaneous decrease in permeability and increase in rejection, as compared to
C60-free membranes (controls) and membranes with 10 % C60 loading. This result points

to smaller skin pores and/or smaller overall porosity of the skin for the membranes with

0.7 % loading.

76

The results can be explained in terms of the properties of the casting mixtures. As

mentioned in section 2.3.1.3, the porosity of the membrane skin will depend on demixing

time (“b”-“c” transition, Fig. 2.2). Measurements of nonsolvent diffusion rate into C60-
ﬁlled polymer blends (Fig. 2.1 a) show that blends with 0.7 % C60 loading gel faster than
C60-free and blends with 10 % C60 loading. The decreased permeability and increased

rejection of membranes prepared from blends with 0.7 % C60 loading can be explained by

shorter demixing periods that result in denser membrane skins (section 2.3.1.3).
Furthermore, the hydrophobic character of the ﬁller would result in an enhancement of
solvent outﬂow towards the polymer-rich phase of the blend, hastening solvent-
nonsolvent exchange and increasing the local polymer concentration. The combined
effect of i) nanoparticle-induced destabilization of the system and ii) hastening of solvent

transfer to the polymer-lean phase due to the hydrophobicity of the nanoﬁller would
explain the accelerated gelation for the C60-ﬁlled membranes with respect to
nanoparticle-free blends. The observed decreases in both viscosity and precipitation front
propagation rate for the casting mixtures with 10% C60 loading point to a mitigation of

these effects due to a decrease in NP-polymer interfacial area upon ﬁller aggregation,

resulting in a reversal of the trend in the observed hydraulic properties for membranes

with 10 % C60 loading.

The observed changes in the demixing behavior of Céo-ﬁlled blends can, in principle, be

described using a phase diagram and the changes in the relative composition of the

casting mixture during the demixing process. To accurately describe the changes,

77

however, a phase diagram for the ﬁve component (polymer, solvent, nonsolvent,
porogen, and ﬁller) system wOuld have to be constructed. In addition to the obvious
difﬁculty of visualizing and interpreting the evolution of the system within the conﬁnes
of a 4-dimensional prism, measurements necessary to track nanoﬁller-induced changes in
binodal and gelation curves may not be always possible. For the polymer blends studied
in this work, cloud point measurements, for example, could only be conducted for the
nanoparticle-free casting mixtures (see SD, Fig. SD.2.2). Although one can measure
values of apparent cloud points [20, 21], in our opinion it is impossible to conclude with
certainty whether the observed polymer precipitation upon nonsolvent addition in the
presence of nanoparticles corresponds to the actual gelation point of the system or to
other nanoparticle-induced effects such as, for example, settling of nanoparticle-polymer

aggregates.

2.3.3 Nanocomposite membranes doped with nanoparticles of different shapes

The effect of nanoﬁller shape on the demixing behavior of nanoparticle-ﬁlled blends was

studied by preparing and characterizing membranes ﬁlled with carbon ﬁllers of different
shapes: C60 (nanospheres) and SWCNTs (nanorods). The model ﬁllers represented

idealized zero- and one- dimensional ﬁllers of very similar surface chemistry.

78

 

 

 

 

 

 

 

l I
it 3r
3.5 100

:3" 3 ‘ 70
g - 80 g.
"‘ $3
>< 2.5 — 5'
TE :3
O
v T 60 "'5
<1) 2 -— ._r
5 L11
8 O
.2 E
m l 5 a a:

0 F- 40
‘3‘ 8‘
.2 5
'3 l — n
g S
_ 20 A
>~. Q
m 0.5 - 3-

O l l l l A 0
Control C60 SWCNT

Figure 2.3.Hydraulic resistance (circles) and rejection (triangles) of 150 kDa dextran for

PSff membranes ﬁlled with C60 and SWCNTs at a ﬁller/polymer weight loading of 0.7
%.

79

The permeabilities and observed rejection of membranes ﬁlled with either C60 or
SWCNTs were similar (Fig. 2.3). Dynamic light scattering measurements indicated that
both C60 and SWCNTs Suspended in NMP existed as aggregates with approximately the
same effective diffusion coefﬁcient that corresponded to the effective hydrodynamic
diameter of ca. 180 nm (section 2.3.1.1). Given the identical surface chemistry of C60 and

SWCNTS, it can be expected that chemical interactions of both types of nanoparticles
with the other mixture components are the same. With the hydrodynamic and chemical
properties matching, the only difference between the two ﬁllers was the morphology of
the primary (as opposed to aggregated) nanoparticles. Evidently, this difference did not
translate into measurable differences in hydraulic properties of C60- and SWCNT-ﬁlled
membranes. This result was also consistent with the hypothesis that nanoparticle

aggregation state within the casting mixtures controls the phase inversion process.

80

2.4. Conclusions

Nanoparticle-ﬁlled polysulfone membranes were synthesized by the wet phase inversion
method. Three types of fullerene nanoparticles - C60 (nanospheres) and single-walled
carbon nanotubes (nanorods) were used as model ﬁllers representing idealized zero-, and
one- dimensional ﬁllers of identical surface chemistry. The hydraulic resistance and

rejection of Cbo-ﬁlled membranes peaked at the relatively low (0.7%) loading and were

similar for C60/PSf and SWCNT/PSf membranes cast from blends with the same

nanoparticle loading. Demixing front propagation rate and casting mixture viscosity were
found to correlate with membrane permeability and observed rejection. The observed

correlations were explained in terms of the following effects:

1. Nanoparticles exert a thermodynamic destabilization effect within the
polymer blend, hastening the onset of demixing thereby reducing the available
time for pore growth and leading to faster gelation and formation of smaller pores.
The destabilization of the blend is further enhanced by the nanoparticles’
hydrophobicity. Hydrophobic nanoparticles enhance solvent transfer to the

polymer-lean phase and accelerate gelation, with smaller pores formed as a result.

2. While C60 and SWCNTs have distinctly different shapes, both

nanoparticles clustered to form aggregates with the similar effective diffusion

coefﬁcient that translated into the effective hydrodynamic diameter of ca. 180 nm.

81

The Observed hydraulic properties for C60/PSf and SWCNT/PSf membranes at

equal nanoparticle loadings were similar. Thus, the phase inversion process
appears to depend on the size, shape (and chemistry) of the nanoparticle aggregate
within the casting mixtures and not on the size, shape (and chemistry) of the

primary nanoparticle.

Demixing front propagation rate and casting mixture viscosity were proposed as indirect
indicators of the degree of aggregation and thermodynamic destabilizing effect of
nanoparticle ﬁller, respectively. Such easy-to-measure indicators can be potentially

applied for the rational design of porous nanocomposites.

82

Acknowledgements

This material is based upon work supported by the National Science Foundation under
grant no. OISE-0530174. We thank Solvay Advanced Polymers, L.L.C. for providing the
polysulfone for membrane synthesis and XG Sciences, Inc for donating exfoliated

graphite nanoplatelets.

83

Supporting Documentation

Fullerene-polysulfone nanocomposite membranes: Effect Of ﬁller shape

and loading on membrane properties

Original version submitted for a possible publication in

Journal of Membrane Science

by

Julian S. Taurozzi, Adam R. Rogensues, and Volodymyr V. Tarabara

Department of Civil and Environmental Engineering, Michigan State University,

East Lansing, MI 48824, USA

84

SD.2.1.SEM sample preparation procedure

For SEM imaging, a droplet of the 1mg/mL nanoparticle/NMP organosols was deposited
on a silica wafer and let to dry. The dry samples were coated with osmium for 20 s at a
current of 10 mA using an osmium plasma coater (model NEOC-AT, Meiwa Shoji CO.,
Osaka, Japan). A Hitachi S-47OOII ﬁeld emission scanning electron microscope operated

in ultra high resolution mode was used to record SEM images (Fig. SD.2.1).

85

 

Figure SD.2.1.SEM images of 1 mg/mL SWCNT/NMP (A) and C60/NMP (B)
organosols.

86

SD.2.2.CIoud point measurements

The cloud points of nanoparticle free casting mixtures were determined following Aerts
et al. method [1], by which mixtures of PSf and NMP of different polymer concentrations
(12.2, 8.8, 6.1, 4.6 and 2.7 % w/w) were titrated with an NMP:Water (66:34 w/w)
solution until appearance of turbidity. The experiments were conducted on a temperature
controlled heat plate at constant 60 °C. The cloud point for each mixture was calculated
from the added amounts of solvent and non solvent at which turbidity remained after 24

hs. of mixing. The obtained cloud points are shown in Fig. SD.2.2

87

PSf

 

NMP

 

Figure SD.2.2.CIoud points for PSf-NMP-water system.

88

SD.2.3.Sonication of silica powders

Table 1 shows the measured effective diameters for silica clusters obtained after
sonication of source silica powders in N, N-Dimethylacetamide (DMA) using a horn
probe ultra-sonicator (Sonicator 3000, Misonix Inc., Newtown, CT). A variety of source
silica powders (Nanoamor Inc., Los Alamos, NM) with different primary particle sizes,
ranging from 10 to 3000 nm, were sonicated at different sonication intensities and

sonication times. In all cases, the size of the resulting silica clusters was at or above ~

 

 

 

 

 

 

 

 

 

 

 

 

 

 

200 nm.
Primary Powder Solvent Sonication Sonication Cluster Size
particle mass Solvent (mL) time (min) Intensrty (nm)
srze (nm) (g) (‘70)
10 0.5 DMA 40 5 100 259.4
10 0.5 DMA 40 15 100 248.3
10 0.05 DMA 40 30 100 212.6
10 0.05 DMA 40 45 100 195.8
10 0.5 DMA 40 30 20 244
10 0.5 DMA 40 5 50 271.8
80 0.5 DMA/PEG 40/6 15 100 222.2
80 0.5 DMA 40 ‘15 100 197
1000 0.5 DMA 40 10 20 691.8
1000 0.5 DMA 40 5 50 559.5
1000 0.5 DMA 4O 5 100 525.7
1000 0.5 DMA 40 ‘15 100 398
3000 0.5 DMA 4O 15 100 521.6

 

 

 

 

 

 

 

 

 

Table SD.2.1.Cluster sizes for different silica powders sonicated at a variety of
intensities and times. 100 % sonication intensity corresponds to a maximum power output
of 600 W. PEG denotes Polyethyleneglycol (400 Mw).

89

SD.2.4.Demixing front propagation rate

Precipitated front propagation rates were obtained from the slopes of front thickness as a
function of the square root of time plots, as shown in Fig. SD.2.3. The thickness of the
propagating front at different times was obtained using Olympus FlroiewTM FV1000

dedicated software.

90

800 ..,L,m

 

 

0 C60 0.7 % w/w .
E 700 h I C6010% w/w . "I ]
:1. g ,
c; 600 _ 0 Control _ ,. . 4
3 ’0' I . o
:a l .' - r "
[:1 400 '. . ~ 'i —-' -
E , ;' 1"" '
L1. 300 l.,_;‘ r. '- —

200 1 1

 

 

 

2 4 6 8110121416

WU?)

Figure SD.2.3.Front thickness vs. the square root of time plots for 0, 0.7 and 10 % C60
loading casting mixtures.

91

References

1. L. Brunet, D. Y. Lyon, K. Zodrow, J. C. Rouch, B. Caussat, P. Serp, J. C.
Remigy, M. R. Wiesner and R]. J. Alvarez, Properties of membranes containing semi-
dispersed carbon nanotubes, Environ. Eng. Sci. 25 (2008) 565-575.

2. K. Ebert, D. Fritsch, J. K011 and C. Tjahjawiguna, Inﬂuence of inorganic ﬁllers on
the compaction behaviour of porous polymer based membranes, J. Membr. Sci. 233
(2004) 71-78.

3. J. B. Li, J. W. Zhu and M. S. Zheng, Morphologies and properties of
poly(phthalazinone ether sulfone ketone) matrix ultraﬁltration membranes with entrapped
Ti02 nanoparticles, J. Appl. Polym. Sci. 103 (2006) 3623-3629.

4. S. P. Nunes, K. V. Peinemann, K. Ohlrogge, A. Alpers, M. Keller and A. T. N.
Pires, Membranes of poly(ether imide) and nanodispersed silica, J. Membr. Sci. 157
(1999) 219-226.

5. W. K. Son, J. H. Youk, T. S. Lee and W. H. Park, Preparation of antimicrobial
ultraﬁne cellulose acetate ﬁbers with silver nanoparticles, Macromol. Rapid Commun. 25
(2004) 1632-1637.

6. N. M. Wara, L. F. Francis and B. V. Velamakanni, Addition of alumina to
cellulose acetate membranes, J. Membr. Sci. 104 (1995) 43-49.

7. L. Yan, Y. S. Li, C. B. Xiang and S. Xianda, Effect of nano-sized Al203-particle
addition on PVDF ultraﬁltration membrane performance, J. Membr. Sci. 276 (2006) 162-
167.

8. Y. Yang, H. Zhang, P. Wang, Q. Zheng and J. Li, The inﬂuence of nano-sized
Ti02 ﬁllers on the morphologies and properties of PSF UF membrane, J. Membr. Sci.
288 (2007) 231-238.

9. M. Mulder, Basic Principles of Membrane Technology. Second ed. 2003,
Dordrecht, Netherlands: Kluwer Academic.

10. P. J. Flory, Principles of Polymer Chemistry. 1953, New York: Cornell Univ.
Press.

11. C. M. Tam, T. A. Tweddle, O. Kutowy and J. D. Hazlett, Polysulfone
membranes. II. Performance comparison of polysulfone-poly(N-vinyl-pyrrolidone)
membranes, Desalination 89 (1993) 275-287.

12. T.A. Tweddle, O. Kutowy, W. L. Thayer and S. Sourirajan, Polysulfone
ultraﬁltration membranes, Ind. Eng. Chem. Prod. Res. Dev. 22 (1983) 320-326.

92

13. P. van de Witte, P. J. Dijkstra, J. W. A. van den Berg and J. Feijen, Phase
separation processes in polymer solutions in relation to membrane formation, J. Membr.
Sci. 117 (1996) 1-31.

14. R. M. Boom, I. M. Wienk, Th. van den Boomgaard and C. A. Smolders,
Microstructures in phase inversion membranes. Part 2. The role of a polymeric additive,
J. Membr. Sci. 73 (1992) 277-292.

15. B. Chakrabarty, A. K. Ghoshal and M. K. Purkait, Preparation, characterization
and performance studies of polysulfone membranes using PVP as an additive, J. Membr.
Sci. 315 (2008) 36-47.

16. A. Idris and L. K. Yet, The effect of different molecular weight PEG additives on
cellulose acetate asymmetric dialysis membrane performance, J. Membr. Sci. 280 (2006)
920-927.

17. J. H. Kim and K. H. Lee, Effect of PEG additive on membrane formation by
phase inversion, J. Membr. Sci. 138 (1998) 153-163.

18. S. R. Kim, K. H. Lee and M. S. Jhon, The effect of ZnC12 on the formation of
polysulfone membrane, J. Membr. Sci. 119 (1996) 59-64.

19. Q. Z. Zheng, P. Wang and Y. N. Yang, Rheological and thermodynamic variation
in polysulfone solution by PEG introduction and its effect on kinetics of membrane
formation via phase-inversion process, J. Membr. Sci. 279 (2006) 230-237.

20. P. Aerts, I. Genne, S. Kuypers, R. Leysen, I. F. J. Vankelecom and P. A. Jacobs,
Polysulfone-aerosil composite membranes: Part 2. The inﬂuence of the addition of aerosil
on the skin characteristics and membrane properties, J. Membr. Sci. 178 (2000) 1-11.

21. P. Aerts, E. Van Hoof, R. Leysen, I. F. J. Vankelecom and P. A. Jacobs,
Polysulfone-aerosil composite membranes: Part 1. The inﬂuence of the addition of
Aerosil on the formation process and membrane morphology, J. Membr. Sci. 176 (2000)
63-73.

22. W. L. Chou, D. G. Yu and M. C. Yang, The preparation and characterization of

silver-loading cellulose acetate hollow ﬁber membrane for water treatment, Polym. Adv.
Technol. 16 (2005) 600 - 607.

23. J. S. Taurozzi, H. Arul, V. Z. Bosak, A. F. Burban, T. C. Voice, M. L. Bruening
and V. V. Tarabara, Effect of ﬁller incorporation route on the properties of polysulfone-
silver nanocomposite membranes of different porosities, J. Membr. Sci. 325 (2008) 58-
68.

24. G. Jonsson, Methods for determining the selectivity of reverse osmosis
membranes, Desalination 24 (1977) 19-37.

93

25. K. M. Persson, V. Gekas and G. Tragardh, Study of membrane compaction and its
inﬂuence on ultraﬁltration water permeability, J. Membr. Sci. 100 (1995) 155-162.

26. J-H Choi, J .onggeon Jegal and W-N Kim, Fabrication and characterization of
multi-walled carbon nanotubes/polymer blend membranes, J. Membr. Sci. 284 (2006)
406-415.

27. A. J. Reuvers, J. W. A. van den Berg and C. A. Smolders, Formation of
membranes by means of immersion precipitation. Part I. A model to describe mass
transfer during immersion precipitation, J. Membr. Sci. 34 (1987) 45-65.

28. A. J. Reuvers and C. A. Smolders, Formation of membranes by means of
immersion precipitation. Part 2. The mechanism of formation of membranes prepared
from the system cellulose acetate—acetone—water, J. Membr. Sci. 34 (1987) 67-86.

29. J. L. White and J. W. Crowder, The inﬂuence of carbon black on the extrusion

characteristics and rheological properties of elastomers: Polybutadiene and butadiene-
styrene copolymer, J. Appl. Polym. Sci. 18 (1974) 1013-1038.

30. V. Pryamitsyn and V. Ganesan, Mechanisms of steady-shear rheology in polymer
nanoparticle composites, J. Rheol. 50 (2006) 655-683.

31. J. F. Douglas and F. W. Starr, Origin of particle clustering in a simulated polymer
nanocomposite and its impact on rheology, J. Chem. Phys. 119 (2003) 1777-1788.

32. J. Lee and H. H. Kim, Effect of encapsulating agents on dispersion status and
photochemical reactivity of C60 in the aqueous phase, Environ. Sci. Technol. 42 (2008)
1552-1557.

33. V. C. Moore, M. S. Strano, E. H. Haroz, R. H. Hauge and R. E. Smalley,
Individually suspended single-walled carbon nanotubes in various surfactants, Nano Lett.
3 (2003) 1379-1382.

34. J. P. DeRocher, B. T. Gettelﬁnger, J. Wang, E. E. Nuxoll and E. L. Cussler,
Barrier membranes with different sizes of aligned ﬂakes, J. Membr. Sci. 254 (2005) 21-
30.

35. E. L. Cussler, S. E. Hughes, W. J. Ward and R. Aris, Barrier membranes, J.
Membr. Sci. 38 (1988) 161-174.

36. D. M. Eitzman, R. R. Melkote and E. L. Cussler, Barrier membranes with tipped
impermeable ﬂakes, AIChE Journal 42 (1996) 2-9.

37. C. Yang, W. H. Smyrl and E. L. Cussler, Flake alignment in composite coatings,
J. Membr. Sci. 231 (2004) 1-12.

38. N. K. Lape, E. E. Nuxoll and E. L. Cussler, Polydisperse ﬂakes in barrier films, J.
Membr. Sci. 236 (2004) 29-37.

94

39. J. Lu, 1. Do, L. T. Drzal, R. M. Worden and 1. Lee, Nanometal-decorated
exfoliated graphite nanoplatelet based glucose biosensors with high sensitivity and fast
response, Acs Nano 2 (2008) 1825-1832.

40. K. Kalaitzidou, H. Fukushima and L. T. Drzal, A new compounding method for
exfoliated graphite-polypropylene nanocomposites with enhanced ﬂexural properties and
lower percolation threshold, Compos. Sci. Technol. 67 (2007) 2045-2051.

95

CHAPTER THREE

Effect of filler incorporation route on the properties of polysulfone-

silver nanocomposite membranes of different porosities
Published in

Journal of Membrane Science

Volume 325, 2008, pages 58-68

by
Julian S. Taurozzil, Hari Arul H , Volodymyr Z. Bosaki, Anatoliy F. Burbanl, Thomas C.
Voicel, Merlin L. BrueningI, and Volodymyr V. Tarabara:r
T Department of Civil and Environmental Engineering, Michigan State University, East
Lansing, MI 48824, USA
II High School Honors Science, Mathematics, Engineering Program, Michigan State
University
1:Department of Chemistry, The National University of "Kyiv-Mohyla Academy", 2,
Skovoroda St., Kyiv, 04070, Ukraine

1Department of Chemistry, Michigan State University, East Lansing, MI 48824, USA

96

CHAPTER THREE

Effect of filler incorporation route on the properties of polysulfone-
silver nanocomposite membranes of different porosities

Abstract

Flat sheet porous polysulfone-silver nanocomposite membranes were synthesized by the
wet phase inversion process. The effects of casting mixture composition and nanoparticle
incorporation route on the morphological and separation properties of prepared
membranes were studied by comparing nanocomposites of different preparations with
silver-free controls. Silver nanoparticles were either synthesized ex-situ and then added to
the casting solution as an organosol or produced in the casting solution via in-situ
reduction of ionic silver by the polymer solvent. N anocomposite membranes of three
types differing in skin porosity and macrovoid structure were prepared. The structure and
properties of nanocomposites were interpreted in terms of the coupling between the
processes of nanoparticle formation and gelling of the polymer-rich phase during phase
inversion. Larger nanoparticles preferentially located in the skin layer were observed in
composites prepared via the ex-situ method while in-situ reduction of silver led to
formation of smaller nanoparticles homogeneously distributed along the membrane cross
section. In some cases, incorporation of nanoscale silver formed ex-situ resulted in
macrovoid widening and an order of magnitude decrease in hydraulic resistance
accompanied by only a moderate decrease in rejection. The accessibility of the silver
nanoparticles embedded in the membrane was quantitatively assessed by the degree of
the growth inhibition of a membrane bioﬁlm due to the ionic silver released by the

nanocomposites and was found to depend on the method of silver incorporation.

97

3.1 Introduction

Porous polymer nanocomposite membranes can be prepared using a number of
nanoparticle incorporation strategies. These strategies can be grouped into four categories
based on whether particles and membranes are pre-formed or synthesized in-situ during
the formation of the nanocomposite. Methods that involve attachment or in-situ synthesis
of nanoparticles on the pore surface of existing membranes [1-7] have the advantages of
improved reproducibility of nanocomposite preparation, the possibility to reuse the
membrane matrix or to regenerate nanoparticles, and better accessibility of immobilized
nanoparticles to reactants in the permeate ﬂow. In contrast, incorporation of nanoparticles
during membrane formation [8-21] is appealing because it should allow the design of
nanocomposite structures with improved mechanical and separation properties along with
embedded nanoparticle-based functionalities. The structure of the resulting membranes is
generally a function of the physical and chemical properties of the polymer matrix and

nanoparticles as well as the method of nanoparticle incorporation.

A number of published studies describe the formation of organic-inorganic porous (10
kDa to 200 kDa) composite membranes by adding particles (A1203 [18, 19], ZrOz [11,
14], Ti02 [1.5, 20], $102 [8, 9, 12, 22], Ago [13, 21, 23]) to the solution of the membrane-

forrning polymer (cellulose acetate (CA) [13, 18, 23], polysulfone (PSf) [8, 9, 14, 17, 20],
polyvinylidene ﬂuoride (PVDF) [11, 12, 19], polyacrylonitrile (PAN) [21], polyamide-

imide (PAI) [24], poly(vinylideneﬂuoride-co-hexaﬂuoropropylene) [22], or

98

poly(phthalazinone ether sulfone ketone) (PPESK) [15]). These studies indicate that skin
layer properties (porosity, surface porosity, thickness) as well as the macrovoid
morphology of the support layer can be affected by nanoparticle incorporation. Both of
these membrane regions are important; the skin layer determines the permeability,
rejection, and selectivity of membranes, while the structure of the support layer affects its
compaction behavior [25, 26]. Although nanoparticle-induced changes in membrane
structure are speciﬁc to the particular ﬁller/matrix combination, certain common trends

can be identiﬁed. Increasing nanoparticle loading led to: (1) increased skin layer

thicknesses (SiOzlPSf [8, 9], Ti02/PSf [20], Ti02/PPESK [15]), (2) higher surface
porosity of the skin (ZrOzlPSf [ l4], Ti02/PSf [20], Ti02/PPESK [15], SiOzlPSf [9]), (3)
suppressed macrovoid formation (SiOz/PSf [8, 9], TiOZ/PSf [20], A1203/CA [18],
Ti02/PPESK [15]) and (4) higher permeability of the membrane (ZrOz/PSf [14],
SiOzlPVDF [12], ZrOz/PVDF [l l], Ale3/PVDF [19], Ag/CA [13], SiOzlPSf [9],
Ti02/PSf [20]. In a few cases, decreased (Ag/PAN [21]) permeability or a maximum
permeability at intermediate nanoparticle loadings (Ti02/PPESK [15]) was also reported.
The rejection either peaked at intermediate loadings (Ti02/PSf [20]), or decreased after a
threshold in ﬁller loading was exceeded (SiOz/PSf [9], SiOZ/PVDF [l2], ZrOzlPVDF

[11], Ti02/PPESK [15]), or remained unchanged (ZrOz/PSf [l4], Ale3/PVDF [19]). In a

recent study, nanotube ﬁllers were used to prepare a nanocomposite membrane when
polysulfone was embedded with (acid-treated) multiwall carbon nanotubes [27]. The

average porosity of the composite membranes peaked at a nanotube loading of 1.5%,

99

which also corresponded to the highest permeate ﬂux. Interestingly, the only pronounced

change in membrane macroscopic morphology was an increase in surface roughness.

Examples of simultaneous formation of the polymer matrix and the nanoparticles include
chemical reduction of metal ions by a component of the membrane casting solution [13,

21, 28]. Bhattacharyya et al. reported on the formation of Fe/N i nanoparticles in porous
cellulose acetate membranes by in-situ reduction wherein the reducing agent (NaBH4)
was a component of the non-solvent bath [28]. Yu et al. prepared silver-loaded cellulose
acetate and polyacrylonitrile hollow ﬁbers by the reduction of Ag+ by N,N-

dimethylformamide, a polymer solvent, with the objective of producing biofouling-

resistant ultraﬁltration membranes [13, 21].

The results of the studies published to-date strongly imply that the microstructural
changes induced by nanoparticle incorporation can potentially be controlled to prepare
membranes with desirable properties. To enable the rational design of porous
nanocomposite membranes, the role of essential process variables such as nanoﬁller size
(relative to the characteristic size of matrix pores), the afﬁnity of the ﬁller material for the
components of the casting mixture, and the effect of the matrix pores functioning as
reactors for forming ﬁller nanoparticles need to be understood. The overall goal of our
research was to understand how the preparation condition (nanoparticle incorporation
route, porosity of the host matrix, etc) inﬂuence the structure of the prepared

nanocomposite membranes. To this end, we attempted to map nanocomposite membrane

100

properties to preparation conditions for a series of membrane morphologies and two
nanoparticle incorporation routes. Three types of membranes differing in skin porosity
and macrovoid structure were prepared, and silver nanoparticles were either synthesized
ex-situ and then added to the casting solution as an organosol or produced in the casting
solution via in-situ reduction of ionic silver during the phase inversion process.
Polysulfone was chosen as the matrix material due to its practical importance and wide
use in the preparation of ultraﬁltration membranes, while nanoscale silver was chosen for
two reasons. First, there are established protocols for the preparation and characterization
of silver sols in various dispersion media, including the organic solvents used in
membrane preparation. Second, the presence and availability of silver can be detected by
the extent of its antibacterial effect; this provides a convenient framework for the
evaluation of the accessibility of silver in polymer matrices and serves as additional
motivation in view of the potential application of silver-ﬁlled nanocomposites as

biofouling-resistant materials.

101

3.2 Experimental

3.2.1 Materials

Silver nitrate (GR ACS crystals), N,N-dimethylformamide (99.8% anhydrous), N,N-
dimethylacetamide (99.8% anhydrous), polyethylene glycol (MW avg. 400), 2-propanol
(99.5%) and polysulfone (Udel P-3500 LCD pellets, MB8, 79 kDa) were all used as
received. The ultrapure water used in all experiments was supplied by a commercial
ultrapure water system equipped with a terminal 0.2 pm capsule microﬁlter; the
resistivity of water was greater than 16 MQ-cm. The standard constituents of the Luria-
Bertani (LB) medium and agar plates used for bacteria incubation were purchased from
Difco Laboratories. Escherichia coli cell lysing agent,(B-PER), bovine serum albumin
protein assay standards, and Bradford assay reagents were all purchased from Pierce
Biotechnology and used according to the manufacturer’s guidelines. For the
measurements of the molecular weight cut-off of prepared membranes, 150 kDa dextran
(DIN certiﬁed standard) was purchased from Sigma-Aldrich. Reagents used for ICP-MS
analysis were ultra-pure nitric acid and single-element plasma grade standards in 1%

nitric acid.

3.2.2 Preparation of polysulfone membranes

All membranes were prepared by the wet phase inversion process wherein cast ﬁlms of

polysulfone (PSf) dissolved in a mixture of N,N-dimethylacetamide (DMAC) and

102

dimethylformarrride (DMF) were immersed in a non-solvent with respect to polysulfone.
DMAC served as the primary polymer solvent while DMF acted as the reducing agent for
silver. To explore the effect of matrix porosity on the distribution and accessibility of
ﬁller nanoparticles, three membrane preparation protocols were adapted to produce
membranes with different porosities. Membranes of three distinct types (Types I, H, and
III) were obtained by varying the composition of the casting mixture and the composition
and temperature of the non-solvent in the immersion bath (Table 1). Either polyethylene
glycol (PEG) or 2-propanol was added to the casting mixture as a pore forming agent

(porogen), and either water or propanol was used as the non-solvent medium.

PSf Type I membranes were prepared according to a known procedure [29-32], where
DMAC or DMF are used as solvents for polysulfone. PEG was added to the casting
mixture to increase the porosity of the cast membrane while maintaining a high polymer
content (15% to 20%) and, therefore, sufﬁciently high viscosity (2 104 cps) of the casting
mixture [33-35]. To prepare Type I membranes, polysulfone pellets were suspended in a
DMAC/DMF/PEG solution and left to dissolve overnight in an oven at 100 °C. The vial
with the polymer solution was sealed to prevent solvent evaporation. Upon complete
dissolution of the polymer, the resulting viscous solution was homogenized by mixing on
a hot plate and was cast as a thin ﬁlm on a 25 cm x 30 cm ﬂat glass support using a
micrometric ﬁlm applicator (model 3570, Elcometer Inc., Rochester Hills, MI). All of the
cast ﬁlms had a drawdown thickness of (300 i 1) pm, which was ensured by adjusting

the applicator’s outlet ﬁlm gap with a micrometric screw.

103

Ial

 

Membrane

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Component PSf I PSf II PSf III PSf/Ag I PSf/Ag II PSf/Ag III
PSf, % 19.97 10.08 10.12 19.20 9.88 9.92
DMAC, % 45.04 70.78 60.87 43.32 69.38 59.66
DMF, % 18.97 19.14 19.22 18.24 18.76 18.84
PEG, % 16.02 --- --- 15.40 --- ---
2-propanol, % --- --- 9.79 --- --- 9.59
AgNO3, % --- --- --- 3.84 1.98 1.98

 

Table 3.1.Components of membrane casting mixture (% mass)

104

 

Immediately after casting, the ﬁlm was immersed in a room-temperature water bath to
induce phase inversion. Free standing membrane sheets were thoroughly rinsed with DI

water to allow for complete solvent removal.

To prepare more porous Type II membranes, the above procedure was modiﬁed in the
following manner: (1) the polysulfone content in the casting mixture was decreased by a
factor of ca. 2, (2) no porogen was used, and (3) the temperature of the water in the

immersion bath was raised to 70 °C.

To prepare Type IH membranes, which are the most porous, the procedure used to
prepare Type I membranes was modiﬁed as follows: (1) the polysulfone content in the
casting mixture was decreased by a factor of ca. 2, (2) 2-propanol instead of PEG was
used as the pore forming agent, and (3) 2-propanol (room temperature) was used instead

of water as the non-solvent.

3.2.3 Preparation of polysulfone-silver nanocomposite membranes

We adopted two silver incorporation approaches. In the first approach, Ag nanoparticles

were synthesized ex-situ and were added to the casting solution as a Ag-DMF organosol.
The organosol was prepared by adding AgNO3 to 5 mL DMF (reducing agent for silver)

and heating the solution under intense stirring to allow for nanoparticle formation [36].
The obtained organosol was then added to the casting mixture, containing DMA, PSf and

a porogen.

105

The second approach involved an in-situ reduction of ionic Ag+ by DMF in the

membrane casting mixture. In this case, AgNO3 was ﬁrst dissolved in DMF at room
temperature to minimize Ag reduction. The polymer solution was subjected to intense

stirring under heating; once the casting mixture became homogeneous, the AgNO3/DMF

solution was added to it to initiate the reduction of silver ions to AgO with the

concomitant formation of silver nanoparticles. Reduction was allowed to proceed for l

min prior to casting the membrane.

Membranes prepared by these two approaches will be denoted as PSf/Agex (Type) and

PSf/Agin (Type), respectively, where “Type” is one of the three casting mixture

compositions described earlier: Type 1, Type H, or Type 111. While the amount of
polysulfone in PSf/Ag Type II and PSf/Ag Type III membranes was half Of that in the
PSf/Ag Type I membranes, the Ag-to-PSf mass ratio was adjusted to remain constant in

all preparations (Table 1).

3.2.4 Characterization of membranes: SEM, TEM, EDS, FTIR, XPS

A Hitachi S-47OOII Field Emission scanning electron microscope operated in ultra high
resolution mode was used to record SEM images. The transmission electron microscope
used for TEM/EDS characterization was a JEM-2200FS with ﬁeld emission gun
operation at 200 kV and a point resolution of 0.19 nm; images were taken using a Gatan
794 MSC digital camera. The instrument was equipped with an X—ray energy dispersive

spectroscopy (X-EDS) system (Oxford Instruments) with an optimal energy resolution of

106

140 eV. SEM and TEM sample preparation procedures are described in the Supporting

Documentation ﬁle.

XPS measurements of Ag-modiﬁed membranes were carried out using a PHI 5400 ESCA
system (Physical Electronics, Chanhassen, MN) equipped with a non-monochromatic Mg

X-ray. In all measurements a take-off angle of 45° was used.

Transmission ATR-FTIR measurements were performed using a Nicolet Magna-IR 550
spectrometer at a rate of 16 scans per second with a resolution of 1.928 cm]. Omnic

software was used to control spectral acquisition and to process the acquired data.

3.2.5 Permeability and rejection tests

The hydraulic resistance of membranes was measured using a stirred Amicon 8010 dead-
end ultraﬁltration cell (Millipore Corp, Billerica, MA) connected to a pressurized
container (100 psi, 5L, Alloy Products Corp., Waukesha, WI) ﬁlled with ultrapure water.
Prior to each test, the membrane coupons were compacted until a steady permeate ﬂow
was attained. The permeate flux was measured continuously by collecting the ﬁltered
water on a digital balance (AV8101C, Ohaus Corp, Pine Brook, NJ) interfaced with a
computer via an RS-232 port. The data from the balance were logged to a computer using

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

107

3.2. I

The 1.
cores
I'm
Pill-iii.

cont:

To evaluate membrane rejection, 20 mg/L aqueous solutions of 150 kDa dextran were
ﬁltered through the compacted membranes using the same dead-end ﬁltration system.
Dextran concentration in the permeate was determined using a total organic carbon

analyzer (model 1010, 01 Analytical, College Station, TX). Calibration standards and

blanks were run for each set of measurements.

3.2.6 Measuring silver losses

The loss of silver during membrane preparation was assessed by measuring its
concentration in the non-solvent immersion bath by atomic absorption (Spectra AA 200,
Varian Inc., Palo Alto, CA) with a hollow cathode Ag lamp at 328.1 nm (Hamamatsu

Photonics, Japan). All of the analyzed samples, calibration blanks, and Ag standards
contained 2% HNO3 v/v to ensure the complete oxidation of nanoparticulate silver to its
ionic, soluble form. Calibration standards and blanks were doped with 100 11L of the
organic solvent to mimic the chemical makeup of the solution in the immersion bath
during membrane casting. In all cases, silver losses to the immersion bath were below

10% of the silver added to the casting mixture.

To measure the amount of silver leached from the membranes during ﬁltration, ultrapure
water was circulated continuously through Type III PSf/Agin and PSf/Agex membranes

for four consecutive days. A variable ﬂow rate gear pump (model EW-74013-20, Cole-
Parmer, Vernon Hills, IL) was used to deliver the feed water (50 mL) into the Amicon

8010 ﬁltration cells containing the membranes. The permeate ﬂux was directed back into

108

 

the feed vessel. Samples (2 mL) of the circulated water were collected every 24 hr for the
4-day period. The permeate samples were then analyzed for silver content with a
Micromass (Thermo Electron Corporation) Platform quadrupole ICP-MS with a hexapole
collision cell using a concentric nebulizer with water-cooled spray chamber and a
CECTAC ASX-500 autosampler. Final sample concentrations were quantiﬁed using a set
of single-element external calibration standards to enable measurements in the (0 to 2.5)

ppb concentration range.
3.2.7 Membrane biofouling characterization

Antibacterial properties of nanocomposite membranes were quantitatively assessed using
the Bradford protein assay [37 , 38]. Two-sample t-test analysis was performed to assess
the statistical signiﬁcance of the observed differences in antibacterial activity (see

Supporting Documentation (SD) ﬁle for tables with t values).
3.2.7.1 Bioﬁlm growth inhibition under diﬁ‘usive conditions

The E. coli culture for membrane inoculation was grown overnight to achieve an optical
density (OD600) of 0.4. This suspension was then diluted 1:100 in LB medium and 2 mL
of this suspension was ﬁltered through membranes. The fouled membranes were placed
on an agar LB plate (feedside up) for incubation at 37 °C for 48 hr. After incubation, the
fouled membranes were cut in thirds, each coupon was immersed into 1 mL of a 1:1 B-

PEszater protein extraction solution to allow for cell disruption and protein

109

solubilization, and the mixture was then vortexed for ca. 1 min. The liquid phase was
then centrifuged for 1 hr using an Adams compact H centrifuge (Becton Dickinson,
Franklin Lakes, NJ) with a 6-place angle rotor at a ﬁxed speed of 3,200 rpm, and the
supernatant containing soluble proteins was decanted. Finally, 50 [IL of the obtained
solution was dissolved in 1.5 mL of Bradford reagent to be analyzed using UV-Vis
spectrophotometry. The intensity of the 595 nm peak in the absorption spectrum was
recorded using the Bradford software module of a NanoDrop ND-IOOO UV-Vis
spectrophotometer (NanoDrop Technologies, Wilmington, DE) equipped with a Xenon
ﬂash lamp light source and a 2048-element linear silicon CCD detector. Measurements
were performed in 10 3 cycles with a wavelength resolution of 3 nm and an absorbance
precision of 0.003. The calibration curve for protein quantiﬁcation was obtained using
BSA protein standards dissolved in the same solvent as that used for the measured
samples (1:1 B-PER:water); this solvent was also used to record the measurement
baseline. In control experiments with bioﬁlm-free nanocomposites, the peak at 595 nm
was not detected, which conﬁrmed that there were no interferences due to silver.
Triplicate results for each membrane type were obtained by analyzing three separate
areas of each membrane, and the protein values were normalized by the area of the tested
membrane sample. For each nanocomposite, its silver-free counterpart was used as the
comparative basis. In all cases the amount of bacteria deposited both on the membrane
surface and inside the membrane pores during ﬁltration was negligible with respect to the

amount subsequently grown during incubation.

3.2. 7.2 Bioﬁlm growth inhibition under convective conditions

110

To assess the performance of the nanocomposites under ﬁltration conditions, ﬁltration

tests were conducted using an Arrricon 8010 dead-end ultraﬁltration cell (Millipore Corp,
Billerica, MA). The convective ﬂow tests were conducted for Type I PSf/Age”, PSf/Agin,

and PSf (control) membranes by continuously ﬁltering LB medium through inoculated
membranes for 48 hr. The inoculation (initial ﬂow of 2 mL of bacteria solution through
the membrane) and quantiﬁcation procedures were conducted as explained in the
previous section, but bacteria were grown while ﬁltering LB broth rather than on an agar

plate.

3.2. 7.3 Assessment of silver leaching

Silver leaching was assessed by ﬁltering ultrapure water through nanocomposite

membranes and quantifying the antibacterial effect of the permeate using the Bradford
assay. In these tests, 10 mL of ultrapure water was ﬁrst ﬁltered through PSf/Agin Type I
nanocomposites. 9 mL of the obtained permeate was mixed with 1 mL of pre-

concentrated (10x) liquid LB medium, and continuously ﬁltered through an inoculated

silver-free membrane for 24 hr.

For each nanocomposite sample, two 9 mL permeate samples were collected: one after
the ﬁrst day and another one after the second day of continuous ﬁltration. After each day,
bioﬁlm growth on PSf membranes was quantiﬁed using the Bradford assay. The

accumulated 48 hr growth was then calculated from the Obtained results. Given that the

111

concentration of silver in the leachate collected for 2 days under convective conditions
was already close to the detection limit of ICP-MS technique (see Results section), and
considering that the intended effect of silver was biofouling mitigation, the degree of

inhibition of the bioﬁlm growth on these membranes was a convenient and appropriate

surrogate measure of the silver content of the leachate.

112

3.3. Results and Discussion

3.3.1 Silver-free polysulfone membranes

Three strategies were used to prepare more porous PSf membranes of Types H and HI: a)
decreasing the polymer content of the casting mixture, b) increasing the temperature of
the non-solvent bath, and c) altering the composition of the non-solvent bath (Table 1).
Relative to the Type I system, the polymer content of the casting solutions was decreased
by a factor of 2 for both Type H and Type IH membranes. The viscosity of the casting
mixture has a direct inﬂuence on the gelation rate of the membranes [33, 35, 39] as the
solvent/non-solvent exchange in lower viscosity mixtures proceeds faster, allowing for
the formation of more porous materials. For the PSf Type H membranes no porogens
were added; instead, the temperature of the non-solvent bath was increased to hasten the

onset of demixing and produce a higher porosity structure [33].

For the PSf Type III membranes, 2-propanol (weak non-solvent) was added as a porogen
to the casting mixture to bring it closer to the limit of its polymer solubility and to induce
instantaneous demixing at lower polymer content thereby producing more porous
membranes. 2-Propanol was used instead of water as the less effective non-solvent in the
immersion bath to synergistically retard gel formation and further increase the porosity.
By weakening the non-solvent character of the immersion bath, the coagulation process
develops as a gradual, less abrupt transition, allowing for the formation of a more porous

membrane [31, 33].

113

ii.

lht
of l
in:
:1

\

let

'4)
J)

c

lt‘t‘t

Ollll

3.3.1.1 Surface morphology

The surface of the silver-free PSf Type I membranes was of low porosity with pore sizes
of less than 10 nm (Fig. 3.1 a). Type H membranes showed a more porous skin, with
interconnected surface pores 30 nm to 100 nm in size (Fig. 3.1 b). In Type HI
membranes, well-deﬁned, larger pores with the typical pore size of 0.5 pm and with
lower surface density were observed (Fig. 3.1 c). In some experiments (see section
3.3.3.1), the opposite (i.e. more porous) side of the PSf Type H membranes faced the
feed. SEM imaging of this surface showed a macroporous structure, with pores on the

order of 20 mm to 50 pm in diameter and of regular shape (Inset A, Fig. 3.1 b).

3.3.1.2 Cross-sectional morphology

Silver-free membranes of all three types had the asymmetric porous structure, with both
cellular pores and macrovoids typical of polysulfone membranes prepared by phase
inversion. In PSf Type I membranes, tear-shaped elongated macrovoids extended from
the ca. 25 [rm-thick lower porosity layer towards the permeate side of the membrane (Fig.
3.1 d). In Type II and HI membranes, a more pronounced cross-sectional asymmetry
could be observed. In the PSf Type H membranes, a hierarchy of three macrovoid types
of dimensions increasing with the distance away from the membrane skin was observed

(Fig. 3.1 e). In PSf Type III membranes, the elongated narrower macrovoids quickly

114

 

 

  

 

 

Figure 3.1. SEM surface (a, b, c) and cross-section (d, e, f) images of PSf Type I (a, (1),
Type H (b, e) and Type III (c, f) membranes. Inset A shows the permeate side surface of

PSf Type II membranes.

115

transitioned into wider macrovoids that spanned the entire cross-section of the membrane

(Fig. 3.1 0.

3.3.2 Polysulfone-Silver nanocomposite membranes

3.3.2.1 Cross-sectional morphology

For membranes of all types, the majority of the observed nanoparticles were located

along the internal pore surface of the polymeric matrix (Fig. 3.1). TEM EDS results
conﬁrmed that the nanoscale inclusions were indeed silver (see SD). For the PSf/Agex

Type I membrane, SEM imaging revealed a high density of crystalline silver particles
with diameters in the 50 nm to 500 nm range along the skin area but nowhere else along
the membrane cross-section (Fig. 3.1 c); smaller (less than 50 nm) nanoparticles were
observed at all locations within the cross-section but at a much lower density. Compared

to corresponding membranes containing no silver (Fig 3.1 a), the cross-sections of both
PSf/Agex Type I (Fig. 3.2 C) and PSf/Agin Type I (Fig. 3.2 b) nanocomposites showed

wider macrovoids of stressed tear-shaped morphology.

In contrast, no differences in cross-sectional morphologies were observed upon silver
modiﬁcation for the Type 11 (Fig. 3.2 e and Fig. 3.2 f versus Fig. 3.2 d) and Type HI (Fig.
3.2 h and Fig. 3.2 i versus Fig. 3.2 g) membranes, and no large (> 50 nm) Silver crystals
were detected along their cross-sections. However, silver nanoparticles in the 20 nm to 50

nm range were scattered throughout the entire cross-section of all of the silver modiﬁed

116

 
     
   
   

       

    
 
       
 

a) PSf Type I b) PSf /Agin Type I c) PSf lAgex Type I

Macrovoid

   

92:15:
|:'

d) PSf Type II c) PSf/Agin TypeH nPsr/Agex Type II

 

I..°..‘e'.':.'."l. I
‘5 n . «Hz-g.“ 3:3?
I , ..

    

 

m
r r

W .
db ”b

g) PSf Type HI h)PSf /Agin Type III i) PSf /Agex Type III

  

WW
IIIIIIIIIIHIHIIIIIII

Milli I I THE];

 

 

 

 

 

 

 

 

m
IIIIIIIIIIIIIIIIIIIIII

 

 

- —
Figure 3. 2. Schematic illustration of the distribution of srlver nanoparticles In the silver-
polysulfone nanocomposites prepared using different methods of Silver incorporation. On
micrographs, the black dots with arrows pointing to them correspond to silver
nanoparticles. TEM micrographs (A, C and D) show membrane cross—sections, while
SEM micrograph (B) shows the top cross section layer, with large silver clusters
embedded into the polysulfone matrix.

117

membranes (Figures 3.2 A, 3.2 C, 3.2 D). While some nanoparticles in PSf/Agex

membranes could be observed embedded within the polymer of the interpore walls, most
nanoparticles were located along the pore surface (Fig 3.2 D). For all three membrane
types, the ex-situ variations appeared to have a higher nanoparticle density with a

relatively higher number of bigger nanoparticles than their in-situ counterparts. This

difference between PSf/Agex and PSf/Agin membranes could arise for several reasons: 1)
longer reduction period for the silver in PSf/Agex nanocomposites; 2) pre-formed silver

nanoparticles in PSf/Agﬁx casting mixtures acting as nucleation sites for further
nanoparticle growth; 3) limited silver availability for nanoparticle growth under

conditions Of developing porosity in PSf/Agin cast ﬁlms; and 4) a heterogeneous

environment for nanoparticle formation in PSf/Agm membranes wherein other
components of the casting mixture could delay the nucleation and retard the growth of

Ago. The apparent higher nanoparticle densities observed in PSf/Agex membranes may
also be due to the fact that the larger particles in the PSf/Agex composites are more
readily observed. The total amount of silver was approximately the same in PSf/Agex and

PSf/Ag1n casting mixtures, and minimal Ag was detected in the non-solvent after casting.

3.3.2.2 Surface morphology

118

PSf/Ag Type I nanocomposite membranes showed a higher density of nanoscale surface

pores as compared to their PSf Type I counterparts, with no signiﬁcant overall

differences in pore size between PSf/Ag“, PSf/Agin or PSf Type I membranes (Fig. 3.3).

119

 

Figure 3.3.Representative SEM images of the surfaces of PSf Type I (a), PSf/Agin Type I
(b) and PSf/Agex Type I (c) membranes.

120

For PSf Type H and Type HI membranes, there were no discernible differences in the
SEM images of membranes prepared with and without silver (see SD). N 0 silver particles

were Observed to protrude from the top surface of membranes, and XPS analysis of
PSf/Agex Type I membranes revealed no Ag on the membrane surface. (The PSf/Agex

Type I system had the highest Ag content of any membrane but still had no detectable
silver on its surface.) FTIR spectra showed no difference between Type I membranes
prepared with and without Ag nanoparticles. The above ﬁndings imply that the Ag does
not form chemical bonds with PSf. XPS, EDS, and FT IR results are included in the SD

ﬁle.

3.3.2.3 Forrnation-structure links for nanocomposite membranes

As illustrated in Fig. 3.2, silver incorporation did not produce visible macroscale
morphological changes for membranes of Types II and HI. For the less porous Type I
membranes, the introduction of silver did induce macrovoid broadening along with an

increase in surface porosity (Fig. 3.3); however, no signiﬁcant morphological differences
could be observed between PSf/Agin Type I and PSf/Agex Type I membranes. These

ﬁndings indicate that the mere presence of silver nanoparticles in the casting mixture,
regardless of their size, is enough to induce changes in the morphology of the polymeric
matrix for relatively dense membranes. We attribute this to an increase in the demixing
rate in the forming membrane when nanoparticles are added to the casting solution.
Because of faster demixing rates in the lower viscosity mixtures used to prepare Type II

and Type IH membranes, the effect of the ﬁller was not noticeable. The absence of an

121

 

ll

r)_)

'5:
It

effect of Ag incorporation on the structure of Type Hand Type HI membranes could be
partly due to the fact that the Ag loading in the casting solution was smaller for Type II
and 1H membranes (Table 1). (The Ag/PSf ratio was constant, but less PSf was used in

casting of Type Hand HI membranes).

In all cases, nanoparticles seemed to be preferentially concentrated along the internal
pore surface (Insets A, C and D, Fig. 3.2), with PSf/Ag6x membranes showing apparent
higher coverage densities, due to the fact that the ex-situ Ag incorporation approach gave
larger nanoparticles. That PSf/Agex Type I membranes showed larger silver crystals (Fig.
3.2 B) concentrated in the denser top cross-section layer can be explained by noting that
two opposing processes take place in the casting mixture: 1) nanoparticle formation via
DMF-assisted reduction of ionic silver, and 2) nanoparticle re-dissolution into the
DMAC, a component of the casting mixture that cannot reduce silver but can dissolve it.

Such interplay between silver reduction and dissolution results in a pseudo bimodal size

distribution of nanoparticles in the cast membranes. The fact that the large silver crystals

could only be observed in PSf/Agex (Type I) membranes can be due to two reasons: 1)
casting mixtures for PSf/Agex and PSf/Agin membranes of Types II and HI have higher

DMACzAg ratios; 2) PSf/Agin of Types 1, II and III incorporate the silver in

predominantly ionic form due to negligible reduction of silver at room temperature [36]
prior to incorporation into the casting mixture. Thus, the enhanced dissolution process

can be responsible for the absence of larger crystals in membranes other than denser

PSf/Agex Type I nanocomposites.

122

 

In summary, the ﬁller distribution and matrix microstructure of nanocomposites can not
be interpreted to arise from changes in the type or growth conditions of the ﬁller or
changes in the composition of the casting solution alone; instead, interactions in the
quaternary system of polymer, solvent, non-solvent, and ﬁller have to be considered to

predict nanocomposite structure.

3.3.2.4 Permeability and rejection of nanocomposite membranes

The more compact Type I membranes showed a much higher (up to two orders of
magnitude) resistance than the other membrane types (Table 2), which correlates with
their lower surface porosity. Interestingly, for Type I membranes an order of magnitude
decrease in membrane resistance upon silver modiﬁcation was not accompanied by a
major change in membrane rejection of 150 kDa dextran (Table 2). This suggests that
while Type I Ag/PSf nanocomposites and their silver-free counterparts have pores of
approximately same size, the density of pores is higher in the nanocomposites. This effect

of silver on porosity was corroborated by SEM imaging as mentioned above (Fig. 3.3).

Membranes of Type 11 showed both higher permeabilities (lower resistances) and higher
rejections than Type III membranes, likely because of the lower pore density in Type HI
membranes (Table 2). The comparison of PSf/Ag membranes of types 11 and IH and their
silver-free counterparts shows that while the inclusion of silver did not affect flux, there

were signiﬁcant differences in rejection.

123

 

 

 

 

 

 

 

 

 

 

 

Membrane Resistance, Rejection , %
x1012 m'1 150 kDa dextran

PSf 58.8 :1: 11.8 82.5 :1: 12.2

g: PSf/Agin 5.61 :1: 1.02 70.9 1 4.0

E psf/Age" 4.03 x 0.75 67.6 :1: 5.1
PSf 0.16 :i: 0.03 35.1 :1: 0.2

E PSf/Agin 0.15 4: 0.01 32.1 :I: 1.8

E psf/Age" 0.15 :l: 0.02 23.0 i- 6.6
PSf 0.64 1 0.16 17.8 i 3.7

E psf/Agi" 0.53 e 0.06 8.8 e 3.4

F psf/Age" 0.66 i 0.09 5.4 :r: 2.8

 

 

 

 

 

 

Table 3.2.Resistance and rejection values for all membrane types. Mass fractions of
silver were the same for PSf/Agex and PSf/AgIn casting mixtures

124

In summary, the introduction of silver nanoparticles into the polymeric matrix induced
noticeable changes in morphological properties and permeate ﬂux only in membranes
with a denser polymeric matrix (Type I membranes). However, rejection was affected for

membranes of all types.

3.3.3 Antibacterial properties of nanocomposites

Biofouling tests were conducted to determine the antibacterial effectiveness of
nanocomposites of different porosities. Given that the nanoparticles are embedded within

the polymeric structure of the membrane, the biocidal capacity of the nanocomposite
membranes relies on their ability to leach out biotoxic silver ions (Ag+) to the solution.

The measured extent of bacteria growth inhibition was then used to assess silver leaching

by nanocomposites and to evaluate accessibility of embedded nanoparticles.

Recently, several comprehensive studies on the bacteriostatic and bactericidal properties
of silver (both in ionic and nanoparticle form) have been published [40-44]. Silver
nanoparticles can attach to the cell membrane to impede or disrupt vital metabolic
processes, such as cellular respiration and membrane mediated transport mechanisms [43,
44]. N anoparticles can also intercalate into the cell membrane [42-44], penetrate into the
cell [43], and interact with proteins [43] or DNA [43] - macromolecules, possessing
functional groups (amine, thiol, phosphate) that are known to have a high afﬁnity for
silver. Similar interactions can occur with silver ions, which can impact critical metabolic

functions of the cell via protein denaturation [41], or cause lethal mutations and affect

125

cellular replication due to silver induced DNA condensation [41]. Ions can either enter
the cell from the solution of the medium surrounding the cell, or can be introduced inside
the cell as a result of leaching of nanoparticles located in the close proximity or within

the cell [43].

The Bradford assay was chosen for the direct measurement of the amount of bacteria in
the membrane bioﬁlm. A signiﬁcant advantage of the Bradford assay over commonly
applied biofouling quantiﬁcation methods such as colony counting, dye-based, and
microscopy methods [45] is that the assay accounts for intrapore fouling. To our
knowledge this is the ﬁrst application of the Bradford assay to quantify membrane

biofouling.

A thick bioﬁlm and a large amount of extracellular polymeric substances (EPS) cOuld be
observed on the surfaces of all PSf membranes (Fig. 3.4 a, for example) after ﬁltration of

a bacterial solution and subsequent growth of the bacteria on agar. Under the same
conditions, there was much less bacterial growth on the surfaces of both PSf/Agin (Fig.
3.4 b) and PSf/Agex (Fig. 3.4 c) nanocomposites with respect to their silver-free

counterparts. To quantify the amount of bacteria grown on a membrane, we lysed the E.

coli cells present inside and on the surface of the membranes and determined the amount
of released protein using a Bradford assay. As shown in Figure 3.5, the PSf/Agin
membranes showed less released protein than the PSf/Agex membranes Of the same type.

This conclusion was conﬁrmed by t-testing, with a level of conﬁdence higher than 95%

for differences between PSf/Agex and PSf/Agin membranes of Types 1, H and H1

126

 

 

Figure _3.4.Typical SEM images of the surface of biofouled PSf Type I silver-free (a),
Psr/Ag‘" (b) and PSf/Agex (c) membranes. Evident is the stark contrast between the thick
bioﬁlm developed on the surface of a silver-free membrane (a), and the absence of
bioﬁlm growth on the nanocomposite surface. The sample preparation procedure is
detailed in section 3.2.4.

127

inoculated using the standard procedure when bacteria were deposited onto the skin
surface of the membrane. The difference between PSf/Agex and PSf/Agin was
signiﬁcantly different at a conﬁdence level of 90% for Type H membranes inoculated
through the porous support side (“Type II reverse” in Fig. 3.5). The observed difference
in antibacterial activity between PSf/Agin and PSf/Agex could result from a higher
number of smaller nanoparticles, and, therefore, higher total surface area of the ﬁller, in

PSf/Agin membranes.

Figure 3.5 shows that the bioﬁlm growth on the more porous membranes (Types H and
HI) was more extensive than that on Type I membranes, presumably because the higher
porosity of the Type II and HI surfaces facilitates initial attachment of bacteria, which
leads to further growth and EPS secretion. The developing bioﬁlm could also act as a
protective barrier for subsequent bacteria by hindering the diffusion of silver ions towards

the bioﬁlm.

With all membranes, the nanoscale silver was somewhat effective in reducing bioﬁlm
growth (Figure 3.5). In all cases, silver modiﬁed membranes showed a signiﬁcant
decrease (higher than 95 % conﬁdence level) in antibacterial efﬁciency with respect to
the silver free counterparts. The t-test analysis indicates that for the very porous
membranes, simulated by Type II membranes operated with the porous support side
facing the feed, the antibacterial effect was less dependent on the method of silver
incorporation; this was probably due to the fact that the access of bacteria to silver was

limited by the silver diffusion rate and not by the availability of silver nanoparticles.

128

 

2.5
e
D
N
T“ 2
E
O
5
E 1.5
8.
E
E 1
m
0.5
0

 

 

 

 

III

I I

 

I

 

 

 

 

Type II
Type IH reverse
Type II

TypeI _
_(

l lﬂl ml 1 l l | l 1 I I I l l l l l l l l l l l |
t... s .s u—i :5 .E u—i 5 .5 c... 5 .s
m on or) m on ‘30 W co co m 60 at)
D- < < 0.. <1 < 9" <11 < Ch < <
in s: m a s: a a a

m m

‘63 E E e- 33 a. o. a.

Figure 3.5.Bradford assay results describing the bioﬁlm growth on the surface of PSf
membranes and PSf/Ag nanocomposites of different porosities under diffusive transport

conditions.

129

 

While the accessibility of silver is a function of both the route of nanoparticle
incorporation and the morphology of the polymeric matrix, these two factors are
interdependent and it is thus difﬁcult to separate their inﬂuence on ﬁller’s accessibility to
the permeate ﬂow. Note that the membrane skin, which in this experiment was on the
permeate side, provided an absolute barrier to the bacteria, so the number of inoculated
bacteria retained by the membrane was the same as that in all other biofouling assays.
Thus a straightforward comparison with the Type H membranes inoculated on the skin

side should be possible.

Finally, consideration must be given to the fact that the mass of silver contained in more
porous nanocomposites (types II and IH) was lower than in the denser Type I membranes
(although the PSf/Ag mass ratio was the same). The lower absolute amount of silver,
coupled with the fact that more porous membranes have a rougher surface and, therefore,
higher surface area, could have also contributed towards the observed higher bacteria

growth in more porous matrices.

3.3.4 Practical aspects of the application of Silver nanocomposites as membrane and

membrane spacer materials

3.3.4.1 Effect of antibacterial activity on the eﬁ‘iciency of hydraulic cleaning

To evaluate the responsiveness of bioﬁlms formed on PSf and PSf/Ag membranes to

hydraulic cleaning, rinsing was included as a part of the biofouling assay. During the

130

 

rinsing step, membranes were washed with ultrapure water for 30 seconds. Even though
in all cases there was an observed, and expected, decrease in the measured protein
content after hydraulic rinsing (see SD), statistical analysis showed that the differences in
the efﬁciency of hydraulic cleaning between nanocomposites and their silver-free

counterparts were not statistically signiﬁcant.

3.3.4.2 Antibacterial efficiency of nanocomposite membranes under convective transport

conditions

As stated in the experimental section, PSf/Agin and PSf/Agex nanocomposites of Type III
were tested in a continuous deionized water recirculation system, and the leached silver
was periodically measured using ICP-MS. Both PSf/Agin HI and PSf/Agex HI leached
comparable amounts of silver with ﬁnal 4-day silver loss values (expressed as a percent
of the calculated mass of silver originally present in the coupons) of 0.0016% and
0.0014% for the PSf/Agin HI and PSf/Agex 111, respectively. One could expect even
slower leaching for nanocomposites of types I and H because of their lower porosity. In
all cases the leachates had a silver concentration around or below 1 jig/L, which is 100
times below the EPA standard (0.1 mg/L) for Ag+ in drinking water. Interestingly, this
concentration is also signiﬁcantly below reported values (4 ug/mL to 85 jig/mL) for the -
concentration of silver (ionic or nanoparticulate form or both) that results in observed
toxicity for E. coli[40-44, 46]. It should be noted, however, that ionic silver released from

the nanocomposites might be enhanced in the presence of exogenous complexing agents

that could be found in more complex water matrices.

I31

While the values of silver concentration were within the range of the calibration curve,
they were close to the ICP-MS detection limit. For this reason, further leaching tests
involved indirect measurement of silver by the degree of bioﬁlm growth inhibition, as
explained in section 3.2.7.2. The bioﬁlm growth inhibition under convective ﬂow
conditions was studied only for Type I membranes. The hypothesis was that the leached
silver would be carried away from the membrane surface with the permeate ﬂow and thus
the growth inhibition capacity of the nanocomposite would be suppressed. To test the
hypothesis under the most conservative conditions, Type I membranes were chosen as
those that showed the highest growth inhibition in preliminary tests without permeate
ﬂow (Fig. 3.5). The results obtained for the inhibition of bioﬁlm growth under conditions
of convective ﬂow through the membrane are shown in Fig. 3.6. The ﬁrst two bars of Fig.
3.6 describe the extent of bioﬁlm growth after continuous 48 hr of ﬁltration of LB
medium through nanocomposite samples that were previously used to ﬁlter 2 mL of an E.

coli solution. There was no signiﬁcant difference between the growth observed on
PSf/Agin Type I membranes and PSf controls. We attribute this to the fact that leached

silver is convectively transported towards the permeate side Of the membrane and away
from bacteria accumulated on the membrane surface facing the feed. This ﬁnding
emphasizes the limited efﬁciency of biocide—releasing membranes in inhibiting the
bioﬁlm development on the upstream side of the membrane due to the fact that leachate
is carried away from the growing bioﬁlm. This conclusion is further corroborated by the

result of a silver leaching test under convective conditions where silver-containing feed

132

(permeate from a bacteria-free silver nanocomposite) is directed to the membrane surface

inoculated

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.6 I r
'l

N»- 1.4 — ._ _.
e
O
as 1.2 — -
E
.s it —
0)
‘5 u
a: 0.8 - _.
'U
0.)
,2. 0.6 -. -
g .l
a 0.4 _ ~
Z

0.2 - " -

O 1 1
PSf PSf/Agi" PSf/Ag‘"

Figure 3.6.Bioﬁlm growth on PSf/Ag Type I nanocomposites after 48 hr of continuous
ﬁltration of a) LB medium (ﬁrst two columns) and b) leachate obtained by continuously
ﬁltering DI water through PSf/Ag Type Inanocomposite for 2 days (third column). The
leachate was Ag-loaded water obtained from the ﬁltration of DI water through a bacteria-
free PSf/Ag Type I nanocomposite; the leachate was then mixed with LB medium to be
ﬁltered through an inoculated Ag-free membrane.

133

with bacteria (section 3.2.7.3.) The third column of Fig. 3.6 illustrates the inhibition of
the bioﬁlm growth on the membrane surface when 2-day nanocomposite leachate is
ﬁltered through a membrane inoculated with bacteria, where a signiﬁcant (90 %

conﬁdence level) decline was observed in this case.

Although convective ﬂow essentially eliminated the biocidal activity of PSf/Ag
membranes, the built-in antibacterial capacity should still be effective in reducing
intrapore fouling in more porous membranes. The silver-releasing nanocomposites could
also be used as materials for macroporous membrane spacers to inhibit the bioﬁlm
growth on downstream membrane surfaces. The leached silver would primarily be
targeting, and absorbed by, bacteria growing on the downstream target membrane. Given
the extremely low concentrations of silver necessary to inhibit the growth of membrane
bioﬁlms, such approaches may be advantageous relative to the use of the biocide as a
bulk water additive wherein the biocide’s ability to reach the membrane could be affected
by the presence of other surfaces and components of the water that could act as sinks for

the biocidal agent.

134

 

3.4. Conclusions

In this study, the effects of casting mixture composition and nanoparticle incorporation
route on the morphological and separation properties of prepared porous nanocomposite

membranes were studied using silver as the model ﬁller material. We demonstrated that:

(1) The reduction of ionic silver by polymer solvents prior to or during the

 

phase inversion process can be employed to synthesize silver-polymer

nanocomposites of a range of porosities and nanoscale silver distributions.

(2) Ex-situ and in-situ routes of ﬁller incorporation result in markedly
different nanocomposite morphologies. The two methods yield differences in both

the size of the ﬁller and its distribution along the cross-section of the matrix.

(3) The effects of ﬁller incorporation are more pronounced for less porous
membranes. For casting mixtures that yield membranes with relatively low
porosity, the induced morphological changes - macrovoid broadening and an
increase in surface pore size and density - translate into an improvement in
separation properties wherein a signiﬁcant decrease in the hydraulic resistance is
accompanied by only a minor decrease in rejection. To predict the structure of
such nanocomposite membranes, interactions in the quaternary system of

polymer, solvent, non-solvent, and ﬁller have to be considered.

135

(4) The built-in antibacterial capacity due to the gradual release of ionic silver
by the prepared nanocomposites can be effective in reducing intrapore biofouling
in porous membranes of a wide range of porosities. Such nanocomposites could in
principle be used as materials for macroporous membrane spacers to inhibit the

bioﬁlm growth on downstream membrane surfaces.

In summary, incorporation of nanoparticles into a forming membrane matrix appears to
be an attractive approach to the design of nanocomposite structures with improved
mechanical and separation properties and with embedded nanoparticle-based

functionalities.

136

 

Acknowledgements

This material is based upon work supported by the National Science Foundation under
Grant No. OISE-0530174. Robin Diets is acknowledged for his assistance with
membrane permeability and rejection measurements. We also thank Solvay Polymers for
providing the polysulfone for membrane synthesis. J ST acknowledges support by
Nordberg and McCowan fellowships. Any opinions, ﬁndings, and conclusions or
recommendations expressed in this material are those of the authors and do not

necessarily reﬂect the views of the National Science Foundation.

137

 

Supporting Documentation

Effect of ﬁller incorporation route on the properties of polysulfone-

silver nanocomposite membranes of different porosities

Published in
Journal of Membrane Science

Volume 325, 2008, pages 58-68

by

Julian S. Taurozzil, Hari ArulII , Volodymyr Z. Bosaki, Anatoliy F. Burbani, Thomas C.
Voicel, Merlin L. BrueningI, and Volodymyr V. TarabaraT

I Department of Civil and Environmental Engineering, Michigan State University, East
Lansing, MI 48824, USA
H High School Honors Science, Mathematics, Engineering Program, Michigan State
University
:5 Department of Chemistry, The National University of "Kyiv-Mohyla Academy", 2,
Skovoroda St., Kyiv, 04070, Ukraine

IDepartment of Chemistry, Michigan State University, East Lansing, MI 48824, USA

138

 

SD.3.1.SEM sample preparation procedure

For SEM imaging, all samples were coated with osmium for 20 s at a current of 10 mA
using an osmium plasma coater (model NEOC-AT, Meiwa Shoji Co., Osaka, Japan). To
image membrane cross-sections, samples were freeze-fractured upon immersion in liquid
nitrogen, coated and mounted on SEM sample holders. For bioﬁlm observation,
membranes were ﬁxed by immersion in a solution of glutaraldehyde and
paraforrnaldehyde (2.5% v/v each) in 0.1 M sodium cacodylate buffer for 30 min
followed by a 10 min immersion in 0.1 M sodium cacodylate buffer. The ﬁxed samples
were then dehydrated by sequential rinsing in solutions of ethanol in water with

increasing ethanol concentrations (25, 50, 75, 95, and 100% ethanol).

SD.3.2.TEM sample preparation procedure

For TEM sample sectioning, the membranes were embedded in LRwhite resin (medium

grade, London Resin Co., UK) for 24 hr and further polymerized for 24 hr in an oven at
60 0C. Ultrathin (70 nm) sections of the embedded samples were cut using an

ultramicrotome (Power Tome XL, RMC, Boeckeler Instruments. Tucson, AZ) equipped
with a diamond knife (Diatome, 35° angle). For ultrathin section observation, a JEOL 100
CX TEM with an accelerating voltage of 100 kV and a maximum theoretical resolution
of 0.2 nm, coupled to a MegaView IH Camera (Soft Imaging System, Lakewood, CO)
was employed. Standard copper grids (300 mesh, Electron Microscopy Sciences) were

used to support samples.

139

 

To verify that silver nanoparticles were not dissolved during embedding, a control test
was performed. Silver nanoparticle hydrosol was prepared [32] with a mean particle
diameter of around 50 nm (ZetaPALS, Brookhaven Instrument Corp., Holtsville, NY).
The obtained hydrosol was then added to the vial with LR white resin, and vortexed to
allow for nanoparticle dispersion within the resin. The nanoparticles were left in the
viscous medium for 24 hr, emulating the membrane embedding process, and the resin
was then polymerized in an oven at 60 °C. Ultrathin sections of the nanoparticle-doped
resin were obtained following the same procedure as that for membrane sample
sectioning, and observed using the same TEM instrument. The TEM micrographs clearly
showed well deﬁned crystalline nanoparticles within the measured 50 nm size range.
Similar images were obtained under the TEM for the hydrosol dried on a formvar coated

TEM grid, verifying that dissolution of silver in the embedding resin was negligible.

SD.3.3.TEM-EDS analysis

The transmission electron microscope used for TEM/EDS characterization was a JEM-
2200FS with ﬁeld emission gun operation at 200 kV and a point resolution of 0.19 nm;
images were taken using a Gatan 794 MSC digital camera. The instrument was equipped
with an X-ray energy dispersive spectroscopy (X-EDS) system (Oxford Instruments) with
an optimal energy resolution of 140 eV. The following image shows the spectra obtained

from the analysis of a nanoparticle loaded portion of a PSf/Agin Type I membrane.

140

 

 

Spectrum 1 (line)

Spectrum 2 (filled)

    

49
“killed“ ‘th-Lnlhtuu amumwteh .l'u- It.
2 4 6

8 10 12 14 16 18 20 22 24 26 28

Figure SD.3.1.TEM image (top) of silver nanoparticles (Spectrum 1) embedded in the
PSf matrix (Spectrum 2) of a PSf/Ag1n Type I membrane, and EDS spectra (bottom) of
both the PSf matrix (yellow ﬁlled spectrum) and the Ag nanoparticles (line spectrum)

showing the silver peaks for the nanoparticle. The x-axis for the EDS spectra is in KeV
units.

141

SD.3.4.ATR-FI‘IR analysis

ATR-FTIR measurements were performed using a Nicolet Magna-IR 550 spectrometer at
a rate of 16 scans per second with a resolution of 1.928 cm]. Omnic software was used to

control spectral acquisition and to process the acquired data.

 

142

 

 

 

 

 

 

 

 

 

 

Absorbance (normalized)

 

 

 

0.4 -

 

 

 

 

w j

0.2 -

 

 

 

O l l l l I L l
950 1 000 1050 1 100 1 1150 1200 1 250 1 300 1 350
CITI'

 

Figure SD.3.2.FTIR spectra of PSf Type I (Top) and PSf/Agin Type I (bottom)
membranes.

143

SD.3.5.XPS analysis
XPS measurements of Ag-modiﬁed membranes were carried out using a PHI 5400 ESCA
system (Physical Electronics, Chanhassen, MN) equipped with a non-monochromatic Mg

X-ray. In all measurements a take-off angle of 45° was used. The following shows the

results obtained from the analysis of a PSf/Agin Type I membrane.

144

 

 

 

 

 

 

 

 

12 l l I I I I I I I I
m
U
I
10— -
8_ C.
:2
Q
_J
TU; 6- SCJ _J T
_.r
U X
C?
4 n
2" .1
(5}
V‘ a)
N r o,
m N
(7.0
I
L N

 

0 I 1 1 I l l J 1 1
1 100 1 000 900 800 700 600 500 400 300 200 100 0
Binding Energy (eV)

 

 

 

 

Cls le 013 Sin 82p Ag3d

 

 

80.50 0.02 15.62 0.83 2.97 0.06

 

 

 

 

 

 

Figure _SD.3.3.XPS survey spectrum and atomic concentration (relative) table for a
PSf/Agm Type I membrane. The silver signal is not visible in the survey spectrum.

145

 

SD.3.6.Effect of antibacterial activity on the efficiency of hydraulic cleaning

To evaluate the responsiveness of bioﬁlms formed on PSf and PSf/Ag membranes to
hydraulic cleaning, rinsing was included as a part of the biofouling assay. During the
rinsing step, membranes were washed with ultrapure water for 30 seconds. The following

ﬁgure shows the measured protein values for the membranes with and without rinsing.

146

 

IFTIII

I

 

.4

Type II

reverse

I Type III

 

 

I Before Rinse

@ After Rinse

 

 

Type H

Type I

_ _ _ _ L

new?”.uouonouhovvnen.

2.3.....35.......3....¢.

ccccccccccccccccccccccccccccccccccccccccccc

_
5

1

nonononononoueuonou.
vooooceoooedoooccdc.

eoooooooooooboooc...

cccccccccccccccccccccccccccccccccccc
orator...none...orator?operebororororeeoboueve-or95.00.555.055...

lv...0.0.0....OOOOOOIOOWOOOOO.

...........................
I. III re.9.55....5.....o..Peasant...5.3.5.9....

tttttttttttttttttttt
TI .6».51.555.016.05nopopopotepfopo.

_ Ir _ _

chNsagocv £289 832m

_
0

eméme
emémm

“mm

emémm
sesame
awe

eméme
emémm
Ems

emﬁmm
emémm
ewe

Figure SD.3.4.Bradford assay results describing the bioﬁlm growth on the surface of PSf
membranes and PSf/Ag nanocomposites of different porosities under diffusive transport

conditions, both with and without hydraulic rinsing.

147

SD.3.7.T-test tables

T-tests were performed to determine whether differences in the biocidal effects for the
tested nanocomposites were statistically signiﬁcant. The following tables summarize the
comparative t values for the analyzed pairs. The more general, two-sample T-test (instead
of the commonly used pooled T-test) was applied, due to the large variance differences

between sample sets.

148

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Mean Error t values t values .
(VS- PSf) (Agex vs Ag m)
PSf 1.000 0.479
é PSf/Ag‘" 0.318 0.092 2.82
{-1
PSf/Agex 0.130 0.058 3.64 3.50
PSf 1.358 0.475
= .
é PSf/Ag'" 0.702 0.159 2.65
E—
PSf/Agex 0.373 0.070 4.15 3.82
PSf 1.618 0.509
El .
a PSf/Ag'" 0.939 0.169 2.56
>~.
{-
PSf/Ag“ 0.442 0.097 4.59 5.15
PSf 1.665 0.555
== ‘8 .
8. a PSf/Ag'" 0.705 0.133 3.40
13‘ e
PSf/Ag” 0.569 0.097 3.93 1.67

 

 

 

 

 

 

 

Table SD.3.1.Statistical data analysis for released protein on diffusive biofouling
experiments without hydraulic rinsing (Figure 3.5 of main document). Mean values are
normalized to that of PSf/Type I membranes, errors are calculated with a 90% conﬁdence
level, the “t values” columns indicate the comparative basis for each pair.

149

 

t values
Mean Error (vs. PSf-I)

 

PSf-I 1.000 0.518 ---

 

 

 

 

 

 

 

PSf-I/Agi" 0.906 0.481 0.344
“5:22 0.576 0.333 1.705

 

 

Table SD.3.2.Statistical data analysis for released protein on convective biofouling
experiments (Figure 3.6 of main document). Mean values are normalized to that of
PSf/Type I membranes, errors are calculated with a 90% conﬁdence level, the “t values”

columns indicate the comparative basis for each pair.

 

150

References

1. T. H. Bae and T. M. Tak, Preparation of Ti02 self-assembled polymeric
nanocomposite membranes and examination of their fouling mitigation effects in a
membrane bioreactor system, J. Membr. Sci. 266 (2005) 1-5.

2. J. Dai and M. L. Bruening, Catalytic nanoparticles formed by reduction of metal
ions in multilayered polyelectrolyte ﬁlms, Nanoletters 2 (2002) 497 - 501.

3. Z. Li, H. Huang, T. Shang, F. Yang, W. Zheng, C. Wang and S. K. Manohar,
Facile synthesis of single-crystal and controllable sized silver nanoparticles on the
surfaces of polyacrylonitrile nanoﬁbres, Nanotechnol. 17 (2006) 917-920.

4. R. Molinari, L. Palmisano, E. Drioli and M. Schiavello, Studies on various reactor
conﬁgurations for coupling photocatalysis and membrane processes in water puriﬁcation,
J. Membr. Sci. 206 (2002) 399-415.

5. L. Wu and S. M. C. Ritchie, Removal of trichloroethylene from water by cellulose
acetate supported bimetallic N i/Fe nanoparticles, Chemosphere 63 (2006) 285-292.

6. L. F. Wu, M. Shamsuzzoha and S. M. C. Ritchie, Preparation of cellulose acetate
supported zero-valent iron nanoparticles for the dechlorination of trichloroethylene in
water, J. Nanopart. Res. 7 (2005) 469-476.

7. J. Xu and D. Bhattacharyya, Fe/Pd nanoparticle immobilization in microﬁltration
membrane pores: Synthesis, characterization, and application in the dechlorination of
polychlorinated biphenyls, Ind. Eng. Chem. Res. 46 (2007) 2348-2359.

8. P. Aerts, I. Genne, S. Kuypers, R. Leysen, I. F. J. Vankelecom and P. A. Jacobs,
Polysulfone-aerosil composite membranes: Part 2. The inﬂuence of the addition of aerosil
on the skin characteristics and membrane properties, J. Membr. Sci. 178 (2000) 1-11.

9. P. Aerts, E. Van Hoof, R. Leysen, I. F. J. Vankelecom and P. A. Jacobs,
Polysulfone-aerosil composite membranes: Part 1. The inﬂuence of the addition of
Aerosil on the formation process and membrane morphology, J. Membr. Sci. 176 (2000)
63-73.

10. L. Ballinas, C. Torras, V. Fierro and R. Garcia-Valls, Factors inﬂuencing
activated carbon-polymeric composite membrane structure and performance, J. Phys.
Chem. Solids 65 (2004) 633-637.

11. A. Bottino, G. Capannelli and A. Comite, Preparation and characterization of
novel porous PVDF-Zr02 composite membranes, Desalination 146 (2002) 35-40.

151

 

 

12. A. Bottino, G. Capannelli, V. D'Asti and P. Piaggio, Preparation and properties of
novel organic-inorganic porous membranes, Separ. Purif. Technol. 22-23 (2001) 269—
275.

13. W. L. Chou, D. G. Yu and M. C. Yang, The preparation and characterization of
silver-loading cellulose acetate hollow ﬁber membrane for water treatment, Polym.
Advan. Technol. 16 (2005) 600-607.

14. I. Genne, S. Kuypers and R. Leysen, Effect of the addition of ZrO2 to polysulfone
based UF membranes, J. Membr. Sci. 113 (1996) 343-350.

15. J. B. Li, J. W. Zhu and M. S. Zheng, Morphologies and properties of
poly(phthalazinone ether sulfone ketone) matrix ultraﬁltration membranes with entrapped
Ti02 nanoparticles, J. Appl. Polym. Sci. 103 (2006) 3623-3629.

16. D. E. Meyer, K. Wood, L. G. Bachas and D. Bhattacharyya, Degradation of
chlorinated organics by membrane-immobilized nanosized metals, Environ. Prog. 23
(2004) 232-242.

17. O. Monticelli, A. Bottino, I. Scandale, G. Capannelli and S. Russo, Preparation
and properties of polysulfone-clay composite membranes, J. Appl. Polym. Sci. 103
(2007) 3637-3644.

18. N. M. Wara, L. F. Francis and B. V. Velamakanni, Addition of alumina to
cellulose acetate membranes, J. Membr. Sci. 104 (1995) 43-49.

19. L. Yan, Y. S. Li, C. B. Xiang and S. Xianda, Effect of nano-sized A1203-particle
addition on PVDF ultraﬁltration membrane performance, J. Membr. Sci. 276 (2006) 162-
167.

20. Y. Yang, H. Zhang, P. Wang, Q. Zheng and J. Li, The inﬂuence of nano-sized
Ti02 ﬁllers on the morphologies and properties of PSF UF membrane, J. Membr. Sci.
288 (2007) 231-238.

21. D. (3. Yu, M. Y. Teng, w. L. Chou and M. C. Yang, Characterization and
inhibitory effect of antibacterial PAN -based hollow ﬁber loaded with silver nitrate, J.
Membr. Sci. 225 (2003) 115-123.

22. X. He, Q. Shi, X. Zhou, C. Wan and C. Jiang, In situ composite of nano Si02-
P(VDF-HFP) porous polymer electrolytes for Li-ion batteries, Electrochim. Acta 51
(2005) 1069-1075.

23. W. K. Son, J. H. Youk, T. S. Lee and W. H. Park, Preparation of antimicrobial
ultraﬁne cellulose acetate ﬁbers with silver nanoparticles, Macromol. Rapid Commun. 25
(2004) 1632-1637.

152

 

24. K. Ebert, D. Fritsch, J. K011 and C. Tjahjawiguna, Inﬂuence of inorganic ﬁllers on
the compaction behaviour of porous polymer based membranes, J. Membr. Sci. 233
(2004) 71-78.

25. G. Jonsson, Methods for determining the selectivity of reverse osmosis
membranes, Desalination 24 (1977) 19-37.

26. K. M. Persson, V. Gekas and G. Tragardh, Study of membrane compaction and its
inﬂuence on ultraﬁltration water permeability, J. Membr. Sci. 100 (1995) 155-162.

27. J. H. Choi, J. legal and W. N. Kim, Fabrication and characterization of multi-
walled carbon nanotubes/polymer blend membranes, J. Membr. Sci. 284 (2006) 406-415.

28. D. E. Meyer and D. Bhattacharyya, Impact of membrane immobilization on
particle formation and trichloroethylene dechlorination for bimetallic Fe/N i nanoparticles
in cellulose acetate membranes, J. Phys. Chem. B 111 (2007) 7142-7154.

29. M. J. Han and D. Bhattacharyya, Changes in morphology and transport
characteristics of polysulfone membranes prepared by different demixing conditions, J.
Membr. Sci. 98 (1995) 191-200.

30. V. Kaiser and C. Stropnik, Membranes from polysulfone/N,N-
dimethylacetamide/water system; Structure and water ﬂux, Acta Chimica Slovenica 47
(2000) 205-213.

31. P. Radovanovic, S. W. Thiel and S. T. Hwang, Formation of asymmetric
polysulfone membranes by immersion precipitation.2. The effects of casting solution and
gelation bath compositions on membrane-structure and skin formation, Journal Of
Membrane Science 65 (1992) 231-246.

32. C. Stropnik, V. Kaiser, V. Musil and M. Brumen, Wet-phase-separation
membranes from the polysulfone/N,N-dimethylacetamide/water ternary system: The
formation and elements of their structure and properties, Journal Of Applied Polymer
Science 96 (2005) 1667-1674.

33. RE. Kesting, Synthetic polymeric membranes, a structural perspective. 2nd ed.
1985: John Wiley & Sons, Inc.

34. J. H. Kim and K. H. Lee, Effect of PEG additive on membrane formation by
phase inversion, Journal of Membrane Science 138 (1998) 153-163.

35. Q. Z. Zheng, P. Wang and Y. N. Yang, Rheological and thermodynamic variation
in polysulfone solution by PEG introduction and its effect on kinetics of membrane
formation via phase-inversion process, J. Membr. Sci. 279 (2006) 230-237.

36. I. Pastoriza-Santos and L. M. Liz-Marzan, Formation and stabilization of silver
nanoparticles through reduction by N,N-Dimethylforrnamide, Langmuir 15 (1999) 948-
951.

153

 

37. M. M. Bradford, A rapid and sensitive method for the quantitation of microgram
quantities of protein utilizing the principle of protein-dye binding, Anal. Biochem. 72
(1976) 248-254.

38. J. D. Gu and L. Pan, Comparing the growth characteristics of three bacteria
involved in degrading rubbers, J. Polym. Environ. 14 (2006) 273-279.

39. K. Boussu, C. Vandecasteele and B. Van der Bruggen, Study of the characteristics
and the performance of self-made nanoporous polyethersulfone membranes, Polymer 47
(2006) 3464-3476.

40. C. Baker, A. Pradhan, L. Pakstis, D. J. Pochan and S. I. Shah, Synthesis and
antibacterial properties of silver nanoparticles, J. Nanosci. N anotechnol. 5 (2005) 244-
249.

41. Q. L. Feng, J. Wu, G. Q. Chen, F. Z. Cui, T. N. Kim and J. 0. Kim, A
mechanistic study of the antibacterial effect of silver ions on Escherichia coli and
Staphylococcus aureus, J. Biomed. Mater. Res. 52 (2000) 662-668.

42. S. K. Gogoi, P. Gopinath, A. Paul, A. Ramesh, S. S. Ghosh and A.
Chattopadhyay, Green ﬂuorescent protein-expressing Escherichia coli as a model system
for investigating the antimicrobial activities of silver nanoparticles, Langmuir 22 (2006)
9322-9328.

43. J. R. Morones, J. L. Elechiguerra, A. Camacho, K. Holt, J. B. Kouri, J. T. Ramirez
and M. J. Yacaman, The bactericidal effect of silver nanoparticles, Nanotechnology 16
(2005) 2346-2353.

44. I. Sondi and B. Salopek-Sondi, Silver nanoparticles as antimicrobial agent: a case
study on E-coli as a model for Gram-negative bacteria, J. Colloid Interface Sci. 275
(2004) 177-182.

45. J. Mallevialle, P. E. Odendaal and M. R. Wiesner, Water Treatment Membrane
Processes. 1996: McGraw-Hill.

46. K. Y. Yoon, J. H. Byeon, J. H. Park and J. Hwang, Susceptibility constants of
Escherichia coli and Bacillus subtilis to silver and copper nanoparticles, Sci. Total
Environ. 373 (2007) 572-575.

154

CHAPTER FOUR

Silver nanoparticle arrays on track etch membrane support as ﬂow-

through optical sensors for water quality control
Published in

Environmental Nanotechnology special issue of

Environmental Engineering Science journal

Volume 24 (1), 2007, pages 122-137

by

Julian S. Taurozzii and Volodymyr V. Tarabarai

1 Department of Civil and Environmental Engineering, Michigan State University,

East Lansing, MI 48824, USA

155

CHAPTER FOUR

Silver nanoparticle arrays on track etch membrane support as ﬂow-
through optical sensors for water quality control

Abstract

Silver nanoparticle arrays were assembled on the surface of polycarbonate track etch
membranes using 3-aminopropyltrimethoxysilane as the chemical linker. The assembled
arrays were sub-monolayer and reproducibly regular. The nanoparticle-modiﬁed
membranes were characterized in terms of their hydraulic and optical properties and were
evaluated as ﬂow-through surface-enhanced Raman scattering (SERS) sensors for water
quality monitoring. The membrane-supported nanoparticle arrays were found to be
SERS-active with slightly lower but signiﬁcantly more reproducible enhancements in
comparison with enhancements afforded by source nanoparticle suspensions. Large
enhancements of up to 105 were demonstrated due to preconcentration by recirculating
analyte solution through nanoparticle-modiﬁed membranes that were shown to retain
their hydraulic properties. The ﬁndings point to the promise of combining high
speciﬁcation SERS-active systems and the ﬂow-through design. for the development of

analytical sensors for the trace detection of aqueous pollutants.

156

4.1. Introduction

Timely detection of water supply contamination, accurate assessment of related risks, and
development of appropriate remediation strategies depend on the availability of reliable
water quality monitoring technologies [1]. At the core of these technologies are
deployable environmental sensors subject to such criteria as ﬁngerprinting capability, low
fouling characteristics, reproducibility, and low detection limits. The lack of sensing
devices that meet these demands is the main limitation of pollution detection methods in

use today.

The detection process can be viewed as consisting of two consecutive stages: (1)
selective preconcentra(ion/separation of pollutants and (2) measurement of pollutant
concentration. To reliably determine the occurrence of molecular contaminants or
microbial pathogens in natural and treated waters, large volumes (up to 1000 liters) of
source water have to be sampled and preconcentrated to much smaller volumes. For
rapid assaying of microbial pathogens, for example, acceptable concentrate volumes are
smaller than 1 liter and often are as small as several milliliters [2, 3] . In addition to
separating pollutants from the system matrix, preconcentration enhances detection
sensitivity, which is especially important for the detection of water pollutants in very low
concentrations. The practical worth of preconcentration is exempliﬁed by the use of

membranes for the detection of microbial pathogens in aqueous systems [3-8].

157

 

At the measurement stage, the employed detection methods have to be sensitive and
selective. The remarkable sensitivity and molecular speciﬁcity afforded by surface
enhanced Raman scattering (SERS) [9-13] makes this spectroscopic technique especially
attractive for those sensing applications, where both low detection limits and
ﬁngerprinting capability are needed [14]. Reports of single-molecule detection using
SERS [15] and the design of ﬂow-through SERS systems [16] have further ampliﬁed

interest in possible uses of this technique for environmental sensing [17-21].

The SERS effect can be observed when a light-scattering molecule is adsorbed on a
SERS-active substrate, typically a roughened surface of certain coinage metals, gold and
silver being primary examples. The choice of the type of the enhancing substrate remains
application-speciﬁc because different types of SERS-active substrates exhibit different
properties; island metal ﬁlms on a solid support, for example, are more stable, while for
aggregated metal colloids higher enhancements are generally expected [22]. What all
SERS-active systems have in common, however, is that the morphology of substrates has
a profound effect on their optical properties. In the case of nanoparticle suspensions, the
relatively poor stability of sols and the dependence of nanoparticle aggregate morphology
on the nature and concentration of the analyte partly offsets the advantage of high
enhancements typical of such substrates. The difﬁculty in fabricating SERS-active
substrates that reproducibly yield high enhancements is the single most signiﬁcant factor
that hampers further applications of the SERS method for sensing. Development of
highly enhancing, but also stable, reproducible, and robust SERS-active substrates is,

therefore, very important for SERS to mature into an “off-the-shelf” detection method.

158

 

 

One approach to improve reproducibility of SERS is the immobilization of nanoparticles
on stable supports [23, 24]. Permeable supports are of particular interest as they may
allow for the added advantage of preconcentration. The idea of using ﬁlters as SERS

substrates has been approached previously from different perspectives by several research

 

groups [25-29]. The studies involved coverage of a permeable substrate - ﬁlter paper or

aluminum based membranes - with silver nanoparticles. The substrate preparation has

 

been based either on the ﬁltration of silver colloids through polymeric membranes [28] or
on the chemical reduction of silver directly on the ﬁlter surface [26, 27]. Since the
ﬁltration-based methods involved retention of silver particles in the membranes, the
ﬁltering ability of such membranes could be expected to be greatly inhibited. Recently,
the method of vacuum deposition of silver onto an aluminum membrane ﬁlter was
reported wherein the ﬁlter permeability could be expected to remain unchanged [29]. By
ﬁltering a small amount of analyte solution through the membrane, an improved

detection capability was demonstrated.

The objective of this work was to develop ﬂow-through SERS-active substrates by
assembling arrays of silver nanoparticles on the surface of track etch polycarbonate
membranes [30, 31]. Previously, nitrogen functionalities have been successfully
introduced to polycarbonate and other polymeric structures by using plasma treatment
[32-36] and by employing the reactivity of polycarbonate with amines [37, 38]. Our
approach was to employ 3-aminopropy1trimethoxysilane and cysteamine as chemical

linkers between polycarbonate surface and silver nanoparticles. SERS enhancing

159

properties of resulting membrane-supported arrays were tested and compared with those
of source nanoparticle suspensions using methylene blue (MB) as an analyte relevant for
pathogen detection and inactivation techniques [39, 40]. Finally, we considered
sensitivity in a broader context that included membrane-enabled preconcentration as the
factor that augments the inherently high sensitivity of nanoparticle-based SERS
substrates and studied the effect of preconcentration on the sensitivity of the developed

SCIISOI‘ S.

160

 

4.2. Experimental

4.2.1 Materials

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 =-
um capsule microﬁlter (PolyCap, Whatman Plc., Sanford, ME). The resistivity of water

was greater than 16 MQ-cm. Boil off nitrogen and 99.9999 % purity argon gases were

used in the plasma experiments.

4.2.2 Synthesis of silver nanoparticles

Aqueous suspensions of silver nanoparticles (Ag hydrosols) were prepared using the

procedure described by Lee and Meisel [41]. 90 mg of silver nitrate were added to 500
mL of ultrapure water and brought to 100 0C. 10 mL of 1 % by weight solution of

trisodium citrate were added dropwise to the boiling water under vigorous stirring. To
prepare sols with a narrower particle size distribution, sodium citrate solution was added
as quickly as possible and in the smallest droplets possible. Gradual color change of the
solution from transparent to greenish gray was indicative of the progress of silver particle
nucleation reaction. The erlenmeyer ﬂask containing the solution was loosely capped
with a ceramic crucible to avoid signiﬁcant solvent loss due to evaporation. After 40

minutes of boiling, the hydrosol was left to cool down. The loss of water to evaporation

161

was compensated by the addition of ultrapure water to restore the original volume of 500

mL. After allowing the hydrosol to cool down to room temperature, it was stored at 4 0C.

To precoagulate sol, 100 uL of silver sol were mixed with 100 “L of ultrapure water and
10 [.11 of a 0.5 mol/L NaClO4 aqueous solution. For SERS measurements in precoagulated
Ag hydrosols, 30 [IL of the precoagulated suspension were removed and 20 11L of the

analyte solution were added, yielding a ﬁnal 1:10 dilution of the sample in the

precoagulated sol. The resulting preparation was used for SERS measurements.

4.2.3 Membranes and membrane surface modification

25 mm diameter, 5 pm nominal pore size hydrophobic polyvinylpyrrolidone(PVP)-free
polycarbonate track etch (PCTE) membrane ﬁlters were purchased from Sterlitech
(Sterlitech Corp, Kent, WA). Based on the value of pore density (4105 pores/cmz)
reported by the manufacture the average distance between pores was estimated to be ca.

16 um.

Two approaches to membrane modiﬁcation were explored. The ﬁrst approach was to
treat membranes with nitrogen plasma to directly introduce nitrogen functionalities on the
membrane surface, or to treat membranes with argon plasma and then immerse them in

allylamine to introduce reactive radicals on the polycarbonate surface that would later

162

react with the unsaturated bond of allylamines. The second approach was the direct

interfacial reaction of polycarbonate with amine containing molecules, APTMS or CysA.
4.2.3.1 Nitrogen plasma and argon/allylamine plasma treatment

Bare (non—modiﬁed) membranes were subjected to pure nitrogen plasma treatment. A
plasma etching system (PX-250, March Instruments, Concord, CA), connected to a pure
nitrogen supply at 9 scfm constant ﬂow was used to generate the plasma. After 5 minutes
of plasma treatment, membranes were immersed in silver hydrosol for 24 hours. In a
separate set of experiments, bare membranes were subjected to pure argon plasma
treatment using the plasma etching system and ﬂow rate described above. After 5 minutes
of plasma treatment, membranes were immersed in a l % by weight allylamine aqueous
solution for 5 minutes, rinsed with ultrapure water for 1 hour and immersed in silver
hydrosol for 24 hours. After each treatment, membranes were separated from the silver
sol and rinsed with ultrapure water to eliminate excess hydrosol. The modified

membranes were kept in ultrapure water until further use.
4.2.3.2 APTMS used as a linker

It has been shown that closely packed arrays of gold and silver particles can be assembled
on a surface of a glass slide modiﬁed with APT MS [42]. Other studies have shown that
aminosilanes react with polycarbonate via the amino group [43]; the proposed reaction

pathway for the APTMS/polycarbonate reaction can be seen in Fig. 4.1. The membranes

163

were immersed in a 25 % by weight solution of APTMS in methanol for 5 minutes. After
incubation in APTMS, membranes were rinsed in pure methanol for 45 minutes with
three changes of the solvent. After the ﬁnal rinse, membranes were immersed in silver
hydrosol for 24 hours. Aggregation of silver suspension in the immediate vicinity of the
derivatized membrane initiated by residual physisorbed APTMS was indicative of poor
rinsing and such samples were discarded. The aggregation could be recognized by an
initial darkening of the suspension, a gradual increase in transparency, and appearance of
a lilac tint in the suspension. Prolonged rinsing in methanol was found to be necessary to

avoid these effects.

4.2.3.3 CysA used as a linker

In the case of CysA, the amino group would preferentially react with the polycarbonate
surface leaving an available thiol group for silver chemisorption. The membranes were
immersed in a 3.24 mol/L aqueous solution of CysA for 60 minutes. After CysA
incubation, membranes were rinsed in ultrapure water for 15 minutes with three changes
of the solvent. After the ﬁnal rinse, membranes were immersed in silver hydros'ol for 24

hours.

4.2.4 XPS, UV-vis, Raman and SERS measurements

XPS measurements of bare and APT MS-modiﬁed membranes were conducted using PHI

5400 ESCA system (Physical Electronics, Chanhassen, MN), equipped with a

164

Polycarbonate chain APTMS

Figure 4.1.Proposed mechanism for the APTMS reaction with polycarbonate and further
silver chemisorption to the incorporated nitrogen.

165

 

 

non-monochromatic Mg X-Ray. Measurements were done using a take off angle of 45

degrees. Raman and SERS spectra were recorded at a spectral resolution of 5 cm'1 and

coverage from 250 cm'1 to 4,000 cm'1 using Kaiser Optical Systems HoloProbe Raman

spectrograph coupled to an Olympus BX-60 optical microscope and equipped with a
motorized sample stage. A frequency doubled YAG laser emitting at 532 nm with a beam
power of 5 mW was employed as the excitation source. The laser spot diameter was less
than 2 um (with 100x objective). Exposure times (5 ms to 5 s) and the number of
accumulations (1 to 25) were adjusted in each measurement to avoid detector saturation.
Measured intensity values were normalized to account for variations in exposure times
and accumulations. To record Raman signal from membrane-based substrates, the
substrates were placed on the sample stage with the SERS-active surface facing the
microscope lens. To record Raman signal from liquid samples and hydrosols, the samples
were transferred to glass capillaries with ca. 2 mm internal diameter; the laser beam was

then focused on the center of the capillary containing the sample.

UV-vis absorption spectra of the silver hydrosol and treated membranes were recorded
using a photodiode array spectrophotometer (Multi-Spec 1501, Shimadzu, Kyoto, Japan).
Spectral coverage was from 180 nm to 800 nm with a spectral sampling interval of 1 nm
and with 100 accumulations per measurement. Absorption spectra of ultrapure water and
bare membranes were used as baselines in measurements of absorption of hydrosols and
treated membranes, correspondingly. Measurements were performed using square quartz
cells of 3 mL nominal capacity (Fisherbrand quartz, Fisher Scientiﬁc International,

Hampton, NH).

166

 

4.2.5 Particle size and zeta potential measurements

Particle size distribution of the prepared silver hydrosol was measured using multi-angle
light scattering apparatus (ZetaPALS, BI_MAS Option, Brookhaven Instrument Corp.,
Holtsville, NY ) equipped with a solid state laser emitting 15 mW beam at 660 nm.
Measurements were recorded in 10 cycles of 2 minutes each, with a 30 % cutoff dust
ﬁlter, using 3 mL sample cells provided by the equipment manufacturer. Samples were
diluted in a 10 mmol/L KCl solution to reach a measurement count rate in the range of 50
kcps to 500 kcps. Zeta potential was measured by phase analysis light scattering
(ZetaPALS, Brookhaven Instrument Corp., Holtsville, NY) using the same laser line as
the particle sizing module described above. Measurements were conducted in 10 runs of
10 cycles each, using 3 mL sample cells provided by the equipment manufacturer.

Samples were diluted in a 1 mmol/L KCl solution prior to measurements.

4.2.6 SEM and TEM imaging

SEM imaging was carried out using Hitachi S-47OOII Field Emission scanning electron
microscope operated in high vacuum, ultra high resolution mode. Samples were mounted
on aluminum SEM specimen stubs, membrane sample. surfaces were made conductive by
sputtering gold on them for 3 minutes at a current of 20 mA (Emscope SC 500 sputter
coater) prior to the SEM imaging. For higher resolution, smaller electron beam spot size
and shorter working distances were employed. Samples were also imaged by

transmission electron microscopy (JEOL JEM 2010, JEOL USA, Inc., Peabody, MA).

167

 

 

Copper (cat. no. 3HGC500, Electron Microscopy-Sciences, Hatﬁeld, PA) and formvar
carbon covered copper (cat. no. 200-CU) TEM grids were used to support the samples.
The grids were drop-coated by Ag hydrosol and examined at an accelerating voltage of

100 kV.
4.2.7 SERS characterization of APTMS modified membranes

To assess the Raman enhancing properties of membrane-based substrates, methylene blue

(MB) was used as a test compound. Aqueous solutions of MB with concentrations
ranging from 10'4 mol/L to 10'7 mol/L were tested on the modiﬁed membranes. 20 uL of

each solution were dropped on the surface of a modiﬁed membrane and let to dry (“drop
and dry” method). For each concentration, measurements were conducted on ﬁve

different membranes, on three different areas of each membrane’s surface.
4.2.8 Membranes filtration performance

Clean water ﬂux tests were conducted to determine intrinsic membrane resistance.
Amicon 8010 ultraﬁltration cell (Millipore Corp, Billerica, MA) was used as the
membrane module. Variable feed water ﬂow rate was achieved by varying the delivery
pressure of the compressed zero air tank. Ultrapure water was stored in a pressurized
container (100 psi, 5L, Alloy Products Corp., Waukesha, WI). Pressurized air from the
air tank pressure regulator (cat no. 8069621, Controls Corp. of America, VA. Beach, VA)

connected to the inlet of the water container was used to deliver feed water to the

168

‘

 

ﬁltration cell. The permeate ﬂux was measured continuously by collecting the ﬁltered
water on a digital weighing balance (AV8101C, Ohaus Corp., Pine Brook, NJ) interfaced
with a computer via a built-in RS-232 port. In all experiments, the feed water ﬂow rate
was sequentially increased by increasing the outlet pressure of the compressed zero air
tank from 0 psi to 28 psi. The data from the weighing balance was logged to a computer
using a program written in LabView (version 7.1, National Instruments). Data were
acquired at a 5 s time interval. Clean water ﬂux tests were conducted both for untreated

membranes and modiﬁed membranes.

4.2.9 Preconcentration studies

A variable ﬂow rate positive displacement pump (model XX80EL000, Millipore Corp,
Billerica, MA) was used to deliver feed aqueous solution of MB into the ﬁltration cell
(Amicon 8010, Millipore Corp, Billerica, MA) containing the modiﬁed membrane. The
permeate ﬂux was directed back into the feed solution container to allow for continuous
recirculation of the solution. Dilute solutions of the tested compound were recirculated
through the APTMS modiﬁed membranes for 2 hours. To evaluate the effect of ﬂow-
induced preconcentration, a modiﬁed membrane was immersed for 2 hours in a solution

of equal concentration to that used for the preconcentration experiments, and then let to

dry.

169

 

4.3. Results and Discussion

4.3.1 Silver nanoparticle hydrosol

Prepared hydrosols had suspended solids fraction of 3 x 10'4 by weight and ionic strength

of ca. 8.6 x 10"3 mol/L, as calculated based on the amount of citrate and nitrate salts used

in the preparation of the hydrosol. No aggregation was observed in the stored hydrosols
over the duration of experiments. Figure 4.2 shows SEM and TEM micrographs of the
silver nanoparticles. Two types of particles were observed — ones that were
approximately spherical and ones that were rod shaped, the former type being clearly

predominant.

Effective diameter of silver nanoparticles in the hydrosol, as measured using light
scattering, was found to be 58.8 um i 0.6 nm with a polydispersivity factor [44] of 0.319
t 0.004. As Fig. 4.3 illustrates, particle size distribution proﬁle of the sol shows two
major peaks at 13.76 nm and 78.86 nm and two minor peaks with effective diameters
higher than 500 nm. The ﬁrst two peaks can be attributed to spherical particles. The
peaks that correspond to larger particles together with the relatively high polydispersivity
of the measurement are likely to be due to the presence of rod-shaped particles that skew
the size distribution pattern to higher diameters. To estimate the length of rod shaped
particles, the following expression for the diffusion coefﬁcient of rod-shaped particles

can be used [45]:

170

 

 

 

 

 

 

 

 

Figure 4.2.Micrographs of Ag nanoparticles: SEM image of ﬁltered hydrosol (A), and
TEM image of Ag nanoparticles deposited onto TEM grids by drop-coating (B).

171

 

 

d = 78.86

 

 

 

0.00 *7 I
b ‘b '1» 1o "‘3 "a ’\ ‘o 01 c8 Q ‘3

O 1
1
100.001 1
1 1 z
1 G 80.00J 1
f E 1 d=13.76 ‘
j g 60.00‘
a) ‘ O L=3492
j fa} 40.00J \
z- I1 1 II we! »
i III_ I ,7 vV_ I -‘- ‘

Particle diameter (nm)

Figure 4.3.Size distribution proﬁle of silver nanoparticle suspension. Also shown are the
predominant spherical diameters for the two major distribution peaks, and the estimated
rod lengths for the two high end size peaks (Schematics not to scale)

172

 

 

D = k,
3n770L ’
where
-1 2 2 2L
7" = 1.27 - 7.4(6 — 0.34) , 7i = 0.19 —4.2(5‘1 —0.39) , 6 = 111(7)

and D is the particle diffusion coefﬁcient; k3 is the Boltzmann constant, T is the absolute
temperature, 770 is the viscosity of the suspending liquid, L is the rod length, and w is the

rod width. By comparing equation (1) and the expression for the diffusion coefﬁcient of

k 3T
3,27700: (2), a relationship between sphere diameter and rod

 

spherical particles, D

length for an equal diffusion coefﬁcient can be obtained:

Using the above analysis, the equivalent rod lengths were calculated to be ca. 2.9 pm and

3.5 pm.

Zeta potential of the silver sol, calculated using the Smoluchowski model, was found to

be -24.27 mV :1: 0.98 mV, which was indicative of a stable suspension.
4.3.2 Modified membranes characterization

4.3.2.1 Surface chemistry

173

To evaluate surface chemistry of membranes derivatized with APT MS, XPS studies were
conducted. Prior to treatment, XPS analysis showed the presence of carbon and oxygen
with peaks C(1s)[0.314] = 84.14 % and O(ls)[0.733] = 15.86 %, in accordance with the
chemical composition of a bare polycarbonate surface. After APTMS treatment, nitrogen

and silica peaks appeared as follows, C(ls)[0.314] = 77.70 %, N(1s)[0.499] = 1.80 %,

 

O(ls) [0.733] = 18.75 % and Si(2p)[0.368] = 1.75 % indicating the presence of APTMS

at the membrane surface.

 

4.3.2.2 Silver nanoparticle array morphology

Under the hydraulic conditions tested in our experiments, particles were not removed
from the membrane surface. However, under mechanical stress, such as tapping or
scratching, the particles could be easily removed; the modiﬁed membranes, therefore,

were handled using tweezers and were stored immersed in ultrapure water in a Petri dish.

The distribution of nanoparticles on the membrane surface was evaluated using SEM
imaging. Figure 4.4 A illustrates the ﬂat surface and narrow pore size distribution of a
bare PCTE membrane. Figure 4.4 (B, D) shows the membrane surface after modiﬁcation
by APTMS and silver. Homogeneous sub-monolayer coverage of the membrane surface
by silver nanoparticles, as well as some internal pore coverage can be observed. The
membrane surface coverage by silver particles was determined using image processing
(ImageJ v1.34s, National Institutes of Health, USA) of SEM micrographs of modiﬁed

membranes (Figures 4.4 B, 4.4 D). The surface coverage was calculated to be 56.6%.

174

 

 

Figure 4.4.SEM micrographs of PCTE membrane ﬁlter that is non-modiﬁed (A),
modiﬁed using APTMS linker (B, C, D), and modiﬁed using CysA linker (E, F).

175

(E, F) shows silver nanoparticle coverage of a CysA-modiﬁed membrane; in this case
surface aggregation is evident and the coverage is less homogeneous in comparison with
that of membranes modiﬁed using API‘ MS as a linker. Interestingly, those membranes
modiﬁed using CysA as a linker did not show Raman enhancement properties while

membranes modiﬁed using APTMS as a linker were SERS-active. This could be

 

attributed to the possibility of chemisorption of remnant cysteamine molecules onto the
silver nanoparticles during the silver immersion stage of membrane modiﬁcation. This

would hamper direct analyte-silver nanoparticle interaction and weaken SERS signal.

 

Plasma treatments with subsequent incubation of treated membranes in the silver
hydrosol did not result in any noticeable coverage of the membranes surface by silver

particles.

4. 3.2.3 Optical properties of modiﬁed membranes

 

Optical properties of nanoparticle arrays were studied with UV-vis absorption and Raman
spectroscopy. UV-vis analysis of the modiﬁed membranes was possible due to the semi-
translucent nature of the membranes. In the absorption spectrum of the Ag hydrosol, a
broad band at 410 nm is observed, which corresponds to the surface plasmon resonance
within primary particles. The same band, but blue shifted by (13.2 i 1.5) nm (95 %
conﬁdence) with respect to the silver sol absorption peak was observed for the silver
nanoparticles attached to the modiﬁed membranes (Fig. 4.5). The shift was attributed to
preferential exclusion of larger particles, including rod—shaped particles, from the

membrane surface, as indicated by SEM results. The fact that a band at larger

176

 

Absorbance

" Modiﬁed membrane

 

 

, Silver hydrosol

 

 

 

0 1 1 1 1 1 1 1 1 1 1 1
340 380 420 460 500 540 580 620 660 700 740 780

Wavebngthmm)

 

 

 

Figure 4.5.UV-vis spectra of a modiﬁed PCTE membrane (bare membrane used as
baseline), and of silver hydrosol (250 uL in 2 mL ultrapure water, ultrapure water used as
baseline).

177

 

wavelengths does not emerge in the absorption spectrum of the modiﬁed membrane
points to the absence of electromagnetic coupling between primary particles that is at the
absence of surface aggregation. This observation is corroborated by SEM imaging

results.

To evaluate Raman enhancing properties of modiﬁed membranes and compare them with
those of source hydrosols, MB was chosen as the test compound. Figure 4.6 shows UV
and Raman spectra of MB with relevant band assignments [46]. The excitation
wavelength used in this work was 532 am that is far from the resonance frequency of
MB. Much lower detection limits, and perhaps higher enhancements factors, could have
been observed had the resonant excitation been chosen. Nonetheless, this laser line is
suitable for the comparison of modiﬁed membranes and silver hydrosol with respect to

their Raman enhancing properties, which was one of the goals of this study.

First, Raman spectra of bare and modiﬁed membranes were recorded. In contrast to the
spectrum of the bare membrane, the spectrum of the modiﬁed membrane was found to be

relatively featureless (Fig. 4.7). Raman spectra of aqueous solutions of MB with
concentrations ranging from 10'7 mol/L to 10'2 mol/L were then measured. The two most
intense peaks on the Raman spectrum, 1401 cm'1 and 1626 cm'], correspond to the
symmetric CN stretch and CC ring stretch, respectively. A linear increase in Raman

intensity for both these peaks was observed for solutions in the 10'5 mol/L to 10'3 mol/L

178

 

 

 

 

 

 

 

 

 

 

 

0‘
10000 A v (CC) 3'
ring .‘2
=1 1
«i 1
>1 1
.. 1
8 -. 1
.5 (CN) <81
v _
é 1
06 11 \ K
11 611/ x. x
f‘ 141 N “K Inf 1) “J/Kﬂ‘w‘r/ “J AM!
400 800 1200 1600
Raman Shift, cm'1
0. 6 664 nm
0.)
U
1:
<6
{3
o
m
.0
<1:
340 500 660 780
Wavelength (nm)

 

Figure 4.6.Absorption (A) and Raman spectra (B) of methylene blue.

179

 

 

 

 

60000-1

\\
2.4-w) 1592.56

1 1372.99
\

 

/ \
~2- dfm
1 a.
C.
10000 -; 5:;
82‘ 1
1 1" F. \
:1- O f." V
g 3 \‘vv,/
1 88 1 = 1 .
1 cc 00 [I 1 [1‘ N/"\ b
1 g ‘3 N (‘3 [J \v j 1' ’ \l'.f‘\~'/‘r/
1 «‘53-. 3‘ .32 ‘ 1 / ' V
1 ° 9" "’
;\\ A f“ A t: ‘ LOVN/
‘f‘wj.\.., M “an/4 \‘4 \ )1 '7' Iw-uv‘rt"
1 ‘J T 1" ’\.
4000 -;
400 800 1 200 1 600

Raman Shift, cm'1

Figure 4.7.Raman spectra of an untreated membrane (bottom) and modiﬁed membrane
(top). Y-axis label: Raman Intensity (a.u).

180

 

range. To avoid saturation effects, Raman intensity recorded for 10‘4 mol/L solutions was

used in the computation of enhancement SERS enhancement factors.

To evaluate the Raman enhancement of assembled nanoparticle arrays, Raman spectra
from solutions of known concentrations of MB on the membrane surface were recorded
and compared to those obtained from the same solution concentrations on the
precoagulated silver hydrosol. Figure 4.8 shows the Raman spectra recorded for a 1

mmol/L solution of MB, and the enhanced peaks from silver sol and modiﬁed membrane
at 1626 cm']. A slight shift of the peaks can be observed, pointing to varying MB/silver

nanoparticles interactions for the different enhancing substrates. To calculate the
enhancement factors of the silver sol and the modiﬁed membranes, the following
equation was used:
4
= 10c, <1
1b Ca

where Ia and 1;, indicate the normalized Raman intensities for a given peak, Ia being the

 

EF

enhanced intensity (from the Ag hydrosol or the modiﬁed membranes) and 1;, the
comparative basis, at MB concentrations Ca and Cb. By comparing intensities of the 1626
cm‘1 peak in the SERS spectrum of a) 0.01 mmol/L solution of MB at the modiﬁed
membrane surface and b) 10'6 mol/L solution of MB enhanced by precoagulated $01 with

. -4 .
Raman Signal from 10 mol/L aqueous MB solution, average enhancement factors of

157.9 (t95%=14l.3) and 111.7 (t95%=11.8) were obtained for silver sol and modiﬁed

membrane, respectively, considering all possible intensity combinations between

181

 

 

 

enhanced peaks and source MB peaks. While the EF of the sol is slightly higher than that

of the modiﬁed membrane, so is the standardized error, indicating that the modiﬁed

182

 

10000.1 1622.31

 

1395 1

1398.45

=
Cd
23
2
2
.s \1
S [1, /\ .
E , 9’ \
ea 1
0‘ ,AM/‘J \\
1 \“

I .

1 - "'1. ‘\.-_,/
5000 : P/,.,r/‘kdf1\ﬂ.ww “\xu—v/‘M /\/\UM-/‘v—.\\/\\—w /"\"f\ _1/—/ \/ \_~_‘__;
*VI 1 —‘

 

 

1

l w"!
I’M“

1 r
:1 /\z./‘1"MJ
, f‘ A/

‘ We‘-

15‘ A V'\ _ Aj‘ "Mag/x4]
1 1} HWY/f“, “’A' \JUU'V‘” 1f ’

1
400 800 1200 1600

Raman Shift, cm'1

Figure 4.8.Raman spectrum of a 1 mmol/L aqueous solution of MB (bottom), a 10'5

mol/L aqueous solution of MB in pre-coagulated silver sol (middle), and a 10'5 mol/L
aqueous solution of methylene blue after “drop and dry” on a modiﬁed membrane (top).

183

 

 

membranes have better SERS reproducibility than the sols. The observed EF values are
summarized in the Fig. 4.9. The main body (box) of the plot encloses the middle 50%
range of the EF measurements (and excludes lower and upper quartiles of the data set)
while the line that crosses the box corresponds to the EF data median. Whiskers indicate
the sample maximum and minimum, and open circles correspond to outliers. As can be
seen from Fig. 4.9; the spread in EF values signiﬁcantly higher for the precoagulated sol
than for sols, indicative of lower reproducibility. A narrower and more symmetric
distribution can be observed for the modiﬁed membranes; also, the median value of EF

from modiﬁed membranes was higher than that for the hydrosol.

4. 3.2.4 Filtration performance of modiﬁed membranes

To study possible changes in the ﬁltration performance of membranes after modiﬁcation,

clean water flux tests were conducted both for untreated membranes and nanoparticle-

modiﬁed membranes (Fig. 4.10). Membrane resistance was calculated to be (1.09-107 :

0.01) m.1 and (1.15107 i 0.02) m'1 for bare and modiﬁed membranes, respectively. As

Fig. 4.10 illustrates, the flux versus pressure trendline was drawn for the ﬁrst three points
only, using linear least square ﬁtting. The subsequent points were not considered since
they did not follow the linear trend, likely due to membrane pore constriction at higher
transmembrane pressure differentials. The above measurements clearly indicate that the
particle attachment to membrane surface did not result in any signiﬁcant changes in the

membrane resistance.

184

 

'L

 

 

 

.95 85582: BEBE
mo coasts—8:895 Baa Ea A3 85.582: 3:608 98 _8 3333805 no m2 mo ﬁcﬁﬂ “5885::0 8m 32a xomdé unsur—

UOuwbﬂooﬂoooum

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

8:89:82 Em
33a . 1 1111111 1 o
a . _
1 1 not: A W .
. ., . _T. .52 m
.1. . . . . mots.— m if _
_ w H
4 not“? w, w . .. .. V, 8N
, H _ ,. w w H
w, 2x3 5 _ , P
f $251 2 W
k 1 ,7 , ‘_ i .84
.4 . 13?? z .
. . _
a m . z < a
r--. 1 -1 1:1 -Lmo_o6_ V ., 1. 1 i 11 111 seen

 

70 a

> y = 3.6509x
R2 = 0.9958 “

   
 

   

y = 3.3972x [:1 Bare Membrane
2 _ a
R — 0.9846 0 Modif. Membrane

Flux (mL/cm2/min)

 

 

O 5 10 15 20 25 30

Pressure (psi)

 

Figure 4.10.Results of clean water ﬂux test for bare and modiﬁed membranes.

 

186

4.3.2.5 Preconcentration

We considered sensitivity in a broader context that included membrane-enabled
preconcentration as the factor that augments the inherently high sensitivity of

nanoparticle-based SERS substrates. To evaluate the effect of preconcentration on the
sensitivity of the developed sensors, an aqueous solution of known concentration (10'7

mol/L) of MB was ﬁltered through the membrane at a constant volume flow of 4 mL/sec
for 2 hours. The membrane was let to dry and the SERS spectrum of MB at the
membrane surface was then recorded. To ensure that any increased enhancement was

solely due to preconcentration, parallel measurements were conducted both for modiﬁed

membranes immersed in a 10'7 mol/L MB(aq) solution for 2 hours, and modiﬁed

membranes inoculated with a 10'7 mol/L MB (aq) solution using “drop and dry” method.

While no MB peaks where observed in the SERS spectrum from the modiﬁed

membranes after “drop and dry” or 2 hours immersion, preconcentration resulted in the
appearance of the 1390 cm'1 peak (Fig. 4.11). For all tested membranes, the
preconcentration experiments have reproducibly resulted in an improved detection limit.
Interestingly, the MB peak at 1627 cm'1 did not appear after preconcentration, instead,
the 1401 cm'1 peak was shifted 10 cm'1 with respect to its original position. Shifts could

also be noticed at the precoagulated sol and modiﬁed membranes; the shifts in SERS are
indicative of the disturbance of the electronic conﬁguration of an analyte due to

chemisorption. We attribute the difference in SERS spectra form MB chemisorbed by the

187

 

“drop and dry” method and chemisorbed during preconcentration experiments to the

different molecule-surface interactions. Convective forces present during

 

12000—
1390.3

Raman Intensity, a.u.

 

 

 

400 800 1200 1600

Raman Shift, cm']

Figure 4.11.Raman spectrum of a 10'7 mol/L aqueous solution of MB after 2 hours of
pre-concentration on a modiﬁed membrane.

188

 

preconcentration together with the natural hydrophobicity of the membrane substrates
might favor MB chemisorption to the silver nanoparticles through the CN bond, rather
than through the less polar CC bond, which could explain the predominance of the CN

peak.

Based on the intensity of the 1401 cm'1 band in the Raman spectrum of a 10'2 mol/L MB

aqueous solution, the enhancement factor (EF) was calculated using (4) to be 1.17 ~ 105
(t95% = 1.92 - 104); the data distribution can be observed in Figure 4.9 B. This result

shows that preconcentration is effective in increasing the enhancement provided by the

membrane substrate and that the detection limit can be signiﬁcantly lowered (since by
“drop and dry” or prolonged immersion no detection of MB 10'7 mol/L was possible).

Additionally, the spread in the measurements (with respect to the EF value) is one order
of magnitude lower than the averaged EF; for precoagulated silver sols the the relative

width of the spread is considerably larger.

The effect of convective ﬂow across the membrane on the improved MB adsorption to
nanoparticles immobilized on the membrane surface can be rationalized in terms of
diffusive and convective transport of analyte molecules. The thickness of the boundary
layer formed by a fluid in motion in the vicinity of a ﬂat plate can be roughly estimated

as [47]:

6:

,

V, (5)
I

189

where 6 is the boundary layer thickness, v is the kinematic viscosity of the ﬂuid, 1 is the
plate length, and u is the velocity of the ﬂuid far from the surface. In our

experiments, u = j / A , where j is the volumetric ﬂowrate and A is the membrane area.
Given that u =O.8-1041 m/s and taking 1 to be the average distance between pores (ca. 16

pm), the thickness of the boundary layer can be estimated from (5) to be ca. 45 um. The
time it takes for a molecule to be transported by Brownian diffusion across the boundary

layer is then given by:

(5)2 (6)
= 20 ’

 

ta

where D is the diffusion coefﬁcient of the solute. By estimating the molar volume of MB
by the LeBas method [48] and using Hayduk-Laudie correlation [49] applicable to small

(MW<1000 Da) non-electrolyte solutes to estimate the diffusion coefﬁcient of MB, we

obtain D 22.96 ~ 10'10 mz/s. Now, using (6) Id is estimated to be ca 3.4 ms, which is of

the same order of magnitude as time tc z [la :2 ms that the analyte molecule spends in

the vicinity of the membrane. This means that by imposing the convective flow, the
boundary layer is compressed sufﬁciently in our experiments, for the MB molecules to be
able to diffuse across the boundary layer to the upstream membrane surface as they are
transported by the convective flow past the membrane surface and into the membrane

pore.

This balance between convective and diffusive transport can also be viewed in terms of
the probability of a Solute molecule to reach (and chemisorb to) the silver coated

membrane surface. Equation (6) can be rewritten as:

190

20 (7)
_ d ,

“d

where ud can be interpreted as the velocity of a solute that travels a distance d due to

Brownian motion. A “pseudo” Peclet number can be obtained as the ratio of convective

to diffusive velocities:

u ud (8)

u d 2D
As can be seen from equation (8), this ratio will be directly proportional to the convective

velocity and the distance from the membrane surface. For a ﬁxed u, for all distances
2D . . . . . .
d > — the ratio (8) Will be higher than one, implylng that more molecules W111 be
u

driven to the membrane per unit time when convection is present, than when only
diffusion occurs, thereby enhancing mass transport of the analyte to the membrane

surface.

191

 

 

4.4. Conclusions

Silver nanoparticle arrays were assembled on track etch membrane supports and
established to be SERS-active with reproducible Raman enhancements. The improved

reproducibility was attributed to the well deﬁned geometry of track etch membranes. The

“L

hydrodynamic control of analyte transport to the permeable SERS-active surface was
demonstrated to allow for a dramatic improvement of the detection limit of the sensor.
Better insight into processes that govern analyte partitioning from dissolved phase onto
the surface of the permeable substrate is needed; boundary layer structure, membrane
morphology, and surface chemistry of membrane and SERS-active sites are likely to
determine the extent of preconcentration. The reported ﬁndings indicate the potential
beneﬁt of combining high speciﬁcation SERS-active systems and the ﬂow-through

design for the development of analytical sensors for the trace detection of pollutants in

 

water.

192

Acknowledgements

This material is based upon work supported by the National Science Foundation under
Grant No. OISE-0530174. Authors thank Eliza Tsui (Rice University) for recording TEM
images of silver hydrosol particles and Dr. Je’réme Rose (Centre Europe’en de Recherche
et d'Enseignement des Géosciences de l'Environnement) for assisting with high resolution
SEM imaging of ﬁltered Ag hydrosol particles. Helpful comments by two anonymous
reviewers are gratefully acknowledged. Any opinions, ﬁndings, and conclusions or
recommendations expressed in this material are those of the authors and do not

necessarily reﬂect the views of the National Science Foundation.

193

 

 

References

1. Water quality monitoring: A practical guide to the design and implementation of
fresh water studies and monitoring programmes, ed. J. Bartram and R. Ballance. 1996:
UNEP/WHO.

2. S. A. Hashsham, L. M. Wick, J. M. Rouillard, E. Gulari and J. M. Tiedje,
Potential of DNA microarrays for developing parallel detection tools (PDTs) for

microorganisms relevant to biodefense and related research needs, Biosens. Bioelectron.
20 (2004) 668-683.

3. MD. Sobsey. Methods to identify and detect microbial contaminants in drinking
water. in Strategies for Drinking Water Contaminants. 1999. Washington, DC:
Commission on Geosciences, Environment and Resources.

4. J. Juliano and MD. Sobsey. Simultaneous concentration of Cryptosporidium,
bacteria and viruses from water by hollow ﬁber ultraﬁltration. in Water quality
technology. 1997. Denver, CO.: American Water Works Association.

5. H. A. Morales-Morales, G. Vidal, J. Olszewski, C. M. Rock, D. Dasgupta, K. H.
Oshima and G. B. Smith, Optimization of a reusable hollow-ﬁber ultraﬁlter for

simultaneous concentration of enteric bacteria, protozoa, and viruses from water, Appl.
Environ. Microbiol. 69 (2003) 4098-4102.

6. W. Quintero-Betancourt, E. R. Peele and J. B. Rose, Cryptosporidium parvum and
Cyclospora cayetanensis: a review of laboratory methods for detection of these
waterborne parasites, J. Microbiol. Methods 49 (2002) 209-224.

7. T. M. Straub and D. P. Chandler, Towards a uniﬁed system for detecting
waterborne pathogens, J. Microbiol. Methods 53 (2003) 185-197.

8. M. R. Wiesner and S. Chellam, The promise of membrane technologies, Environ.
Sci. Technol. 33 (1999) 360.

9. M. G. Albrecht and J. A. Creighton, Anomalously intense Raman-spectra of
pyridine at a silver electrode, J. Am. Chem. Soc. 99 (1977) 5215-5217.

10. A. Campion and P. Kambhampati, Surface-enhanced Raman scattering, Chem.
Soc. Rev. 27 (1998) 241-250.

11. R. K. Chang and T. E. Furtak, Surface Enhanced Raman Scattering. 1982, New
York: Plenum Press.

12. D. L. Jeanmaire and R. P. Vanduyne, Surface Raman spectroelectrochemistry.1.
Heterocyclic, aromatic, and aliphatic-amines adsorbed on anodized silver electrode, J.
Electroanal. Chem. 84 (1977) 1-20.

194

13. M. Moscovits, Surface—enhanced spectroscopy, Rev. Modern Phys. 57 (1985)
783-826.

14. T. Vo-Dinh, SERS chemical sensors and biosensors: New tools for environmental
and biological analysis, Sens. Actuators, B 29 (1995) 183-189.

15. K. Kneipp, Y. Wang, H. Kneipp, I. Itzkan, R. R. Dasari and M. S. Feld,
Population pumping of excited vibrational states by spontaneous surface-enhanced
Raman scattering, Phys. Rev. Lett. 76 (1996) 2444-2447.

16. N. Weissenbacher, B. Lend], J. Frank, H. D. Wanzenbock and R. Kellner, Surface
enhanced Raman spectroscopy as a molecular speciﬁc detection system in aqueous ﬂow-
through systems, Analyst 123 (1998) 1057—1060.

17. P. A. Mosier-Boss and S. H. Lieberman, Detection of volatile organic compounds
using surface enhanced Raman spectroscopy substrates mounted on a thermoelectric
cooler, Anal. Chim. Acta 488 (2003) 15-23.

18. K.I. Mullen, L.G. Crane, D.X. Wang and K.T. Carron, Trace detection of ionic
species with surface enhanced raman spectroscopy, Spectroscopy 7 (1992) 24-32.

19. S. Sanchez-Cortes, C. Domingo, J. V. Garcia-Ramos and J. A. Aznarez, Surface-
enhanced vibrational study (SEIR and SERS) of dithiocarbamate pesticides on gold ﬁlms,
Langmuir 17(2001) 1157-1162.

20. T. Vo-Dinh, Surface-enhanced Raman spectroscopy using metallic
nanostructures, Trends Anal. Chem. 17 (1998) 557-582.

21. N. Weissenbacher, B. Lendl, J. Frank, H. D. Wanzenbock, B. Mizaikoff and R.
Kellner, Continuous surface enhanced Raman spectroscopy for the detection of trace
organic pollutants in aqueous systems, J. Mol. Struct. 410 (1997) 539-542.

22. K. Kneipp, H. Kneipp, R. Manoharan, E. B. Hanlon, I. Itzkan, R. R. Dasari and
M. S. Feld, Extremely large enhancement factors in surface-enhanced Raman scattering
for molecules on colloidal gold clusters, Appl. Spectrosc. 52 (1998) 1493-1497.

23. Z. X. Luo and Y. Fang, SERS of C-60/C-70 on gold-coated ﬁlter paper or ﬁlter
ﬁlm inﬂuenced by the gold thickness, J. Colloid Interface Sci. 283 (2005) 459-463.

24. M. Volkan, D. L. Stokes and T. Vo-Dinh, A sol-gel derived AgCl photochromic
coating on glass for SERS chemical sensor application, Sens. Actuators, B - Chem. 106
(2005) 660-667.

25. Y. Kurokawa and Y. Imai, Surface-enhanced Raman-scattering (SERS) using

polymer (cellulose-acetate and Naﬁon) membranes impregnated with ﬁne silver particles,
Journal of Membrane Science 55 (1991) 227-233.

195

26. J. J. Lasema, A. D. Campiglia and J. D. Winefordner, Surface-enhanced Raman
spectrometry on a silver-coated ﬁlter-paper substrate, Anal. Chim. Acta 208 (1988) 21-
30.

27. A. S. L. Lee and Y. S. Li, Surface-enhanced Raman-spectra using silver-coated
paper substrates, J. Raman Spectrosc. 25 (1994) 209-214.

28. M. Muniz-Miranda and M. Innocenti, AFM and micro-Raman investigation on
ﬁlters coated with silver colloidal nanoparticles, Appl. Surf. Sci. 226 (2004) 125-130.

29. R. J. Walsh and G. Chumanov, Silver coated porous alumina as a new substrate
for surface-enhanced Raman scattering, Appl. Spectrosc. 55 (2001) 1695-1700.

30. P. Apel, Track etching technique in membrane technology, Radiat. Meas. 34
(2001) 559-566.

31. K. P. Han, W. D. Xu, A. Ruiz, P. Ruchhoeft and S. Chellam, Fabrication and
characterization of polymeric microﬁltration membranes using aperture array
lithography, Journal of Membrane Science 249 (2005) 193-206.

32. M. Bryjak, I. Gancarz, G. Pozniak and W. Tylus, Modiﬁcation of polysulfone
membranes 4. Ammonia plasma treatment, Eur. Polym. J. 38 (2002) 717-726.

33. I. Gancarz, G. Pozniak and M. Bryjak, Modiﬁcation of polysulfone membranes 3.
Effect of nitrogen plasma, Eur. Polym. J. 36 (2000) 1563-1569.

34. C. J ama, O. Dessaux, P. Goudmand, B. Mutel, L. Gengembre, B. Drevillon, S.
Vallon and J. Grimblot, Surface modiﬁcations of polycarbonate (PC) and polyethylene
terephtalate (PET) by cold remote nitrogen plasma (CRNP), Surf. Sci. 352 (1996) 490-
494.

35. K. R. Kull, M. L. Steen and E. R. Fisher, Surface modiﬁcation with nitrogen-
containing plasmas to produce hydrophilic, low-fouling membranes, Journal of
Membrane Science 246 (2005) 203-215.

36. K. Schroder, A. Meyer-Plath, D. Keller, W. Besch, G. Babucke and A. Oh],
Plasma-induced surface functionalization of polymeric biomaterials in ammonia plasma,
Contrib. Plasma Phys. 41 (2001) 562-572.

37. V. S. Foldi and T. W. Campbell, Preparation of copoly(carbonate-urethanes) from
polycarbonates, J. Polym. Sci. 56 (1962) 1-9.

38. H. Ishida and Y. H. Lee, Study of exchange reaction in polycarbonate-modiﬁed
polybenzoxazine via model compound, J. Appl. Polym. Sci. 83 (2002) 1848-1855.

39. S. E. Murinda, L. T. Nguyen, S. J. Ivey, B. E. Gillespie, R. A. Almeida, F. A.
Draughon and S. P. Oliver, Molecular characterization of Salmonella spp. isolated from
bulk tank milk and cull dairy cow fecal samples, J. Food Prot. 65 (2002) 1100-1105.

196

40. J. P. R. Pelletier, S. Transue and E. L. Snyder, Pathogen inactivation techniques,
Best Pract. Res. Clin. Haematol. 19 (2006) 205-242.

41. P. C. Lee and D. Meisel, Adsorption and surface-enhanced Raman of dyes on
silver and gold sols, J. Phys. Chem. 86 (1982) 3391-3395.

42. K. C. Grabar, R. G. Freeman, M. B. Hommer and M. J. Natan, Preparation and
characterization of Au colloid monolayers, Anal. Chem. 67 (1995) 735-743.

43. G. L. Wilkes and C. Li, The mechanism for 3-Aminopropyltriethoxysilane to
strengthen the interface of polycarbonate substrates with hybrid organic-inorganic sol-gel
coatings, J. Inorg. Organomet. Polym. 7 (1997).

44. J. C. Brown, P. N. Pusey and R. Dietz, Photon Correlation Study Of Polydisperse
Samples Of Polystyrene In Cyclohexane, Journal Of Chemical Physics 62 (1975) 1136—
1144.

45. S. R. Mishra, H. S. Rawat, S. C. Mehendale, K. C. Rustagi, A. K. Sood, Ranjini
Bandyopadhyay, A. Govindaraj and C. N. R. Rao, Optical limiting in single-walled
carbon nanotube suspensions, Chemical Physics Letters 317 (2000) 510. ‘

46. R. R. Naujok, R. V. Duevel and R. M. Corn, Fluorescence and Fourier-transform
surface-enhanced Raman-scattering measurements of Methylene-Blue adsorbed onto a
sulfur-modiﬁed gold electrode, Langmuir 9 (1993) 1771-1774.

47. H. Schlichting, Boundary Layer Theory. 1968, New York: McGraw Hill.

48. LeBas, The molecular volumes of liquid chemical compounds, ed. Longmans.
1915, London.

49. W. Hayduk and H. Laudie, Prediction of diffusion-coefﬁcients for nonelectrolytes
in dilute aqueous-solutions, AIChE J. 20 (1974) 611-615.

197