You are on page 1of 41

MEMBRANE CHARACTERIZATION

Roy Bernstein
KU Leuven, Heverlee, Belgium

Yair Kaufman
Ben Gurion University of the Negev, Zuckerberg Institute for Water Research (ZIWR), Sde-Boqer,
Israel

Viatcheslav Freger
Technion–Israel Institute of Technology Haifa, Israel

1 OVERVIEW

Optimal functioning of separation membranes is usually achieved through a combination


of many factors, such as membrane chemistry, physical properties, surface characteristics,
morphology, etc. Some of these characteristics are fully determined by the manufactur-
ing process and optimized to meet a limited number of performance-related specification
characteristics, for example, flux, rejection and molecular weight cutoff (MWCO). Nev-
ertheless, this information often needs to be supplemented with other a priori unknown
data in order to address additional aspects such as fouling, cleaning, and modeling. This
requires various characterization methods that have been a crucial part in membrane
research, development, and engineering.
Given that most membranes possess an asymmetric structure, often comprising several
distinct layers, the number of independent parameters may become large. This requires a
range of characterization techniques varying in specificity to various chemical moieties
and a spacial resolution spanning many orders of magnitude. Membranes characterization
methods may be roughly subdivided into several main groups based on the type of
information they provide. The main differentiation is between chemical (structure and
composition) and physical characteristics (morphology, mechanical properties, charge,
etc.) of the layers. In addition, in many cases, differentiation between bulk and surface
characteristics and between porous and dense structure is necessary because distinct
methods may be required in each case. This article largely follows this classification.
Given the small thickness of the membranes, especially of the selective layer, techniques

Encyclopedia of Membrane Science and Technology. Edited by Eric M.V. Hoek and Volodymyr V. Tarabara.
Copyright © 2013 John Wiley & Sons, Inc.

1
2 MEMBRANE CHARACTERIZATION

offering highest spacial resolution and minimal probing depth are of special interest to
membrane characterization and are given emphasis.

2 CHEMICAL CHARACTERISTICS

Membranes may utilize a wide variety of chemical building blocks, both organic and inor-
ganic. Polymers are likely to remain in the near future the primary materials of choice for
most synthetic membranes. Nevertheless, alternative materials, especially, ceramics, zeo-
lites, metals and carbon nanomaterials, metal-organic frameworks (MOFs), biomaterials,
as well as various composites including these materials are extensively developed and
their performance, stability, and cost are steadily improving. Chemical analyses or appro-
priate spectroscopic tools are widely used to verify the presence of desired constituents
and examine their local content and distribution within the membrane.
Usually, the nominal chemical structure and composition indicated by the manufac-
turers do not uniquely determine the chemical characteristics of the membrane. Polymers
of the same nominal chemistry may widely differ in homogeneity, content of terminal
groups, cross-link density, branching, chain packing, etc. Further differences may develop
as a result of conditioning or after modification. Membrane chemistry can undergo fur-
ther reversible and irreversible changes as a result of exposure to various feeds, fouling,
or cleaning. Such variations are of significant interest to membrane manufacturers and
users. To obtain a comprehensive picture at different spacial scale, it is often beneficial
to combine information obtained by several complementary techniques.
A number of techniques are presently available; they may significantly differ in cost,
resolution, sensitivity, dimensions of the probed region and specificity to different atoms,
oxidation states, or chemical moieties. The most popular techniques are reviewed here,
information on other microanalysis techniques emerging in membrane research such
as electron energy loss spectroscopy (EELS) or X-ray microscopy (1), may be found
elsewhere (2).

2.1 Methods Based on Backscattered Radiation and Particles


This group of methods requires a high energy beam of electromagnetic radiation or parti-
cles such as electrons or ions. When the beam impinges on a specimen surface, it interacts
with the atoms in the specimen and generates various types of scattered or secondary
radiation, which can be both particles and electromagnetic waves (1). Backscattered
radiation and particles may be used for identifying, quantifying, and mapping elements
present in the specimen.
Energy-dispersive X-ray spectroscopy (EDS or EDX) usually comes as an integral
part of most scanning electron microscopes (SEMs) or transmission electron microscopes
(TEMs) (Section 3.1) (2). It uses the secondary X-rays produced by the interaction of
high energy electrons with the atoms in the specimen (Fig. 1a). A focused electron beam
hitting the specimen usually produces an excited onion-shaped region just under the
specimen surface that is about a micrometer large. Each element in this region generates
a distinct energy spectrum of characteristic X-rays, with intensity directly proportional to
the atomic content. Modern instruments are capable of detecting elements with atomic
number down to 3 (boron) but the intensity of the atomic band increases with the atomic
number; therefore, EDS is far more sensitive toward heavier elements. Using well-defined
intensity factors, local atomic compositions may be determined and mapped onto SEM
MEMBRANE CHARACTERIZATION 3

(a) Primary (b)


Secondary beam
emission region ~10 nm Secondary
electrons
Characteristic
X-rays Backscattered
electrons

Interaction
volume
~1 µm
Sample

(c)
cps/eV
2.0

1.8
Si
1.6 C O

1.4
AI
1.2

1.0

0.8

0.6

0.4

0.2

0.0
0.5 1.0 1.5 2.0 2.5
KeV

FIGURE 1 (a) Interaction volume of an electron beam with a sample and types of backscattered
radiation. (b) SEM images of a fouled surface of a reverse osmosis (RO) membrane obtained using
backscattered electrons detector (left) and secondary electron detector (right); lighter regions on
the left are indicative of heavier atoms of inorganic deposits, also revealed by EDS, and deposits
in right bottom part of the image are predominantly organic. (c) EDS spectra obtained for different
deposits on fouled RO membrane surface. The three spectra correspond to inorganic alumosilicate
particles (top spectrum), organic deposits (middle) and a thin layer of silica with some clay (bottom)
on polyamide membrane surface. (Please refer to the online version for the color representation of
the figure.)
4 MEMBRANE CHARACTERIZATION

or TEM images of the same region for any relevant elements. The lateral resolution of
elemental maps is usually limited to about a micrometer, that is, the size of the excited
onion-shaped region. This lateral resolution size for films of submicrometer thickness,
such as the ones used in TEM, may be better, 50–100 nm. In membrane research, EDX
has been particularly useful in combination with SEM imaging for examining chemically
distinct regions within composite membranes and on the surface of fouled membranes.
Figure 1b and c shows an example of fouled membranes examined using SEM–EDS
containing various types of organic and inorganic deposits on the surface.
X-ray photoelectron spectroscopy (XPS), also known as electron spectroscopy for
chemical analysis (ESCA), records the energy spectrum of photoelectrons escaping from
the surface region of a sample hit by a focused monochromatic beam of X-rays (using Al
Kα X-rays of an energy 1.5 keV) (1). The incident beam is usually tilted, and escaping
electrons are detected at a certain take-off angle that can be varied. The region, from
which electrons can escape to reach the detector, usually Si, is much thinner than in
EDS, about 1–10 nm.
The energy of photoelectrons is uniquely related to their binding energy within the
atoms, for example, 1s, 2s, and 2p. The intensity of binding energy bands of a specific
element within the energy spectrum is proportional to the atomic content in the probed
region. A remarkable feature of XPS is that the bands in high resolution spectra exhibit
chemical shifts that allow distinguishing of different oxidation states of elements (e.g.,
C in –CH2− , –CF2− , –COOR, and –COH groups). This capability and extremely small
probing depth make XPS uniquely suitable for chemical analysis of surfaces. In principle,
XPS may also be used for depth profiling by varying the penetration depth through
varying the take-off angle, but, more often, depth profiling is carried out by combining
XPS with Ar sputtering that etches away the surface material in a layer-by-layer manner.
This method applied to polymers is, however, prone to artifacts, as species from the
upper layers readily contaminate the underlying layers during sputtering.
The small probing depth of XPS makes it particularly suitable for analyzing modified
surfaces. Examples include plasma-treated or oxidized surfaces, self-assembled mono-
layers, polymer brushes or grafted polymer layers, and thin coating layers.
Rutherford backscattering spectroscopy (RBS) uses the elastic scattering of a beam
of ions (typically, 2 MeV He ions from an accelerator) to probe the composition and
structure of materials (3, 4). A Si surface-barrier device is used to detect backscattered
ions with a typical energy resolution of 20 keV. The scattering cross sections of atoms
are known with great accuracy; therefore, RBS is a quantitative analytical technique
and does not require the use of standards. Depth resolution is enabled by the fact that
an energetic ion loses energy continuously as the ion traverses a solid, the “stopping-
power” of polymers is on the order of 150 eV/nm. Typically, the specimen is tilted at
some angle to trade-off between depth resolution and uncertainties created by surface
roughness. Data analysis of layered structure (e.g., a composite membrane) is performed
using modeling software, which yields the number of atoms per unit area. Knowledge of
the atomic density is needed to convert this real density to thickness.
RBS has been successfully applied to determine characteristics of the top layer of
several types of commercial reverse osmosis (RO) and nanofiltration (NF) membranes
(4, 5). As the support layer contains S atoms and the top layer does not, the shape and
position of the S edge of the RBS spectrum is particularly sensitive to the thickness
and thickness variations of the polyamide layer. Furthermore, Coronell and coworkers
recently used labeling by Ba+ and I− ions to quantify the density and ionization behavior
MEMBRANE CHARACTERIZATION 5

of positive and negative functional groups in RO and NF membranes (6) as well as ion
partitioning (7) using RBS. As in EDS and XPS, the sensitivity is greatly enhanced with
the use of high Z atoms. By comparing atomic composition obtained by RBS and XPS
having very different probing depths, it was also possible to assess nonuniformity of ion
distribution within the top layer (8).

2.2 Vibrational Spectroscopies


This group of methods utilizes inherent resonant vibrations within molecules. These
vibrational modes may be activated, (subject to the so-called selection rules), by elec-
tromagnetic waves of certain frequencies, resulting in frequency-dependent adsorption
of radiation, that is, the absorption spectrum. Such spectra, or specific bands, are often
characteristic of certain bonds or chemical substructures, which may be used for iden-
tifying and quantifying the content of particular components in a mixture or chemical
groups within a complex structure. The most common of these techniques are infrared
(IR) spectroscopy and Raman spectroscopy (RS). One may add to this group UV and
nuclear magnetic resonance (NMR) spectroscopies, sensitive, respectively, to conjugated
π-electron structure and magnetic interaction of atomic nuclei within molecules (9).
Unfortunately, these widely used chemical analysis methods are bulk techniques and have
no surface-sensitive versions; thus, they have found so far a limited use in membrane
characterization. A notable exception is pulsed field gradient (PFG)-NMR, an important
method for measuring self-diffusion of solutes in membranes (10) (Section 4.1).
IR spectroscopy utilizes transitions between vibrational states associated with
chemical bonds as well as some collective vibrations, for example, within aromatic
rings, polymers, or crystals. The selection rule requires that the vibration result
in a change in dipole moment so that certain vibrational modes within symmetric
molecules (e.g., CO2 ) do not show up in IR spectra. The region of largest utility for
chemical identification is middle infrared (MIR) between about 200 and 4000 cm−1 (or
wavelengths 2.5 and 50 μm). Modern Fourier transform infrared spectroscopy (FTIR)
instruments record an interferogram, which is processed using a fast Fourier transform
(FFT) to yield the complete IR spectrum (11).
As the attenuation length of IR does not exceed a few micrometers in most materials,
including water and all polymers, surface-sensitive reflection-based FTIR techniques,
in particular, attenuated total reflection (ATR) has been the most popular in membrane
research. In ATR–FTIR, the sample, a film or solution, is brought in tight contact with
a high refractive-index crystal. A totally reflected IR beam travels inside the crystal;
however, the so-called evanescent radiation penetrates and may be absorbed within a
layer adjacent to the crystal (Fig. 2a). The short penetration depth dp of evanescent
wave (dp ∼1 μm) makes ATR–FTIR a surface-sensitive method (12).
Owing to the specificity of IR bands, ATR–FTIR is highly suitable for examining
chemical changes in the top layer, for example, as a result of chlorination, irradiation
or hydrolysis, or for comparing different membranes and analyzing the thickness and
chemistry of the top layer (4, 15). Figure 2a (bottom) presents typical ATR–FTIR spec-
tra of RO polyamide membranes, original composite membrane and one modified with
hydroxyethylmethacrylate (HEMA) at different monomer concentrations. The increase
in modification is seen by the increase in the intensity of the carbonyl band of acry-
lates relative to a characteristic band of the supporting membrane (13). The method is
also widely used for identification and semiquantitative analysis of extra layers attached
6 MEMBRANE CHARACTERIZATION

(a) (b) Inlet Outlet

Support membrane

Solution
Evanescent Grafting layer
Film
wave

Incident IR ATR crystal ATR crystal


beam Reflected
IR beam

ESPA-1
20 mM 20 bar 50 mM 20 bar
0.05 75 mM 0 bar 100 mM 20 bar
1.2
Normalized intenstiy
Supporting
0.04 membrane 1
Absorbance

Carbonyl 1586 cm−1 OH bend 0.8


0.03
1724 cm−1
0.6
0.02
0.4
0.01 0.2
0 0
1750 1700 1650 1600 1550 1500 1450 0 2 4 6 8 10
Wavenumber (cm−1) Time (min)

FIGURE 2 (a) Top: An ATR–FTIR-based setup typically used to acquire spectra of the active
part of composite membranes including active layer and, possibly, an added layer for surface-
modified or fouled membranes; bottom: a typical spectrum of a pristine RO membrane (ESPA-1
by Hydranautics) and several ESPA-1 membranes modified by grafting an acrylic monomer. The
band at 1720 cm−1 is indicative of grafted acrylate layer, whose thickness is much lower than
the penetration depth (dp ) (13). Source: Adapted with permission from Reference 13. Copyright
(2010) American Chemical Society. (b) Top: A setup used to measure partitioning and diffusion
of a solute between a solution and a thin membrane film; bottom: effect of in-plane oriented clay
flakes on water diffusion in a poly(vinyl alcohol)-clay nanocomposite quantified by ATR–FTIR
(14). Source: Reprinted from Reference 14, Copyright (2012), with permission from Elsevier.
The maximum (plateau) intensity corresponds to equilibrium partitioning and effective diffusion
coefficient in the direction normal to the films may be deduced from the transient similar to the
“time-lag” method. (Please refer to the online version for the color representation of the figure.)

to the membranes surface, for example, after surface modification by polymer graft-
ing or coating or as a result of fouling or adsorption. Relative thickness of thin layers
(dp ) may be estimated by comparing the intensities of characteristic bands of the
deposited material and support or, for thicker layers (≥dp ), by attenuation of the bands
of the support. The dependence of dp on the refractive index of the crystal (e.g., Ge
vs ZnSe) and on the angle of incidence enables depth profiling through use of different
crystals (12).
ATR–FTIR uniquely combines the depth resolution and chemical sensitivity with the
ability to work in a liquid environment. This feature of ATR–FTIR may be employed
for measuring polymer swelling as well as partitioning and diffusivity of solutes in
thin polymer films (Fig. 2b), for example, in the top layer of composite membranes.
MEMBRANE CHARACTERIZATION 7

Partitioning of different organics was derived from variations in the intensity of solute
bands for an ATR crystal with an attached free polyamide film in contact with the
solution (16, 17).
RS is based on inelastic scattering of photons activated by the interaction of light
with molecular or lattice (phonon) vibrations (11). If interaction occurs, Raman spectral
lines appear characteristically shifted in frequency relative to the elastically scattered
(Rayleigh) line. Detection of the Raman lines is possible only for a strictly monochromatic
light source; therefore, a typical Raman setup usually includes a laser and a confocal
microscope that focuses the beam on a particular microvolume in three dimensions.
The appearance of Raman lines does not depend on the laser wavelength; however,
a particular wavelength may be preferred in order to minimize fluorescence that may
formidably interfere with the Raman signal for polymeric and biological samples.
RS may be viewed as a technique complementary to IR. The selection rules for Raman
absorption requires that, rather than changing the dipole moment, vibrations cause a
change in the polarizability of the vibrating group of atoms. Thus, symmetric modes,
inactive in IR, become visible in Raman; on the other hand, certain antisymmetric modes
are inactive in Raman but active in IR. This is beneficial in some cases: water strongly
absorbs IR in the important finger-print region but it is transparent for Raman, which
makes RS particularly useful for biological samples in water, for example, biofilms
developed on membranes (18, 19). Another important and rapidly developing field is
carbon-based nanomaterials, such as nanotubes, fullerenes, and graphene that possess a
highly symmetric structure and thus featureless IR spectra, but show distinct Raman bands
useful for differentiation between different allotropic forms (20). As Raman spectrometers
become more affordable and more advanced nano- and biomaterials enter the membrane
research, a wider use of RS may be foreseen in the future.

3 PHYSICAL CHARACTERISTICS AND MORPHOLOGY

Most membrane manufacturing techniques, such as phase inversion, coating, or interfacial


polymerization, have evolved in a largely empirical way and are often too complex
to model. For instance, the composition of the casting solution and precipitation bath
in the phase inversion process may strongly affect porosity, tortuosity, and pore size
distribution of porous films and lead to a variety of structures and features, for example,
macrovoids or sponges. Even if the specifications such as permeability, pore size, or
MWCO are identical, significant variations of morphology, microstructure, mechanical
properties, etc. are often a rule rather than exception. Microscopic examination may
then greatly assist developing optimal membranes and manufacturing procedures for a
particular application.
Physical characteristics of membranes may also change, reversibly or irreversibly,
when membranes age, for example, as a result of conditioning, fouling, cleaning, exposure
to chemicals, disintegration, or loss of certain membrane components. In such cases,
physical and morphological examination provides insights into mechanisms that lead to
loss or restoration of performance. Similar methods are employed to quantify the effect of
modification (e.g., surface grafting) in studies of modified membranes. The need to focus
on the topmost selective part of the membrane and on surface-related phenomena gives
the advantage to modern high resolution surface specific techniques that increasingly
play an important role in such studies.
8 MEMBRANE CHARACTERIZATION

3.1 Electron Microscopy


Electron microscopy (EM) employs electrons scattered from a specimen hit with a finely
focused electron beam (21). Most EM instruments require a high vacuum, and for this
reason the specimen should usually be completely dry or deeply frozen (cryo-EM). The
beam causes local heating and charging that may be reduced using a few nanometer
thick conductive coating (e.g., Au), cryotechniques or lower beam energy, usually, at
the expense of resolution. The electrons penetrate into the specimen producing an onion-
shaped region of excitation (Fig. 1a). A variety of signals is produced from this zone
(secondary electrons, backscattered electrons, Auger electrons, characteristic X-rays, and
cathode luminescence) and may be collected with suitable detectors (1). For instance,
characteristic X-rays allow surface microanalysis (EDS), which is an important and useful
feature of EM methods (Section 2.1).
In SEM, the beam scans across the specimen surface. SEM images usually combine
the signals by elastically backscattered electrons and inelastically scattered secondary
electrons. The large depth of focus allows imaging of wide areas and large, irregular-
shaped objects. SEM microscopes routinely operate between the magnification range
of up to 100,000, with a resolution limit of about 0.1 μm for regular instruments and
2.5 nm for the most advanced high resolution instruments employing a field-emission
gun (field-emission scanning electron microscopy, FESEM). Using separate detectors for
backscattered electrons that penetrate more deeply and are more sensitive to composition
and secondary electrons, more sensitive to surface topography, compositional and topo-
graphic images of the same area may be produced. Figure 1b and c shows an example of
such images along with EDS of a fouled RO membrane that allows an easy differentiation
between organic and inorganic deposits.
In environmental scanning electron microscope (ESEM), the specimen chamber is
separated from the electron optical system by a system of apertures. This allows a working
pressure of up to 0.19 Pa and thus gentler specimen pretreatment and fewer artifacts for
soft hydrated samples such as biofilms. The ESEM work is however generally more
difficult and less straightforward than in regular high vacuum instruments (2).
As opposed to SEM, transmission electron microscopy (TEM) employs electrons scat-
tered mainly in the beam direction, that is, after passing the specimen (2, 21). This
imposes a critical limitation on the specimen thickness, which should not significantly
exceed 100 nm. Direct examination is then only possible for nano-objects, for example,
nanoparticles or nanotubes. Otherwise, bulk materials or films embedded in a cross-
linked resin have to be sliced using an ultramicrotome to prepare thin cross sections that
can be mounted on a standard TEM grid for examination. In addition, contrast between
regions of interest requires differences in electron density, which are usually too small
within organic materials and need to be enhanced using staining, that is, incorporation of
heavy elements such as Os or U into selected regions. The numerous limitations of TEM
are, however, offset by a significantly higher resolution, which may reach subatomic
dimensions for inorganic materials and a few nanometers for polymers. For example,
TEM was used to reveal the nanostructure of the top layer of a composite NF membrane
modified with a grafted layer of negatively charged acrylic acid that was stained with
uranyl cations for enhanced contrast (22).
MEMBRANE CHARACTERIZATION 9

3.2 Scanning Probe Microscopy (SPM)


Scanning probe microscopy (SPM) is a group of methods that employs various interac-
tions (e.g., mechanical or electromagnetic) between the surface and a sharp tip, mounted
on a flexible cantilever, to examine topography and map various physical characteristics
of the material making up the surface (21). The characteristics depend on the specific
method and include viscoelastic, optical and magnetic properties, conductivity, surface
potential, chemical forces, etc. Atomic force microscopy (AFM) is the most popular
version of SPM, in which the tip mechanically interacts with the surface (23). A piezo-
electric scanner is used to move the tip relative to the sample in three dimensions with
high resolution. The force is monitored by measuring the cantilever deflection using a
laser beam focused on the free end of the cantilever and detecting the position of the
reflected beam with a position-sensitive photodiode. In this way, one may either measure
how the force varies with distance from the surface (in this case, a larger object, e.g., a
solid sphere or even a living cell may replace the sharp tip (24)) or scan the topography
using a feedback loop that maintains a constant force between tip and sample. The lateral
resolution is in general lower than vertical, but may be nearly as high as in EM, being
largest on smooth hard surfaces but decreasing with the tip curvature and on softer and
rougher surfaces.
AFM may perform topographic scans in several modes. In the contact mode, a fixed
repulsive force between tip and sample is maintained; and in the tapping mode, the
cantilever oscillates at near-resonant frequency and the oscillation amplitude is maintained
somewhat below the value corresponding to free oscillations. The contact mode yields
somewhat more robust topography scans and tip-surface friction-based images, but the
tapping mode is usually less damaging and, in addition, yields the so-called phase images,
indicative of local mechanical dissipation, that is, softness or hardness of the material.
A great advantage of AFM is that surfaces and forces may be examined under liquid,
for example, in conditions in which the membrane actually works. This feature as well
as the relative ease of operation and moderate cost makes AFM one of the most popular
methods for surface examination in membrane science.

4 TRANSPORT CHARACTERISTICS

Transport properties are key characteristics of any membrane, commercial or state of


the art. For commercial membranes, limited data reported by the manufacturer may
be used to compare different product lines, yet, in real practice, membrane research
and development, and process design, more elaborate data are often required. Certain
issues such as concentration polarization need thorough attention (25) but, otherwise,
the performance measurements are straightforward and not addressed here. However, for
many purposes, especially for studies focused on separation mechanisms or modeling,
other independently measured characteristics are required along with regular performance
measurements. They are briefly reviewed here.

4.1 Transport and Diffusion of Neutral Solutes


If the membrane thickness is known, regular performance tests may yield the permeance
of the membrane material (P = measured permeability × thickness). It is however of fun-
damental interest that P may be split into diffusivity (mobility) and sorption (partitioning)
10 MEMBRANE CHARACTERIZATION

factors D and S (P = D × S ). Complementary measurements are then required and may


be accomplished in three principal ways.
First, equilibrium solute sorption can be measured to yield S and diffusivity may be
evaluated as D = P /S . In cases when the sample is too small or too thin for conventional
techniques such as dilatometry, gravimetry, titration, and extraction, the measurements
may be accomplished using modern microanalytical methods (4).
Second, diffusivity of the solute can be measured independently and sorption cal-
culated as S = P/D. Diffusivity may be measured using tracer-based methods such as
PFG-NMR (26) or radiolabeled solutes (e.g., tritiated water) (27). PFG-NMR, also known
as pulse-gradient spin-echo (PGSE)-NMR, is among the most accurate methods to mea-
sure self-diffusion of molecules in solutions and polymers (28). In this method, the spins
of specific atoms in the sample are made to change orientation using a sequence of highly
synchronized radio frequency and magnetic field-gradient pulses that produce mutually
canceling changes of orientation and position-dependent rotational phase of the spins.
The synchronized cancelation of orientation and phase changes results in the so-called
“spin echo” signal that would be of maximal amplitude if the atom remained in the
same position throughout the sequence. However, as a result of diffusion of the atoms,
the synchronization of the phase changes produced by radio-frequency pulses is partly
lost during field-gradient pulses and the spin-echo signal is attenuated. The attenuation
is exponentially related to the time between field-gradient pulses  and spin diffusivity
D. The field-gradient pulses may then be viewed as a “labeling” of atoms or molecules
and by measuring spin-echo attenuation versus  the diffusion coefficient may be accu-
rately determined (10). Having all three characteristics, P, D, and S , available from
independent measurements, for example, permeation, sorption, and PFG-NMR, allows
a consistency test that may supply useful insights (29). Some care is required as the
radio tracer and NMR methods produce self-diffusion coefficient D ∗ that is related to
the Fickian diffusivity D through the following relation (27)
 
d ln a d ln γ
D = D∗ = D∗ 1 + (1)
d ln C d ln C
where C , a, and γ are the concentration, activity, and activity coefficient of the solute,
respectively. It is seen that D ∗ and D are identical only in an ideal solution and, in a
general case, knowledge of sorption isotherm C (a) or activity coefficient γ = a/C is
required to connect the two diffusivities.
A third approach to evaluation of S and D, not requiring permeance data, is to
determine them simultaneously using time-resolved transient sorption or diffusion mea-
surements for a sample of well-defined geometry, for example, a film or a spherical
particle. For materials available as macroscopic bulk samples, the classical realization
of this approach is the time-lag method (30), whereby a sample is exposed to a steplike
change of the environmental conditions (solution or vapor) and sorption is monitored as
a variation of the sample weight or a dilation or disappearance of the solute from the
environment. The time-lag method requires that the diffusion time τ ∼d 2/D, where d is
the characteristic size of the sample (e.g., thickness of a film or radius of a particle),
exceeds the minimal timescale of the experiment; otherwise, the transient will be unob-
servable. Sample handling and sorption measurements for thin films and dilute solutes
may be facilitated by using time-resolved spectroscopic methods, such as ATR–FTIR for
in situ monitoring of concentration changes (31). The obtained dependence, normalized
to equilibrium sorption, is fitted to a theoretical model, D being a fitting parameter.
MEMBRANE CHARACTERIZATION 11

In a slight variation of this approach, D and S may be as well deduced from the
transient rate of diffusion rather than transient concentrations. For instance, in elec-
trochemical experiments with reversibly oxidizable and reducible species the transient
diffusion flux is measurable as a faradaic (electron-transfer) current. Using a setup con-
sisting of a film attached to a solid electrode and immersed in solution, the concentration
gradient across the film may be varied stepwise and the diffusion rate monitored as a tran-
sient current (chronoamperometry). A similar setup may be used to measure D and S of
species-forming redox couples using electrochemical impedance spectroscopy (EIS). EIS
is conceptually similar to time-resolved experiments, but it relies on a small sinusoidal
perturbation, rather than a steplike one, and analyzes the response in the frequency rather
than time domain. Despite the fact that the method is strictly limited to redox couples, the
remarkable advantage of EIS is that τ may be as short as 1 μs (frequency up to 1 MHz).
Thus, EIS suits very thin film, <1 μm, for example, those making the top layer of com-
posite membranes. Another notable feature is that EIS may examine and separate many
simultaneously occurring charge transfer phenomena, faradaic and non-faradaic (32–34).

4.2 Transport of Charged Species and Salts


Charged solutes, that is, ions, fundamentally differ from neutral ones in that their trans-
port is strongly coupled through the requirement of electroneutrality, which results in
development of a priori unknown gradient of potential across the membrane. As a result,
permeation or filtration experiments alone only yield permeability of a salt ωs related to
permeabilities of cation and anion ω+ and ω− as follows (35)

(z+ + z− )ω+ ω−
ωs = (2)
z+ ω+ + z− ω−

where z+ and z− are respective ion charges. To evaluate ω+ and ω− , the value of ωs has
to be supplemented with some electrical measurement, which may be of the following
types:

1. Transmembrane potential , which may result from salt gradient (diffusion potential)
or solvent flow across the membrane [streaming potential (SP)]. The former is
usually measured by placing the membrane between solutions with different salt
concentrations C1 and C2 , which results in an electrical potential difference (36)
ω− − ω+ RT C1 t − t+ RT C1
ϕ12 = ϕ1 − ϕ2 = ln = − ln (3)
ω− + ω+ F C2 t− + t+ F C2

where t+ and t− are respective transport numbers. An identical relation holds for
the SP if convection contributes negligibly to salt permeation, as is the case for
many modern thin-film composites. If convection contributes substantially, both
Equations 2 and 3 will be more elaborate and less useful, as they will contain
more a priori unknown parameters.
2. Electrical conductivity of the membrane per unit area (G/A), which actually sums
up the conductivities consisting of anions and cations. The latter are related to ionic
permeabilities through Einstein-like relation, which yields (32)

G F 2 Cs
= (ω− + ω+ ) (4)
A RT
12 MEMBRANE CHARACTERIZATION

where Cs is the salt concentration of solution, in which the conductivity is mea-


sured. Conductivity G or resistance R = 1/G are usually measured in AC con-
ditions, using a membrane or a stack of membranes pressed between flat solid
electrodes or separating two electrolyte solutions. EIS is the most suitable tech-
nique for such measurements, which clearly distinguishes membrane resistance
from other AC responses.

Equations 2–4 are easily generalized to the case of salts containing multivalent and
polyelectrolyte ions. Combination of Equation 2 with either Equation 3 or 4 will yield
ω+ and ω− .
These relations assume steady-state conditions; thus, only permeabilities may be
obtained (ωs , ω+ , and ω− ). However, the permeabilities of salts or ions may be split
to D and S factors (37). This may be done, similar to neutral solutes, by measuring var-
ious transient, that is, time- or frequency-resolved transport measurements, for samples
of defined geometry (e.g., see Fig. 2b).
Such schemes may be applied in a straightforward way to thick membranes, for
example, those used in electromembrane applications (Section 5.4). Unfortunately, it
is usually inapplicable to the selective top layer in thin-film composites, such as the
ones found in commercial NF or RO membranes, which are very thin (15–100 nm) and
unavailable as macroscopic thick films. The major impediment in this case is the large and
a priori unknown resistance of the support layer, which is 3–4 orders of magnitude thicker
than the top layer. One proposed remedy is the use of non-steady-state measurements
employing very short times and hence small spatial scales. This was shown to work
in a few cases, for example, for relatively resistant inorganic NF membranes; however,
for commercial polymeric composites the required time scale is exceedingly short (38).
Another promising approach, eliminating interferences by the support, is to separate the
top layer and transfer it to a solid electrode whenever possible. Such a supported top
layer is amenable to examination by EIS, which yields its resistance along with a few
other transport and physical characteristics (39). Figure 3a displays representative EIS
results for polyamide films isolated from an RO membrane; the spectrum clearly shows
several interfering contributions that may be separated from the film resistance (Fig. 3b)
by using appropriate models, usually presented as equivalent circuits (inset in Fig. 3a).

5 BULK CHARACTERISTICS AND POROSITY

In the context of bulk characterization, all membranes may be classified as either porous
or dense. In porous membranes, the pores are voids permanently formed within a dense
solid matrix, so that pore dimensions significantly exceed the characteristic scale of
molecular dimensions or structural fluctuations. Such membranes are used in microfil-
tration (MF) and ultrafiltration (UF), as well as hemodialysis and membrane distillation.
While exposed to the feed solution, the dense matrix undergoes little change and the
transport mainly occurs within the pores. The fixed morphology and negligible trans-
port within the matrix differentiates porous membranes from dense ones, in which the
transport occurs directly in a homogeneous dense material, for example, a polymer or an
inorganic solid, through dynamic pores of molecular dimensions.
Porosity may be defined as the fraction of the void space. If the void space of a
porous membrane is completely filled with a liquid until a constant weight is obtained
MEMBRANE CHARACTERIZATION 13

(a) 1.000 Mohm


Cm

Cdl Rs Cdl

100.0 kohm
Rm

Rm
Cm
10.00 kohm
Rs
Zmod

1.000 kohm
C Rm Cm
(M) (kΩ) (μF)
0.001 9.5 0.13
0.01 2.2 0.12
100.0 ohm
0.1 0.58 0.09
0.5 0.15 0.13
1 0.093 0.17

10.00 ohm
100.0 mHz 1.000 Hz 10.00 Hz 100.0 Hz 1.000 kHz 10.00 kHz 100.0 kHz
Freq (Hz)

(b) 1.E+04

1.E+03
Rm (Ohm)

1.E+02
ESPA1
SWC1
1.E+01
IP

1.E+00
0.001 0.01 0.1 1
KCl concentration (M)

FIGURE 3 (a) Bode plots of absolute value of impedance for a polyamide film isolated from an
RO composite membrane and attached to a glassy carbon electrode in solutions of KCl of different
concentrations (C = 10−3 to 1 M). The middle region in the impedance spectra yields membrane
electrical resistance Rm . Note the capacitive contributions Cdl and Cm marked by the two dashed
lines would preclude determination of Rm anywhere outside the middle region, emphasizing the
importance of recording the whole spectra. Insets show the equivalent circuit and the values of
Rm and Cm for different salt concentrations. (b) Dependence of Rm on C ; the dependence levels
off as C decreases, indicative of a dominant contribution by the fixed charge at low C . Source:
Reprinted from Reference 39, Copyright (2007), with permission from Elsevier.

and excess liquid is carefully removed from the external surface, fractional porosity is
determined from the weight change as follows:

Ww − Wd
Porosity = (5)
V ×ρ
14 MEMBRANE CHARACTERIZATION

where Ww and Wd are the weights of the membrane in the wet and dry states, ρw the
density of the liquid, and V is the volume of the membrane in the wet state.
In the context of pore transport, the pore size is essentially the effective radius of bottle-
necks, sometimes referred to as pore “throats” (Fig. 4). It is common to consider macro-
pores (r > 50 μm), mesopores (2 μm ≤ r ≤ 50 μm) and micropores (r < 2 μm) (40).
Some types of pores, such as “dead” pores (i.e., totally disconnected from the active
transport routs) and “dead-end” pores (i.e., connected to active pores only at one end),
do not contribute to transport but they do contribute to porosity. These types of pores
may be differentiated from one another by measuring porosity using different methods
(see subsequent text).
Pores within a membrane are rarely uniform; therefore, the pore size distribution needs
to be addressed. The size of large pores may be directly measured using microscopic
imaging, such as optical microscopy (OM) (macro- and mesopores) and EM (42), as
well as by a range of indirect methods. Many of the indirect techniques are relatively
old but remain in use because of simplicity and good accuracy, and today may be run
automatically on commercial instruments. The most common methods are summarized
in Table 1 and reviewed in subsequent text.
The last common parameter pertinent to transport through porous media is tortuosity.
It is defined as the factor by which the observed transport rate would increase if the pores
were perfectly straight for the same porosity and pore size. Typical values of tortuosity
for porous membrane are in the range of 1.5–5 (43). Even though tortuosity and porosity
are in general independent parameters, simulations of transport through random porous

Pore mouth

Pore throat

FIGURE 4 Schematic of a pore tunnel. Source: Reprinted from Reference 41, Copyright (2010),
with permission from Elsevier.

TABLE 1 Common Methods for Assessing Membranes Porosity, Pore Size, and Pore Size
Distribution

Transport/Flow Techniques Equilibrium Techniques Advanced Techniques

Solute permeation (MWCO) Gas adsorption–desorption PALS


Bubble point with gas transport Thermometry SAXS and SANS
Gas permeability Permporometry TEM and SEM
Liquid–liquid displacement Bubble point RBS
Mercury intrusion porosimetry Nanoparticles
MEMBRANE CHARACTERIZATION 15

media and percolation theory (44) show that a “master” relation between tortuosity and
porosity may hold for a wide class of anisotropic randomly porous media and may be
used in modeling (e.g., Reference 45).

5.1 Porosity and Pore Size of Membrane for Pressure-driven Separations


Solute permeation stands somewhat alone among the indirect methods listed in Table 1
because it uses transport of solutes along with a solvent rather than a pure fluid. MWCO
is nominally defined for UF, NF, and dialysis membranes as the molecular weight (MW)
of a neutral solute that is rejected to 90%. This is a widely accepted method used by
the membrane manufacturers as well. Macromolecules such as dextran or poly(ethylene
glycol) (PEG) are most common probe solutes. Rejection of several probe molecules
within different MW is measured in standard filtration experiments (46) and the rejection
is plotted versus MW on a log scale (Fig. 5). MWCO is determined by interpolating the
dependence to 90% rejection or, equivalently, sieving coefficient 0.1.
To minimize the influence of solute–solute interactions, a low solute concentration
(typically 200–1000 mg/l) is used in rejection experiments. Such experiments often
employ a multicomponent solution (a range of MW) and feed and permeate compo-
sitions are determined in a single run using gel permeation chromatography (GPC). The
concentration polarization must be either calculated or minimized. One should also pre-
vent fouling and, especially, pore blocking by molecules larger than the MWCO of the
membrane. The pore radius of the membrane can be estimated by using a model of pore
transport and an appropriate correlation between the hydrodynamic radius and MW of
the specific type of solute (e.g., Reference 47 for PEGs). Pore size distribution can be
similarly estimated using the solute transport results (48).
The other indirect methods for the estimation of the average pore size and size distri-
bution mainly utilize either capillary equilibrium of two fluids (e.g., a gas and a liquid)
in a pore or pore flow (Table 1). A major difference between the two groups is that
equilibrium methods count in the “dead-end pores,” while the pore flow does not. The
capillary equilibrium methods rely on the relation of the capillary (Laplace) pressure to

1
Observed sieving coefficient, so

0.1

0.01

PLCGC
PBHK

0.001
1000 10,000 100,000 1,000,000
Dextran molecular weight (Da)

FIGURE 5 Dextran sieving coefficients for two membranes: PBHK (100 kDa polyethersulfone)
and PLCGC (10 kDa regenerated cellulose). Source: Reprinted from Reference 46, Copyright
(2007), with permission from Elsevier.
16 MEMBRANE CHARACTERIZATION

the pore radius


2γ cos θ
rp = (6)
P
where γ is the surface tension coefficient of the liquid, θ the contact angle of the liquid
or liquid–liquid interface on the pore wall, and P the critical pressure required to open
or fill the pore. The pore flow methods measure transmembrane flow of a fluid that may
be analyzed using the Hagen–Poiseuille equation:

π rp4 N P
Q = (7)
8μL

where Q is the change in the volume flow rate following a change in the transmem-
brane pressure (TMP) along the pore P , N the number of pores per unit area, μ the
fluid viscosity, and L is the pore length. The common assumptions are that all pores are
cylindrical, straight, parallel to each other, perpendicular to the membrane surface, and
of the same length equal to the membrane thickness. These assumptions obviously over-
simplify the pore structure. However, if used consistently, the results may well represent
the differences between membranes. A more adequate representation of the complex pore
geometry, for example, as a random composite, may only require a minor adjustment
using appropriate pre-factors in Equations 6 or 7 (45). For most methods, measurements
can be conducted today automatically using commercial instruments.
Bubble point was the first method for measuring a membrane pore size (49). A gas
is forced through the prewetted membrane at increasing pressure until it overcomes the
capillary pressure (the breakthrough point) and the gas bubble is recorded. This method
measures the Laplace pressure required to just start displacing the liquid in the pores. The
pore size at the breakthrough pressure is calculated using Equation 6 assuming θ = 0. In
order to reduce the Laplace pressure, a liquid with a low γ is usually used. The obtained
pore size corresponds to the largest pore (lowest Laplace pressure).
Bubble point with gas transport (50) is a development of the bubble point in order to
evaluate pore size distribution as well. After the breakthrough pressure is detected, the
pressure continues to increase in incremental steps and the flow rate of gas, considered as
a nonwetting fluid (θ = 180◦ ), is recorded at each pressure after it reaches the steady state.
Using the Hagen–Poiseuille equation, the number of pores of the corresponding pore
size at the corresponding pressure is calculated. Because observation of the breakthrough
point is delayed by the time required for the gas to fully displace pore liquid, the method
usually overestimates the smallest pore size compared to porosimetry (see below).
Liquid–liquid displacement (LLD), also known as bi-liquid permporometry, was intro-
duced in 1933 by Erbe and further developed by Capannelli et al. (51). This method is
similar to the bubble point with gas transport but the penetrating gas is replaced with a
second liquid. The two liquids should be immiscible and have a small difference in their
surface tensions (e.g., distilled water/water-saturated isobutanol or distilled water/15 :
7 : 25 v/v water–isobutanol–methanol mixture (52). The membrane is saturated under
vacuum with the wetting solvent. Then, the second solvent is pressurized in incremen-
tal steps across the membrane and its flow is recorded versus pressure (Fig. 6). The
advantages are that dead-end pores are excluded, the membrane is characterized in a wet
state, and the required pressure is usually low. The main drawback is swelling of the
membrane. In addition, the correct contact angle between the membrane and liquids is
sometimes difficult to obtain.
MEMBRANE CHARACTERIZATION 17

Liquid flux
J2

J1

P1 P2
Pressure

FIGURE 6 A typical dependence of flow on applied pressure in the gas flow and liquid–liquid
displacement tests: an applied pressure (P ) causes the corresponding liquid flow (J ). The dashed
line gives the dependency of the fluid flow on the pressure for the membrane with pores com-
pletely filled with displacing fluid. Source: Reprinted from Reference 53, Copyright (1997), with
permission from Elsevier.

The mercury intrusion porosimetry (MIP) (54) is based on capillary flow of mercury,
a nonreactive nonwetting fluid in most materials, which is forced through a prevacuumed
dry membrane. The pressure is elevated in increment steps and the pressure required for
the mercury to overcome the capillary pressure and intrude the pores is recorded versus
the volume fraction of the mercury intruding the pores. Then, using Equation 6, pore
size distribution is evaluated. The method enables calculating pore size in the range 5 nm
to 200 μm. Supplementary information can be obtained through a subsequent extraction
experiment by reducing the pressure. The drawbacks of this method are that it is per-
formed in dry state and the membrane can undergo compaction; thus, the pores might
change or collapse under high pressure. Moreover, it also measures the dead-end pores
irrelevant in membrane filtration (41).
Permporometry was developed by Eyraud et al. (55) for gas–liquid permporometry
and expanded to gas–vapor permporometry by Mey-Marom and Katz (56). It is based
on capillary condensation of a fluid and on the assumption that vapor pressure inside the
pores is governed by the pore radius, as described by the Kelvin–Laplace equation:
 
RT P 2γ cos θ
ln = (8)
Vm ps rp

where P is the vapor pressure, Ps the saturated vapor pressure, rp the pore radius, γ the
surface tension, Vm the molar volume, and θ is the contact angle. A fully wetted mem-
brane is brought in equilibrium with a known mixture of gas (e.g., nitrogen) and vapor of
a condensable wetting liquid (e.g., alcohols such as methanol, hydrocarbons such as hex-
ane, mineral oil, or special liquids provided by the instrument manufacturers). The pores
in which the capillary pressure is lower than the vapor pressure will be evacuated and the
gas flow can be measured, while pores having vapor pressure below the vapor pressure
will remain filled with a liquid. The vapor pressure is decreased stepwise until all pores
are evacuated. The measurement can be carried out by starting with a dry membrane.
18 MEMBRANE CHARACTERIZATION

Then, at low pressure only adsorption of vapor molecules occurs—at-layer with a thick-
ness of a few molecules. As the vapor pressure rises, the liquid condenses and blocks an
increasing fraction of pores, progressively restricting the gas flow. In order to shorten the
time of equilibration, the experiment can be carried out in an atmosphere of reduced total
pressure. It is important to evaluate the t-layer thickness in order to get an accurate result.
Gas adsorption desorption is used for determination of pore size and pore size dis-
tribution as well as the surface area using the Brunauer–Emmett–Teller (BET) method,
mainly for inorganic membranes but also for organic ones (57). This method is similar to
permporometry, only the liquid is replaced with a condensable gas, usually, nitrogen at
its boiling temperature at 1 atm. The result is fitted to the BET isotherm, which yields the
total number of nitrogen molecules filling the primary monolayer and, using the known
area per nitrogen molecule, the total pore area. The pore radius is then determined from
the total pore area and pore volume. In addition, using the adsorption–desorption hys-
teresis, the pore geometry and the uniformity of pore sizes can be evaluated. Unlike
permporometry, this method measures the “dead pores” as well.
Gas permeability measurements are based on gas transport through dry open pores of
a porous membrane. Gas permeability through a membrane is measured using Darcy’s
law at a prescribed TMP drop. The permeability, that is, the slope of the flux versus
TMP, is related to the (average) pore size using the standard relation of pure Knudsen
or Poiseuille flow (58).
Thermometry is based on the decrease of the freezing or, more often, melting point of a
liquid (usually water) with the pore size recorded using a differential scanning calorimetry
(DSC) (59). A membrane soaked in water is placed in DCS and the temperature is
ramped at a constant rate while the heat flow is measured at different segments. The pore
size distribution is then deduced from the DSC thermogram (Fig. 7) using appropriate
thermodynamic relations, analogous to Equation 8 (60).

1 kDa

5 kDa
Heat flow (W/s)

Pore

50 kDa

Bulk

Temperature ( °C)

FIGURE 7 Thermoporometric heating curves (0.1 ◦ C/min) for water confined within the selective
layer of 1, 5, and 50 kDa titania membranes. Pore water refers to the water confined within the
membrane pores and bulk water refers to the excess water outside of the membrane. The sample
was a powder obtained by scraping off the selective layer. Source: Reprinted from Reference 61,
Copyright (2008), with permission from Elsevier.
MEMBRANE CHARACTERIZATION 19

5.1.1 Advanced Methods for Measuring Porosity and Pore Size. In addition to the
above-mentioned classical techniques, there are a number of advanced methods, direct
and nondirect, that may be used for characterization of porosity and pore size in mem-
branes. In positron annihilation lifetime spectroscopy (PALS), the sample is irradiated
with positrons generated using positron-emitting atoms such as 22 Na (4). Emission of a
positron by 22 Na is detected by means of simultaneously released 511 keV γ-rays. When
positron is implanted in the matter, it may first form a relatively stable ortho-Positronium
(o-Ps) complex with electron and eventually annihilates releasing 1.28 MeV γ-rays. The
1.28 MeV and 511 keV γ-rays may both be detected to determine positron lifetime and,
using many emitted position, lifetime distribution. The lifetime increases with the size of
cavity in which o-Ps is formed and the assumption that cavities are spherical is typically
used to convert lifetime distribution to pore size distribution and molecular-scale porosity
(fractional free volume, FFV). This method is most useful for molecular pores in dense
membranes rather than regular porous membranes such as MF or UF. Most advanced
PALS setups allow depth profiling of FFV within the top layer of composite membranes
with a resolution of a few nanometer (62).
Porous materials may scatter various types of radiation, for example, visible light,
X-rays and neutrons, depending on the pore size; therefore, scattering techniques [X-ray
diffraction (XRD), small angle X-ray scattering (SAXS), small angle neutron scattering
(SANS), light scattering, etc.] may, in principle, be used for examining porosity and
pore size. Unfortunately, nanopores in polymeric membranes are often too large and
irregular for these methods. Therefore, scattering techniques are commonly applied to
highly regular membrane materials (e.g., zeolite and MOFs (63, 64)), which produce
sharp XRD spectra, and to pseudoporous microphase-separated polymers such as block
copolymers and ionomers, most notably, Nafion (Section 5.4) (65).
The utility of a few other techniques may be illustrated for the case of the topmost
(“skin”) part of tight asymmetric membranes (UF and NF) and the supporting structure
of RO, NF, and gas separation (GS) composites just under the active layer (see next
sections). Cross-sectional images of these membranes may be obtained by TEM and SEM
(Section 3.1); however, resolution and contrast become insufficient for pores less than a
few nanometers. A further complication is the large spacial gradient of porosity and pore
size in the under-surface region. Top view SEM images can often be misleading as the
underlying porous structure is hidden under a less porous “crust” (in UF or asymmetric
NF) or a dense active layer (in composite membranes).
To analyze the size of nanopores within the skin part of asymmetric NF membranes,
that is, just under the “crust,” Stawikowska and Livingston (42, 66) devised an ingenious
method to quantify it. Large Os nanoparticles, a few nanometers in size, were injected
into the skin region from the back by backflow and the region was imaged using TEM.
The high contrast nanoparticles trapped within the finest nanopores in the skin region
could be easily observed and counted by TEM. Using particles of several sizes, consistent
estimates of pore size and comparison of several membranes could be made.
The porosity of the undercrust region can also be determined using RBS (Section 2.1).
RBS, being an absolute atom-counting method, yields the stoichiometric ratio of differ-
ent atoms within surface layers, which may be depth profiled. Zhang et al. estimated
the porosity of the top 500 nm of polysulfone support (structurally identical to UF mem-
branes) in RO composite membranes after equilibrating the membrane with a salt solution
and measuring underlayer atomic composition by RBS. Assuming the salt (i.e., Na and
20 MEMBRANE CHARACTERIZATION

Cl atoms) concentration in the pore was identical to that in external solution, the porosity
was found to be 0.45 ± 0.05 (7).

5.2 Dense Membranes for Pressure-driven Separations


Pressure-driven membrane processes of RO and NF employ composite membranes with a
dense top layer, 15–300 nm thick, to retain salts and organic molecules smaller than about
200–1000 Da. In most commercial RO and NF membranes, this layer is prepared in situ
via the process of interfacial polymerization (IP) and is not available as a bulk polymer.
Theoretical models indicate that IP results in films possessing a complex and irregular
structure, in which significant nonuniformities of polymer density, chemistry, and charge
extend down to nanoscale (67). Even if the selective layer originates from a bulk polymer,
for example, as for the few types of RO and NF prepared by phase inversion or coating,
its structure may be altered by the proximity to interface, preferential orientation of the
polymer molecules in the course of preparation, or chemical changes induced by post-
treatment. To reveal the actual characteristics of the ultrathin top layer, conventional
examination techniques have to be replaced by surface-sensitive high resolution methods.
The parameters of greatest interest are usually those that can be related to transport or
rejection of relevant permeants via known models, as reviewed in subsequent text.

5.2.1 Thickness, Morphology, Swelling. Effective thickness of the top layer ultimately
determines permeability and flux for a given material. Normally, one would like to
keep the thickness to the minimum, as long as it is robust enough and defect-free.
Thickness may be directly read from cross-sectional SEM or TEM images using freeze-
fractured or microtomed specimens. In addition, SEM and, especially, TEM reveal various
morphological features, such as thickness variations, voids, asperities, or bubble-like
protrusions (68). For all-polymer samples, lack of contrast and featureless morphology at
the boundary between the top layer and support may be a problem. In TEM, the contrast
between the top and support layers may be enhanced by staining the top layer with heavy
atoms such as uranium or osmium. The disadvantage of EM is that a very small area is
visualized, requiring many samples and images to obtain a representative general picture.
Average top layer thickness over macroscopic areas (millimeters to centimeters) may
be directly obtained by examining the composition of the membrane surface using a
surface-sensitive spectroscopy with a probing thickness about an order of magnitude
larger than the thickness of the top layer. ATR–FTIR, SEM–EDX, and RBS are suitable
methods with penetration depth of the order of a micrometer (4). The acquired spec-
trum will then contain signals from both the top layer and the underlying support. In
ATR–FTIR, the relevant signals are IR bands characteristic of specific chemical units.
In SEM–EDX, the signals are bands characteristic of specific atoms in the X-ray energy
spectrum and in RBS they are for bands in the energy spectrum of backscattered He++
ions. After appropriate calibration, the absolute intensities of the bands may yield the
thickness of the top layer. In the case of ATR–FTIR and SEM–EDX, the intensities
show a good linear correlation with thickness, while for RBS modeling is required; how-
ever, it yields depth profiles of various elements as a significant extra benefit (69). To
increase sensitivity and specificity, the top layer may be separated form the support and
transferred onto a different substrate. For instance, a Si wafer as a substrate does not
contain C atoms and thus SEM–EDX may use the strong C band of the top layer to
correlate thickness with the C/Si band ratio in the EDX spectrum (70).
MEMBRANE CHARACTERIZATION 21

Transfer of the top layer onto a different substrate often facilitates examination of
thickness along with other characteristics, such as morphology or swelling. On a solid
smooth substrate, such as a Si wafer, AFM may be used to scan across stripes of a
thin film both in air and under water (a unique feature of AFM) and thus determine
its thickness in dry and swollen states. This, along with thickness, this yields swelling
of the film, highly important in transport modeling. For instance, a higher swelling of
polyamide NF membranes (20–30%) compared to polyamide RO membranes (5–10%)
correlates well with the much higher salt rejection of the latter (71).
Voids, nonuniformities, and large surface roughness may significantly overestimate
the effective barrier thickness, that is, the completely dense selective part of the top
layer, which may be only a small fraction of the total thickness measured by one of the
above-mentioned methods. A distinct dense barrier within the top layer may be observed
in TEM cross sections of some RO membranes, but its thickness is hard to determine. In
such cases, estimates based on transport measurements must be preferred. Transfer of the
top layer on a solid electrode allows measuring transport characteristics such as diffusion
or electrical resistance of the film, which—as well as the hydraulic resistance—are all
inversely related to barrier thickness and correlate with one another (39).

5.2.2 Fixed Charge, Ion Partitioning and Surface Charge/Potential. Knowledge of


fixed-charge density and ion partitioning is of special interest for modeling salt separations
by RO and, especially, NF. Owing to strong exclusion of ions, their content in the ultrathin
active layer is very small and quantitative measurements are challenging. The content of
ionic groups within the top layer of NF membranes was first systematically examined by
Schaep and Vandecasteele (72) using several methods. However, the results by different
methods showed discrepancies and some required knowledge of (yet unknown) top layer
thickness in order to be converted to fixed-charge density (72). The use of RBS brought
about a major breakthrough and provided accurate partitioning data for a number of
mono- and divalent cations and anions. Moreover, by performing equilibration at different
pH, the neutralization behavior, that is, pKa values of the carboxylic charges within the
active layer could be determined. They were found to be a few pH units above the pKa
of carboxylic acid value in water, indicating the effect of low dielectric environment on
dissociation constant within the active layer.
Recently, Perry and Coronell (73) demonstrated that results nearly identical to RBS but
at a much lower cost could be obtained using a bench-top quartz crystal microbalance
(QCM; see Section 7.3). The unique sensitivity of QCM was sufficient to sense the
changes of the mass upon exchange of light cations within the active layer to heavier
Cs+ ions from solution and the mass change could be converted to the ion uptake
and fixed-charge density for a given solution composition. The method required that
the active layer be separated from the support and attached to the sensor surface, but
excellent agreement with RBS data obtained for intact membranes confirmed that the
procedure was fully adequate.
Surface charge or potential are of particular importance in fouling and biofouling
phenomena. For homogeneous materials surface charge may also serve as an indicator
of the fixed charge density within the membrane (8). Perhaps the first estimate of
amount of fixed charged groups at the surface of a dry RO composite membrane was
made by Koo et al. (74) using XPS. Currently, the most popular method is steaming
potential, estimating surface potential of membranes in solutions resembling the working
conditions (Section 6.2).
22 MEMBRANE CHARACTERIZATION

Chemical composition and polarity of RO and NF composites are of great interest for
examination of fouled, damaged (e.g., by chlorine), and chemically modified membranes.
The most common techniques that possess both surface- and chemistry-sensitive are
ATR–FTIR, XPS, RBS, SEM–EDX. ATR–FTIR and XPS are usually the most infor-
mative methods, as they allow identification and quantitative assessment of the content
of specific chemical groups (rather than just atomic composition) with the surface layers.
A remarkably simple, yet highly useful, technique for assessing surface chemistry
is measurement of contact angle (CA) at the three-phase line between membrane, liq-
uid, and air (vapor) (75). CA is related to interfacial energies of a surface and thereby,
indirectly, to polarity of the membrane. In the “sessile drop” method, a liquid drop of
water or an organic liquid is placed on a surface. An alternative is the “captive bubble”
method, in which a bubble is attached from below to the downfacing surface of a mem-
brane immersed in a liquid. The latter method is usually recommended, as the membrane
is properly equilibrated in practically relevant conditions. Accurate assessment of true
equilibrium CA is usually difficult, particularly, on rough, chemically heterogeneous, and
nonequilibrated surfaces and produces a large hysteresis between receding and advancing
angles. However, if performed in a consistent manner, CA is a facile way of comparing
different membranes, for example, modified or fouled. Moreover, evaluation of CA using
several appropriately chosen liquids allows assessment of various contributions to molec-
ular interactions such as dispersive, polar, H-bonding, and acid–base. (76). These data can
aid modeling of transport and partitioning of organic solutes within the membrane (77).

5.3 Gas Separation Membranes


GS membranes are structurally similar to RO and NF ones; they have an integrally
skinned or composite structure with an ultrathin nonporous selective layer on top of a
porous support. The above-mentioned microscopy and microanalysis applicable to RO
and NF membranes may then be as beneficial for GS. However, materials used in GS are
often available as bulk materials and are amenable to more conventional characterization.
Of particular interest to GS is the state of polymer, that is, it being glassy or rubbery,
which is closely related to other molecular characteristics such as FFV (78, 79) and
segmental mobility. The latter are known to profoundly affect transport characteristics
of the polymer, in particular, permeability, sorption, and diffusion of individual gases,
as well as selectivity for gas mixtures. The temperature and composition dependence of
these parameters is also highly important, as gas mixtures are often separated at elevated
temperatures (80, 81). Characterization of the state of the polymer and glass-transition
temperature Tg using dilatometry, thermal, and mechanical analyses has been an important
part of studies of GS materials and membranes.
Dilatometry measures the thermal expansion of polymers, that is, variation of specific
volume with temperature, which changes the trend and the value of thermal expansion
coefficient near Tg and may employ different types of instruments. In “pushrod” dilatome-
ters, the polymer is heated and the pressure it exerts on a pressure sensor is measured.
Another approach is to measure the volumetric expansion directly by immersing the sam-
ple in a liquid of constant expansion coefficient (e.g., mercury) and monitor the volume
change with a rise of capillary tube. For very accurate expansion measurements, the poly-
mer can be placed between two capacitor plates and the capacitance can be correlated
with the change in the polymer dimension. A change in thermal expansion coefficient, that
is, the slope of the polymer volume versus temperature, signifies a glass transition, where
segmental mobility is activated and free volume increases, and is used to evaluate Tg .
MEMBRANE CHARACTERIZATION 23

Notably, Tg is not sharply defined and attention to various experimental details such as
the rate of temperature change is crucial, as they may significantly alter the value Tg (82).
Thermal analysis using DSC or differential thermal analysis (DTA) is another way to
observe glass transition and evaluate Tg as well as melting (Tm ), and crystallization (Tc )
temperatures. Both techniques perform measurements for a sample of interest versus a
reference sample of well-known heat capacity, which are heated or cooled in adiabatic
conditions. In DSC, both the sample and the reference of the same geometry are heated
at the same rate while the heat flow is measured; in DTA measurements, the heat flow
is kept constant and the temperature difference between the sample and the reference is
measured. The increase in heat flow, that is, slope of heat consumed or released versus
temperature, is indicative of a change in thermal capacity of the sample and signifies glass
transition. It is distinctly different from phase transitions such as melting or crystallization
that show up as positive (endothermic) or negative (exothermic) peaks (Fig. 7), associated
with latent heat.
The free volume in polymers is often estimated using the relationship FFV =
(V − V0 )/V (80), where V is the polymer specific volume, calculated from bulk
polymer density, and V0 is the occupied volume. V0 may be calculated from correlations;
a popular one is V0 = 1.3Vw , where Vw is the van der Waals volume, often estimated
using Bondi’s group contribution method.
A more accurate method that measures both the size and concentration of the free
volume cavities in condensed nonconductive materials, such as polymers for GS, is
PALS (Section 5.1.1). For instance, PALS was applied to elucidate free-volume-increasing
effect of nanoparticles dispersed in a polymer that resulted in an unusual flux-selectivity
behavior of this composite material (78).

5.4 Ion-Selective Membranes for Electro-driven Separations and Fuel Cells


The distinct feature of ion-selective membranes (ISMs), also often termed ion-conductive
or ion-exchange membranes, is a very large difference in the permeability of anions and
cations. This is achieved by using materials with a high content of fixed charged groups,
either positive or negative, covalently bound to the matrix. The large fixed charge results
in a strong exclusion of co-ions, precluding nonselective transport of salts, and a high
ionic conductivity (36).
ISMs are used in electro-driven processes such as electrodialysis, membrane electroly-
sis (ME), and fuel cells (FCs), where selective transport of a certain type of ions (cations
or anions) is required (83). Compared to most separation membranes, commercial ISMs
are usually nonporous symmetric films, a few tens to a few hundreds micrometers thick,
homogeneous or heterogeneous (50–70% ion-exchange resin embedded in a hydrophobic
matrix). The main characteristics of an ISM are ion-exchange capacity (i.e., fixed-charge
density), electrical conductivity (or resistance), and permselectivity. Because most ISM
membranes are thick and self-supporting, these measurements are relatively straightfor-
ward. The ion-exchange capacity is most easily determined by converting the membrane
to either H+ or OH− or other appropriate ionic form, exchanging counter ions out, and
measuring their amount by titration, inductively coupled plasma, or ion chromatography.
Transport characteristics of ISMs are usually determined using a two-compartment
cell, in which a membrane separates two electrolyte solutions. An adequate agitation
is crucial to minimize nonselective contributions of unstirred layers at each side of the
membrane. To avoid polarization, electrical resistance of the membrane is measured in
24 MEMBRANE CHARACTERIZATION

AC mode, preferably using EIS (Section 4.2), when each compartment contains identical
solutions and is fitted with one or two electrodes. A four-electrode arrangement is most
recommended, whereby each compartment contains one reference electrode (e.g., calomel
or Ag/AgCl) and one working electrode that passes the AC current. In some cases, for
example, for low-resistance membranes, measurements employ a stack of membranes
placed between electrodes.
Membrane permselectivity, that is, absolute value of transport number difference
|t+ − t− |, may be derived from diffusion permeability of salt and electrical resistance
of the membrane, provided both are measured for similar salt concentrations (cf. Eqs.
2 and 4). However, most often, the permselectivity is directly assessed from the diffu-
sion potential measured between two salt solutions of different concentrations C1 and C2
with C1 /C2 between 2 and 10 (Eq. 3). KCl is the most commonly used electrolyte as
tK+ ≈ tCl− in aqueous solutions, which minimizes potential gradients outside the mem-
brane. The permselectivity is usually reported as the ratio of measured diffusion potential
to that of an ideally selective membrane, that is, when either t+ or t− is zero. Note that
measured potential difference for simple Ag/AgCl reference electrodes will include chem-
ical potential differences of chloride (RT/F ln[C1 /C2 ]), which will approximately double
(for t+ ≈ 0) or cancel out (for t− ≈ 0) the membrane potential. The problem is eliminated
by using references electrodes with a salt junction, for example, Ag/AgCl/KCl(sat).
Applications such as FCs and ME impose a few more requirements on the membranes
(84). In addition to the high conductivity necessary to minimize ohmic losses in the FC,
a low permeability to pertinent uncharged species is required. For instance, in ME the
membrane should be impermeable to molecular chlorine and in an FC the permeability
to fuels and oxidants should be as low as possible. Practically, this means that the ratio
of membrane conductivity to the neutral solute permeability should be maximized, for
example, the ratio of proton conductivity to methanol permeability in direct methanol FC.
Membranes working in FC and ME are usually exposed to high osmotic and oxidative
stresses. The oxidative stress brought about by formation of radicals in electrode reaction
may cause rapid degradation of polymer and poses strict requirements on selection of
membrane materials. Characterization of chemical changes within polymer structure,
for example, by ATR–FTIR and NMR (85), may add to understanding of degradation
mechanisms and development of more chemically stable materials.
The osmotic stress may be caused by high osmotic pressure of solution (in ME) or by
electro-osmosis that tend to dehydrate the anode side of working FC membranes; thus,
their conductivity, selectivity, and stability drop sharply. These effects motivated many
characterization studies, especially, for FC membranes. Nafion, the current benchmark
membrane material for ME and FC applications, has been characterized most exten-
sively. The main purpose was to understand the relation between microstructure and
“macroscopic” characteristics such as water sorption isotherm, proton conductivity and,
ultimately, FC performance. The unusual microphase-separated nanostructure of Nafion,
composed of bundles of elongated micelles, has been elucidated by SAXS, SANS, and
TEM (86). The evidence of more complex arrangements in the interfacial regions was
obtained by AFM, grazing angle SAXS, and contact angle (Fig. 8) (87, 88), as well as
neutron reflectivity (89). Important insights into water and proton mobility and distri-
bution within Nafion were obtained by NMR, and space- and time-resolved scattering
techniques and IR and μ-Raman spectroscopy (90). These methods were also helpful in
comparing Nafion with alternative state-of-the-art membrane materials for FCs.
MEMBRANE CHARACTERIZATION 25

(a) (b)

(c) (d)

(e) 0.3 2.5 (f) 0.3 1.5

log10(Intensity) (a. u)
0.25 2 0.25
Qz (1/A)

Qz (1/A)

0.2 0.2 1
1.5

0.15 0.15
1
0.1 0.1 0.5
0.5
0.05 0.05
0.05 0.1 0.15 0.2 0.25 0.3 0.05 0.1 0.15 0.2 0.25 0.3
Qy (1/A) Qy (1/A)

FIGURE 8 The surface of cast Nafion films equilibrated with water vapor (a, c, e) and liquid
water (b, d, f). The surface structure was characterized by three different methods: AFM (a and b),
contact angle (c, by sessile drop method, and d, by captive bubble method), and by grazing incidence
SAXS (e and f). The AFM results show the “surface-roughening” transition from a smooth (a) and
hydrophobic (c) surface in vapor to a hydrophilic (d) surface with random asperities (b) formed in
water. The highly anisotropic patterns observed in grazing incidence SAXS spectra, sensitive to the
structure of the outmost region, indicate that the orientation of the Nafion micelles adjacent to the
surface changes from parallel to surface in vapor (e) to predominantly perpendicular to the surface
in water (f). Source: (a, b, and f) Reprinted with permission from Reference 87. Copyright (2011)
American Chemical Society. Parts (c, d, and e) Reprinted with permission from Reference 88.
Copyright (2010) American Chemical Society. (Please refer to the online version for the color
representation of the figure.)
26 MEMBRANE CHARACTERIZATION

6 SURFACE CHARACTERISTICS

6.1 Surface Morphology and Roughness


Numerous studies have shown that surface morphology can significantly affect the mem-
brane permeability, rejection, and propensity to colloidal fouling (91–96). For instance,
Hoek et al. (97) concluded that colloidal particles may preferentially deposit in the val-
leys of a rough membrane, as the energy well predicted by DLVO theory becomes deeper
compared to a smooth surface. SEM and SPM are the most common methods for exam-
ining morphology and roughness of membrane surfaces. In SEM, information on surface
topography is mainly contained in the signal coming from secondary electrons (Fig. 1a).
In SPM, the topography and other physical characteristics of the surface are mapped
using lateral and vertical movements of a sharp tip (Section 3.2).
Note that SEM images are essentially 2D projections in which some information on
surface topography is inevitably lost; therefore, SEM is used mainly for imaging and
qualitative comparison. In contrast, AFM or other SPM techniques produce a genuine
3D topographic map of the surface; thus, various quantitative statistical parameters may
be calculated, for instance:

• Root mean squared (RMS) roughness, defined as standard deviation of height;


• Average roughness (AR), defined as the average absolute deviation of height from
the mean height;
• Surface area difference (SAD) is the increase of surface area over the projected area
due to roughness;
• Peak count (PC) is the number of peaks in the scanned area.

More information may be gained from a full height distribution, for example, a height
histogram or the so-called bearing area plot that reveals fractional surface that lies above
or below a given height. These plots may indicate uniformity of height distribution
and presence of one or more dominant heights, which cannot be done using any single
parameter such as RMS, AR, SAD, and PC. Still, a height histogram or bearing plot do
not address spacial correlations between peaks and valleys, for example, the presence
and amplitude of periodical patterns or objects of characteristic size on the surface. A
more comprehensive picture can be obtained by spectral analysis tools, such as FFT or
a related power spectral density (PSD). They may usually be generated using built-in
software of AFM instruments along a 1D (line or stripelike) or for a 2D regions, yielding
a 1D spectrum or a 2D spectral map. The PSD intensity is correlated with amplitude
of height variations, while the presence of characteristically spaced features is seen as
maxima in intensity at particular spacial frequency. For instance, anisotropic 2D maps
indicate that characteristic spacing is different in X - and Y -directions. An example shown
in Figure 9 demonstrates how different surfaces (imaged using phase signal of AFM in
tapping mode) may be transparently compared using PSD (66).
Ba and Li (98) have shown that the pore configuration of nanosize pore array on
an aluminum membrane can be simply analyzed using FFT. Bowen et al. (99) have
shown that FFT can be also used to screen out AFM artifacts and “correct” membrane
pore images. Notably, FFT or PSD analysis is not limited to SPM images; for instance,
Stawikowska and Livingston (42) applied it for both AFM surface images and TEM
MEMBRANE CHARACTERIZATION 27

(a) (b)

100 nm 100 nm

10°
(c) (d)

100 nm 100 nm

–20°

(e)
4
M1
M2
M3
3 Dense film
PSD [log (deg2 nm2)]

1
qmax

FIGURE 9 Phase images of the surface of three nanofiltration (NF) membranes fabricated in
different conditions (a–c) and a dense film of the same material (d). Insets show correspond-
ing FFT (scale bar: 0.1 nm−1 ). (e) 2D power spectra density (PSD) plots corresponding to the
four samples; qmax yields ξ , the characteristic spacing of asperities (ξ ∼ 2π/qmax ), and the PSD
amplitude is indicative of the amplitude of phase variations. Source: Reprinted from Reference 66,
Copyright (2013), with permission from Elsevier. (Please refer to the online version for the color
representation of the figure.)
28 MEMBRANE CHARACTERIZATION

cross-section images of nanoparticles trapped within a subsurface nanoporous structure


to analyze the pore size.
Measured surface roughness under water may be affected by surface swelling that may
change with the ionic strength and pH of the solution and both decrease and increase
roughness depending on degree of swelling and shape and spacing of peaks and valleys
(100). The size of the scanned samples is another important parameter. Koyuncu et al.
(101) combined AFM with vertical scanning interferometry microscopy to show that
the average roughness monotonically increases with the scan area up to a critical size
of about 500 × 500 μm2 , which significantly exceeds the typical size of AFM scans.
However, in many cases, AFM images prove sufficient, for example, for correlating
surface morphology with colloidal fouling (102).

6.2 Surface Charge and Electrical Properties


Most surfaces acquire an electric charge when brought in contact with an aqueous solu-
tion. The surface charge originates from dissociation of ionizable groups at the surface,
for example, carboxylic groups in many commercial polyamide NF and RO membranes,
and/or preferential adsorption of ions or polyelectrolytes. For instance, small anions
are in general more polarizable than cations and more readily adsorb on hydrophobic
surfaces, making them negatively charged. The surface charge and electric potential
are pH-dependent, turning from positive to negative as pH increases from acidic to
basic. The importance of surface charge in membrane characterization comes from the
fact that it usually correlates with the propensity to fouling or may indicate the state
of fouling of membranes (103). It may also be indicative of the fixed-charge density
within the membrane, provided the surface and bulk of the membrane have similar
chemistry (5, 8).
Electrokinetic measurements of surface charge and potential have been most popular.
They utilize the fact that a charged surface attracts counterions from solutions to form
an electric double layer (EDL) containing a total space charge that neutralizes the sur-
face. When the solution flows along the surface, more counterions are dragged by the
fluid compared to co-ions, which induces a net surface current and a potential gradient.
(The phenomenon may be likened to potential drop across a membrane with unequal
permeabilities of cations and anions; see Eq. 3.) The induced SP is usually measured in
conditions of zero current, where the flow-induced surface current is exactly compen-
sated by the opposite current in the bulk solution. It is common to express the result as
ζ -potential, which is formally the potential at shear plane located several angstroms away
from the surface. A common setup includes two identical membrane samples forming
walls of a flat channel about 0.1 mm thick. An ion solution, typically, 1–100 mM KCl
at defined temperature and pH, is pumped through the channel using a fixed pressure P
applied at the inlet to obtain a steady-state flow. In the case of a hollow fiber membrane,
the solution is simply pumped through the fiber. For porous membranes, such as MF or
UF, the ζ -potential of the inner pore surface is measured by pumping the solution across
the membrane (104). Measurements at several values of inlet pressure P are used and
the potential drop U along the channel is measured for each pressure. The slope dU/dP
is then used to calculate ζ -potential using the Helmholtz-Smoluchowski (105)

dU μ L
ζ = (9)
dP εε0 AR
MEMBRANE CHARACTERIZATION 29

TABLE 2 ζ -Potential of Commercial Membranes

Membrane Commercial Name Active Layer ζ -Potential at References


Type (Manufacturer) pH 7, mV

NF NF-270 (Dow-Filmtec) Semiaromatic Polyamide −20a 106


NF NTR-7450 (Hydranautics/ Sulfonated-polysulfone −5a 106
Nitto Denko)
Brackish XLE (Dow-Filmtec) Polyamide −30b 107
water RO
Brackish ESPA-3 (Hydranautics) Polyamide −17b 107
water RO
Seawater RO SWC4 (Hydranautics) Polyamide −10b 107
RO (Desalination Systems, Cellulose acetate −15b 108
Escondido)
a In 1 mM KCl.
b In 10 mM NaCl.

where L, A, and R are the length, cross-sectional area, and electrical (AC) resistance
of the channel, respectively, and ε0 = 8.85 × 10−12 F/m, and ε and μ are the dielectric
constant and viscosity of the solution, respectively. As an alternative to SP, one can
measure in the same setup for several values of P the streaming current, which is the
current that has to be passed along the channel in order to exactly cancel SP. The slope
of current versus P may be similarly used to calculate ζ -potential.
ζ -Potential often approximates the actual surface potential, as both have the same
sign and vanish at the same pH (the isoelectric point). Table 2 shows ζ -potential of
several representative commercial membranes. It must be emphasized that ζ -potential
will change with pH (see preceding text). The effect of ionic strength on ζ -potential is
not as strong as pH; however, measurements often fail at high ionic strengths, as the
current and potential become immeasurable owing to low channel resistance R.
Apart from electrokinetic methods, the surface charge can be evaluated by several
other techniques. For a thick homogeneous membrane, the surface charge may be esti-
mated from the bulk fixed-charge density (Section 5.4). Assumptions about the surface
structure are required to convert bulk charge density to surface charge density. A few
other techniques are available for estimating surface charge or potential for model sur-
faces that are often used in membrane research. Shan et al. (109) used surface plasmon
resonance (SPR) to measure the surface potential of gold, a popular model surface. Force
measurements between a colloidal probe and the surface using AFM fitted to an appropri-
ate theoretical model can estimate the surface charge or potential and other characteristics
of EDL in solution (110, 111).

7 PROPENSITY TO FOULING AND BIOFOULING

“Real world” conditions faced by membranes during scaling, organic or colloidal fouling,
and biofouling tend to be highly diverse, poorly defined, and constantly fluctuating. In
general, it is difficult to follow them in situ using advanced analytical tools and costly and
time consuming to reproduce in the laboratory. Many fouling studies, in particular those
aimed at the physical mechanism of fouling and at narrowing down the dominant factors
30 MEMBRANE CHARACTERIZATION

affecting fouling are performed at laboratory scale or in small dedicated inline units
using well-defined conditions and suitable analytical tools. Depending on the specific
characterization technique, one may use either genuine membranes or representative
model systems and employ either genuine complex feeds or well-defined synthetic media.

7.1 Membrane Fouling Tests


Two principal testing modes, dynamic and static, are employed for fouling experiments
with real membranes. Static experiments best suit the cases of fast reversible organic
fouling (e.g., with proteins), biofouling, and/or scaling. A membrane coupon is immersed
in a solution containing foulants of interest or bacteria and nutrients in well-defined
conditions, possibly under shaking or stirring. After the prescribed time, the membrane
is removed, rinsed, and examined for the amount of adsorbed foulant or grown biofilm.
This approach is less representative than the dynamic tests; however, it is easy to perform
and the conditions are well controlled. It normally employs small membrane samples
and experiments are thus easily parallelized, for example, for testing reproducibility or
performing multiple comparative experiments with many different conditions in a high
throughput format (112, 113). The latter may be particularly useful for optimization of
membrane materials or preparation or modification procedures with respect to propensity
to fouling (114).
In dynamic experiments, the feed is continuously pumped along the membrane in a
cross-flow mode, which more closely approaches the genuine conditions of membrane
filtration, even though recovery in a small laboratory cell is small, that is, the solution
composition changes insignificantly along the cell. Some examples of fouling and bio-
fouling studies with different types of membranes can be found in MF (115), UF (116),
NF (117) and RO (118), forward osmosis (FO) (119), MBR (120), and membrane dis-
tillation (121). The attention to well-defined hydrodynamic conditions is crucial in such
experiments, in particular, fluid velocity, shear rate, and, if the tests involve a permeate
flow (i.e., filtration under pressure), the permeate flux and concentration polarization.
Filtration tests usually employ multistage protocols and various model foulants (see
next). A typical protocol is shown in Figure 10. After a membrane is mounted inside
a filtration cell, it is first compacted by filtering clean deionized (DI) water until a
constant flux is obtained. Thereafter, feed solution containing foulants is pumped through
the cell for a prescribed time and thereafter the feed is again replaced with DI water.
Additional chemical or mechanical cleaning and washing steps before and after fouling
may be included. During the whole sequence, the flux (and, preferably, rejection as well)
is monitored. The difference between the membrane permeability and rejection before
and after the filtration, washing, and cleaning steps serves as indicators of fouling, its
reversibility and efficiency of cleaning (122). Another indicator is the rate of permeability
change due to fouling, which may be analyzed using the mathematical models to conclude
about fouling mechanism, for example, pore blocking, intermediate pore blocking, cake
filtration, and pore constriction (123).

7.2 Characterization of Fouled Membranes


Characterization of fouled membranes is used to observe morphological changes due to
fouling, identify the type of fouling (organic, inorganic, biological, etc.), and quantify the
amount of specific foulants on the surface. Most techniques commonly used to examine
MEMBRANE CHARACTERIZATION 31

Clean H2O Solution Clean H2O

ra2
rm

ra1 NaOH
Jv cleaning

rg Water flush

Tc Pressure release

Time

FIGURE 10 A typical filtration protocol used to evaluate the extent and irreversibility of fouling
and the efficiency of washing and cleaning. Source: Reprinted from Reference 122, Copyright
(1999), with permission from Elsevier.

chemistry and morphology of the surface of membranes may just as well apply to fouled
membranes, such as SEM–EDS, XPS, ATR–FTIR, SP, and contact angle. One of the
techniques that shows particular benefits for characterization of fouled and biofouled
membranes is OM, as briefly reviewed here.
Regular OM is most suitable for foulants that form morphologically distinct objects of
a size exceeding about a few tenths of a micrometer and have optical properties different
from the surrounding medium, such as inorganic crystals or particles in air or water.
OM offers a facile noninvasive examination of dry samples as well as through a layer
of transparent fluid, which makes it a convenient technique for real-time monitoring of
fouling and scaling in water. Figure 11 shows OM images of an RO membrane subject
to inorganic fouling (scaling) with and without concurrent biofouling (124). The contrast
may be improved using various contrast-enhancing modes of OM such as phase contrast,
differential interference (Nomarski) contrast, and dark field mode.
OM may be difficult to use for biopolymers, living cells, and biofilms that have
refractive indices insignificantly different from that of water. Furthermore, the use of
OM contrast-enhancing modes may be also precluded by membrane porosity and surface
roughness that produce diffuse scattering. For these types of organic and biofoulants,
fluorescence microscopy provides a solution. Foulants may either have a native fluores-
cence (e.g., via genetic tagging with a green fluorescent protein, GFP) or be stained with
various types of fluorescent markers bound to specific groups. Apart from information
on morphology, with a proper calibration of fluorescence intensity, the method may be
used to simultaneously quantify the amount of different foulants emitting at different
wavelengths; more information can be found in Reference 125. The method is widely
used for examining model surfaces (126), as well as genuine membranes (127) and can
conducted in real-time and in situ in specific locations on the surface. For thick transpar-
ent samples, more information on the structure and morphology may be obtained using
confocal scanning laser microscopy (CSLM), which, by moving the focal plane, collects
images of horizontal slices within a 3D structure. These slices can be subsequently recon-
structed to yield a 3D image. The method has been most popular for studying biofouling
and biofilms. It can operate, just as regular OM, in transmission and reflection modes,
but mostly it is operated in fluorescence mode, making use of native fluorescence of
32 MEMBRANE CHARACTERIZATION

t=0h t=0h

t=5h t=5h

t = 15 h t = 15 h

t = 20 h t = 20 h

t = 25 h t = 25 h
(a) (b)
Flow direction

FIGURE 11 Optical images of the RO membrane surface during gypsum scaling without (a)
and with (b) a biofilm. Source: Reprinted from Reference 124, Copyright (2012), with permission
from Elsevier.

the sample or fluorescent markers. Usually, CSLM provides the highest resolution of all
OM techniques approaching the diffraction limit. Advanced techniques are being devel-
oped to further improve its resolution such as two-photon CLSM. Figure 12 displays
CLSM images showing deposition and growth of a biofouling layer on a polyamide RO
membrane in biofouling experiments terminated at different times (118).

7.3 Model Systems: Materials and Instrumentation


Sufficient sensitivity and control of specific factors, important for elucidating physi-
cal mechanisms, is rarely possible using genuine membranes and feed solution. In this
situation model foulants and surfaces are commonly employed. A number of advanced
techniques are available for studying adhesion, deposition, and adsorption phenomena
forming the physical basic of fouling. The list of useful materials and methods is obvi-
ously long; here, a few modern techniques that offer exceptional sensitivity and unique
physical insights are addressed.
The most common model foulants used in filtration tests and sorption on membranes
and model surfaces include

• single biomolecules such as protein or polysaccharides;


• standardized mixtures of organic molecules, such as purified humic acid fraction
of natural organic matter (NOM), alginates, or extracellular polymeric substances
(EPSs) extracted from real biofilms;
• suspension of colloidal particles such as silica, clay, or polystyrene nanoparticles;
• suspensions of single-strain bacteria and native bacterial consortia for initial depo-
sition experiments and biofilm growth;
MEMBRANE CHARACTERIZATION 33

Cell deposition and micro-colony formation (4 h after PA01 inoculation)


(a) (b) (c)

Biofouling layer maturation (19 h after PA01 inoculation)

Cell detachment (24 h after PA01 inoculation)

FIGURE 12 3D reconstructions of biofilms of Pseudomonas aeruginosa PA01 formed on the


surface of an RO membrane in the course of biofouling runs terminated at different times. 3D
images were reconstructed using Imaris software from planar images acquired at depth intervals of
1 μm. Light gray (a) and gray (b, c) colors correspond to live and dead fluorescent stains, respec-
tively. Source: Reprinted from Reference 118, Copyright (2007), with permission from Elsevier.
(Please refer to the online version for the color representation of the figure.)

• model industrial effluents such as oil–water emulsions.

Proteins such as bovine serum albumin (BSA) or lysozyme are usually used, with well-
known characteristics, which include isoelectric charge, size, amino acid composition, etc.
The protein absorption on membranes can be quantified by ATR–FTIR (using specific
characteristic bands such as Amide I and II), extraction of the protein from the membrane
and measuring its concentration with standard techniques (e.g., Bradford or Lowry assays
or chromatography), fluorescence microscopy (for fluorescent or fluorescence marker-
labeled protein), and other methods.
Genuine assessment of the propensity to biofouling requires relatively long and tedious
growth of biofilm in realistic conditions and quantification of the amount of biofilm and
biovolume of bacteria per unit area and their correlation to loss of performance (118).
Biofilm growth may be carried out either by using model cultures (most often single
strain) or native consortia containing naturally present or added minerals and nutrients.
In experiments with model solution, the membranes are often exposed to bacteria only in
the beginning of the experiment for a predetermined time followed by circulation of clean
nutrient solution. Experiments with natural consortia are usually carried out simply by
exposing the membrane to the natural feed solution in controlled conditions until a biofilm
is grown. The amount of bacteria and biofilm on the membrane may be determined using
by fluorescent OM or, preferably, CLSM after appropriate fluorescence staining (e.g.,
live–dead stains). An alternative methods is SEM, which requires an appropriate fixation
(e.g., with glutaraldehyde) and drying of biofilm. The bacterial community composition
and bacterial population may be also analyzed using DNA extraction or similar molecular
biology techniques.
34 MEMBRANE CHARACTERIZATION

Faster alternative methods that provide useful insights include determination of depo-
sition rates using suspended model strains (128), adsorption of and fouling with EPS
extracts and various EPS fractions (129), and accelerated biofilm growth on membrane
coupons (analogous to static fouling tests) (130). Apparently, all methods are indica-
tive of different stages of biofouling, affinity of the surface to attachment of bacteria
and biofilms, but their correlation to overall propensity to biofouling has not been yet
established conclusively (130, 131).
Bacteria deposition experiments employ single-strain bacteria grown in specific condi-
tions to the desired growth phase, which determines their physicochemical characteristic
(ζ -potential, hydrophilicity, etc.). Grown bacteria needs to be harvested, thoroughly
washed of nutrients and resuspended in the testing solution at the chosen concentration,
which is usually determined by the optical density with UV–Vis. Strains genetically
tagged with GFP have an advantage for microscopic observation on opaque membrane
surfaces. The deposition experiment can be run in a static mode (usually with shaking)
(112) or in dynamic mode, that is, in flow cells with or without concurrent filtration
(127). The number of bacteria or colony-forming units that have adhered to the surface
is counted (usually starting from a clean surface) at prescribed time(s). The mass transfer
(deposition) coefficient (kd ) is calculated using the following equation:

1 dN 1 N
kd = ≈ (10)
CA dt CA t
where N is the number of bacteria in the image, A the image area, and C is the concen-
tration of bacteria in the feed solution. When dynamic experiments are carried out in a
flow cell with a transparent window, GFP-tagged bacteria can be counted in situ in real
time using a fluorescent microscope with a camera.
EPSs are a mixture of secreted biopolymers forming a biofilm matrix, which comprise
a wide variety of polysaccharides, proteins, glycoproteins, glycolipids, and DNA (132).
EPS or its components may be extracted from biofilms in purified form (133). Since EPS
plays a crucial role in adhesion of biofilms to surfaces and their impact on membrane
performance, EPS adsorption on membranes and model surfaces is of significant interest
and has been studied as another indicator of propensity to biofouling.
QCM utilizes resonance oscillations of a thin piezoelectric crystal that is highly sen-
sitive to minute quantities of foulants or particles absorbing or adhering to the crystal.
When the foulant is adsorbed to or removed from the crystal, its resonance frequency
shift (f ) is related to added mass (m), as given by the Sauerbrey equation:

CQCM
m = f (11)
n
where CQCM is the mass sensitivity constant at the fundamental frequency and n is
the overtone number (n = 1, 3, 5, . . .). The validity of this relation is based on a few
assumptions (134): (i) the adsorbed mass is evenly distributed on the crystal, (ii) m
is much smaller than the mass of the crystal, and (iii) the added layer is rigid, does
not deform during the oscillations and has a negligible mechanical interaction with the
medium. More general relations are available when these assumptions are not valid,
for example, for viscoelastic materials or crystal oscillating in a viscous liquid (134).
The QCM crystal can be modified with specific functional groups (135) or coated with
various materials to mimic membrane surfaces (136). Figure 13 shows an example of
MEMBRANE CHARACTERIZATION 35

EPS

O
H2C C
C O
CH3
O
H2C C
C O
EPS
CH3
PVDF

(a) PVDF-coated graft copolymers

A B CD E
0
200
Frequency shift (Hz)

Dissipation factor
150
Dissipation factor
Frequency shift 100

50

0
0 50 100 150
(b) Time (min)

FIGURE 13 (a) The concept of grafting a zwitterionic copolymer to polyvinylidene fluoride


(PVDF) ultrafiltration membranes for reduced adsorption of extracellular polymeric substances
(EPSs) examined by quartz crystal microbalance with dissipation (QCM-D). PVDF membrane sur-
face is mimicked by a PVDF-coated QCM sensor, onto which two oppositely charged monomers are
cografted in an aqueous solution. (b) Kinetics of graft polymerization on PVDF-coated surface mon-
itored by QCM-D: (A) washing with double-distilled water; (B) feeding the monomers, QCM-D
response showing monomer adsorption; (C) feeding redox initiators to start the polymerization
reaction, QCM-D response showing further increase in the weight of the crystal and increasing
dissipation due viscoelasticity of the grafted polymer layer; (D) quenching the reaction with ethanol
showing termination of the reaction; (E) washing with 5 mM EDTA in 100 mM NaCl at pH 11.5;
the crystal weight and dissipation stabilize. Source: Reprinted with permission from Reference
100. Copyright (2011) American Chemical Society. (Please refer to the online version for the color
representation of the figure.)

using quartz crystal microbalance with dissipation (QCM-D) for studying the grafting of
the antifouling layer onto a model polyvinylidene fluoride (PVDF) surface, mimicking the
modification of PVDF UF membranes, which was subsequently demonstrated to reduce
EPS adsorption (100).
QCM-D is an advanced version of QCM that allows obtaining significantly more infor-
mation on the viscoelastic properties of the adhering layer and its interaction with the
medium. Along with the frequency shift, QCM-D measures the changes in the dissipation
36 MEMBRANE CHARACTERIZATION

factor (D), which is the fraction of energy lost (dissipated) per oscillation deduced from
the rate of decay of oscillation amplitude (ring-down). In addition, QCM-D instruments
are capable of analyzing simultaneously several overtones (up to about n = 15 or 17).
These features yield a complete description of the system in terms of complex mechanical
impedance spectrum, form which viscous, elastic, and inertial (added mass) terms may
be deduced. In particular, the ratio f/D may serve a simple criterion while compar-
ing viscoelestic properties of adsorbed layers, larger ratio indicating a more rigid layer
(134, 137). When applied to adhesion of particles and bacteria, this ratio may indicate
the strength of adhesion between the bacteria and the surface (138).
In the context of membrane fouling and biofouling, QCM and QCM-D were used to
study absorption kinetics of proteins (139), polysaccharides (134), mineral scaling (140),
EPS of biofilms (100), and adhesion of particles and bacteria (138). In the latter case, the
QCM may be combined with direct OM observations to establish quantitative relations
between the surface density of adhered bacteria and the QCM-D parameters f and D
for different overtones. When applied to analysis of detachment of adhered polymers,
particles, and bacteria from model surfaces, QCM may also be used as a tool to study
mechanisms underlying cleaning membranes (135).

(a) 0.4
Without alginate
0.2 With alginate (1) (2)
Force (mN/m)

0.0
–0.2 (2) CA CA
–0.4 In contact Rupture
–0.6
–0.8 (1)
–1.0
0 100 200 300 400 500
Extension distance (nm) (3) (4)

(b) 0.4 PA PA
Without alginate
0.2 With alginate In contact Long-range attraction
Force (mN/m)

0.0
–0.2 (5)
–0.4
Alginate
–0.6 (4) (5)
(3) Ca2+
–0.8
–1.0
0 100 200 300 400 500 PA
Extension distance (nm) Rupture

FIGURE 14 Representative adhesion force curves obtained for the cellulose acetate (CA) and
polyamide (PA) membranes in the presence or absence of alginate on the membrane surface. The
solutions contained 50 mM NaCl and 0.5 mM CaCl2 , with or without 20 mg/l alginate. (a) Force
curves for CA membrane. (b) Force curves for PA membrane. (1 and 2) schematically illustrate
the process of retracting the probe from the CA membrane surface in the presence of alginate
and calcium ions. (3–5) Schematically illustrate the process of retracting the probe from the PA
membrane surface in the presence of alginate and calcium ions. Source: Reprinted from Reference
146, Copyright (2010), with permission from Elsevier. (Please refer to the online version for the
color representation of the figure.)
MEMBRANE CHARACTERIZATION 37

SPR is a spectroscopic method for investigating adsorption and specific interactions of


biomolecules, especially, proteins with surfaces. It is common in biology and biochem-
istry and is increasingly used in membrane science, especially, for assessing benefits
of surface modification as a means to reduce fouling. It has the advantage of being
a surface-sensitive method, which does not require labeling and can produce real-time
kinetics data. More information regarding the method can be found in References 141
and 142 and about new SPR techniques (such as fluorescence SPR or glassified SPR) in
Reference 143. The analysis is conducted using a glass sensor chip covered with a thin
gold layer, in which SPR occurs, that is, collective oscillation of the valence band of
the metal, whose frequency is highly sensitive to the dielectric environment and hence
to adsorption of organic molecules. The gold chip can be functionalized with specific
groups (e.g., via thiol chemistry) and SPR results can be directly correlated with their
affinity to organic foulants and propensity to fouling (144, 145).
Besides its use as a versatile imaging technique suitable for both dry and under liquid
conditions, AFM is a powerful tool for studying interactions between foulants and mem-
brane surfaces, clean or fouled, through direct measurements of the interaction force as a
function of distance between an AFM probe and surface (“force curves”). The measure-
ments are performed using a standard AFM tip or surrogate micrometer-sized particle
mounted on microcantilever and containing different functional groups mimicking the
surface of foulants or bacteria. The particle can also be an actual protein or a bacterium.
Typical force curves demonstrating the effect of fouling on the interaction between a car-
boxylated latex particle (a surrogate of bacteria) to a membrane is presented in Figure 14.

8 CONCLUSIONS AND OUTLOOK

Membrane characterization is an important part of membrane research, development,


and engineering. It provides a crucial link between the preparation and performance
of the membranes and their structure, chemistry, morphology, transport properties, and
other characteristics, with the ultimate goal of understanding how to make the best
membrane and use it in the best way. To keep up with the rapid development of new
types of membranes and a wider range of materials, characterization employs a large
number of techniques that is steadily expanding. This article by no means presents
an entire picture, which would be impossible. Rather, it provides an overview of the
methods and aspects that seem to be most common and relevant in today’s membrane
science. Along with old conventional techniques, many of which are covered here,
highly sensitive high resolution methods are taking the central place and provide
insights down to the molecular level. It may be expected that in future, many methods
that are considered today as too sophisticated or costly may become more affordable
and common in membrane research. Still more methods are expected to appear as a
result of novel developments in other fields of science. They will contribute to a better
understanding of the underlying transport mechanisms, structure-performance relations,
and the development of new membranes with superior performance and selectivity.

REFERENCES

1. Brune D, Hellborg R, Whitlow HJ, Hunderi O. Surface Characterization. Weinheim: Wiley-


VCH; 2008.
38 MEMBRANE CHARACTERIZATION

2. Hoppert M. Microscopic Techniques in Biotechnology. Weinheim: Wiley-VCH; 2006.


3. Oura K, Lifshits V, Saranin A, Zotov A. Surface Science: An Introduction (Advanced Texts
in Physics). Springer; 2003.
4. Cahill DG, Freger V, Kwak SY. MRS Bull 2008;33:27–32.
5. Bartels CR. J Membr Sci 1989;45:225–245.
6. Coronell O, González MI, Mariñas BJ, Cahill DG. Environ Sci Technol 2010;44:6808–6814.
7. Zhang X, Cahill DG, Coronell O, Mariñas BJ. Appl Phys Lett 2007;91:181904–181904.
8. Coronell O, Mariñas BJ, Cahill DG. Environ Sci Technol 2011;45:4513–4520.
9. Silverstein R, Webster F. Spectrometric Identification of Organic Compounds. John Wiley &
Sons; 2006.
10. Stilbs P. Prog Nucl Magn Reson Spectrosc 1987;19:1–45.
11. Koenig JL. Spectroscopy of Polymers. New York: Elsevier Science; 1999.
12. Mirabella FM. Internal Reflection Spectroscopy: Theory and Applications. New York: CRC;
1992.
13. Bernstein R, Belfer S, Freger V. Langmuir 2010;26:12358–12365.
14. Breen AF, Breen C, Clegg F, Döppers L, Labet M, Sammon C, Yarwood J. Polymer
2012;53:4420–4428.
15. Do VT, Tang CY, Reinhard M, Leckie JO. Environ Sci Technol 2012;46:13184–13192.
16. Freger V, Ben-David A. Anal Chem 2005;77:6019–6025.
17. Ben-David A, Oren Y, Freger V. Environ Sci Technol 2006;40:7023–7028.
18. Mallevialle J, des eaux-Dumez L, Odendaal PE, Wiesner MR. Water Treatment Membrane
Processes. USA: McGraw-Hill; 1996.
19. Ridgway HF, Flemming H-C. Biofouling of membranes. In: Mallevialle J, Odendaal PE,
Wiesner MR (eds): Water treatment membrane processes. McGraw-Hill, New York 1996. p
6.1–6.62.
20. Dresselhaus MS, Jorio A, Hofmann M, Dresselhaus G, Saito R. Nano Lett 2010;10:751–758.
21. Sawyer L, Grubb DT, Meyers GF. Polymer Microscopy. New York: Springer; 2008.
22. Freger V, Gilron J, Belfer S. J Membr Sci 2002;209:283–292.
23. Rugar D, Hansma P. Phys Today 1990;43:23–30.
24. Butt H, Cappella B, Kappl M. Surf Sci Rep 2005;59:1–152.
25. Bason S, Freger V. J Membr Sci 2010;360:389–396.
26. Freger V, Korin E, Wisniak J, Korngold E, Ise M, Kreuer K. J Membr Sci 1999;160:213–224.
27. Park GS, Crank J. Diffusion in Polymers. London: Academic Press; 1968.
28. Blum F. Spectroscopy 1986;1:32–38.
29. Freger V, Korin E, Wisniak J, Korngold E. J Membr Sci 2000;164:251–256.
30. Crank J. The Mathematics of Diffusion. Oxford: Clarendon press; 1979.
31. Hajatdoost S, Yarwood J. J Chem Soc, Faraday Trans 1997;93:1613–1620.
32. Freger V, Bason S. J Membr Sci 2007;302:1–9.
33. Barsoukov E, Macdonald JR. Impedance Spectroscopy: Theory, Experiment, and Applications.
Hoboken (NJ): John Wiley & Sons, Inc.; 2005.
34. Sistat P, Kozmai A, Pismenskaya N, Larchet C, Pourcelly G, Nikonenko V. Electrochim Acta
2008;53:6380–6390.
35. Bason S, Oren Y, Freger V. J Membr Sci 2011;367:119–126.
36. Helfferich F. Ion Exchange. New York: McGraw-Hill; 1962.
37. Geise GM, Park HB, Sagle AC, Freeman BD, McGrath JE. J Membr Sci 2011;369:130–138.
38. Yaroshchuk A, Karpenko L, Ribitsch V. J Phys Chem B 2005;109:7834–7842.
MEMBRANE CHARACTERIZATION 39

39. Bason S, Oren Y, Freger V. J Membr Sci 2007;302:10–19.


40. Mulder M. Basic Principles of Membrane Technology. Dordrecht: Kluwer Academic Pub-
lishers; 1996.
41. Yu J, Hu X, Huang Y. Sep Purif Technol 2010;70:314–319.
42. Stawikowska J, Livingston AG. J Membr Sci 2012;413–414:1–16.
43. Cuperus F, Bargeman D, Smolders C. J Membr Sci 1992;71:57–67.
44. Torquato S. Random Heterogeneous Materials: Microstructure and Macroscopic Properties.
New York: Springer; 2001.
45. Bason S, Kaufman Y, Freger V. J Phys Chem B 2010;114:3510–3517.
46. Zydney AL, Xenopoulos A. J Membr Sci 2007;291:180–190.
47. Sarbolouki M. Sep Sci Technol 1982;17:381–386.
48. Mochizuki S, Zydney AL. J Membr Sci 1992;68:21–41.
49. Bechhold H. Z Elektrochem Angew Phys Chem 1907;13:527–533.
50. Nunes SP, Peinemann KV. J Membr Sci 1992;73:25–35.
51. Capannelli G, Vigo F, Munari S. J Membr Sci 1983;15:289–313.
52. Nakao S. J Membr Sci 1994;96:131–165.
53. Germic L, Ebert K, Bouma R, Borneman Z, Mulder M, Strathmann H. J Membr Sci 1997;
132:131–145.
54. Washburn EW. Proc Natl Acad Sci U S A 1921;7:115–116.
55. Eyraud C, Betemps M, Quinson J, Chatelut F, Brun M, Rasneur B. Bull Soc Chim France
1984;9–10:237–244.
56. Mey-Marom A, Katz M. J Membr Sci 1986;27:119–130.
57. Prádanos P, Rodriguez ML, Calvo J, Hernández A, Tejerina F, de Saja J. J Membr Sci
1996;117:291–302.
58. Yasuda H, Tsai J. J Appl Polym Sci 1974;18:805–819.
59. Brun M, Lallemand A, Quinson JF, Eyraud C. Thermochim Acta 1977;21:59–88.
60. Landry MR. Thermochim Acta 2005;433:27–50.
61. Bothun GD, Ni Q, Ilias S. J Membr Sci 2008;325:982–988.
62. Chen H, Hung W, Lo C, Huang S, Cheng M, Liu G, Lee K, Lai J, Sun Y, Hu C. Macro-
molecules 2007;40:7542–7557.
63. Kallus S, Condre J, Hahn A, Golemme G, Algieri C, Dieudonné P, Timmins P, Ramsay J. J
Mater Chem 2002;12:3343–3350.
64. Zornoza B, Tellez C, Coronas J, Gascon J, Kapteijn F. Microporous Mesoporous Mater
2012;166:67–78.
65. Gebel G, Diat O. Fuel Cells 2005;5:261–276.
66. Stawikowska J, Livingston AG. J Membr Sci 2013;425–426:58–70.
67. Freger V. Langmuir 2003;19:4791–4797.
68. Pacheco FA, Pinnau I, Reinhard M, Leckie JO. J Membr Sci 2010;358:51–59.
69. Mi B, Coronell O, Mariñas BJ, Watanabe F, Cahill DG, Petrov I. J Membr Sci 2006;
282:71–81.
70. Ben-David A, Bason S, Jopp J, Oren Y, Freger V. J Membr Sci 2006;281:480–490.
71. Freger V. Environ Sci Technol 2004;38:3168–3175.
72. Schaep J, Vandecasteele C. J Membr Sci 2001;188:129–136.
73. Perry LA, Coronell O. J Membr Sci 2013;429:23–33.
74. Koo J, Petersen R, Cadotte J. Polym Prepr 1986;27:391.
75. Israelachvili JN. Intermolecular and Surface Forces. London: Academic press; 1992.
40 MEMBRANE CHARACTERIZATION

76. Van Oss CJ. Interfacial Forces in Aqueous Media. CRC; 2006.
77. Verliefde ARD, Cornelissen ER, Heijman SGJ, Hoek EMV, Amy GL, Bruggen BV, van Dijk
JC. Environ Sci Technol 2009;43:2400–2406.
78. Merkel T, Freeman B, Spontak R, He Z, Pinnau I, Meakin P, Hill A. Science 2002;296:
519–522.
79. Kobayashi Y, Haraya K, Hattori S, Sasuga T. Polymer 1994;35:925–928.
80. Ghosal K, Freeman BD. Polym Adv Technol 2003;5:673–697.
81. Mohamed H, Russo AP, Szarowski DH, McDonnell E, Lepak LA, Spencer MG, Martin DL,
Caggana M, Turner JN. Sep Sci Technol 2007;42:25–41.
82. Donth E. The Glass Transition: Relaxation Dynamics in Liquids and Disordered Materials.
Springer; 2001.
83. Strathmann H. Ion-exchange Membrane Separation Processes. Amsterdam: Elsevier; 2004.
84. Larminie J, Dicks A, McDonald MS. Fuel Cell Systems Explained . New York: John Wiley
& Sons, Inc.; 2003.
85. Collette FM, Lorentz C, Gebel G, Thominette F. J Membr Sci 2009;330:21–29.
86. Mauritz KA, Moore RB. Chem Rev 2004;104:4535–4585.
87. Bass M, Berman A, Singh A, Konovalov O, Freger V. Macromolecules 2011;44:2893–2899.
88. Bass M, Berman A, Singh A, Konovalov O, Freger V. J Phys Chem B 2010;114:3784–3790.
89. Dura JA, Murthi VS, Hartman M, Satija SK, Majkrzak CF. Macromolecules 2009;
42:4769–4774.
90. Deabate S, Gebel G, Huguet H, Morin A, Pourcelly G. Energy Environ Sci 2012;5:
8824–8847.
91. Kim JY, Lee HK, Kim SC. J Membr Sci 1999;163:159–166.
92. Khulbe K, Matsuura T. Polymer 2000;41:1917–1935.
93. Knoell T, Safarik J, Cormack T, Riley R, Lin S, Ridgway H. J Membr Sci 1999;157:117–138.
94. Kwak S, Woo Ihm D. J Membr Sci 1999;158:143–153.
95. Kwak S, Yeom M, Roh IJ, Kim DY, Kim J. J Membr Sci 1997;132:183–191.
96. Riedl K, Girard B, Lencki RW. J Membr Sci 1998;139:155–166.
97. Hoek EM, Bhattacharjee S, Elimelech M. Langmuir 2003;19:4836–4847.
98. Ba L, Li WS. J Phys D: Appl Phys 2000;33:2527–2531.
99. Bowen W, Doneva TA. J Membr Sci 2000;171:141–147.
100. Herzberg M, Sweity A, Brami M, Kaufman Y, Freger V, Oron G, Belfer S, Kasher R.
Biomacromolecules 2011;12:1169–1177.
101. Koyuncu I, Brant J, Lüttge A, Wiesner MR. J Membr Sci 2006;278:410–417.
102. Vrijenhoek EM, Hong S, Elimelech M. J Membr Sci 2001;188:115–128.
103. Childress AE, Elimelech M. J Membr Sci 1996;119:253–268.
104. Huisman IH, Trägårdh G, Trägårdh C, Pihlajamäki A. J Membr Sci 1998;147:187–194.
105. Kirby BJ, Hasselbrink EF. Electrophoresis 2004;25:187–202.
106. Camesano TA, Logan BE. Environ Sci Technol 2000;34:3354–3362.
107. Poortinga AT, Bos R, Norde W, Busscher HJ. Surf Sci Rep 2002;47:1–32.
108. Redman JA, Walker SL, Elimelech M. Environ Sci Technol 2004;38:1777–1785.
109. Shan X, Huang X, Foley KJ, Zhang P, Chen K, Wang S, Tao N. Anal Chem 2009;
82:234–240.
110. Campbell SD, Hillier AC. Langmuir 1999;15:891–899.
111. Xu S, Arnsdorf MF. Proc Natl Acad Sci U S A 1995;92:10384–10388.
112. Peng F, Hoek EM, Damoiseaux R. J Biomol Screen 2010;15:748–754.
MEMBRANE CHARACTERIZATION 41

113. Bilad MR, Declerck P, Piasecka A, Vanysacker L, Yan X, Vankelecom IF. J Membr Sci
2011;379:146–153.
114. Zhou M, Liu H, Kilduff JE, Langer R, Anderson DG, Belfort G. Environ Sci Technol
2009;43:3865–3871.
115. Li H, Fane A, Coster H, Vigneswaran S. J Membr Sci 1998;149:83–97.
116. Lee H, Amy G, Cho J, Yoon Y, Moon SH, Kim IS. Water Res 2001;35:3301–3308.
117. Hoek EMV, Kim AS, Elimelech M. Environ Eng Sci 2002;19:357–372.
118. Herzberg M, Elimelech M. J Membr Sci 2007;295:11–20.
119. Wang Y, Wicaksana F, Tang CY, Fane AG. Environ Sci Technol 2010;44:7102–7109.
120. Zsirai T, Buzatu P, Aerts P, Judd S. Water Res 2012;46:4499–4507.
121. Krivorot M, Kushmaro A, Oren Y, Gilron J. J Membr Sci 2011;376:15–24.
122. Cho J, Amy G, Pellegrino J. Water Res 1999;33:2517–2526.
123. Zeman LJ, Zydney AL. Microfiltration and Ultrafiltration: Principles and Applications.
New York: Marcel Dekker Inc.; 1996.
124. Thompson J, Lin N, Lyster E, Arbel R, Knoell T, Gilron J, Cohen Y. J Membr Sci
2012;415–416:181–191.
125. Lakowicz JR, Masters BR. J Biomed Opt 2008;13:029901.
126. Eshet I, Freger V, Kasher R, Herzberg M, Lei J, Ulbricht M. Biomacromolecules 2011;12:
1169–1177.
127. Bernstein R, Belfer S, Freger V. Environ Sci Technol 2011;45:5973–5980.
128. Subramani A, Hoek E. J Membr Sci 2008;319:111–125.
129. Herzberg M, Kang S, Elimelech M. Environ Sci Technol 2009;43:4393–4398.
130. Lee W, Ahn CH, Hong S, Kim S, Lee S, Baek Y, Yoon J. J Membr Sci 2010;351:112–122.
131. Miller DJ, Araújo PA, Correia P, Ramsey MM, Kruithof JC, van Loosdrecht M, Freeman
BD, Paul DR, Whiteley M, Vrouwenvelder JS. Water Res 2012;46:3737–3753.
132. Flemming H, Wingender J. Water Sci Technol 2001;43:1.
133. Liu H, Fang HHP. J Biotechnol 2002;95:249–256.
134. Kwon KD, Green H, Bjoeoern P, Kubicki JD. Environ Sci Technol 2006;40:7739–7744.
135. Contreras AE, Steiner Z, Miao J, Kasher R, Li Q. Environ Sci Technol 2011;45:6309–6315.
136. Liu SX, Kim JT. Desalination 2009;247:355–361.
137. Sweity A, Ying W, Ali-Shtayeh MS, Yang F, Bick A, Oron G, Herzberg M. Water Res
2011;45:6430–6440.
138. Marcus IM, Herzberg M, Walker SL, Freger V. Langmuir 2012;28:6396–6402.
139. Kim J, Weber N, Shin G, Huang Q, Liu S. J Food Sci 2007;72:E214–E221.
140. Lin NH, Cohen Y. J Membr Sci 2011;379:426–433.
141. Homola J, Yee SS, Gauglitz G. Sens Actuators B Chem 1999;54:3–15.
142. Homola J. Surface Plasmon Resonance Based Sensors. Berlin: Springer-Verlag; 2006.
143. Phillips KS, Cheng Q. Anal Bioanal Chem 2007;387:1831–1840.
144. Yang Q, Kaul C, Ulbricht M. Langmuir 2010;26:5746–5752.
145. Li G, Cheng G, Xue H, Chen S, Zhang F, Jiang S. Biomaterials 2008;29:4592–4597.
146. Mi B, Elimelech M. J Membr Sci 2010;348:337–345.

You might also like