Karl W.

Böddeker
Liquid Separations with Membranes
Karl W. Böddeker

Liquid Separations
with Membranes
An Introduction to Barrier Interference

With 40 Figures and 9 Tables

123
Karl Wilhelm Böddeker
Groothegen 2e
21509 Glinde
Germany

Library of Congress Control Number: 2007934042

ISBN-13 978-3-540-47451-7 Springer Berlin Heidelberg New York

This work is subject to copyright. All rights are reserved, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilm or in any other way,
and storage in data banks. Duplication of this publication or parts thereof is permitted
only under the provisions of the German Copyright Law of September 9, 1965, in its
current version, and permission for use must always be obtained from Springer.
Violations are liable for prosecution under the German Copyright Law.
Springer is a part of Springer Science+Business Media
springer.com
© Springer-Verlag Berlin Heidelberg 2008
The use of general descriptive names, registered names, trademarks, etc. in this publi-
cation does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for
general use.

Typesetting and Production: LE-TEX Jelonek, Schmidt & Vöckler GbR, Leipzig
Cover: WMXDesign, Heidelberg
Printed on acid-free paper 68/3180 YL – 5 4 3 2 1 0
Preface

When a fluid mixture (liquid, vapor, gas) is forced to traverse a per-

meable partition, such as a membrane or septum, its components
are likely to move at different speed. The ensuing spread in rate and
composition is a barrier separation effect. Unlike equilibrium sepa-
rations, which depend on the thermodynamic condition of the fluid
mixtures alone, barrier separations additionally are subject to speci-
fic interactions of the mixture components with the barrier. While
the thermodynamics of fluid mixtures is predictable and open to
adjustment, barrier interference adds another dimension to the re-
pertoire of separation effects. Exploiting barrier interference is the
challenge of membrane separation science and technology. This
book is about the principles behind.
As membrane processes, barrier separations independently have
acquired their peculiar identities and colorful diversity, each fond-
ling its own tradition, each adhering to its own terminology (down
to defying SI units), each earning different public attention and sup-
port. By way of illustration: Gas leakage through everything inflat-
able has been observed since the time the elemental gases were iden-
tified; gaseous diffusion of UF6 through a microporous barrier, for
better or worse, gave access to nuclear energy; osmotic phenomena,
originally of academic interest to botanists, were engineered into
providing “fresh water from the sea” by reverse osmosis; electro-
membrane processes, again of academic origin, now offer the vision
of a clean mobility based on fuel cells; hemodialysis, starting from
little known beginnings among pharmacologists, may well be viewed
as the most benevolent of membrane processes; evaporation across
a suitable barrier breaks azeotropes and tends to favor the higher
boiling species, aqueous aromas being a case in point; microfiltra-
tion got its start at stabilizing wine through removal of microorgan-
isms, foreshadowing the use of membranes in biotechnology.
VI Preface

Attempting to formally – let alone retrospectively – unify all this

would be a disservice to grown variety. Yet, ignoring the fundamen-
tal kinship limits information exchange, as frequently it has in the
past, and, for no good reason, burdens information dissemination as
in teaching. The feature in common, and ordering principle, is the
formal structure of mass transfer in barrier separation, being con-
strued of a driving force (intrinsic to the fluid mixtures to be separ-
ated) and a permeability (summarily describing the interaction of
the mixture components with the barrier). The plan of this book, ac-
cordingly, is to present the relevant thermodynamic features of fluid
mixtures in contact with semipermeable barrieres, then to apply this
information in deriving the working principles and design require-
ments of individual membrane separation processes. The mem-
branes, by this approach, are introduced by way of the mass trans-
port and selectivity demands which they are to meet, with due
reference to the separation effects which they inspire.
The approach is made specific by examining the information
needed, (a) to interpret a membrane effect (hindsight), or (b) to de-
sign a membrane separation process (foresight). In practical terms,
three independent sources of information are available.
• The thermodynamic condition of the fluid mixtures to be sepa-
rated, both upstream and downstream of the barrier to identify
gradients. Constituting the driving force for mass transfer, this in-
formation translates into the operating conditions of the respec-
tive membrane separation process.
• The barrier (membrane) itself, its chemical nature and physical
characteristic. This information comes from material science,
foremost from polymer science, with additional emphasis on the
art of creating thin films and microporous structures.
• The permeability of the barrier (membrane), which effectively
introduces a barrier selectivity differing from equilibrium selec-
tivity. A conglomerate of many influences, membrane permeabil-
ity acquires meaning only in the context of specific applications.
Independent information on polymer permeability comes from
sorption and diffusion studies (“small molecule meets big mole-
cule”) originally designed to elucidate polymer structure.
 Preface VII

This is not a compilation of expert knowledge, nor a universe of

citations. Rather, an attempt is made to survey, in systematic order,
the terms and concepts by which barrier separations operate, and
through which practical membrane separation processes are de-
signed.
If there is to be a motto: The only meaningful perspective is that of
context.

Karl Wilhelm Böddeker

Technische Universität Hamburg-Harburg
Contents

1 An Introduction to Barrier Separation .................................. 1

1.1 Separation is … .............................................................. 1
1.2 Barrier separation is … ................................................. 3
1.3 Membranes, economy of size and affinity................... 5
1.4 Driving force, actuating barrier interference .............. 8
1.5 Dynamics of barrier separation .................................... 10
1.6 On units and dimensions .............................................. 12
Bibliography............................................................................. 15
References....................................................................... 15
Text books....................................................................... 15
Background reading....................................................... 15

2 The Thermodynamic Connection ........................................... 17

2.1 Mixtures and solutions.................................................. 17
2.1.1 The solute has no vapor pressure ..................... 17
2.1.2 The solute has vapor pressure........................... 19
2.1.3 On thermodynamic activity .............................. 23
2.2 The driving force in barrier separation........................ 25
2.2.1 The chemical potential, no barrier ................... 26
2.2.2 The chemical potential, barrier inclusive......... 27
2.2.3 Chemical potential and separation................... 29
2.3 The master flux equation .............................................. 31
2.3.1 Sorption............................................................... 32
2.3.2 Diffusivity............................................................ 34
2.3.3 Membrane thickness.......................................... 36
Bibliography............................................................................. 37
References....................................................................... 37
Background..................................................................... 37
X Contents

3 Osmosis et cetera ....................................................................... 39

3.1 Osmosis............................................................................ 39
3.1.1 Osmotic investigations ....................................... 39
3.1.2 The law of osmotic pressure .............................. 41
3.1.3 Osmotic pressure illustrated .............................. 43
3.2 Reverse osmosis .............................................................. 46
3.2.1 Solvent flux and solute rejection ....................... 46
3.2.2 Model implications ............................................. 48
3.2.3 Dynamic reverse osmosis................................... 51
3.2.4 Energy considerations ........................................ 54
3.2.5 Reverse osmosis in the real world ..................... 56
3.3 The power of osmosis..................................................... 59
3.3.1 The osmotic pump .............................................. 59
3.3.2 Osmotic power generation ................................. 60
Bibliography.............................................................................. 62

4 Membrane Filtration................................................................. 63
4.1 On size and size exclusion.............................................. 63
4.2 Liquid transport in membrane filtration...................... 66
4.2.1 Concentration polarization................................ 68
4.2.2 Gel polarization ................................................... 71
4.3 Solute transport in membrane filtration ...................... 74
4.4 Rating porous membranes............................................. 76
4.5 Notable applications....................................................... 78
Bibliography.............................................................................. 80

5 Pervaporation versus Evaporation ......................................... 81

5.1 Phenomenon and realization......................................... 81
5.2 Mass transport and selectivity ....................................... 82
5.3 The capability of pervaporation .................................... 87
5.4 Hydrophilic pervaporation ............................................ 88
5.4.1 General observations .......................................... 88
5.4.2 Pervaporation versus reverse osmosis .............. 91
5.5 Organophilic pervaporation .......................................... 93
5.5.1 General observations .......................................... 93
5.5.2 High boiler pervaporation.................................. 96
5.5.3 Butanol, a glimpse at bioseparations ................ 99
 Contents XI

5.6 Pervaporation in perspective ........................................ 103

Bibliography............................................................................. 103
References....................................................................... 103
Reviews............................................................................ 104

6 What Membranes are About.................................................... 105

6.1 Prelude: Collodion membranes .................................... 105
6.2 Membrane polymers – polymer membranes .............. 106
6.3 Like dissolves like........................................................... 109
6.4 Microporous barriers..................................................... 111
Bibliography............................................................................. 112

7 Tracing Membrane Science, an Historical Account ............. 113

Literature cited......................................................................... 119

Appendix 121

A Properties of Aqueous Solutions............................................. 123

B Criteria of Technical Water Quality........................................ 125

Hardness................................................................................... 125
Alkalinity .................................................................................. 126
Corrosiveness........................................................................... 127

C Marker Molecules ...................................................................... 129

Synthetic macromolecules...................................................... 130

D Membrane Polymers ................................................................. 131

Natural polymeric materials................................................... 131
Synthetic polymers .................................................................. 131
Synthetic copolymers (elastomers)........................................ 132

E Microporous Structures............................................................ 133

Subject Index........................................................................................ 141

Symbols and Abbreviations

Symbols
ai activity of i [concentration]
bp. boiling point [°C]
c concentration (generic) [various units]
ci mass concentration [kg/m3]; [mol/m3]
D diffusion coefficient [m2/s]
d diameter; pore diameter [m]; [μm]
E recovery (Entnahme) [%]
G Gibbs free energy [J/mol]
g acceleration of gravity [9.81 m/s2]
gfd gallons per square foot & day [flux]
J mass flux (flow density) [kg/s m2]; [mol/s m2]
Jp permeance [kg/s m2 bar]
Jv volume flux (practical) [L/h m2]
J (Joule) energy; work (Nm) [Ws]; [kWh]
k mass transfer coefficient [m/s]
L permeability (Leitfähigkeit) [various units]
Lp hydraulic permeability [various units]
m molality of solute [mol/kg]
n number of mols [–]
P total pressure [Pa]; [bar]
p pressure [Pa]; [bar]
pi partial pressure of i [Pa]; [bar]
pi° pure component vapor pressure [Pa]; [bar]
ppm parts per million [mg/L]
psi pounds per square inch [pressure]
Q flow rate [L/h]
R gas constant [8.314 J/mol K]
R (x 100) retention; rejection [0–1] or (%)
S sorption coefficient [various units]
XIV Symbols and Abbreviations

T temperature [°C]
Tg glass transition temperature [°C]
V volume [m3]; [L]; [mL]
Vi partial molar volume [m3/mol]
W (Watt) power (rate of work) [J/s]
wi weight fraction of i [kg/kg]
xi mol fraction of i (feed) [mol/mol]
yi mol fraction of i (permeate) [mol/mol]
z distance coordinate [m]
z membrane thickness [μm]

Greek
αij separation factor [–]
βi enrichment factor [–]
γi activity coefficient [–]
δ thickness of polarization layer [μm]
δ solubility parameter [(cal/cm3)0.5]
ε porosity (surface or volume) [–]
η viscosity [kg/s m]
μi chemical potential of i [J/mol]
π osmotic pressure [bar]
ρ mass density [kg/m3]
σ reflection coefficient [–]
τ tortuosity factor [–]

Indices
i, j components i, j
1, 2 components 1 = solvent; 2 = solute
' feed; feed side
" permeate; permeate side
° standard or reference state
b bulk (well mixed feed)
liq liquid
m (as index) membrane; membrane phase
org organic
 Polymer notation XV

p (as index) permeate

v (as index) volume
vap vapor
w wall (interface)
w water

Abbreviations
ABE acetone-butanol-ethanol
CA cellulose acetate (generic)
CED cohesive energy density
ED electrodialysis
HD hemodialysis
HF hollow fiber
IX ion exchange
MBR membrane bioreactor
MD membrane distillation
MF microfiltration
MW molecular weight
NF nanofiltration
NOM natural organic matter
PA polyamide (generic)
PRO pressure retarded osmosis
PV pervaporation
RO reverse osmosis
SDI silt density index
SLM supported liquid membrane
SW seawater; spiral wound
TDS total dissolved solids
UF ultrafiltration
VLE vapor-liquid equilibrium
VOC volatile organic compound

Polymer notation
See Appendix D.
1 An Introduction to Barrier Separation

1.1 Separation is …
Separation is the key to the uses of nature. – Gathering, harvesting,
mining are elementary manifestations of selection, typifying the ob-
jective of all separation, which is added value to the product pro-
cured.
Generic categories of separation are:
• Enrichment, enhancing the proportion of a target component;
• Isolation, recovering a target product from unwanted material;
• Extraction, same when employing a liquid extractant;
• Depletion, refers to the target product in the residue of isolation;
• Purification, removing impurities from the wanted product;
• Refining, purification in specific industries or circumstances;
• Fractionation, dividing into components or component groups;
• Phase separation, parting into mutually immiscible liquid phases;
• Precipitation, rendering a solution component insoluble;
• Volume reduction, concentrating dissolved species by removal of
solvent;
• Dehydration, concentrating foods and biomass by removal of
water.
Membranes are instrumental in many of these.
As to technical categories of separation, King [1] lists 54 separa-
tion processes in 11 categories. Different separation processes often
are applied to the same separation task, the merits of one approach
then having to be assessed in comparison to others. As membrane
processes, barrier separations add to the inventory of separation
science, showing specific advantages in some applications (for ex-
ample hemodialysis; azeotrope splitting; bioseparations; ultrapure
water; fuel cells), while competing on equal terms with traditional
2 1 An Introduction to Barrier Separation

Fig. 1.1. The Sherwood plot: Selling prices of materials correlate with their
degree of dilution in the initial matrix from which they are being separated.
Taken from [2].

processes in many others. As a conspicuous example, membrane

processes compete with distillation in water demineralization.
More often than not, separation is focused an the minority com-
ponent(s) of mixtures: As wanted product to be recovered from
a low-valued matrix or, conversely, as impurity to be removed to
upgrade the matrix. In either mode, the expenditure to separate the
minority component increases with dilution; dilution, in turn, in-
creases with depletion. Specifically, recovery of valuable solutes from
dilute liquid solutions is dominated by the cost of processing large
masses of unwanted solvent. Sherwood plots illustrate a linear corre-
lation between selling price of materials and their degree of dilution
in the initial matrix when presented on a logarithmic scale (Fig. 1.1).
Solvent removal from dilute solutions by membrane filtration effec-
tively leads to solute enrichment, but just as well may serve as
a means to purify the solvent.
The mechanism of separation is mass transfer. – Any mass trans-
fer operation which produces a change in composition of a given
feed mixture without permanently altering the identity of its com-
ponents inherently is a separation. Any such operation yields – at
 1.2 Barrier separation is … 3

least – two product mixtures which differ in composition from one

another and from the original feed. If one of the products is consid-
ered the target fraction of the separation, the other, by necessity, is
the original feed devoid of the target fraction. The separation effect
or selectivity of the process is assessed by comparing the analytical
composition of the two products, or by relating the composition of
either one of the products to that of the original feed. The objective
of separation process design usually is to render one of the products
as pure as possible.
Separation is demixing. – Selective mass transfer within a multi-
component system enhances the degree of order, counteracting the
natural tendency to uniform mixing, and thus requires energy. Ac-
cording to the thermodynamics of mixtures, the minimum energy to
isolate a pure component species from a mixture or solution is pro-
portional to (− ln xi), where xi is the mol fraction of that species in
the feed mixture (Sect. 2.2.3). In terms of ordinary concentration,
this proportionality is the reference coordinate of the Sherwood plots
depicting cost of product recovery as function of initial product
concentration. Conversely, the minimum energy to recover pure sol-
vent from a given solution increases in proportion to solute concen-
tration, affirming that the solute disturbs the thermodynamic condi-
tion of the solvent. Actual energy requirements may exceed the
theoretical minimum by an order of magnitude, providing ample
incentive for separation process development.

1.2 Barrier separation is …

Barrier separation is rate controlled mass transfer. – Barrier separa-
tions rely on mass transport across semipermeable physical parti-
tions, selectivity coming about by differences in permeability of the
barrier towards the feed components resulting in the rates of mass
transfer to differ. The operative distinction of rate governed versus
equilibrium separation is dynamics: Mass transfer through a barrier
is slowed by molecular interaction with the barrier matrix (figura-
tively viewed as friction on a molecular level), and likely is affected
by encounters between the permeating species en route (loosely
referred to as coupling); this is the essence of barrier interference. By
comparison, mass transfer across a liquid-vapor interface (VLE
4 1 An Introduction to Barrier Separation

= vapor-liquid equilibrium) is considered instantaneous, and inter-

action between the vaporized species is negligible.
The two generic products of barrier separation are the permeate
(= the fraction transported through the barrier), and the retentate
(= the fraction retained or rejected by the barrier), Fig. 1.2. Although
either one may be the target fraction of the process, analysis of bar-
rier separation is by relating the permeating fraction to the feed,
thereby registering the influences of barrier interference and process
conditions. Feed components present within the barrier at any time
are the permeants (penetrants to some).
The term semipermeable membrane was introduced by van’t Hoff
(1887) [3], originally denoting an ideal barrier permeable to solvent
(water) only while being completely impermeable to dissolved spe-
cies (Sect. 3.1.2). Such a membrane would stabilize the osmotic equi-
librium between a liquid solution and its own pure solvent. The
contrary limit is a freely permeable, nonselective barrier yielding
a permeate identical in composition to that of the feed, – in effect
a throttle. Real barriers, even though selected or designed for high
selectivity, are “leaky” in that, in principle, they are permeable to all
species encountered. The ultimate state of a system of fluid mixtures
in contact with any real membrane would be complete uniform mix-
ing, if only one would wait long enough. There is thus no absolute
barrier separation on two counts: The process is self-quenching, the
energy to remove the selectively permeating species increasing with
depletion; and, real membranes are leaky.
The earliest barrier on record is a section of moist pig’s bladder
stretched over the mouth of medicine bottles before cork stoppers
came into use, hence the terms membrane and, in due course, mem-

Fig. 1.2. Pictograph of a barrier separation stage.

 1.3 Membranes, economy of size and affinity 5

Fig. 1.3. A sketchbook impression of Nollet‘s chance discovery of semiper-

meability: Water entering a membrane-capped vial containing “spirit of
wine” creates pressure (courtesy Anne Böddeker).

brane process. Selective permeability came as a surprise to Nollet

(1748) when he discovered that pig’s bladder is more permeable to
water than to “spirit of wine” (ethanol), resulting in a pressure phe-
nomenon seemingly out of nowhere (Chapter 7).

1.3 Membranes, economy of size and affinity

Membranes are defined by what they do, rather than what they are. –
Nature and man’s ingenuity provide an abundant variety of barrier
materials, both organic and inorganic, having the capacity of being
permeable to individual fluids (liquids, vapors, gases), and semi-
permeable (selectively permeable) to fluid mixtures (Appendices D
and E). The models describing membrane mass transport seek to
relate structure and function of the barriers, reducing the material
variety to a few phenotypes as follows.
Porous barriers, operating on size discrimination, conform to the
notion of “filters”: The solvent moves more or less freely, dissolved
species are discriminated upon. The criterion distinguishing mem-
brane filtration from ordinary (particle) filtration is solute size,
smaller solutes requiring narrower pores to be retained. While grav-
ity is all it needs to drive ordinary filtration, narrow pores require
6 1 An Introduction to Barrier Separation

a pressure head to overcome the hydraulic resistance of the pore

structure; the filtration spectrum of pressure versus solute diameter,
shown in Fig. 4.2, covers the operative range of membrane filtration.
Since specific solute species tend to be more uniform in size (respec-
tively mass) than is met by the pore size distribution of most porous
membrane materials, membrane characterization is in terms of re-
jection functions with respect to given solute size (Sect. 4.4). A pore
size commensurate with a solute diameter of 0.2 μm is noteworthy in
that it nominally excludes bacteria from water by microfiltration.
Gaseous diffusion through porous (inorganic or metallic) barriers
follows a different mechanism, being governed by considerations of
pore geometry versus mean free path (= pressure) of the gas or
gaseous mixture components.
A survey of microporous structures is presented in Appendix E.
Homogeneous barriers (nonporous or “dense”) discriminate ac-
cording to relative solubilites and diffusivities of the feed compo-
nents in the membrane phase. Unlike porous barriers, solution-
diffusion type barriers rely on specific interactions of the permeants
with the membrane material, its chemistry and molecular morph-
ology. With a view at performance, more than on principle, mem-
brane polymers (Appendix D) are assigned to one or both of the
following categories,
• as glassy (crystalline) versus rubbery (elastomeric) by physical
nature,
• as hydrophilic versus hydrophobic by interactive preference.
Attempting for guidance in diversity, glassy polymers generally
show lower permeability and higher selectivity than rubbery ones.
Liquid (aqueous-organic) separations are dominated by the sorption
capacity of the membrane polymers, attended by swelling. By sorp-
tion preference, glassy polymers are hydrophilic, responding to
water as being the smallest of liquid molecules at room temperature,
whereas rubbery polymers tend to be organophilic (Sect. 6.2).
In practical gas separation, sorption of gases into polymers being
low, the higher diffusive selectivity of glassy polymers outweighs the
higher permeability of rubbery ones.
Liquid membranes function as solution-diffusion barriers, pro-
viding the very high diffusivity to permeants characteristic of the
liquid state. Consequently, selective mass transport is expected to be
 1.3 Membranes, economy of size and affinity 7

governed by the rules of solute distribution (partition) between im-

miscible liquid phases in contact. Facilitated transport makes use of
mobile carriers incorporated in the liquid membrane phase to pro-
vide species-specific selectivity.
Functionalized membranes, adding chemistry to polymer science,
attempt to modify the barrier as a whole or the barrier surface to
facilitate selective sorption, or else to counteract undesired mem-
brane fouling by chemical means. A prime objective is to convey
hydrophilicity to membranes used in aqueous separations, notably
to reduce fouling by proteins. Another objective is resistivity to-
wards oxidizing agents, chlorine in particular, widely employed to
disinfect feed waters in water treatment.
A category of functionalized membranes of their own are charged
membranes coming as anion exchangers (positive fixed charges) and
cation exchangers (negative fixed charges). As “immobilized electro-
lytes”, charged membranes are anticipated to be highly hydrophilic.
When employed in electrodialysis, mass transport, pertaining to
charged species only, is by combined action of ionic conduction and
Donnan exclusion under the driving force of an electric potential.
The role of water. Water is a key component in liquid barrier
separation, as is water vapor in gas separation. Not surprisingly, the
presence of water within a membrane is a telltale piece of informa-
tion on the nature of that membrane. With reference to the above
phenotypes:
• Water in porous barriers is pore fluid. Indeed, as long as water
sorption by the membrane (polymer) material itself is negligible,
the difference in weight between “wet” and “dry” should equal the
void space within the membrane structure (then termed volume
porosity as against surface porosity, Sect. 4.2). Mass transport of
solutes smaller than pore dimension is by convection (as in mem-
brane filtration) and/or by diffusion within the pore fluid (as in
dialysis). Even though mass transport is confined to the pores, the
nature of the polymer matrix does matter. For example, a hydro-
phobic porous barrier like a microporous PTFE (Teflon) mem-
brane may prevent liquid water to enter but will allow water vapor
to pass (as in membrane distillation – and breathable textiles).
• Water absorbed (dissolved) by homogeneous polymers may be
considered as a molecular solute in a polymeric solvent, causing
8 1 An Introduction to Barrier Separation

the polymer to swell. Sorption capacity depends on the relevant

interactive forces (hydrogen bonds and polarity, Sect. 6.3), but
also on the “stiffness” of the polymer matrix (glassy versus rub-
bery) resisting polymer swelling. As an orientational aid, the
dense salt rejecting layer of a composite hydrophilic membrane as
employed in water desalination by reverse osmosis typically con-
tains 10% of dissolved water, whereas the porous support of such
a membrane may have a “porosity” (water as pore fluid) exceed-
ing 60%.
• Charged polymers (ion exchange membranes), by both their fixed
charges and mobile counter ions, provide ample ion-dipole at-
traction for water storage. With up to 30% of water their consis-
tency is that of a swollen gel with restricted water mobility. How-
ever, when modeling solute mass transfer (as of ions in electro-
dialysis), ion exchange membranes are pictured as porous with
the charges lining the pore walls.

1.4 Driving force, actuating barrier

interference
Maxwell’s demon needs help. – Next to the membranes, agents of
barrier separation are the operating conditions which provide the
driving force for selective mass transport
• against the inherent resistance of any mixture to demixing (this is
where Maxwell’s demon comes in);
• against the cohesive energy of fluid mixtures (this is where mo-
lecular interaction comes in);
• against the dynamic (transport) resistance of the barrier (this is
where barrier interference comes in).
In form of the respective gradients, the driving force is composed
of the very same variables which describe the thermodynamic condi-
tion of the fluid mixtures contacting the membrane, – temperature,
pressure, and composition. Between them, these intensive properties
(independent of total mass) constitute the Gibbs free energy or free
enthalpy of the mixture (G). The free energy of any individual mix-
ture component, its partial molar free energy, after Gibbs is named
the chemical potential of that component species within the mixture
 1.4 Driving force, actuating barrier interference 9

(μi). It becomes manifest as change in free energy of the mixture as

the concentration of the component under consideration varies, as,
for example, upon its removal in a separation process.
In actual practice, there is no need to explicitely include a tem-
perature gradient among the driving forces since barrier separations
for the most part are isothermal, usually operating at ambient (in-
cluding bio-ambient) temperature. A case of exception is membrane
distillation, which requires a thermal gradient across the porous
barrier. – Likewise, an electrochemical potential is not included in
the general treatment, electromembrane processes being confined to
a class entirely of their own [12]. – On the whole, therefore, the rele-
vant driving forces in barrier separation derive from pressure and
composition of the fluid mixtures to be separated.
Pressure is the “natural” driving potential in all filtration opera-
tions, which are characterized by preferential transport of solvent
(water) over solute, hence the nominal inclusion of reverse osmosis
as “hyperfiltration” (the common expression “desalination by re-
verse osmosis” is misleading, “dewatering” is called forth). The up-
per reach of pressure encountered in membrane filtration is 100 bar
(10 MPa); at this pressure ordinary liquids are incompressible, how-
ever, porous or swollen polymers are not, neither are microorga-
nisms. – Gas permeation through, and gas separation by, homogene-
ous polymer membranes likewise is pressure driven, as is gaseous
diffusion across microporous barriers.
Akin to pressure, vapor pressure is a driving force in barrier sepa-
ration. Depending on how the vapor pressure gradient is created, the
relevant membrane processes are:
• Membrane distillation, the only membrane process operating on
a temperature gradient between liquid feed and liquid permeate.
The membrane is a porous hydrophobic (water-repellent) barrier
permeable to water vapor only; water transport is by evaporation
into the pore space followed by re-condensation on the permeate
side. – In osmotic distillation the vapor pressure gradient is cre-
ated, not by temperature, but by a difference in solute concentra-
tion, a high solute concentration creating a vapor pressure “sink”
on the permeate side, irrespective of the nature of the solute used
(Sect. 2.1.1). Gentle dehydration is the usual objective of both
process variants.
10 1 An Introduction to Barrier Separation

• Pervaporation is a hybrid, operating on a drastic reduction of

vapor pressure (of partial pressures in case of volatile mixtures)
by causing the permeants to evaporate as they emerge from the
membrane. In effect, pervaporation may be viewed as nonequilib-
rium vacuum distillation across interacting (solution-diffusion
type) barriers, usually applied to “difficult” liquid separations:
Separation of narrow boiling or constant boiling (azeotropic)
mixtures; separation of high boiling organics from aqueous solu-
tion (Sect. 5.3).
Composition. While pressure as driving force for mass transport
conforms to intuition, concentration gradients do not. In fact, na-
ture’s urge to establish and maintain uniform mixing within fluid
mixtures at all cost represents a powerful driving force for mass
movement. It is a virtual force, it is the motor of diffusion. If, given
a concentration imbalance, diffusive mixing is intercepted by a per-
meable barrier, mass flows will adjust themselves predictably to the
permeability situation:
• A porous membrane will allow “small” solute species (including
the solvent itself) to equilibrate more or less freely while retaining
macromolecules. This is the operating principle of dialysis, hemo-
dialysis as an example. – Electrodialysis is a namesake in that it,
too, relocates the solute.
• With a homogeneous (“dense”) membrane, if at all permeable to
solvent (water), there is only one way to comply with nature’s call
to mitigate concentration differences: By allowing water to cross
from the dilute to the concentrated side of the membrane. This is
the phenomenon of osmosis (Sect. 3.1.2).

1.5 Dynamics of barrier separation

Mass transport is molecular motion with a directional bias. – It is
slow motion, as a simple calculation will illustrate: At a throughput
(flux) of 1000 L/d m2 (low for ultrafiltration, high for reverse osmo-
sis) the apparent linear velocity of mass transport within the mem-
brane is about 4 cm/h or little more than 10−3 cm/s. To be sure, ex-
cept for revealing a net relocation, this is no information on the
 1.5 Dynamics of barrier separation 11

actual random motion of the permeants in the membrane phase

(which is a subject of molecular modeling).
Performance. – The formal relation between mass flux and driv-
ing force has the structure of a generalized Ohmic law: Flux is pro-
portional to driving force. The coefficient of proportionality (a re-
ciprocal resistance in the Ohmic analogy) has two meanings depen-
ding on how the driving force is introduced:
• It is a permeability when flux follows a gradient of the potential;
by confining the gradient to within the membrane boundaries
(“difference approximation”, Sect. 2.2.2), membrane thickness
becomes part of the permeability format.
• It is a permeance when, for a given membrane, the causality be-
tween observed flux and applied potential (as pressure or indi-
vidual feed concentration) matters; it is thereby a record of per-
formance.
Flux = Permeability × Potential gradient
Flux = Permeance × Potential
Permeability characterizes the transport capability of the barrier
material itself; it thus allows for membrane material evaluation. The
permeance of a given membrane (sometimes called its “productiv-
ity”) is the experimentally observed flux as function of operating
conditions (see Figs. 3.3; 4.4; 5.3; 5.6). If the thickness of the mem-
brane is known, permeability and permeance correspond, perme-
ability appearing as thickness-normalized permeance.
Barrier separations coming about through differences in trans-
port rate of the permeants, the ratio of individual permeabilities (or
permeances) suggests itself as a measure of the separation effect:
Selectivity (ij) = Permeability i (high) ⁄ Permeability j (low)
While this relation is formally correct, it is no recipe to estimate,
much less to predict practical membrane separations, for two reasons:
Individual (single component) permeabilities often are inaccessible
(imagine pure salt permeability); if they are, their numerical ratio
misjudges the interactions which make barrier separations interest-
ing. It is only with true (“permanent”) gases that the ratio of pure
component permeabilities, individually established, quantitatively
predicts the separation effect (then referred to as ideal separation).
12 1 An Introduction to Barrier Separation

Nevertheless, where accessible, single component permeability

(or permeance) provides information on the intrinsic transport be-
havior of the barrier; pure water permeability of microporous mem-
branes, in particular, is a key criterion in membrane filtration.
Concentration polarization. – The most influential effect of proc-
ess dynamics on rate-governed separations by far. Referring to a gra-
dient in composition within the feed phase next to the membrane
surface, concentration polarization is a consequence of the slower
permeating feed component accumulating near the solution-
membrane interface as the faster permeating component moves on.
As a result, the feed mixture as “seen” by the membrane differs in
composition from the bulk feed, aggravating the separation task. If it
is the solvent to permeate preferentially (as in reverse osmosis and all
membrane filtrations), the solute being retained, concentration po-
larization requires conditions to be adjusted to a higher than bulk
solute concentration. Conversely, if the solute or minority compo-
nent permeates preferentially (as in pervaporation and dialysis),
solute depletion near the membrane boundary effectively causes
a lower than bulk concentration. lt is to alleviate these effects that
barrier separations almost always operate in the cross flow (tangen-
tial flow) mode, to be contrasted with dead end filtration.
Concentration polarization is a phenomenon to be reckoned with
in liquid barrier separations. In the limit of perfect mixing of the feed
components, as is generally the case when handling gas mixtures, the
effect is irrelevant.
Whereas concentration polarization is a boundary layer effect
readily rationalized, the mutual influence of permeating species on
their transport behavior, referred to as coupling, is not easily predic-
ted and needs case by case attention. By tendency, coupling would
be expected to impair selectivity by leveling differences in mobility
of the permeants, – reminiscent of the individual freedom of ions in
solution being restricted by the condition of electroneutrality.

1.6 On units and dimensions

Permeability and selectivity are categories of performance rather
than units by themselves. Reduction to practical needs is by identify-
ing the parameters involved, both by their physical meaning and
 1.6 On units and dimensions 13

by dimension, then assigning appropriate units to the parameters

identified.
It is noted that true SI units (the system dating back to 1960), be-
sides not being universally accepted, rarely answer the needs of
practical separation processing. Examples for unwieldy SI units
are: Pascal (Pa) for pressure [replaced in this text by bar;
1 bar = 105 Pa = 0.1 MPa]; second (s) for time [in most cases replaced
by hour (h) or day (d)]. Both kg (for “mass”) and mol (for “amount
of substance”) are SI base units; yet, a mol of a specified substance is
still a mass to be expressed in kg/mol. As an aside it is observed that
industrial output is not normally reported in mols of product, – and
if so, it would have to be number of mols (n) which, when multiplied
with the respective molecular weight, is a true mass again (kg).
In the following some key parameters of barrier separation are
discussed, using SI base units and hinting at SI derived units. It is
noted that volume, a preeminent parameter in fluid mass transfer, is
3
not a base unit in the SI system, although m and L (liter) belong as
SI derived.
Flux (J) [kg/s m2] or [mol/s m2]. – Flux is the quantity of permeant
collected in a time (flow rate) at given membrane area, hence a flow
density by dimension. Total flux in multiple component permeation
is the sum of individual fluxes, established retrospectively by analyz-
ing the permeate composition. Adaption to practical units, including
to volume flux, is self-evident; for example, the common flux unit
[L/h m2] passes as SI derived. Dimensionally reducing a volume flux
to a velocity [m3/s m2 → m/s], except for implementing mass transfer
coefficients (Sect. 4.2), in most cases distracts from the physical mea-
ning of the compound unit.
Permeability (L). – Permeability has many faces, all of the same
dimensional configuration: Flux as function of driving force.
• The driving force for each component is a gradient of its chemical
potential in terms of pressure or concentration (Sect. 1.4), hence
the SI unit [kg m/s m2 bar] when considering pressure-driven
processes.
• Phenomenologically, permeability covers the sequence of events
as a permeating mixture component makes its way from bulk feed
into membrane (sorption) and thence across the membrane (dif-
fusion), boundary layer influences and coupling effects inclusive;
14 1 An Introduction to Barrier Separation

it is thus a record of barrier interference. – Gas permeation is

characterized by a low level of molecular interaction; individual
gas permeabilities are still recorded in Barrer units as a semi-
standard (using “cmHg” for pressure).
• Liquid permeation through porous membranes (as in membrane
filtration) is described as hydraulic permeability (Lp); it is convec-
tive – as opposed to diffusive – volume flux ( Jv) driven by a hy-
draulic pressure gradient [bar/m]. Pure water hydraulic perme-
ability is one of the parameters characterizing a porous mem-
brane. Analysis of hydraulic permeability, true to the Ohmic law
analogy, is in terms of the resistance of the barrier to liquid trans-
port; solute deposited on the membrane surface adds to the over-
all resistance (gel polarization, Sect. 4.2.2).
Permeance. – Rather than to a potential gradient, permeance re-
lates the flux to the potential itself, to pressure or concentration of
the permeating species. When referring to a constant pressure as
driving force, permeance appears as pressure-normalized flux,
2
[kg/s m bar] in SI units. Concentration-normalized flux (having the
dimension of a mass transfer coefficient), besides applying to con-
trolled laboratory conditions, refers to separations at constant com-
position feed supply (seawater, for example). In batch operation,
which is identical to plant operation under conditions of recovery,
there is a methodical concentration dependence of flux instead
(Sect. 2.2.2). – A decidedly non-SI unit of permeance is the concoc-
tion [gfd/psi], encountered in water treatment (refer to list of abbre-
viations).
Since sorption is prerequisite to solution-diffusion governed mass
transfer, a correspondence between permeance and sorption iso-
therms (Sect. 2.3.1) is anticipated.
Selectivity. – Selectivity is a statement of separation performance
based on a comparison of analytical compositions of feed (“bulk”)
and permeate. Practical needs dictate which form is used to express
selectivity (Sect. 5.2 has examples). Intrinsic selectivity refers to the
true separation capability of the barrier under undisturbed condi-
tions, – absence of concentration polarization in particular.
 Bibliography 15

Bibliography
References
[1] C. J. King: Separation Processes. Second edition, McGraw-Hill Book Com-
pany, New York, 1980.
[2] Separation & Purification, Critical Needs and Opportunities. National Re-
search Council, National Academy Press, Washington, 1987.
[3] J. H. van’t Hoff, loc. cit. Chap. 7 Ref. [9].

Text books
[4] M. C. Porter: Handbook of Industrial Membrane Technology. Noyes Publi-
cations, Park Ridge, New Jersey, 1990.
[5] W. S. Ho, K. K. Sirkar (eds.): Membrane Handbook. Van Nostrand Rein-
hold, New York, 1992.
[6] J. A. Howell, V. Sanchez, R. W. Field (eds.): Membranes in Bioprocessing,
Theory and Applications. Chapman & Hall, London etc., 1993.
[7] R. D. Noble, S. A. Stern (eds.): Membrane Separations Technology, Princi-
ples and Applications. Elsevier, Amsterdam etc., 1995.
[8] M. Mulder: Basic Principles of Membrane Technology. Second edition,
Kluwer, Dordrecht, 1996.
[9] M. Cheryan, Ultrafiltration and Microfiltration Handbook, Technomic
Publishing Company, Lancaster, Pa., 1998.
[10] T. Melin, R. Rautenbach: Membranverfahren, Grundlagen der Modul- und
Anlagenauslegung. Second edition, Springer-Verlag Berlin Heidelberg,
2004.
[11] R. W. Baker: Membrane Technology and Applications. Second edition,
Wiley, Chichester, West Sussex, England, 2004.
[12] H. Strathmann: Ion-Exchange Membrane Separation Processes. Elsevier,
Amsterdam etc., 2004.

Background reading
[13] V. Stannett, The transport of gases in synthetic polymeric membranes: An
historical perspective. J. Membrane Sci. 3 (1978) 97–115.
[14] S. Loeb, The Loeb-Sourirajan membrane, how it came about. In: Synthetic
Membranes (A. F. Turbak, ed.), Vol. 1, American Chemical Society, Wash-
ington, 1981.
[15] H. K. Lonsdale, The growth of membrane technology. J. Membrane Sci. 10
(1982) 81–181.
[16] E. N. Lightfoot, M. C. M. Cockrem, What are dilute solutions? Sep. Sci.
Technol. 22 (1987) 165–189.
16 1 An Introduction to Barrier Separation

[17] E. A. Mason, From pig bladders and cracked jars to polysulfones: An

historical perspective on membrane transport. J. Membrane Sci. 60 (1991)
125–145.
[18] W. J. Kolff, The beginning of the artificial kidney. Artif. Organs 17 (1993)
293–299.
[19] K. S. Spiegler, Y. M. El-Sayed: A Desalination Primer, Introductory Book for
Students and Newcomers to Desalination. Balaban Desalination Publica-
tions, Rehovot, 1994.
2 The Thermodynamic Connection

2.1 Mixtures and solutions

A mixture, in dictionary parlance, is a commingling of two or more
substances in varying proportion in which the components retain
their individual chemical identity. Solutions, for the purpose of fluid
separations, are homogeneous mixtures of solid, liquid or gaseous
solutes in a liquid solvent. With uneven mixtures of two or more
miscible liquids, the majority component is considered the solvent,
the minority component(s) assuming the role of the solute. Even
(equimolar) mixtures are exceptional.
The behavior of liquid solutions is governed by molecular inter-
actions: Solvent-solute, solvent-solvent, and solute-solute. It is these
interactions which separation has to deal with. In the following it is
appropriate to distinguish between two types of solution behavior,
• the solute has no vapor pressure;
• the solute is itself volatile.
Water is the common solvent. Old alembic teaching has it that no
solute will be found in the steam evolving from a boiling aqueous
solution once that solute boils higher than water by upwards of
130°C. Polymer-solvent and polymer-solute interaction changes all
that: Evaporation across a polymeric membrane (pervaporation,
Sect. 5.5.2) will enrich even vanillin having a normal boiling point of
285°C from its aqueous solution, testifying to barrier interference.

2.1.1 The solute has no vapor pressure

Solutions of this type are aqueous electrolyte solutions including
solutions of inorganic and higher organic acids, sugar solutions (of
Pfeffer’s osmotic cell fame, Sect. 3.1.1), but also true to colloidal
18 2 The Thermodynamic Connection

solutions of macromolecules from proteins to microorganisms. Ta-

ble 2.1 presents the range of molecular mass encountered in medical
membrane use. Solubility, though ranging widely with molecular
mass, always has an upper limit, solvent removal (“volume reduc-
tion”) invariably leading to saturated, occasionally super-saturated
solutions. True saturation would require contact with precipitated
solute, a situation not normally attained with dissolved macromole-
cules (Sect. 4.2). If there is a solubility limit to common salts in
water, it is because “free” water to effect ion hydration is no longer
available: The Dead Sea, rated at 26% total dissolved solids (TDS)
and lined with mineral precipitate, has the gluey consistency of
glycerine.

Table 2.1. Molecular mass of nonvolatile solutes used for in-vitro clearance
studies to characterize hemodialysis membranes (Sect. 4.4). After Gerner.

Solute Molecular mass

g/mol
sodium chloride 00058 small molecules
urea 00060
creatinine 00113
uric acid 00168
glucose (dextrose) 00180

sucrose 00342 middle molecules

EDTA 00380
raffinose 00504
vitamin B12 01355
inulin 05200

β2 microglobulin 11800 large molecules

cytochrome C 13000
hemoglobin 68000
serum albumin 69000
 2.1 Mixtures and solutions 19

What is observed when nonvolatile solutes are dissolved in water

are the colligative properties, which are recognized as deviations of
solution properties from those of the pure solvent, namely
• lowering of vapor pressure;
• elevation of boiling point;
• lowering of freezing point;
• increase of osmotic pressure.
Strictly speaking, it is solvent properties which are affected. The
colligative properties are number effects, depending on the molar
concentration of dissolved species – ions in case of electrolytes –
irrespective of their kind (and thus may be drawn upon to determine
solute molecular weight). It is noted in advance that number effects
rely on the statistical presence of the mixture components, whereas
activity effects moreover account for molecular interactions between
solute and solvent (Sect. 2.1.3).
While the first three of the colligative properties can be measured
directly, osmotic pressure requires a semipermeable membrane to
become evident. The observed effects tend to diminish with increas-
ing molecular mass mainly because larger solute species tend to have
lower solubility and thus are less “numerous”. Small solutes to about
MW 500 are viewed as “osmotically relevant” (Sect. 3.1.3), most
common salts belonging into this category.
Considering barrier separations, solvent osmotic pressure due to
the presence of nonvolatile solutes matters in osmosis and reverse
osmosis. In osmotic distillation, a gradient in vapor pressure is gen-
erated by deliberate action of nonvolatile solutes. That action also
accounts for the salting-out effects of organic chemistry (Sect. 5.5.2).
The complex solutions of biochemical origin often contain both
electrolytes and macromolecules, the separation task typically being
to demineralize a macromolecular solution either by dialysis (as in
hemodialysis) or by ultrafiltration.

2.1.2 The solute has vapor pressure

When two volatile liquids are mixed, the noticeable effects of solute-
solvent interaction apply mutually to both. Under ideal (noninter-
acting) conditions, as witnessed by the absence of temperature and
volume effects on mixing, the partial pressures of the two compo-
20 2 The Thermodynamic Connection

nents would vary linearly with molar composition; this is Raoult’s

law for ideal solutions, and again a number effect. Real liquid mix-
tures deviate from Raoult’s law reflecting the likes and dislikes of
molecular interaction, formally accounted for by introducing a liq-
uid phase activity coefficient ( γ ≠ 1 ) into Raoult’s law. For ideal
solutions, therefore, activity coefficients are unity.

Raoult’s law, ideal solutions (

pi = xi po = 1 − x j pio )
Raoult’s law, real solutions pi = xi γ i pio = ai pio (2.1)

Consequently xi resp. ai = pi pio

pi° = pure component or saturation vapor pressure; pi pio = norma-

lized vapor pressure. It is noted that mass action (as number effect
or as activity effect) may be expressed in terms of molar concentra-
tion as well as partial pressure of the volatile solute.
The activity (ai = xi γi) replacing the analytical concentration xi in
Eq. 2.1 is an effective concentration, understood to represent the
“vigor” of the component under consideration under the influence
of its molecular surrounding in the mixture (Sect. 2.1.3). This influ-
ence takes one of two directions depending on the nature of the
molecular interaction between solute and solvent.
Positive deviation from Raoult’s law [γi > 1]. – On a molecular
level, positive nonideality indicates repulsive interaction between
dissimilar (solvent-solute) species, reflected by the solute activity
coefficient to increase with dilution up to the limit of infinite dilu-
tion ( γ ∞ ) , in which limit the solute encounters “alien” solvent mo-
lecules only. The solvent, by then encountering “kin” only, has no
reason not to behave ideally, implying γsolvent ≈ 1 at low solute concen-
tration.
What is observed? Across the composition range (when going
from xi = 0 to xi = 1) the partial pressure of any one liquid solution
component increases more than proportional with concentration; so
does the total vapor pressure as sum of the partial pressure contri-
butions. Corollaries of positive solution nonideality may be a limited
miscibility of the components, observed as phase separation or
a miscibility gap in the phase diagram, and the occurence of positive
azeotropes (this is where the term “positive” deviation has its ori-
 2.1 Mixtures and solutions 21

gin). Positive azeotropes are constant boiling liquid mixtures of

higher vapor pressure (lower boiling point) than either of the pure
components. As constant boiling mixtures, azeotropes can not be
separated by ordinary distillation. Pervaporation, a nonequilibrium
membrane process, is capable of “splitting” azeotropes (Sect. 5.3).
Figure 2.1 illustrates the dependence of the activity coefficients on
mixture composition of a moderately nonideal, partially immiscible
aqueous-organic solution system, water-nitromethane. The system
forms a positive azeotrope at 76.4 w-% nitromethane (bp. 83.6°C). It
is observed that solvent (= either majority component) activity coef-
ficients remain close to unity well into the equimolar composition
range, increasing towards γ ∞ for either component with progressive
dilution as shown. In the vicinity of equimolar composition, that is,

Fig. 2.1. Activity coefficients as function of mixture composition for the “posi-
tively” nonideal, partially immiscible system water-nitromethane at 50°C,
indicating solubility limits [1].
22 2 The Thermodynamic Connection

even likelihood of encounter, the distinction between concentration

and activity looses significance.
Of special separation concern are aqueous-organic solutions or
mixtures containing sparingly soluble organic solutes, such as occur
as wanted bioproducts (for example aroma compounds) or as indus-
trial pollutants (summarily referred to as volatile organic com-
pounds, VOC’s). If phase-separated, a bulk aqueous phase saturated
with the organic solute is in contact with a minor organic phase
saturated with water, the organic phase being either distinctly sepa-
rated or dispersed into “globules”. At sufficiently high dissimilarity
between the components, solute activity coefficient and solute solu-
bility (= concentration) correspond inversely,
γ org ≈ 1 xorg and xorg ≈ 1 γ org (2.2)
For a derivation consider the organic component in equilibrium
across the phase boundary, implying equal organic activity in both
environments (aorg = xorgγorg). When viewing the organic minority
phase to be itself a dilute solution (of water in organic), organic
activity in that phase is unity by xorg ≈ 1 and γorg = 1. Equilibrium
stipulates organic activity in the aqueous phase to be unity as well,
hence Eq. 2.2.
The relation hints at an answer to the question “what is a dilute
solution?” The answer suggesting itself at this point is: at γ solute
∞
> 100 .
Diminishing interaction between water and organic solute is evi-
denced by decreasing mutual solubility, resulting in the partial vapor
pressures to become independent of each other. As a result, the total
vapor pressure approaches the sum of pure component vapor pres-
sures. Moreover, given a qualitative relationship between volatility
and solubility, high boilers being less soluble than low boilers, the
vapor phase mol fraction of a high boiling organic species in contact
with its aqueous solution is expected to be close to the ratio of pure
component vapor pressures, Eq. 2.3:
p ≈ pwater
o
+ porg
o
⎫
⎪
⎬ (x ) org
≈ porg
o o
pwater (2.3)
⎪
vapor

po
org
 po
water ⎭
Steam distillation to isolate low volatile organic compounds, notably
nature products such as essential oils, operates on these principles,
 2.1 Mixtures and solutions 23

aided by the fact that vapor pressure is an “intensive” variable:

Eq. 2.3 applies independently of liquid composition, and regardless
of whether the feed components are in a state of colloidal solution,
microscopic dispersion, or visible phase separation, – in principle
until the organic solute species is exhausted.
Negative deviation from Raoult’s law [γi < 1]. – What is observed
is a lowering of partial and total vapor pressures below those pro-
portionate to liquid composition, qualitatively corresponding to the
colligative vapor pressure lowering observed with nonvolatile sol-
utes. On a molecular level, negative nonideality is associated with
preferential interactive forces between solvent and solute, – for the
most part hydrogen bonding and dipole interaction forces. As seen
by the solute, these forces are most effective when no like molecules
are encountered, that is, under conditions of dilution. Solute activity
coefficients therefore decrease with dilution. Negatively nonideal
liquid solutions always are miscible without limit, and may be asso-
ciated with the occurence of “negative” azeotropes having a lower
vapor pressure (higher boiling point) than either of the pure com-
ponents. Again, solvent activity coefficients approach unity with
decreasing solute concentration; again, deviation from ideal mixture
behavior is smallest in the vicinity of equimolar composition.
In solution reality, positive deviation from Raoult’s law is wide-
spread, negative deviation being confined to cases of predominating
solvent-solute interaction. Nearly all common aqueous-organic solu-
tion systems exhibit positive nonideality with γ org
∞
ranging from less
10
than 2 (methanol) to a fictitious 10 for nonpolar species (the range
is smaller in nonaqueous systems). In separation reality, the nature
of the molecular interaction bearing on the ease of separation, it is
anticipated that positively nonideal liquid mixtures are easier to se-
parate than negatively nonideal ones, including negative azeotropes.
Prominent examples of negative nonideality are the aqueous solu-
tions of simple carboxylic acids, which are notoriously difficult to
separate (formic acid, above all, forming a negative azeotrope).

2.1.3 On thermodynamic activity

When an arbitrary solute species is given a chance to roam freely
between two immiscible phases in contact, its analytical presence in
each of the two phases is likely to differ, one environment proving
24 2 The Thermodynamic Connection

more accommodating to the solute than the other. (One of the pha-
ses may well be a polymeric membrane). While this phenomenon,
conforming to expectation, presents no problem to intuition, its in-
terpretation in terms of the thermodynamic activity does. The condi-
tion of a distribution equilibrium across a phase boundary presup-
poses random movement of the solute, however, at equal rate of
passage to and fro, thus maintaining the uneven distribution at any
moment. The ratio of analytical (“number”) concentrations is the
partition or distribution coefficient of the solute (after Nernst). To
account for the influence of the molecular environment on the dy-
namic behavior of a solute species (its “vigor”), a solute activity is
introduced to replace solute concentration in such a way that, at
equilibrium, activities on both sides of the phase boundary are
equal. The dimensionless activity coefficients by which the number
concentrations are modified ( γ ≠ 1 ; Eq. 2.1) in this capacity are
taken to characterize solute-solvent interaction in any given solu-
tion. Since activity coefficients are typically concentration depend-
ent (see Fig. 2.1), they are reported in the limit of infinite dilution, it
being understood that the “tendency to relocate” of a solute species
is highest (at γ > 1 ) respectively lowest (at γ < 1 ) when it finds itself
isolated in a surrounding of solvent.
There is little practical use of assigning activity coefficients to in-
dividual inorganic ions in water. Charge-dipole interaction being the
strongest among fluid systems, highly negative deviation from ideal
solution behavior is anticipated. Moreover, though osmotically
counting as individual solute species, ions are subject to the condi-
tion of electroneutrality restricting their activity. Thus when a sym-
metrical electrolyte dissociates, the oppositely charged ions are
bound to exist at near-equal activity within their hydration shells,
and mean activity coefficients γ ± = γ + ⋅ γ − are invoked to repre-
sent the activity of the salt (Appendix A).
Activity acquires a somewhat special meaning when one of the
system components is largely immobile, as, for example, a mem-
brane polymer. Mass distribution between fluid components and
polymers is rationalized in terms of sorption isotherms, which reveal
the structural identity of the polymer phase (Sect. 2.3.1).
Being associated with molecular interaction, the concept of activ-
ity looses its meaning when there are no interactive forces to speak
of, as in gas mixtures under ordinary conditions of temperature and
 2.2 The driving force in barrier separation 25

pressure. Such mixtures (which include the permeate of pervapora-

tion) are described by Dalton’s law, which states that partial pressure
directly correlates with mol fraction, summing up to the total pres-
sure of the gaseous mixture:
Dalton’s law pi = xiP P = ∑ pi (2.4)
Osmotic pressures and activity coefficients of two prototype aqueous
solution systems are compiled in Appendix A. One is H2O-NaCl
(a nonvolatile solute up to saturation), alluding to membranes in
water demineralization; the other is H2O-EtOH (both components
volatile and miscible in all proportions), alluding to membranes in
biotechnology. The data are referred to in Chapters 3 (reverse osmo-
sis) and 5 (pervaporation).

2.2 The driving force in barrier separation

Separation is demixing, overcoming all of the tendencies which sta-
bilize the mixture or solution, and which are mirrored by the en-
thalpy of mixing. The thermodynamic condition of a fluid mixture,
its state, is completely described by three independent variables:
Temperature, pressure, composition, – the latter in terms of mol
numbers ni . Between them, these state variables amount to the Gibbs
free energy (or free enthalpy) of the mixture, G(T, p, n). Mol numbers
to express mixture composition, according to Gibbs, are state vari-
ables in that any one of them may vary independently without affect-
ing the presence of all others (as a percentage would do). A measure
of concentration in terms of mol numbers is each component’s mol
fraction xi. For a binary mixture (i = 1, 2) these relations hold:
n1 n2
x1 = x2 = x1 + x2 = 1
n1 + n2 n1 + n2
n1 x n2
= 1 x2 ≈ at n2  n1 (2.5)
n2 x2 n1

dx1 = − dx2 dx = x d lnx

What makes these “intensive” variables special is that they are capa-
ble of forming gradients, which, by their natural tendency to level,
incite transport processes in the manner of a force. The same vari-
26 2 The Thermodynamic Connection

ables which define the tangible properties of a mixture also are the
influence variables in separation process design, seen as change of
state.
“Extensive” variables, by contrast, are volume and total free en-
ergy, but also the individual mol numbers; they depend on the
“amount” present.

2.2.1 The chemical potential, no barrier

To describe a general change of state of the mixture, pictured as
gradient of the Gibbs free energy (free enthalpy) over a distance
coordinate z (Fig. 2.2), the total differential over the relevant vari-
ables G(T, p, n1, ..., nn) is formed:
∂G ∂G ∂G ∂G
dG = dT + dp + dn1 + dn + ... (2.6)
∂T ∂p ∂n1 ∂n2 2

This is the Gibbs fundamental equation for mixtures. Inspection

of its terms with a view at practical barrier separations reveals:
• As a rule, liquid barrier separations operate isothermally, hence
dT = 0 in the simplified treatment. The term thusly eliminated is
the entropy contribution to the free energy.
• The pressure dependence of the free energy is a volume, establish-
ing the mechanical link to the thermodynamic free energy,
∂G / ∂p = V (the bar denoting a partial molar quantity). As “pV
energy”, this term reiterates the dimension of the free energy as
energy or work.
• The variation of the free energy of the mixture with a change in
mol number of any one of its components (∂G/∂ni) is the partial
molar free energy of that component. Representing the contribu-
tion of mixture composition to the total free energy, Gibbs as-
signed the name chemical potential to the partial molar free en-
ergy, each component thus having its own chemical potential, μi .
It is noted that selective removal of an individual component
from a mixture or solution is the essence of separation.
∂G
= μi (T , p,xi ) (2.7)
∂ni
 2.2 The driving force in barrier separation 27

In its reduced form the fundamental equation (2.6) now reads:

dG = V dp + ∑ μ dn
i i
(2.8)

As in case of the total free energy, to describe a change of state of

an individual component the total differential is again formed, now
in terms of partial molar quantities:
∂μ i
d μi = Vi dp + dx (2.9)
∂xi i
To formally relate the chemical potential of an individual compo-
nent to mixture composition (∂μi/∂xi), Lewis introduced the concept
of the ideal mixture, Eq. 2.10. Mixture composition is completely
described by the mol fraction of the component in question; the state
of reference is the chemical potential of the pure component (μ = μ°
at x = 1 and ln x = 0). The Lewis concept and its implications on
separation are discussed in Sect. 2.2.3.
μi = μio + RT ln xi ⎫
⎪
and d μi d ln xi ⎬ (2.10)
= RT ⎪
dxi dxi ⎭
When introduced into Eq. 2.9 the following elemental relation is
obtained, to which reference will be made time and again when ana-
lyzing the driving force in barrier separation:
d μi = Vi dp + RT d ln xi (2.11)
In terms of experimentally accessible variables, pressure and molar
concentration (activity where applicable), Eq. 2.11 formulates a ge-
neral (isothermal) variation of the chemical potential of an arbitrary
component within an open fluid mixture. In Fig. 2.2 that variation is
depicted as a linear gradient over an unconfined distance coordi-
nate, dμi / dz.

2.2.2 The chemical potential, barrier inclusive

While it needs potential gradients to move fluid mixture compo-
nents, it takes barrier interference to sort them. As indicated in
Fig. 2.2, the membrane is introduced as a partition dividing the free
28 2 The Thermodynamic Connection

Fig. 2.2. Free energy gradient and difference approximation [2]. Within the
tangible confinement of the membrane the gradient may assume different
shapes as schematically indicated.

energy continuum into two realms of distinct potential, thereby

transforming the potential gradient into a potential difference: The
difference approximation [2].
dμ Δμ
≈ (2.12)
dz z
The system now consists of three phases with two phase boundaries
inbetween, marking, as phase boundaries do, abrupt changes of pro-
perty. The potential gradient now is confined to within the bounda-
ries of the barrier phase of thickness z, while the potential difference
to drive mass transfer across the barrier is localized in the two exter-
nal phases representing feed and permeate in membrane separation.
With reference to Eq. 2.11, that difference takes one of two forms,
depending on whether mol fractions or partial pressures (Eq. 2.1)
are used to express individual concentrations,
xi′
Δμi = Vi Δ p + RT ln
xi′′
(2.13)
p′
Δμi = Vi Δ p + RT ln i
pi′′
 2.2 The driving force in barrier separation 29

wherein the superscripts ( ′ ) and ( ′′ ) now indicate the feed (up-

stream) and permeate (downstream) phases of the membrane sys-
tem. – When likened to a concentration gradient, the shape of the
potential gradient within the membrane reflects the swelling profile
of the membrane (Fig. 2.2).
By the prevailing driving forces, practical barrier separations fall
into one of two classes (not counting electromembrane processes):
• pressure driven,
• concentration (activity) driven.
Applied (external) pressure is the driving force in all membrane
filtration processes including reverse osmosis (Fig. 4.2) as well as in
membrane gas separation. Concentration (respectively partial pres-
sure) is the driving force in dialytic separations as well as in per-
vaporation. Salt passage in reverse osmosis, as in dialysis, follows its
own concentration gradient.
When comprehended as influence parameters in separation pro-
cess design, a basic difference between the two kinds of driving force
is noted: While pressure usually is maintained constant throughout
the operation, the composition of the feed mixture undergoing sepa-
ration varies systematically, as is the objective of selective mass
transfer. Permeance in pressure driven membrane processes is the
response of flux to operating conditions; hence the term “pressure
normalized flux” (Figs. 3.3 and 4.4). Permeance in concentration
dependent mass transfer is the response of flux to feed composition,
usually with focus on the flux of the preferentially transported target
species (Figs. 5.3 and 5.6).

2.2.3 Chemical potential and separation

The Lewis concept of the ideal mixture correlates the chemical po-
tential (free energy) of each mixture component with its molar con-
centration, Eq. 2.10. The term “ideal” envisions dilute solutions of
real fluids, in which both components behave ideally: The solvent as
being nearly pure (x1 ≈ 1); the solute as having little statistical chance
to become evident ( x2  1) .
30 2 The Thermodynamic Connection

Addition of a solute to a solvent (which itself may be a solution)

inevitably lowers the chemical potential of the solvent, as evidenced
by a lowering of vapor pressure (colligative property):
μio − μi = Δμi = − RT ( ln xi ) (2.14)

Δμi (index i = 1 for solvent) is the difference in free energy between

pure solvent and solvent containing solute (note that mol fractions
< 1 render ln xi negative). Reversing the perspective, this energy
difference is the minimum isothermal work required to isolate pure
solvent from the solution.
These are the questions of separation concern:
• How does the presence (concentration) of a solute (x2 = 1 − x1)
affect the energy of separation of pure solvent from its solution
(presumed dilute)?
Answer: Δμi is propotional to solute mol fraction since, for dilute
solutions, ln x1 ≈ − x2 . A case in point is “water desalination” by
reverse osmosis, a process actually involving dewatering of the saline
solution (Chapter 3.2).
• How does the energy of separation of a solute species from
a given solution depend on its own concentration?
Answer: Δμ 2 is proportional to (− ln x2). By the arithmetic of
logarithms that quantity is negative, causing Δμ 2 to increase with
dilution (respectively depletion), the limit being ln x2 = − ∞ at x2 =
0. Sherwood diagrams (Fig. 1.1) illustrate this correlation. Volume
reduction to precede isolation would be the method of choice to
recover wanted materials from dilute solutions.
As real solutions deviate from ideal mixture behavior, mol frac-
tions no longer truly represent the concentration dependence of the
free energy. Introducing the thermodynamic activity in place of the
analytical concentration (Sect. 2.1.3) has its origin in the desire to
retain the formal beauty of the Lewis concept for fluid mixtures in
general. This poses the need for a convention on the condition “di-
lute”. As implied above, aqueous solutions are considered dilute as
far as the approximation x2 ≈ − ln x1 is acceptable (x2 for solute mol
fraction). At this level, the difference between the molar volume of
the solvent and its partial molar volume in solution looses meaning,
 2.3 The master flux equation 31

too. For many practical purposes, even seawater passes as dilute

solution (Sect. 3.1.3); the vapor pressure of seawater at room tem-
perature is 1.84% below that of pure water (Spiegler).

2.3 The master flux equation

Mass transport is relocation with a directional preference under the
influence of a potential gradient. One difference between fluid mix-
tures (such as feed and permeate phases in barrier separations) and
solid solutions (such as of permeants in polymeric membranes) is
that in the fluid phase all components in principle are free to move,
whereas in the solid phase only the permeants have mobility, the
polymer matrix remaining stationary. The segmental mobility of
polymer chains, although obviously instrumental in allowing per-
meant relocation, has no directional bias.
Mass transport through homogeneous (“dense”) membranes is
by diffusion only. Mass transport within the pore space of porous
barriers is by diffusion as long as the liquid phase remains stationary
(dialysis); on applying pressure, convection superimposes diffusion
(diafiltration).
The original statement of diffusive mass transport is attributed to
Nernst: The rate of migration ( Ji) of a species through a homogene-
ous fluid medium is given by the concentration of that species in the
medium (cim) times its mobility in the medium (ui) under the influ-
ence of a potential gradient,
J i = cim ui grad μi (2.15)
Introducing the diffusion coefficient to represent mobility, u = D/RT
(Nernst-Einstein) and applying the difference approximation (Eq.
2.12), a working expression for solution-diffusion mass transfer
across a barrier of thickness z is obtained: The master flux equation.
cim Di
Ji = Δμi (2.16)
RT z
By structure analogy, Eq. 2.16 is akin to an Ohmic law linking a cur-
rent (mass or volume flux) to a potential (chemical potential diffe-
rence) by way of a conductor (the permeable barrier). The permea-
32 2 The Thermodynamic Connection

bility of the barrier is seen to be compounded of three contributing

factors, unrelated in their physical nature, yet interrelated in their
influence on mass transport: Sorption (cim), diffusivity (Di), mem-
brane thickness (z). All three, properly identified as they are in the
flux equation, require detailing under the circumstances of actual
barrier situations, as indicated below.
By dimension, the flux in membrane operations is a flow density,
expressed in terms of mass or volume per time and membrane area
(Sect. 1.6). The Journal of Membrane Science lists 4 SI units and 15
“practical” units, all conforming to this dimension, to present flux.
Separation coming about by differences in the rates of mass
transfer (the message of barrier interference), selectivity is defined
by the ratio of partial fluxes. Within a self-contained membrane
system that ratio reduces to the ratio of permeabilities,
Ji / Jj = αij = (cim/cjm) (Dim/Djm) (2.17)
wherein the arrangement of terms points to the two possible mecha-
nisms by which differentiation in mass transport according to the
solution-diffusion model occurs, namely
• sorption selectivity, and (or)
• diffusion (= mobility) selectivity.
Establishing membrane selectivity (as function of feed composi-
tion and operating conditions) as a rule requires recording the sepa-
ration effect on actual feed mixtures, inferring on individual (par-
tial) flux rates from the composition of the permeate. The only
exception appears to be the membrane separation of “permanent”
gases, where the ratio of single gas permeabilities actually predicts
the observed separation effect.

2.3.1 Sorption
Sorption (absorption) refers to the solubility of fluids (liquids or
gases) in a contacting liquid or solid phase, – a polymeric membrane
as a case in point. Sorption isotherms are a pictorial record of the
equilibrium concentration of a sorbed species as function of its con-
centration in the external phase (external pressure in case of gases).
The simplest sorption system pictures the solubility of gases in
liquids, for which Henry’s law states that the concentration of sorbed
 2.3 The master flux equation 33

gas is proportional to gas pressure (partial pressures in case of gase-

ous mixtures), ci = Si pi′ . This is the statement of a linear or “Henry-
type” sorption isotherm (with Si = sorption or solubility coefficient
of component i). An example of gaseous sorption selectivity as ratio
of sorption coefficients, crucial to aquatic life, is water exposed to
air, the oxygen-to-nitrogen ratio in water being considerably higher
than in the air above.
The approach may be generalized to apply to gas as well as to liq-
uid sorption by polymers. In general, linear sorption is exceptional
and referred to as “ideal”. As illustrated schematically in Fig. 2.3, de-
viations from linear sorption behavior occur in both directions: Lan-
gmuir isotherms indicating a saturation situation, Flory-Huggins iso-
therms indicating polymer swelling (a plasticizing effect), – “ideal”
sorption occuring at low sorbed concentration only.
Sorption equilibrium means equality of activity of the species un-
der consideration between feed and membrane, xi′ γ i′ = xim γ im .
Discriminating (preferential) sorption of minority solutes is the
predominant mechanism of selection in liquid membrane separa-
tion, implying xim > xi′ and γ im < γ i′ . The “isotherm” linking
sorbed concentration with feed concentration reads
γ i′
xim = x′ and Si = γ i′ γ im (2.18)
γ im i

Fig. 2.3. Principal shapes of sorption isotherms (schematic). Ideal (Henry-

type) sorption is found at low sorbed concentration only.
34 2 The Thermodynamic Connection

wherein the ratio of activity coefficients assumes the role of a sorp-

tion coefficient. Eq. 2.18 conveys the following information:
• Sorbed concentration – and consequently the flux – of a solute
diminishes along with its concentration in the feed; hence there is
a practical lower limit to recovering minority solutes, the process
“slows down”.
• Sorbed solute concentration increases with the degree of (posi-
tive) nonideality of the feed solution ( γ i′ > 1) , resulting in im-
proved separation selectivity (Table 5.1 as example).
• Sorbed concentration approaches feed concentration as the ratio
of activity coefficients approaches unity (and vice versa); this is
a statement of solvent-polymer compatibiliy otherwise known as
the “like dissolves like” principle (Sect. 6.3, solubility parameters).
It is a trivial observation that sorption is prerequisite to permea-
tion ( Ji = 0 at cim = 0, Eq. 2.16), and that flux increases with the
sorption capacity of the membrane. A correlation between flux (per-
meance) and sorption isotherm, both functions of feed composition,
is therefore expected. Examples for kinship are presented in Fig. 5.3
(Langmuir isotherms) and Fig. 5.6 (Flory-Huggins isotherms).
To the dismay of purists, liquid sorption occasionally is found to
be higher than sorption from saturated vapor. The phenomenon,
known as Schroeder’s paradox [3], points to water clustering as
a possible contribution to the (unwanted) salt passage through
“dense” hydrophilic reverse osmosis membranes (Sect. 3.2.2).

2.3.2 Diffusivity
Diffusion contributing to membrane permeability is the mechanism
of permeant transport within the barrier, the relevant concentration
gradient being that of the sorbed species between the internal mem-
brane boundaries. Diffusion is a kinetic phenomenon actuated by
random thermal motion of sorbed species which are actually pre-
sent, hence depending on true local concentration (a number effect).
Influence factors, conforming to intuition, are
• size and shape of the permeants;
• the structural identity of the polymer phase; and
• permeant-polymer interaction.
 2.3 The master flux equation 35

Fig. 2.4. Correlation diagram of diffusivity versus molecular size (as van der
Waals volume) of low molecular weight permeants in a rubbery and
a glassy polymer [4].

The illustration of Fig. 2.4, oft-quoted [4], summarizes the situa-

tion. Shown is the correlation between diffusivity and permeant size
(as van der Waals molecular volume; alternative size indicators
would have served equally well) for two polymers representing the
two prototype classes of polymeric behavior described as rubbery
and glassy. The considerable range of diffusion coefficients in case of
the glassy (“stiff”) polyvinylchloride is contrasted with the higher
and less discriminating diffusivity in case of natural rubber. Diffu-
sion selectivity (Di / Dj) is inferred from the steepness of the slope of
the tie lines between permeant pairs.
A key parameter is polymer swelling attendant to liquid sorption.
As again suggested by intuition, swelling enhances permeant mobi-
lity, thereby reducing diffusion selectivity. The effect is formally
accounted for by a plasticizing parameter χ (“Flory-Huggins interac-
36 2 The Thermodynamic Connection

tion parameter”) which renders the diffusivity of any one permeat-

ing species dependent on the local concentration of all permeants
present apt to cause swelling. Local permeant concentration, in turn,
is depicted as sorption profile across the membrane under operation
conditions. In Fig. 2.2, two sorption profiles are indicated, schemati-
cally illustrating the situations of “low” and “high” swelling.
Diffusion, convection, conduction? The picture of a drop of water
spreading on a piece of blotting paper is familiar. Once soaked, the
wet paper will transmit water at the slightest pressure head, – gravity
suffices. A stack of many wet papers will need more of a pressure; its
permeance is reduced. This is the naive model of a hydraulic conduc-
tor of uniform water content whose hydraulic resistance is expected
to increase with the length of the duct, eventually to the point of
closure. Applying the picture to liquid barrier separations, if water is
to be the preferred permeant (as in reverse osmosis, ultrafiltration,
and hydrophilic pervaporation), the membrane needs to be “thin”;
on the other hand, if water is to be retained (as in organophilic per-
vaporation), a “thick” membrane may be desirable.

2.3.3 Membrane thickness

How thick is “thin”? In a 1936 review of the state of the art of ultrafil-
tration, Ferry [5] ascribed the difference in behavior between then
available “ultrafilter membranes” and an “ideal mechanical sieve” to
the high ratio of pore length (= film thickness) to pore diameter,
– which he lamented to be seldom below a thousand (alluding to μm
pores in mm film).
The situation changed decisively when, around 1960, Loeb and
Sourirajan discovered the “high flux” cellulose acetate membrane
[6], whose structural principle was unraveled soon after by Riley:
A microporous barrier integrally covered by a “dense” skin of typi-
cally 0.2 μm (200 nm) thickness which functions as the membrane
proper, – the asymmetric membrane. Almost immediately, the dis-
covery elevated membranes from a laboratory tool to a technical
appliance, the first aimed-for application being water demineraliza-
tion by reverse osmosis. Although the original Loeb-Sourirajan mem-
brane (of cellulose diacetate) was not yet capable to “desalt” sea-
water in a single pass, falling short in salt rejection, it did establish
 Bibliography 37

the lower limit of commercial viability of reverse osmosis in terms of

permeate flux: 400 L/d m2 (the “10 gfd criterion”).
Benefiting from advances in reverse osmosis process design, both
membranes and apparatus, asymmetric (effectively thin) membra-
nes have subsequently transformed ultrafiltration (since 1965) and
membrane gas separation (since 1980) into industrial separation
processes as well, – strongly supported by the now legendary United
States Office of Saline Water (OSW).

Bibliography
References
[1] S. R. Sherman, D. B. Trampe, D. M. Bush, M. Schiller, C. A. Eckert, A. J.
Dallas, J. Li, P. W. Carr, Compilation and correlation of limiting activity
coefficients of nonelectrolytes in water. Ind. Eng. Chem. Res. 35 (1996)
1044–1058.
[2] J. A. Wesselingh, R. Krishna: Mass Transfer. Ellis Horwood, New York etc.,
1990.
[3] C. Vallieres, D. Winkelmann, D. Roizard, E. Favre, P. Scharfer, M. Kind, On
Schroeder’s paradox. J. Membrane Sci. 278 (2006) 357–364.
[4] R. T. Chern, W. J. Koros, H. B. Hopfenberg, V. T. Stannett, in: Material
Science of Synthetic Membranes (D. Lloyd, ed.), Chap. 3, ACS Symp. Ser.
269 (1985).
[5] J. D. Ferry, Ultrafilter membranes and ultrafiltration. Chemical Reviews 18
(1936) 373–455.
[6] S. Loeb, S. Sourirajan, Sea water demineralization by means of an osmotic
membrane. In: R. F. Gould (ed.): Saline Water Conversion, Adv. Chem. Se-
ries 38 (1962) 117. – See also Chap. 1, Ref. [14].

Background
[7] A. S. Michaels, H. J. Bixler, Membrane permeation: Theory and practice. In:
Progress in Separation and Purification (E. S. Perry, ed.), Vol. 1, Wiley, New
York, 1968, 143–186.
[8] P. Meares (ed.): Membrane Separation Processes. Elsevier, Amsterdam etc.,
1976.
[9] E. L. Cussler: Diffusion. Mass Transfer in Fluid Systems. Second edition,
Cambridge University Press, Cambridge, 1997.
[10] B. E. Poling, J. M. Prausnitz, J. P. O’Connell: The Properties of Gases and
Liquids. Fifth edition, McGraw-Hill, New York etc., 2001.
[11] R. B. Bird, W. E. Stewart, E. N. Lightfoot: Transport Phenomena. Second
edition, John Wiley & Sons, New York etc., 2002.
3 Osmosis et cetera

3.1 Osmosis
Van’t Hoff’s semipermeable membrane, postulated to advance the
theory of dilute aqueous solutions, is a barrier permeable to water
(solvent), while completely impermeable to dissolved solutes. It is
thus a model barrier for all membrane filtration operations in which
solutes are being retained (concentrated) by removal of solvent
(Chapter 4). Osmotic effects, like all colligative properties, are con-
fined to liquid solutions. Since nature as we know it is an aqueous
system, the solvent in the following is water.
When pure water and an arbitrary aqueous solution are in con-
tact through a semipermeable membrane at ambient pressure, pure
water is “drawn” into the solution as if to dilute it: Osmosis. As is
well known, osmosis is of utmost importance to life’s functioning
when comprehended as transport phenomenon on a molecular level.
Living cell walls are osmotic barriers with sophisticated selectivity
towards inorganic and organic solutes (“biological membranes”).
The direction of osmotic water transport indicates that the solution
has a lower free energy (potential) than pure water, irrespective of
the nature of the solute. Specifically, it must be the activity of the
solvent being lowered by influence of the solute(s), since the model
barrier is presumed to communicate by way of the solvent only.

3.1.1 Osmotic investigations

When Pfeffer devised the osmotic cell named after him, he had plant
cells in mind (1887). The original osmotic cell is an unglazed ceramic
vessel of about 10 mL capacity (A in Fig. 3.1), to which is applied
a membrane by interfacial precipitation as follows: The vessel, its
40 3 Osmosis et cetera

Fig. 3.1. Pfeffer's demonstration of osmotic pressure. A porous clay vessel

A, lined with an osmotic membrane, is filled with aqueous sugar solution of
known concentration; when immersed in pure water in flask F, readings of
the osmotic pressure can be taken. From a 1909 chemistry textbook [1].
 3.1 Osmosis 41

pore space soaked with water to exclude air bubbles, is briefly rinsed
with a solution of copper sulfate {CuSO4}, then filled with a solution
of potassium ferrocyanide {K4[Fe(CN)6]} – vulgo yellow prussiate –,
whereupon a precipitate of water-insoluble copper ferrocyanide
{Cu2[Fe(CN)6]} forms on the inside surface. The procedure bears
close resemblance to that of interfacial polymerization, by which
today’s composite polyamide membranes are manufactured
(Sect. 6.2). By adding a minute amount of potassium prussiate to the
aqueous solution being investigated, Pfeffer’s membrane even has
a self-mending quality to it.
The results obtained with the osmotic cell using sugar solutions in
contact with pure water are as straightforward as they are puzzling:
The osmotic pressure is equal to the gas pressure which would pre-
vail if the dissolved species would fill the cell volume as an ideal gas.
Thus a 0.01 molar aqueous sugar solution at room temperature ex-
erts an osmotic pressure of 0.224 bar, increasing by 1/273 per degree
of warming as postulated by Gay-Lussac’s law for ideal gases.

3.1.2 The law of osmotic pressure

To quantify the effect of a solute on the free energy of the solvent, an
“osmotic experiment” is visualized in which the model solution is
confined in volume. The ensuing solvent influx then produces
a pressure increase in the solution until a dynamic equilibrium is
reached. The effect, which is in fact related to Nollet’s original dis-
covery of semipermeability (Fig. 1.3), is readily observed when ripe
fruit bursts after a rain, the skin acting as semipermeable barrier
enclosing the fruit tissue. The equilibrium pressure, by definition, is
the osmotic pressure π of the solvent, being one of the colligative
properties of liquid solutions.
At osmotic equilibrium there is no net flux, hence no discernible
driving force. For d μi = 0 at p = π Eq. 2.11 yields (i = 1 for solvent;
i = 2 for solute):
RT
V1π = − RT ln x1 and π = − ln x1 (3.1)
V1
42 3 Osmosis et cetera

For dilute solutions ( x2  x1 ) the mol fraction of solute respon-

sible for the osmotic pressure of the solvent is approximated by
x2 = − ln x1 (Sect. 2.2.3), which, when introduced into Eq. 3.1, gives
RT
π = x or πV1 = RT x2 (3.2)
V1 2

for the osmotic pressure of the solvent as function of the molar con-
centration of a solute, irrespective of the nature of the solute (“num-
ber effect”).
Equation 3.2 is known as van’t Hoff’s limiting law of osmotic
pressure (1886), the attribute “limiting” alluding to the concept of an
ideally semipermeable membrane in contact with a dilute solution.
Intrigued by Pfeffer’s osmotic cell results, the semblance of van’t
Hoff’s law with the ideal gas law gave rise to the notion that osmotic
pressure has the same kinetic origin as an ordinary gas pressure:
Momentum transfer of thermally agitated species bouncing against
the wall of their containment. On closer inspection, this is a spe-
culative notion at best. In fact, as follows from a surface tension
argument, electrolytes in aqueous solution even tend to withdraw
from the phase boundary. (A proposition to recover desalted water
by surface skimming of seawater, based on this argument, did not
materialize, however.)
The relevant influences on the free energy of the solvent of liquid
solutions are now seen to be reduced to two formally related, yet
fundamentally counteracting variables: External pressure (the pV
term) raises the potential of the solvent (of virtually anything), while
the presence of a solute (the πV term) lowers it, Fig. 3.2. When, in
an osmotic experiment, external pressure outweighs the effect of
osmotic pressure, the direction of solvent flow is reversed: Reverse
osmosis. Solvent flow in reverse osmosis thus relies on the increment
of external pressure over and above the osmotic pressure of the feed
solution.
While practical reverse osmosis is concerned with solutions of
solutes having no vapor pressure (such as H2O-NaCl, Appendix A
upper), there is no difference in principle when turning to mixtures
both components having vapor pressure (such as H2O-EtOH, Ap-
pendix A lower). The additional aspect is that by then both mixture
components mutually exert solute effects resulting in osmotic pres-
 3.1 Osmosis 43

Fig. 3.2. Hydrostatic pressure (p) and osmotic pressure (π) counteracting in
their effect on solvent free energy as pV energy (mechanical) versus π V
energy (chemical).

sure (colligative properties) for each. Referring to Raoult’s law,

Eq. 2.1, a rendition of osmotic pressure in terms of vapor pressure is

RT ⎛ Pi o ⎞
π (i ) = ln ⎜ ⎟ (3.3)
Vi ⎝ Pi ⎠
where i now is solvent or solute depending on proportion. Given
sufficient mutual miscibility, osmotic pressures of aqueous-organic
solutions may reach extreme values, see Appendix A.

3.1.3 Osmotic pressure illustrated

Nature provides two semi-standards of osmotic pressure, the pre-
eminence of which remains a matter of speculation.
• One is the equivalent salinity of the body fluids of warm-blooded
mammals, which is osmotically matched by an isotonic solution of
0.9 w-% NaCl at about 7.5 bar osmotic pressure (referred to as
“saline” in Fig. 4.4). A physiological blood infusion (Ringer solu-
tion) is prepared, for example, by dissolving NaCl (8.0 g); KCl
(0.2 g); CaCl2 (0.2 g); MgCl2 (0.1 g); NaHCO3 (1.0 g); NaH2PO4
(0.05 g); glucose (1.0 g) in 1 L water. Whole blood plasma, in ad-
dition to “salts”, contains macromolecular species (proteins).
Therapeutic protein substitution is by hydrophilic (soluble)
polymers in concentration to match the incremental osmotic
pressure of high molecular weight blood components. A telltale
44 3 Osmosis et cetera

example is polyvinylpyrrolidone (PVP) which, it is noted, plays

a key role in continuing efforts to convey hydrophilicity to syn-
thetic membranes, including those employed in hemodialysis
(Sect. 6.2).
• The other is seawater, which covers three quarters of the globe at
remarkably uniform mineral content, mineral composition and
density (Sect. 3.3.1). A concentration of 3.45 w-% total dissolved
solids (34500 ppm TDS of “sea salt”) associated with 25.5 bar os-
motic pressure sometimes is taken as standard seawater. Com-
puted in terms of the ionic concentrations of the major inorganic
constituents, seawater has a salinity of approximately 1 mol/L.
Local variations in the salinity of real seawater do occur, however,
and have a significant effect on the performance (= economics) of
water desalination: The majority of seawater desalination plants
operate from confined seawater bodies such as the Persian Gulf
and the Red Sea having higher than open-ocean salinity.
More osmotic pressure. – For the following observations the data
tabulated in Appendix A serve as illustration. According to Eq. 3.2,
solvent (water) osmotic pressure increases with solute concentra-
tion, a linear dependence being observed well into an intermediate
composition range for both NaCl and EtOH as solutes. The same is
true for water as solute in ethanolic solution, the lower slope indicat-
ing a difference in solute activity. At comparable mol fraction of
solute (for example, x2 = 0.01), osmotic pressure in case of NaCl is
twice that observed with EtOH (26 versus 13 bar), electrolytic disso-
ciation doubling the number of “osmotically relevant” species with
the salt.
Turning to the completely miscible system H2O-EtOH, it is ob-
served that the osmotic pressure of either component increases with
progressive dilution, Eq. 3.4. To verify the data would require two
separate “osmotic experiments” using membranes of contrasting
permeability description,
• one a membrane preferentially permeable to water to ascertain
the effect of an organic solute (a hydrophilic membrane);
• the other preferentially permeable to organic solvents to ascertain
the effect of water as solute (an organophilic or hydrophobic
membrane).
 3.1 Osmosis 45

Reverse osmosis is imaginable in either direction, provided semi-

permeable membranes as described are available. The principal and
practical limitation is the osmotic pressure which needs to be over-
come (referred to as osmotic pressure limitation), as again revealed
by the data of Appendix A. For example, the minimum pressure to
dehydrate wine (11.9% EtOH) by an ideally hydrophilic barrier is
64 bar; conversely, at least 860 bar is needed to remove pure ethanol
from wine through an organophilic barrier. Removing water from the
ethanol-water azeotrope (4 w-% H2O at 78°C) by reverse osmosis
would require pressures in excess of 2000 bar; the other way around
is meaningless. – There is in fact a membrane process capable of
splitting azeotropic mixtures, which relies on a drastic reduction of
the activity of the permeate by causing it to evaporate: Pervaporation.
As examples encountered in food processing, Table 3.1 lists os-
motic pressures of various juices. Membranes have the capacity to
concentrate bioorganic solutions under “mild” (low temperature)
conditions, retaining aroma compounds; however, as the figures
indicate, dehydration by reverse osmosis faces the osmotic pressure
limitation. A suitable dehydration process independent of this limi-
tation is membrane distillation and its counterpart, osmotic distilla-
tion (Sect. 1.4).

Table 3.1. Osmotic pressure of fruit juices and milk [2]

Juice Concentration Osmotic pressure

(degree Brix) (bar)
Sugar beet juice ~ 20 034.5
Cane sugar juice ~ 44 069
Tomato juice ~ 33 069
Lemon juice ~ 10 015
Lemon juice ~ 45 103
Orange juice ~ 10–12 017–20
Orange juice ~ 42 103
Orange juice ~ 60 207
Milk (for comparison) ~ 12% TS 006
Brix: An industrial measure related to density (10 Brix ≈ 1 w-% sugar at
20°C). TS: Total solids (lactose and salts).
46 3 Osmosis et cetera

3.2 Reverse osmosis

By generic category, reverse osmosis is volume reduction through
selective removal of solvent, the driving force being an external pres-
sure over-compensating the osmotic pressure of the feed solution.
Because of the osmotic pressure limitation, reverse osmosis separa-
tions focus on solvent recovery (as permeate) more than on solute
enrichment (in the retentate).
Practical reverse osmosis (RO) aims at recovering demineralized
water from natural saline solutions – seawater and brackish waters –
by pressure-driven permeation through hydrophilic polymer mem-
branes (slogan “fresh water from the sea”). Central to design and
operation of the process is the osmotic pressure of the feed solution,
the pressure of reference being the osmotic pressure of “standard
seawater” of 25.5 bar at 25°C.

3.2.1 Solvent flux and solute rejection

Real membranes are “leaky”, never completely excluding unwanted
feed components. Modeling reverse osmosis desalination therefore
requires to consider both water transport and salt passage. The
Merten model of mass transport in reverse osmosis [3] is based on
these premises:
• Mass transport is by a solution-diffusion mechanism (Sect. 1.3.2);
• there is no coupling between water and salt transported, permit-
ting the master flux equation (Eq. 2.16) to be applied independ-
ently to each;
• the membrane is highly salt rejecting, implying the difference in
salinity between feed and permeate (Δc2) to be high; and
• the model holds as far as the approximation “dilute” applies
(Sect. 2.2.3), implying ideal solution behavior for the solvent, and
negligible difference between the molar volume of water and its
(
partial molar volume in solution V1 ≈ V1 .)
Solute concentration (c2 in the following) is summarily given as
w-% or ppm (= mg/L) of “salt”, linked in practice to electrical con-
ductivity. Justification comes from the fact that, even though natural
saline waters (seawater) contain a multitude of inorganic salts, re-
maining salt in the permeate of reverse osmosis (and by implication
 3.2 Reverse osmosis 47

within the membrane) predominantly is NaCl; the permeate is “soft”

(Appendix B).
Solvent (water, index 1). – The driving force for solvent flow
is obtained by substituting the concentration term in Eq. 2.10 by
the corresponding osmotic pressure (Eq. 3.1) to obtain Δμ1 =
V1 ( Δ p − Δπ ) . With this rendition of the solvent driving force the
flux equation, Eq. 2.16, becomes
c1 D1
J1 = V ( Δ p − Δπ ) (3.4)
RT z 1

To picture the simplest situation, permeate pressure is ambient and

permeate solute concentration is low, in which case both pressure
terms in Eq. 3.4 refer to the condition of the feed.
Solute (“salt”, index 2). – Since the solute has no vapor pressure,
there is no osmotic pressure equivalent to concentration. Instead,
the total differential of its chemical potential with respect to the
variables pressure and concentration is again formed, μ 2 (p; c2).
Observing that ∂μ 2 / ∂p = V2 (the partial molar volume of the sol-
ute), and noting that dlnc/dc = 1/c, the following relation for the
driving potential for the solute is obtained:
RT
Δμ2 = V2 Δ p + Δc 2 (3.5)
c2

Adaptation to the premises of the Merten model is by observing that

the effect of pressure on salt passage is small compared to the influ-
ence of the concentration gradient between feed and permeate, Δc 2 .
Furthermore, the concentration cim of the master flux equation
(Eq. 2.16) is that of the permeant within the diffusion barrier, while
the salinity which determines the driving force is that of the external
feed solution, c′2 . The distribution coefficient relating the two,
c 2m = Kc 2′ , makes salt uptake by the membrane a function of the
salinity of the feed solution contacting it. All things considered, the
flux equation for salt passage in reverse osmosis reduces to
KD
J2 = Δc 2 (3.6)
z
48 3 Osmosis et cetera

Salt rejection. – The selectivity of reverse osmosis desalination is

expressed as salt rejection R in terms of the analytical (bulk) solute
concentrations of feed ( c′2 ) and permeate ( c′′2 ) as
c 2′ − c 2′′ c ′′
R = = 1− 2 R × 100 = [%] (3.7)
c 2′ c 2′

For example, desalting standard seawater (34500 ppm TDS) to pota-

ble water (< 500 ppm TDS) nominally would require a salt rejection
of 98.6%. Under dynamic process conditions a higher than nominal
rejection is needed (Sect. 3.2.3). As the model implies, rejection re-
quirements depend on the desalting task: Less severe on lower salin-
ity feed waters; more stringent when a high degree of demineraliza-
tion is needed.
In place of percentage rejection, the salt passage (100 – R) is used
in industry. In ultrafiltration a sieving coefficient is used, similarly
describing selectivity in terms of solute whereabouts (Sect. 4.4).

3.2.2 Model implications

Although based on low-recovery laboratory conditions (which yield
“intrinsic” membrane performance data), the Merten model repre-
sents actual reverse osmosis performance using available mem-
branes to a fair degree. The principal model predictions, illustrated
schematically in Fig. 3.3, are as follows.
Mass transport. – Water flux (in kg/d m2), commencing at the os-
motic pressure of the feed solution, increases linearly with external
pressure, Eq. 3.4. Salt flux (in g/d m2; note a factor of 103 between
fluxes) is governed by the difference in salinity between feed and
permeate, Eq. 3.6, and is not affected by pressure. Salt transport in
reverse osmosis thus is a dialysis phenomenon, the level of salt pas-
sage depending on the nature of the membrane polymer. The resul-
tant solute rejection R is a function of pressure simply because
a higher rate of solvent flux relative to a constant salt flux results in
enhanced permeate dilution. As a consequence, to achieve the salt
reduction desired, applied reverse osmosis operates at pressures of
typically twice the osmotic pressure of the feed solution, p ≥ 2π. This
condition contributes to the osmotic pressure limitation of practical
reverse osmosis towards high feed salinities.
 3.2 Reverse osmosis 49

Fig. 3.3. Water flux (J1), salt transport (J2) and resultant salt rejection (R) in
reverse osmosis desalination of seawater in the approximation of the
Merten model.

Adjustable parameters of membrane performance, according to

the model, are the water content of the membrane ( cim in Eq. 3.4)
and its thickness (z). For favorable selectivity, water sorption should
be high, salt sorption (as distribution coefficient, K) low. Commonly
used hydrophilic membrane polymers, in the main aromatic poly-
amides and cellulose esters, have sorption capacities for liquid water
of the order of 10% (“primary water”), equivalent to an increase in
volume (swelling) of the same order. At higher water uptake than
allowable to sustain a solution-diffusion mechanism of mass trans-
port, sorbed water tends to aggregate into liquid domains (droplets),
causing a breakdown of salt rejection. There is little variance of
water content across a swollen reverse osmosis membrane in opera-
tion, providing an isotropic passageway for diffusive mass transport
across (Sect. 5.4.2). Water is “faster” than salt by orders of mag-
nitude.
The membrane. – The permeation rates of solvent and solute are
both inversely related to membrane thickness, implying that the
selectivity – as ratio of the two rates – is independent of membrane
thickness. High solvent flux (permeance) at unimpaired selectivity
thus demands the membranes to be thin. It was the discovery of the
50 3 Osmosis et cetera

asymmetrically structured (“skinned”) cellulose acetate membrane

by Loeb and Sourirajan (1960) which paved the way to high flux
membranes (Sect. 2.3.3), (and on which the Merten model is based).
Today’s composite membranes are asymmetric as well, the active
“dense” skin being produced by interfacial polymerization onto
a porous support. A semi-standard in seawater reverse osmosis is
a composite membrane of an aromatic polyamide (< 0.2 μm) on
a polysulfone microporous support (40 μm), mechanically stabilized
by a polyester fabric, Fig. 3.4. An electron micrograph of the skin
section of such a membrane is shown in Appendix E, Fig. E.6.
The development of seawater reverse osmosis membranes in
terms of “intrinsic” performance is summarized by the data of Ta-
ble 3.2 [5].
Historically, the original asymmetric cellulose acetate membrane
of Loeb-Sourirajan attained an intrinsic flux of 400 L/d m2 (= 10 gfd)
at salt rejection of 96%. To this day, that flux marks the lower limit
of economic viability in reverse osmosis desalination (the “10 gfd
criterion”).

Fig. 3.4. Schematic representation of the make-up of a thin-film composite

membrane as employed in seawater desalination (FT-30; FilmTec/Dow).
The active layer is an aromatic polyamide produced by in situ polymeriza-
tion [4].
 3.2 Reverse osmosis 51

Table 3.2. Representative “intrinsic” performance record of seawater reverse

osmosis membranes. CA = cellulose acetate; PA = polyamide [5].

Year Flux Salt Rejection

2
L/d m %
CA 1978 0650 98.9
PA 1986 1300 99.4
1995 1300 99.7
2004 1500 99.8

Test conditions: 32000 ppm NaCl; 55 bar; 25°C; recovery < 10%

Two comments are in order:

• Membrane performance under actual production conditions ne-
cessarily is below intrinsic: Dynamic reverse osmosis, Sect. 3.2.3.
• A major improvement of membrane performance along present
lines of development seems unlikely; improvement of overall pro-
cess economy therefore focuses on auxiliary technology: Energy
recovery (Sect. 3.2.4) and feed water conditioning (Sect. 3.2.5).

3.2.3 Dynamic reverse osmosis

With reference to the flow scheme of Fig. 3.5, reverse osmosis in
operation is described as follows. The feed solution of analytical
(bulk) concentration co = c 2′ and corresponding osmotic pressure πo
is pumped to the membrane stage (a single “module” or an align-
ment of modules) at feed flow rate Qo (L/h) and feed pressure (oper-
ating pressure) po consistent with the osmotic pressure of the feed,
po ≥ 2π . System pressure is established by a back-pressure regulator
(adjustable valve) in the retentate (reject) stream, Qs . En route
through the membrane stage of specified membrane area, deminer-
alized permeate (the low salinity product water) is withdrawn from
the feed stream at an integral flow rate Qp , resulting in a gradual
build-up of solute concentration from co (inlet) to cs (outlet), along
with a decrease in volume feed flow commensurate with the overall
flow balance, Qo = Qs + Qp . Membrane performance (permeance),
being a function of local feed concentration (increasing) and local
feed pressure (decreasing), declines systematically from inlet to out-
let of the membrane stage.
52 3 Osmosis et cetera

Fig. 3.5. Dynamic reverse osmosis: Flow pattern and process parameters
schematically representing a single membrane stage in operation. Q = flow
rate (L/h); P = hydrostatic pressure; E = recovery.

Process design (including membrane selection) depends on the

osmotic pressure of the feed solution, being a function of solute
concentration, by
• determining the operating (= inlet) pressure of the process;
• limiting the permeate (product) recovery from a given feed
stream;
• limiting the degree of solute concentration build-up in the reten-
tate; and
• determining the energy requirements of the separation process.
Process analysis is by inspecting the two – interrelated – mecha-
nisms by which solute concentration builds up: Permeate recovery
and concentration polarization.
Recovery. – Permeate recovery (also termed recovery ratio or con-
version) represents the yield of the separation process, that is, its
purpose. Recovery is the ratio of permeate flow rate to feed flow rate,
usually given as percentage of the feed volume,
E = Qp Qo E × 100 = [ %] (3.8)

Solute concentration build-up as a result of permeate recovery

reduces the available pressure head for solvent flux (Eq. 3.4) and
increases the permeation rate of solute (Eq. 3.6), both effects, even
 3.2 Reverse osmosis 53

though intended, predictably to the disadvantage of process per-

formance.
There are two modes of reverse osmosis processing:
• In solvent recovery, operation is out of an infinite feed reservoir
(such as a seawater supply), implying fixed inlet conditions. Plant
(stage) recovery is a function of feed flow rate and membrane area
provided. Within every stage (such as schematically depicted in
Fig. 3.5) conditions correspond to those of volume reduction. Sol-
vent recovery is equivalent to once-through batch processing,
– as long as the batch lasts.
• In volume reduction, operation is out of a finite reservoir into
which the retentate stream is returned (batch); solute concentra-
tion of the feed thus increases systematically as permeate is with-
drawn. Recovery is based on permeate volume relative to the
original feed volume.
Intrinsic membrane performance (Table 3.2) is established under
conditions of low recovery, operating at high feed flow rate and lim-
ited membrane area while restoring permeate and retentate streams
into the feed reservoir.
Concentration polarization. – Less readily assessed in its effect on
performance than permeate recovery, concentration polarization
refers to the accumulation of rejected solute(s) near the feed-mem-
brane interface, as pictured by the concentration profile shown in
Fig. 4.3 (upper). Solute concentration build-up (wall concentration
over bulk concentration, cw /cb) is represented by the thickness of the
laminar boundary layer (δ ), taken to be the distance over which
a concentration gradient exists to effect back diffusion of the “trap-
ped” solute(s) into the bulk feed stream. The adverse effects of con-
centration polarization are again inferred from the flux equations:
Declining solvent flux and increasing solute passage on account of
a higher than bulk concentration at the membrane surface.
As back diffusion out of the laminar boundary layer depends on
the size of the diffusing species (Table 4.2), the effects of concentra-
tion polarization become more severe as the molecular size (molecu-
lar mass) of the rejected solute(s) increases. Since solubility at the
same time decreases, the upper limit of concentration polarization is
solute precipitation on the membrane surface (Fig. 4.3, lower). Salt
precipitation when demineralizing “hard” feed waters (gypsum as
54 3 Osmosis et cetera

a solute in point), referred to as scaling, may require to restrict per-

meate recovery. By fouling is meant the process of irreversible depo-
sition of macromolecular matter on the membrane surface; biofoul-
ing refers especially to “natural organic matter”, NOM (Table 3.3).
– Fouling is a fact of life in ultra- and microfiltration, Sect. 4.2.
Concentration polarization in barrier separation cannot be avoi-
ded. When dealing with truly dissolved solutes, the only means to
alleviate the problem is to influence the thickness of the laminar
boundary layer through appropriate hydraulic process and appara-
tus (module) design. This is the origin of the cross flow (“tangential
flow”) mode of operation in barrier separation. When having to deal
with particulate matter, a clarifying pretreatment of the feed stream
is called for, which in itself may be a case of membrane filtration
(Fig. 3.6). It is noted that waterborne pathogens tend to cling to sus-
pended particulate matter.

3.2.4 Energy considerations

Pressure being the driving force in reverse osmosis, the energy ex-
pended is electrical energy to drive the high pressure feed pumps
(recorded as kWh/m3 of product water). The thermodynamic mini-
mum energy required to separate pure water from a saline solution
is readily established as the difference in free energy between pure
water and the reduced free energy of water containing solute, that
difference increasing with increasing solute (salt) concentration
(Section 2.2.3). Everything beyond the minimum is “reality”, to be
assessed in terms of the three interrelated influence parameters –
• operating pressure (p0 ≥ 2π);
• permeate recovery (QP / Q0);
• concentration polarization (cw / cb > 1).
Operating pressure (as inlet or feed pressure, p0) is the only ad-
justable influence paramter and, as “pV work”, is the principal en-
ergy consumer; regardless of recovery, all of the feed needs to be
pressurized. Specific energy consumption refers to the fraction of the
total which leads to product water as defined by the recovery of the
process. The remaining energy is stored in the reject stream to be
dissipated at the back pressure regulator (Fig. 3.5), which in effect is
a throttle. Hydraulic energy recovery by means of special pumps or
 3.2 Reverse osmosis 55

“pressure exchange” devices therefore is standard practice in high

pressure reverse osmosis, and is in fact one of the few means to im-
prove reverse osmosis economics.
Recovery has two implications in practical water desalination:
• When desalted permeate is withdrawn from the feed stream, the
salinity of the remaining feed increases (volume reduction). Con-
sequently, the minimum free energy of separation, even though
defined to imply reversible (non-producing) conditions, increases
with recovery. Clearly, this aspect is independent of the separation
process used. Taking again “seawater” as example, the minimum
free energy to produce pure water at room temperature is
0.7 kWh/m3 (Eq. 2.14 with x2 = 0.018 for “salt”). At a recovery of
50%, which is close to reality in seawater desalination, salinity of
the reject stream about doubles; the minimum free energy of
separation at this point would be 1 kWh/m3.
• Under actual high pressure conditions and optimal recovery of
product water, the energy of separation becomes specific energy
of water production, outweighing the solute effect on the thermo-
dynamic condition of the feed by far, and decreases with recovery.
This aspect is peculiar to reverse osmosis in that practical recov-
ery not only depends on feed water salinity, but can also be ad-
justed within limits: At low initial salinity, allowable recovery is
high and required operating pressure low (and vice versa). This
explains the advantage of reverse osmosis over distillation in
demineralizing low salinity (brackish) feed waters.
Eq. 3.9 is an exercise in accomodating the dual effects of pressure
and recovery on energy consumption in reverse osmosis. It is ob-
tained by allocating pV work as equivalent electrical energy to the
permeate fraction actually recovered, hence a specific energy con-
3
sumption in kWh/m :
p0V = p0 Q0 h = p0 QP h / E
kWh P [ bar ] (3.9)
≈ 2.78
m 3
E [% ]

Complete recovery (E = 100%) at osmotic pressure p = π = 26 bar

formally registers as the minimum energy of separation for seawater,
0.7 kWh/m3. Practical seawater reverse osmosis operates at pressures
56 3 Osmosis et cetera

of 60+ bar and recovery of 40%, at a predicted energy consumption

of 4+ kWh/m3, – provided, that is, no hydraulic energy recovery is
employed to reduce that figure. Again, lower feed salinity leads to
lower energy consumption on both counts: Required pressure down,
allowable recovery up.

3.2.5 Reverse osmosis in the real world

Reality is where the extra money goes. The undispensables of real-
life reverse osmosis are sketched in Fig. 3.6, grouped into the three
sections –
• feed water pretreatment;
• membrane section with pump and energy recovery;
• product water posttreatment.
Not considered are the feed water intake (well construction), nor
local power supply, nor measures to dispose of the brine. Actual
plant size, reported as daily product water capacity, ranges from less
than 95 m3/d = 25 000 gpd (innumerable units, not covered by inter-
national Desalting Plants Inventory Reports) to singular plants of
capacity of the order of 50 000 m3/d. There is, supposedly, little “eco-
nomy of scale” to be realized, large plants being composed of lesser
units in repetition, using common pre- and posttreatment.
Pretreatment. – Pretreatment has to meet with the adversities of
local raw water conditions. The objective is to clear the feed stream
from everything potentially harmful to membrane performance as
evidenced by flux decline and/or undue limitation of membrane life.
The measures taken are mechanical pre-filtration; acidification to
reduce carbonate hardness, followed, if necessary, by aeration to
reduce CO2 ; addition of anti-scalants to keep divalent salts (sulfates)
from precipitating; chlorination (or alternative oxidation) to deacti-
vate microorganisms, usually followed by de-chlorination to safe-
guard against membrane degradation. In Fig. 3.6 two different pre-
treatment schemes are indicated, one of them using membrane
filtration (MF or UF) in a design to eliminate pretreatment chemi-
cals; membrane pretreatment would also result in substantial bacte-
rial reduction.
 3.2 Reverse osmosis 57

Fig. 3.6. The principal components of a seawater reverse osmosis plant,

grouped in feedwater pretreatment; membrane plant including pump and
energy recovery; product water posttreatment [6].
58 3 Osmosis et cetera

About half of the fouling deposit identified on “autopsy” of spent

membrane elements, reported summarily in Table 3.3, is organic by
nature (NOM), the other half being composed of low solubility inor-
ganic species, – essentially everything mineral except monovalent
salts. Distribution of foulants within a reverse osmosis stage (as
schematized in Fig. 3.5) is uneven: While biofouling is most severe in
the entrance region, danger of inorganic precipitation (scaling) in-
creases with increasing recovery, that is, towards the end of the
stage.
Posttreatment. – This is a lesson in water chemistry, Appendix B.
Devoid of all hardness components yet containing CO2 from the
acid pretreatment, demineralized water is soft and acidic, with the
unpleasant and downright corrosive properties of that state. There is
no principal difference in this regard between the permeate of rever-
se osmosis and the distillate from thermal desalination, except for
the residual salinity of the permeate as “artifact” of membrane per-
meability, – mostly NaCl.
When the steam engine took to sea, seawater desalination became
a necessity; the requirement was boiler feed, and that needed to be
soft (to prevent scaling of heat exchangers). Today, desalination
means drinking water and – increasingly so – irrigation water for
specialized agriculture. In both uses the water has to meet certain
regulations set forth, for example, by the World Health Organization
(WHO). Posttreatment centers around remineralization with lime as
it balances with CO2 according to Ca(OH)2 + 2 CO2 → Ca(HCO3)2.
Excess CO2 beyond that needed to keep Ca++ (Mg++) ions in solution

Table 3.3. Membrane fouling on a polyamide rejecting surface (spiral wound

configuration): Foulants observed with 150 spent membrane elements from
around the world. Data derived from [7].

Organics/biofouling 48.6 to 60.6 %

SiO2 03.4 20.4
Al2O3 01.4 06.3
Fe2O3 (“oxide hydrate”) 06.2 07.6
Ca3(PO4)2 01.6 13.4
CaCO3 01.5 04.8
CaSO4 01.4 03.4
Unaccounted 11.1 15.1
 3.3 The power of osmosis 59

is aggressive, acting much like a mineral acid. Reverse osmosis

membranes are permeable to dissolved CO2 ; the permeate (like any
soft water) tends to retain some excess CO2 even after deacidification
(by aeration). Lime treatment, therefore, not only conveys taste and
compatibility with soap, but also controls aggressive CO2 in protec-
tion of water distribution systems. – In case acidification is circum-
vented by relying on membrane filtration as pretreatment (Fig. 3.6,
lower), CO2 actually may have to be added along with lime to sustain
the desired level of hardness.

3.3 The power of osmosis

3.3.1 The osmotic pump
Pressure increases the potential energy of anything, even if “any-
thing” is incompressible for all practical purposes. Gravity increases
the pressure of seawater with ocean depth at the well-known rate of
about 1 bar for every 10 m of water column, nominally matching the
osmotic pressure at a depth of 258 m (referring to “standard sea-
water”: 3.45 w-% TDS; osmotic pressure 25.5 bar; 25°C). Salinity not
only accounts for the osmotic pressure but also for an increase in
density: Seawater is about 3% denser than fresh water at all pressu-
res (testifying to the structuring power of ion hydration). Also seen
as independent of pressure, that is, of gravity is the composition of
seawater (= ratio of salts to water), implying constant osmotic pres-
sure throughout. The model sketched is that of a uniform ocean, an
off-shore approximation by which deep-sea fish seem to flourish.
The osmotic pump is a lesson in sportive science based on the
above scenario [8]. A sturdy pipe, capped at one end by an ideally
semipermeable membrane, is lowered membrane-first into the
ocean. At a depth of some 258 m, fresh water starts to seep into the
pipe by reverse osmosis. On further lowering, a fresh water column
will rise inside the pipe. Now, if the osmotic pressure equilibrium
across the membrane is to be maintained, the fresh water inside the
pipe will have to rise faster than the pipe is gaining in depth, to ac-
comodate the pressure gradient due to the difference in densities
between fresh water (above the membrane) and saline water (under-
neath). Eventually, fresh water will reach the surface. The depth
60 3 Osmosis et cetera

(length of pipe immersed) at which this happens may be estimated

as follows (index 1 = fresh water; index 2 = seawater).
By fluid statics, the pressure gradient of a vertical liquid column
(dp / dz) is a function of its density ( ρ) and of given gravity (g = ac-
celeration of gravity),
dp = ρ1 g dz (3.10)

Let z be the height of the fresh water column at point of overflow,

equal to the total vertical length of pipe immersed. At this point, the
seawater pressure deep down balances the pressure of the fresh
water column plus the osmotic pressure increment, visualized as vir-
tual addition ( z′ = 258 m) to the fresh water column:
ρ2 g z = ρ1 g z + ρ1 g z ′
ρ1 (3.11)
z = z′
ρ2 − ρ1
Accepting, for simplicity’s sake, seawater to be more dense than
fresh water by a flat 3% throughout, osmotic pumping will com-
mence at a depth of z = 8600 m.

3.3.2 Osmotic power generation

Osmotic power is hydroelectric power gained by utilizing direct
osmosis to enlarge the volume flow of pressurized saline water deliv-
ered to a turbine. Operating at the junction of saline water (sea-
water) and a fresh water source, the hydraulics of the process are
described as follows (Fig. 3.7). A saline feed stream (inlet flow
rate Qo) is pumped in cross flow fashion through a membrane instal-
lation at an hydraulic pressure of approximately half its osmotic
pressure (meaning 10 to 15 bar in case of seawater), paralleled across
the membrane by a fresh water stream at ambient pressure. Osmosis
(osmotic influx) increases the “pV energy” of the saline feed stream
by increasing its volume, diluting it in the process. The exit stream
(exit flow rate Qe > Qo), still under pressure, is divided: One fraction
is directed to the turbine to generate power; the second fraction, on
its way to being returned to the sea, passes through a pressure ex-
changer, there to convey its pressure to the incoming seawater feed
stream. The balance between liquid streams is such that it is the
 3.3 The power of osmosis 61

Fig. 3.7. Osmotic power generation, process scheme. P = pressure ex-

changer; M = membrane osmotic plant; T = turbine. Flow rates indicated
are normalized to a saline water intake of Qo = 100.

volume of water gained by osmosis which – at the preselected pres-

sure – is available for power generation. The return flow thus is of
equal magnitude, albeit lower salinity, as the seawater intake, remi-
niscent of a “feed-and-bleed” loop. The fresh water return, except
for having lost half its volume, is as fresh as before.
Other than the osmotic pump, osmotic power generation may yet
come true. A basis for feasibility studies is the preselected pressure
head of 10 to 15 bar translating into a water column of 100 to 150 m
which, when utilized in a hydroelectric plant, would amount to
about 1 MWs of power per m2 cross section of water column. Conse-
quently, osmotic power is expressed in terms of W/m2 of installed
membrane area, thereby linking the effect to the flux performance of
the membrane. At a goal of 5 W/m2 – using seawater as the saline
feed –, 200 000 m2 of membrane area are slated for every MW of “in-
stalled” osmotic power.
Osmosis as a means for energy production was proposed by Loeb
(1975), who used the term pressure retarded osmosis (PRO) to indi-
cate that osmotic volume flux is designed to function against pre-
pressurized saline water, that pressure constituting the operating
pressure for the hydraulic power scheme [9]. Clearly, the net poten-
tial for osmotic power generation increases with the osmotic pres-
sure of the saline feed solution. Small wonder, Loeb (who is also co-
inventor of the asymmetric cellulose acetate membrane) had the
Dead Sea in mind when proposing PRO.
62 3 Osmosis et cetera

Bibliography
[1] F. Rüdorff: Grundriss der Chemie, 15. Auflage. Verlag H. W. Müller, Berlin
1909.
[2] R. W. Field, loc. cit. Chap. 1, Ref. [11]. – S. Cross, Achieving 60 °Brix with
membrane technology. 49th Annual Meeting, Institute of Food Technolo-
gists, New Orleans, 1988.
[3] H. K. Lonsdale, U. Merten, R. L. Riley, Transport properties of cellulose
acetate osmotic membranes. J. Appl. Polymer Sci. 9 (1965) 1341–1362. –
U. Merten (ed.): Desalination by Reverse Osmosis. MIT Press, Cambridge,
MA, 1966.
[4] J. E. Cadotte, Evolution of composite reverse osmosis membranes. In: Mate-
rials Science of Synthetic Membranes (D. R. Lloyd, ed.), ACS Symposium
Series No. 269, American Chemical Society, 1982.
[5] Data courtesy M. Wilf. See also M. Wilf, C. Bartels, Optimization of seawater
RO systems design. Desalination 173 (2005) 1–12.
[6] International Atomic Energy Agency, IAEA–TECDOC–574, Vienna, 1990.
[7] M. Fazel, E. G. Darton, Performing a membrane autopsy. Desalination &
Water Reuse 11 (2002) 40–46.
[8] O. Levenspiel, N. de Nevers, The osmotic pump. Science 183 (1974) 157–160.
[9] S. Loeb, Osmotic power plants. Science 189 (1975) 654–655.
4 Membrane Filtration

There is good reason to believe that filtering (= straining) is as old as

brewing, Fig. 4.1. The medieval verb filtrare relates to feltrum, mean-
ing anything compacted to serve as filter medium; felt is compacted
wool or hair (“nonwovens” in filtration parlance).

4.1 On size and size exclusion

Filtration is convective discriminating mass transport of liquid mix-
tures or gaseous dispersions (aerosols) through porous barriers,
mass transport ideally being confined to the void space of the barri-
ers. Sieving refers to filtration of particulate matter; gaseous diffusion
is the term used when all components are gases. Discrimination is by
size. The common permeant in liquid membrane filtration is the sol-
vent: Water.
In aqueous membrane filtration effective (observed) solute size
usually differs from geometric (predicted) size as a result of interac-
tions between the solutes (which may be charged) and the barrier

Fig. 4.1. Time-tested: Gravity assisted dead end filtration in ancient Egypt.
64 4 Membrane Filtration

surface contacted (which may include the pore walls). Rarely there is
a snug fit between solute and pore on purely geometrical terms.
In terms of molecular mass, solute size encountered in membrane
filtration extends over five orders of magnitude, Table 2.1 and Ap-
pendix C present examples. Whether truly dissolved or microscopi-
cally dispersed (oftentimes a matter of semantics), pore sizes to
restrain solute passage are such that the force of gravity (Fig. 4.1) no
longer suffices to overcome the hydraulic resistance of the barrier.
Membrane filtration, accordingly, is pressure driven barrier separa-
tion of aqueous solutions, loosely grouped into a number of process
variants with reference to the size brackets of the solutes handled:
• nanofiltration (NF) 0.01−0.001 μm (< 10 nm);
• ultrafiltration (UF) 0.2−0.005 μm (5−200 nm);
• microfiltration (MF) 10−0.1 μm (> 100 nm).
Classifying solutes follows these categories. By a geometrical ar-
gument based on solute diffusivity in aqueous solution (Table 4.2),
solute diameter of nonelectrolytes relates to solute molecular mass
approximately thus:
d [ nm ] ≈ 0.13 × MW 1 3 (4.1)

As a token of reference, a solute diameter of 1 nm is roughly equi-

valent to a molecular weight of 500 g/mol; microsolutes below this
order register as “osmotically relevant”. Macromolecules, represen-
ted in the main by proteins, polysaccharides, natural rubber, synthe-
tic polymers, are characterized by their molecular mass. The spec-
trum of marker molecules presented in Appendix C describes the
mass range of applied ultrafiltration. Colloids are macromolecules in
aqueous dispersion, and are described by their effective size; in fact,
most anything dispersed in water before becoming optically detect-
able, that is, up to 200 nm (0.2 μm), is termed colloidal. If freely mo-
bile (dilute), dispersed macromolecules are referred to as sol; when
becoming “entangled” (concentrated), they turn into a gel. Still big-
ger by another order of magnitude, and definitely particulate, are
microorganisms and biological cells, living or dead (cell debris),
assigned to the 1 to 20 μm size range.
It is mostly proteins which are to be retained or fractionated by
ultrafiltration. Proteins are charged biopolymers; when viewed as
 4.1 On size and size exclusion 65

aqueous solutes, they are pictured as composed of a hydrophobic

“core” into which the sequence of constituent amino acids is folded,
outwardly studded with amino groups (positively charged) and car-
boxylic acid groups (negatively charged). Charge interaction with
the membrane tends to increase the effective size, resulting in elec-
trostatic retention to be higher than size retention.
Figure 4.2 depicts the operative range of pressure driven mem-
brane processes from reverse osmosis to ordinary particle filtration:
Operating pressure (indicative of hydraulic barrier resistance) rela-
tive to solute size to be retained. Reverse osmosis, phenomenologi-
cally a filtration effect (“hyperfiltration”), more appropriately is in-
terpreted as a solution-diffusion process on account of the prevai-
ling mode of barrier interference (Chapter 3.2).
What may appear like an orderly development towards ever more
delicate filtration capability in fact conceals a very different genesis.
Up to the time of Ferry’s 1936 review on ultrafiltration (Chapter 2,
Ref. [5]) the only “synthetic” film forming polymers were cellulose

Fig. 4.2. Operative range of pressure driven membrane processes: The

filtration spectrum. Increasing operating pressure signifies increasing hy-
draulic resistance of the barriers employed. From [1] with permission.
66 4 Membrane Filtration

esters (collodion, Sect. 6.1) and regenerated cellulose (cellophane),

artfully prepared into membranes of graded porosity for laboratory
use. Pore sizes from 5 μm down to 5 nm are reported, filtration of
pathogens was of foremost interest, flux was of no concern. Water
flux came into focus when, in the 1950’s, Reid, using dense cellulosic
films as thin as laboratory art could make them (< 10 μm), demon-
strated that strong electrolytes (salts) could be pressure-filtered, the
filtrate being demineralized water. Water flux in what was to become
reverse osmosis, however, remained too low to be of economic inte-
rest, Reid gave up.
Two events fostered the subsequent development:
• The discovery of the asymmetrically structured and thereby effec-
tively thin cellulose acetate membrane (Loeb and Sourirajan,
1960; Sect. 2.3.3), which opened the way to practicable water
fluxes under reverse osmosis conditions.
• The surge of polymer technology after World War II, accompa-
nied by a surge in polymer film technology. In particular, the
structural principle of the asymmetric membrane soon was ap-
plied to other polymers as well, creating synthetic microporous
membranes with a favorable aspect ratio (pore length to pore di-
ameter) in the process. What cellulose acetate was for reverse os-
mosis came to be polysulfone for ultrafiltration (Chapter 6).
Practical ultrafiltration (since 1965) thus is a follow-up to practical
reverse osmosis, as is practical gas separation (since 1980). Nanofil-
tration, originally an ill-appreciated foundling, now bridges the gap
between ultrafiltration and high retention reverse osmosis (a size
bracket into which virus and multivalent ions belong).

4.2 Liquid transport in membrane filtration

The model image of a porous barrier is a perforated sheet with
straight cylindrical pores of uniform diameter extending vertically to
the plane of the sheet: an “isoporous sieve”. The surface porosity
(open area) of such a barrier is π r2n, its volume porosity (void vol-
ume) is π r2n z (n = number of like pores; z = length of pore, equal to
barrier thickness). It is noted that, in terms of void fraction, surface
porosity conforms to volume porosity.
 4.2 Liquid transport in membrane filtration 67

Nowhere is reality so much more interesting than the model as in

the case of microporous structures. A pictorial record of the struc-
tural diversity of real porous barriers – polymeric, mineral, metallic
– is presented in Appendix E. As to polymers, it is noted that only
“glassy” or else highly crystalline polymers provide the structural
integrity required to maintain pores in liquid contact (Sect. 6.2). The
message at this point is that fairly well-defined molecular solutes are
being matched with the pore size distribution of actual barriers.
Nevertheless, assessing mass transport in and about porous bar-
riers retains the descriptive simplicity of the model barrier, adding
hindsight refinement to accommodate reality as needed. Central to
interpretation is the convective liquid volume flux ( Jv) as it is influ-
enced by the presence of solutes. The following situations need to be
considered:
• volume flux of pure water;
• sub-critical flux from aqueous solutions (partial solute retention);
• critical flux from aqueous solutions (complete solute retention);
and
• transport behavior of permeable solutes.
Pure water flux. – As hydraulic permeability pure water flux is
one of the parameters characterizing porous barriers. The range is
considerable; water flux commensurate with the porosities covered
by the filtration spectrum ranges from 10 to > 1000 L/h m2 (corre-
sponding to convection velocities of between 1 and > 100 cm/h). For
perspective, water flux at the 10 gfd threshold of reverse osmosis is
17 L/h m2 at prevailing pressure (Table 4.1).
An expression for water flux through pores is derived from the
Hagen-Poisseuille equation describing hydraulic pressure loss with-
in a capillary duct (laminar flow) as
ε d2 Δ p Δp
Jv = = Lp
32ητ z z
with geometrical porosity (4.2)
nπ d2
ε =
4
the trivial message being that there is no convective flux at Δp = 0
(and that the flux diminishes with increasing length of the duct,
68 4 Membrane Filtration

Table 4.1. Performance of ultrafiltration membranes: Water flux and rejection

of marker molecules as function of pore size rating. Reproduced from [2].

Nominal Apparent Water flux Rejection (%)

MW cutoff pore diameter at 3.7 bar ---------------------------------------------
2
(nm) (L/h m ) D-Alanin Sucrose Myoglobin IgM
000500 02.1 0017 15 70 > 95 > 98
001000 02.4 0034 00 50 > 95 > 98
010000 03.0 0102 00 25 > 95 > 98
010000 03.8 0935 00 00 > 80 > 98
030000 04.7 0850 0− 00 > 35 > 98
050000 06.6 0425 0− 0− > 20 > 98
100000 11 1105 0− 0− >− > 98
300000 48 2215 0− 0− >− > 98
D-Alanin, MW 89
Sucrose, MW 342
Myoglobin, MW 17500
IgM (Immunoglobulin), MW > 900000

which is the original teaching of Hagen-Poisseuille). Parameters

determining hydraulic permeability Lp according to Eq. 4.1 are: The
surface porosity of the model membrane (ε = n π d2 /4); the viscosity
of the liquid feed within the pore space, η ; an adjustable parameter τ
which symbolizes the fact that real pores are neither straight nor
uniform (“tortuosity factor”). Flux is seen to depend on pore diame-
ter to the fourth power, greatly amplifying the influence of the pore
size distribution towards higher than mean pore rating.

4.2.1 Concentration polarization

The scenario, unfolded in Table 4.1, is one of increasing pore size
rating (a vocabulary such as “nominal” and “apparent” indicating
leeway) meeting specified solutes in order of increasing molecular
mass. The aim of the following discussion is to describe liquid vol-
ume flux in membrane filtration ( Jv) by considering the transport
behavior of the solutes ( Ji), which thereby assume the role of mar-
ker species: The polarization model (also known as stagnant film
model). Refer to Fig. 4.3 for illustration and symbols used.
The smallest solutes encountered are ordinary salts, – and urea
(Table 2.1). They are presumed to pass freely in ultrafiltration
 4.2 Liquid transport in membrane filtration 69

Fig. 4.3. The polarization model (film model) in cross flow membrane filtra-
tion. Upper: concentration polarization; lower: gel polarization.

( ci′′= ci′ ), while being somewhat retained by nanofiltration mem-

branes ( ci′′< ci′ ). Practical implications hinted at are: The use of
70 4 Membrane Filtration

microporous membranes to demineralize macromolecular solutions,

– and to treat uremia by hemodialysis (Sect. 4.5).
As soon as some fraction of the solute is denied passage (by what-
ever mechanism invoked), several interrelated things happen
(Fig. 4.3, upper).
• Solute concentration in the vicinity of the rejecting surface in-
creases above feed level: Concentration polarization. As a result
an osmotic pressure gradient develops across the membrane, re-
ducing the available pressure head for liquid flux. The effect is ba-
sically similar to what is observed in reverse osmosis (Sect. 3.2.3),
and is represented by an analogous flux equation as
Jv = Lp (Δp − σ Δπ) (4.3)
(compare with Eq. 3.4 of the Merten model). Equation 4.3 describes
partial solute rejection by adjusting the influence of osmotic pres-
sure through a reflection coefficient (σ = 0 to 1), which is patterned
after the salt rejection of reverse osmosis. Even though osmotic pres-
sures created by macromolecules are small, when accumulating they
may reach the same order as the pressure applied in ultrafiltration.
• As a further result, the concentration gradient developing bet-
ween wall concentration (cw) and feed ( c′i ) incites a process of
back diffusion counter to the direction of convective solute flow.
The realm of back diffusion into the turbulent feed stream defines
the laminar boundary layer δ , its thickness taken as indicative of
the degree of concentration polarization, cw /cb > 1. The term bulk
concentration cb is introduced to identify the turbulent feed flow
regime under conditions of concentration polarization; it corre-
sponds to the concentration of the “well mixed” feed stream.
Steady state is characterized by a balance of solute flow comple-
ments, in which the total convective solute transport towards the
membrane (= Jvcb) is balanced by the fraction permeating ( = J v ci′′ )
plus the fraction returning to the bulk stream by (Fickian) diffusion
(= − Di dcw /dz). On integration the polarization equation is obtained,
derived in the early days of reverse osmosis [3],
cw − ci′′ ⎛Jδ⎞
= exp ⎜ v ⎟ (4.4)
cb − ci′′ ⎝ D ⎠
 4.2 Liquid transport in membrane filtration 71

Table 4.2. Indicative diffusion coefficients of nonelectrolytes in aqueous

solution as function of molecular size. Derived from [4].

Molecular weight Diameter Diffusion coefficient

2
(g/mol) (nm) (cm /s)
2.20 x 10−
5
0000010 00.29
0000100 00.62 0.70
0001000 01.32 0.25
0010000 02.85 0.11
0100000 06.2 0.05
1000000 13 0.025
d [nm] ≈ 0.13 x MW1/3

The dimensionless composite Jv δ /D represents the ratio of con-

vective to diffusive solute transport in the boundary layer. In a nut-
shell it contains all the information needed to describe the dynamics
of the membrane-solute system under consideration:
• The volume flux Jv relates to membrane porosity, high porosity
leading to high convective flux, in turn causing the wall concen-
tration cw to increase, in turn causing the thickness of the bound-
ary layer δ to grow.
• The diffusion coefficient D indicates solute molecular size, which
(a) determines the diffusive mobility, and (b) limits the solubility
of the solute species in water. Table 4.2 lists indicative diffusion
coefficients of nonelectrolytes over the size range encountered. It
is noted that the molecular diameter merely doubles as the mo-
lecular mass increases by an order of magnitude.
However: what is the solubility limit of colloidal macromolecules?

4.2.2 Gel polarization

There is no ordinary limit to the concentration of macromolecules in
aqueous solution (as would entail a separation into precipitate and
supernatant), however, there is a limiting situation: Gelation. Gelati-
on is dewatering of the macromolecular solution to the point where
it solidifies. In ultrafiltration that point is reached next to the reject-
ing surface (the “wall”) in the limit of complete solute rejection,
producing a gelatinous deposit which constitutes a secondary barrier
72 4 Membrane Filtration

to liquid volume flux on top of the original membrane (Fig. 4.3, lo-
wer). In practical ultrafiltration of macromolecular solutions (ima-
gine cheese whey), gel formation occurs within minutes, progressing
from entrance to exit region of the membrane stage, and is largely
irreversible.
In the formalism of the polarization model (Eq. 4.4) gelation
shows up as zero permeate concentration of the gel forming solute
( ci′′= 0 ), and is signified by replacing the wall concentration by
a hypothetical gel concentration (cw → cg), – gel polarization [5]:
cg ⎛Jδ ⎞
= exp ⎜ v ⎟
cb ⎝ D ⎠
and (4.5)
D ⎛ cg ⎞
Jv = ln ⎜ ⎟
δ ⎝ cb ⎠
Assessing gel polarization again is guided by the volume flux Jv ,
observing that – as a rule – the deposit is less permeable to water
than the porous barrier itself (the notion of a “dynamic mem-
brane”). The key quantity is the critical flux, defined as the lowest
flux at given solute concentration to result in an irreversible deposit
on the barrier surface. When followed as function of gradually in-
creasing pressure, flux increases until, at the occurence of complete
gel layer coverage, becoming independent of pressure (Fig. 4.4). If
pressurized further, the flux remains at the critical level, either by
the deposit growing thicker or by mechanical compaction; compac-
tion of the gelatinous deposit eventually may even cause the flux to
decline with pressure.
Equation 4.5 summarizes the parameters which determine critical
flux (respectively the conditions to avoid gel layer coverage): Solu-
tion-related (cb and D); process-related (δ ). In turn:
• Volume flux is predicted to decrease with the logarithm of bulk
feed concentration, cb . That dependence reflects the effect of in-
creasing permeate recovery along the length of a membrane stage
(Fig. 3.5), or else in batch dewatering, – up to the fictitious limit of
bulk gelation.
 4.2 Liquid transport in membrane filtration 73

Fig. 4.4. Ultrafiltration: Effect of pressure, solute bulk concentration (cb) and
hydrodynamic condition (k) on flux. Saline: isotonic salt solution
(Sect. 3.1.3); protein: albumin. Adapted from [5].

• The dependence of flux on solute diffusivity is a consequence

both of solute solubility affecting its inclination to gel and of sol-
ute mobility affecting back diffusion across the dynamic bound-
ary layer.
• The ratio of diffusion coefficient and boundary layer thickness,
k = D/δ (a “velocity” by dimension), is a mass transfer coefficient
describing the combined influence of solute dynamics (diffusion
coefficient) and fluid dynamics (cross flow rate) on solute trans-
port. Mass transfer coefficients are obtained summarily as slopes
of the flux versus log-bulk concentration curves, without specific
reference to solution properties (Eq. 4.5).
Microfiltration is different. – As solute size (and incumbent
pore size) increases beyond the realm of bona fide solubility, the
scenery changes. Within a dimensional reach of approximately 100
to 1000 nm (0.1 to 1 μm) a transition occurs from colloidal solution
to particulate dispersion. Indeed, it is not only biological material,
74 4 Membrane Filtration

witnessing the transition from macromolecules to microorganisms,

but also waterborne inorganic species like clay components, silica,
oxide-hydrates of heavy metals, and hitchhiking humic acids, which
make their appearance as filterable solutes. The silt density index
(SDI), which routinely records reverse osmosis feed water turbidity,
is based on microfiltration through a 0.45 μm standard membrane
filter.
As the nature of the solute changes, so does the conformation of
the membrane deposit. What is observed is a higher permeate (fil-
trate) flux than predicted by simple gel polarization as solute size
increases and solute response to hydrodynamic conditions changes.
In terms of the parameters of the gel polarization model (Eq. 4.5) it
appears that mass transfer coefficients increase abnormally, consid-
ering that aqueous diffusivities are supposed to even decrease with
molecular mass (Table 4.2). Several mechanisms to explain the en-
hanced transport of high molecular weight solutes away from the
membrane surface are being considered, among them hydrodynamic
lift forces (primarily affecting D) and shear-induced diffusion (pri-
marily affecting δ ), Eq. 4.5.

4.3 Solute transport in membrane filtration

The issue is partial solute rejection in ultrafiltration (as exemplified
in Table 4.1); it is a subject matter of considerable research attention
and never-ending speculation. Key terms in the discussion are pore
blockage and pore constriction [6]. When a singular solute species is
partially rejected, irregularities in membrane pore structure (pore
size distribution) and site-specific solute-polymer interactions are
held responsible. If solutes differing in size are to be fractionated,
barrier irregularities and molecular interactions add to size discri-
mination, limiting the fractionating prowess of ultrafiltration. As
a rule, to effect membrane protein fractionation requires molecular
sizes to differ by nearly an order of magnitude.
The gross solute separation capability of membrane filtration is
addressed thus: Ultrafiltration retains macromolecules while being
freely permeable to microsolutes. Microfiltration, in turn, permeates
macromolecules while retaining microorganisms and cells. Specific
applications follow the gross pattern.
 4.3 Solute transport in membrane filtration 75

Dialysis. – Dialysis is diffusive mass transport transmitted by the

pore fluid, and thus pertains to the fraction of solute which actually
enters the pore space of the membrane. The driving potential is lin-
ked to the solute concentration gradient between “entrance” and
“exit” of the pores, the exit concentration being identical to that of
the receiving phase (now called dialysate). Undesired osmotic flux of
solvent (in either direction) is quenched by maintaining isotonic
conditions on both sides of the membrane.
If the dialysate is continually renewed, the feed solution gradually
becomes depleted of the permeating solute, – it is “washed out”. Not
to exhaustion, to be sure, as the diffusion gradient diminishes along
with depletion; dialysis, as any concentration-driven mass transfer,
slows down asymptotically.
The obvious application of depletion dialysis is to free macromo-
lecular solutions of unwanted microsolutes under gentle operating
conditions; a conspicuous application is hemodialysis.
Diafiltration. – As witnessed by the duration of hemodialysis
treatment, dialysis is slow, and confined to low molecular weight so-
lutes. When a pressure is imparted on the feed solution, a convective
flow is added which is considerably more effective in transporting
solutes than diffusion alone. The process, now a hybrid of dialysis
and ultrafiltration (Eq. 4.6), concentrates the feed solution through
volume reduction, which, unless wanted, may require judiciously
replenishing the solvent lost into the dialysate.

d ci
J i = J v ci − Di (4.6)
dz
(the bar to denote an average solute concentration within the pore
fluid). As the feed pressure is raised, pore flow turns into straight
convection and the concentration gradient within the membrane
disappears: Plug flow. Solute separation in diafiltration under plug
flow conditions is equivalent to sieving at the porous membrane
surface. A necessary corollary in case of partial solute rejection is
again concentration polarization.
76 4 Membrane Filtration

4.4 Rating porous membranes

By a rule of error analysis, uncertainties multiply. When micropo-
rous barrieres are to be characterized (rated), the uncertainties in
question are the morphology of the barrier itself and the size assay of
the macrosolutes with which the membranes are challenged. If rarely
there is a rational fit between pore size and solute size, it is the influ-
ence of solute-polymer interaction in addition (it is recalled that
proteins are charged). Nevertheless, it is the observed ability to re-
tain macromolecules (by ultrafiltration) or microorganisms (by
microfiltration) which characterizes porous membranes.
Ultrafiltration membranes are described by nominal ratings. The
molecular weight cutoff (MWCO) is defined as the molecular weight
of a test solute (preferably a globular protein) which is 90% retained
by the membrane in question. The molecular weight range of “mar-
ker molecules” (Appendix C) conforms to the range of applied ul-
trafiltration, which extends from glucose (MW 180; permeable) to
immunoglobulins (MW > 900000; impermeable). To establish the re-
jection profile of an ultrafiltration membrane, a number of marker
molecules is selected to cover the entire rejection range of that mem-
brane; from the resulting curves of rejection versus molecular weight
the MWCO at 90% rejection is obtained graphically (and absolute
rejection is inferred), Fig. 4.5. As would be expected, the sharpness
of the rejection profile is a reflection on the narrowness of the pore
size distribution; further, at comparable mean rejection, the nominal
(and absolute) MWCO assumes higher values as the rejection profile
becomes more diffuse.
Sieving and clearance. Filtration seeks to separate solute from
solvent, ultrafiltration moreover solute i from solute j in the process.
If the solute is unwanted (as salts in reverse osmosis), the term is
“rejection”; if the solute is wanted (as proteins in ultrafiltration), the
term is “retention”. Whereas the MWCO characterizes individual
membranes, the sieving coefficient addresses individual solutes as
ratio of solute concentration in permeate (filtrate) and feed:
Si = ci′′/ ci′ = 1 − R (compare with Eq. 3.7). When applying ultrafiltra-
tion to discriminate between differently sized solutes, a separation
factor is formed: α = Si S j (analogous to Eq. 5.5). If the smaller
solute is freely permeable (marked to be removed), Si = 1, and the
separation factor reduces to the reciprocal of the sieving coefficient
 4.4 Rating porous membranes 77

Fig. 4.5. Rejection profiles of two ultrafiltration membranes with indication of

their nominal rating in terms of the 90% molecular weight cutoff. After [7].

of the retained macromolecule: α = 1/Sj . In typical ultrafiltration

practice, solutions (dispersions) of proteins and cells are to be freed
of microsolutes.
A graphical proof of consistency of ultrafiltration performance is
obtained by plotting the selectivity (as 1/Sj) of a given marker mole-
cule against the hydraulic permeability (Lp) of the membranes under
consideration. When falling at or near the “trade-off” curve, mem-
brane performance is considered adequate [8].
Hemodialysis (depletion dialysis) is highly selective solute-solute
separation at constant feed volume (Sect. 4.5). In addition to selecti-
vity, the mass transfer capacity of the membrane or hemodialyzer is
needed. It is expressed as clearance (plasma clearance) in terms of
the volume of blood which is completely cleared of a given uremic
toxin per unit time (mL/min). Not only is clearance specific to each
toxic species (to be established using appropriate marker molecules,
Table 2.1), clearance values also depend on the mode of operation:
Purely dialytic (diffusion) to purely plug flow (convection), Sect. 4.3.
Microfiltration membranes are described by absolute ratings.
This rating is based on the notion of complete rejection of microor-
ganisms in dead-end filtration, and thus specifically observes the
largest pore. Presentation is in terms of graded pore diameters rela-
78 4 Membrane Filtration

tive to the sizes of selected “marker microorganisms” (or substitute

marker species) which are (completely) retained; standard denota-
tions of microfiltration membranes based on marker filtration are
0.1 / 0.2 / 0.45 / 0.8 / 1.2 μm of (largest) pore diameter.
Microbial filtration, followed by cultivation to facilitate identify-
ing and counting of the microorganisms rounded up, has become
standard practice in drinking water sanitary assay. The method was
developed in Germany during World War II, when city water sup-
plies, endangered by bombing, needed rapid safety assessment.

4.5 Notable applications

Most membrane filtration operations are busily and profitably at
work, and present no more challenge than available membranes
(Appendix E) and supporting technology are able to cope with. Es-
tablished applications are: Sterile filtration in medical and beverage
operations; recovery of electrocoat paint; processing of milk prod-
ucts and fruit juices; process-specific wastewater treatment.
As to prevailing technique, ultrafiltration (like reverse osmosis) is
operated in the cross flow mode (high ratio of feed flow rate over
permeation rate). In microfiltration, the solids content of the feed
determines process configuration: Cross flow at solids content
> 0.5%, dead-end flow (the literal filtration mode) at solids content
below that rule of thumb figure (Eykamp). Accordingly, sterile filtra-
tion operates as dead-end filtration.
Two ramifications of applied membrane filtration deserve atten-
tion beyond: The membrane bioreactor (MBR), its future role in bio-
organic synthesis and wastewater management as yet unfathomed;
the artificial kidney (hemodialysis, HD), its phenomenal achieve-
ment encouraging continued research on membranes in life-suppor-
ting systems.
Membrane bioreactor. – The essence of a membrane bioreactor is
to conduct bioreactions (as synthesis or degradation) in direct prox-
imity to a semipermeable barrier, technically combining the steps of
reaction and product recovery with a hint at continuous operation.
Although not a priori confined to any particular type of barrier,
current membrane bioreactor development focuses on microporous
membranes in their capacity to contain microorganisms. The two
 4.5 Notable applications 79

directions of membrane biotechnology feasible are bioorganic syn-

thesis and biological wastewater treatment, respectively, – controlled
fermentation of biomass both.
• Bioorganic synthesis. It is a matter of anticipation whether mem-
brane biotechnology eventually will compete with petrochemistry
in providing organic base chemicals. The aim is to improve on
bioreactor productivity through removal of biotoxic metabolites,
facilitating product recovery (“downstream processing”) at the
same time (Sect. 5.6).
• Wastewater treatment. Microbial decomposition of organic sew-
age, replacing the conventional activated sludge/settling treatment
by an activated sludge bioreactor in combination with membrane
filtration [9]. In addition to high quality effluent (water reuse), ob-
jectives of the MBR technology in progress are to reduce treatment
time (reactor efficiency), quantity of sludge to be disposed of, and
plant acreage (“foot print”). As would be expected, the single most
pressing problem is membrane fouling [10].
Hemodialysis. – The human kidney processes about 1000 L of
aqueous solution every week. Even though falling behind in exchange
capacity, artificial membrane devices have come close to mimicking
the clearance function of the kidney, which is to remove biotoxic
metabolites from the blood stream. As a membrane separation proc-
ess, hemodialysis is governed by the molecular mass (size) of the
uremic toxins to be eliminated relative to the mass of serum proteins
to be retained. Diffusion alone transports solutes up to MW 1000
(hemodialysis proper), and is slow. On applying pressure, convective
transport is added, extending the mass transfer capacity to higher
MW solutes (like β2-microglobulin, MW 11800), ultimately turning
dialysis into ultrafiltration (hemofiltration) characterized by sub-
stantial elimination of solutes to MW 40000 while retaining essential
proteins (serum albumin, MW 69000). Required solute fractionation
thus is between MW 10000 and lower (permeating) and MW 60000
(retained), operating in an environment liable to protein fouling.
The classical membrane material for hemodialysis is regenerated
cellulose, almost as nature provides it (Sect. 6.2). More recent devel-
opments use synthetic high Tg polymers (PES, PSU; Appendix D),
which are hydrophilized by blending with polyvinylpyrrolidone
(PVP). It appears that a barrier surface simultaneously exhibiting
80 4 Membrane Filtration

hydrophilic and hydrophobic functions is a viable answer to the pro-

blems of biocompatibility and protein fouling in general [11]. As in
reverse osmosis and ultrafiltration, synthetic hemofiltration mem-
branes are asymmetrically structured for flux (Fig. E.8); pore size
distribution is narrow, centering around a pore diameter below
10 nm, which is in the nanofiltration range.
A typical hollow fiber dialyzer has a membrane area of 1.5 m2;
a typical patient sufferung from chronic kidney failure requires 150
treatments per year. At an estimated one million individuals so af-
flicted the membrane area to come into contact with life blood
amounts to well over 200 million m2 annually, at a price. It deserves
mention that the proceeds of this beneficial endeavor have inspired
a scientific award (the Crafoord Prize) to promote basic research in
areas which the Nobel Price does not recognize: Mathematics, geo-
sciences, biosciences, astronomy.

Bibliography
[1] T. Melin, R. Rautenbach, loc. cit. Chap.1, Ref. [10].
[2] Source: Amicon Corporation. Reproduced by M. Cheryan, M. A. Mehaia in:
Membrane Separations in Biotechnology (W. C. McGregor, ed.), Marcel
Dekker, New York, 1986.
[3] P. L. T. Brian in U. Merten (ed.): Desalination by Reverse Osmosis. MIT
Press, Cambridge, MA, 1966.
[4] R. H. Perry, C. H. Chilton: Chemical Engineer’s Handbook, 5th Edition.
McGraw-Hill, Auckland, 1974.
[5] W. F. Blatt, A. Dravid, A. S. Michaels, L. Nelson, Solute polarization and
cake formation in membrane ultrafiltration. In: Membrane Science and
Technology (J. E. Flynn, ed.), Plenum Press, New York, 1970.
[6] For example: C. C. Ho, A. L. Zydney, A combined pore blockage and cake
filtration model for protein fouling during microfiltration. J. Colloid Inter-
face Sci. 232 (2000) 389.
[7] M. Mulder, loc. cit. Chap. 1, Ref. [8].
[8] A. Mehta, A. L. Zydney, Permeability and selectivity analysis for ultrafiltra-
tion membranes. J. Membrane Sci. 249 (2005) 245–249.
[9] S. Judd: The MBR Book, Principles and Applications of Membrane Bioreac-
tors in Water and Wastewater Treatment. Elsevier, Oxford, 2006.
[10] P. Le-Clech, V. Chen, A. G. Fane, Fouling in membrane bioreactors used in
wastewater treatment. J. Membrane Sci. 284 (2006) 17–53.
[11] M. Storr, R. Deppisch, R. Buck, H. Goehl, The evolution of membranes for
hemodialysis. In: Biomedical Science and Technology (Hincal, Kas, eds.),
Plenum Press, New York, 1998.
5 Pervaporation versus Evaporation

5.1 Phenomenon and realization

Quote: In the course of some experiments on dialyzation, my assis-
tant, Mr. C. W. Eberlein, called my attention to the fact that a liquid
in a collodion bag, which was suspended in the air, evaporated, al-
though the bag was tightly closed. This is the original observation
reported by Kober in 1917 (Chap. 7, Ref. [12]). – Collodion is cellu-
lose nitrate, permeable to water (Sect. 6.1). A parallel to the loss of
water through packaging film or, for that matter, contact lenses
comes to mind.
Pervaporation is mass transfer from liquid to vapor across inter-
active permeable barriers. When applied to volatile liquid mixtures,
pervaporation results in a separation effect to be likened to that of
distillation. However, whereas selectivity in distillation is predictable
by the rules of equilibrium evaporation alone (vapor-liquid equilib-
rium, VLE), pervaporation additionally is influenced by specific
membrane-solvent interaction (barrier interference) which provides
access to unconventional (“difficult”) liquid separation effects.
The diagram of an experimental set-up in Fig. 5.1 illustrates the
working principle. The thermostated liquid feed is pumped at ambi-
ent pressure (p') across a membrane module, the reject stream being
recirculated into the reservoir. Pervaporation is effected by main-
taining a reduced pressure at the downstream side of the membrane
by means of a combination of cold trap and vacuum pump, causing
the permeate to evaporate as it emerges from the membrane (at
downstream pressure p"). The vaporized permeate is recondensed in
the cold trap.
In Kober’s experiment, a reduction of partial pressure of the va-
porizing permeate is achieved through dilution with air; “carrier gas
pervaporation” would be the corresponding process realization.
82 5 Pervaporation versus Evaporation

Fig. 5.1. Experimental set-up of vacuum pervaporation. T = temperature;

Q = feed flow rate; P = downstream pressure. Recirculating the reject
stream systematically extracts the “faster” feed component: Batch opera-
tion (akin to depletion dialysis).

5.2 Mass transport and selectivity

Candidate liquid mixtures are aqueous-organic solutions as charac-
terized in Sect. 2.1.2, the variant component being the multitude of
organic “solvents” (VOC’s = volatile organic compounds) coming as
high boilers or low boilers relative to the normal boiling temperature
of water. For transport analysis, the solution-diffusion model
(Sect. 2.3) is invoked, presupposing homogeneous polymer mem-
branes. This model is not wholly representative, as follows from the
fact that water, the smallest of liquid molecules at room tempera-
ture, effectively pervaporates through inorganic microporous barri-
ers (zeolites; Fig. 5.3). Also, Kober’s collodion bag usually counts as
mildly swollen microporous (cellulose nitrate, Sect. 6.1).
 5.2 Mass transport and selectivity 83

Single component mass transport in terms of the solution-

diffusion model is represented by the following equations, repeated
from Chapter 2:
cim Di
Flux Ji = Δμi (5.1/2.16)
RT z

Pi ′
Driving force Δμi = Vi Δ p + RT ln (5.2/2.13)
Pi ′′

It is noted that single component flux may refer to a pure liquid

permeating ( pio = saturation vapor pressure) or to the partial flux of
a specified component of a liquid mixture ( p′i = partial pressure in
feed mixture), – turning into vapor when emerging from the mem-
brane. Adaption of the solution-diffusion scheme to the peculiar
circumstances of pervaporation is as follows.
Permeability. – The membrane finds itself exposed to liquid feed
and gaseous permeate. Accordingly, sorption and swelling is fully
developed at the feed side only, whereas the permeate side is essen-
tially “dry”. Sorption and the sorption profile across the membrane
depend on polymer-solvent interaction. Under steady state operat-
ing conditions, the parameters determining membrane permeability
m
– permeant solubility (ci ), permeant diffusive mobility (Di), and
membrane thickness (z) – are no longer predictable without ad-
ditional information on the sorption situation. Two limiting si-
tuations, represented by two classes of permeable barriers, can be
identified:
• Low total sorption (linear or Henry sorption isotherm). The poly-
mers are highly crosslinked (“stiff ”), swelling and permeability
are correspondingly low with a definite preference for small per-
meant molecules, – water in particular. The polymers are referred
to as glassy or semicrystalline, their ideal realization being inor-
ganic molecular sieves (zeolites), which do not swell at all. – The
hydrophilic CA and PA membranes employed in reverse osmosis
(Table 3.2) are of the glassy type.
• High total sorption (positively nonlinear or Flory-Huggins sorp-
tion isotherm). A low degree of cross-linkage facilitates mem-
brane swelling extending far into the membrane, in turn enhan-
cing permeant mobility on the whole (Sect. 2.3). The polymers are
84 5 Pervaporation versus Evaporation

referred to as rubbery or elastomeric, preferential interaction is

with organic solutes. The ideal realization are liquid membranes
(in the form of supported liquid membranes, SLM) which, by vir-
tue of their being water insoluble, are designed to provide exclu-
sive passage for organics (Table 5.3).
Driving force. – The relevant contributions to the free energy as
driving force for membrane transport are the pressure difference
between liquid feed and vaporized permeate and the difference in
partial pressures of the permeating species, Eq. 5.2. Of these, the
external pressure gradient is of little concern as long as ambient feed
pressure is maintained, the difference between “ambient” (feed) and
“vacuum” (permeate) amounting to 1 bar at most, Sect. 5.5. The
partial pressure of the permeant species in liquid feed and vaporized
permeate is given, respectively, by Raoult’s law (Eq. 2.1) and Dal-
ton’s law (Eq. 2.4), – presuming that the downstream pressure is
sufficiently low to allow the permeate vapor to be treated as
a “permanent” gas. The following scheme summarizes the condition:
Feed (l) Permeate (v)
pi′ = xi γ i p
o
i
> pi′′ = yi p ′′ (5.3)

(xi = mol fraction of target component in feed solution; yi = same in

permeate vapor). It is noted that the phase change, which commonly
is assumed to be localized at the downstream interface of the memb-
rane, as in ordinary distillation requires heat of evaporation. That
heat is replenished from an external heat source via the thermo-
stated feed stream as indicated in Fig. 5.1.
The master flux equation (Sect. 2.3) when incorporating partial
pressures as driving force takes this form:

cim Di ⎛ x γ po ⎞
Ji = ln ⎜ i i i ⎟
z ⎝ yi p′′ ⎠
or, on rearrangement, (5.4)
c m D ⎛ γ po y ⎞
J i = i i ⎜ ln i i − ln i ⎟
z ⎝ p ′′ xi ⎠

valid with attention to the boundary conditions concerning mem-

brane permeability outlined above. Flux is seen to increase with in-
 5.2 Mass transport and selectivity 85

creasing activity of the liquid feed component (presuming positively

nonideal solution behavior, γi > 1), as well as with decreasing total
(gaseous) permeate pressure (p"). It is noted that the permeate pres-
sure is technically influenced by the rate of vapor transfer from
membrane to condenser, that rate thus constituting an influence
parameter in practical pervaporation.
In single component (pure liquid) pervaporation, γ = 1 and the
mol fractions in feed and permeate each are unity, reducing the dri-
ving potential to the difference between saturation vapor pressure
(liquid) and permeate pressure (gas), Δ p = pio − p ′′ .
Selectivity. – As a measure of the “success” of the separation op-
eration, recording selectivity relies on a comparison of the analytical
compositions of feed and permeate, – both changing systematically
in the process: As batch operation on a time scale (as in Fig. 5.1), or
along the extent of a membrane separation stage (as in Fig. 3.5).
Observed selectivity is influenced by process dynamics. In perva-
poration it is concentration polarization and the swelling state of the
membrane affecting transport rates. It is therefore necessary to dis-
tinguish between the “intrinsic” selectivity of the barrier, and its
(lesser) performance under operating conditions. As in cross flow
membrane filtration, the feed mixture outside the reach of process
dynamics is referred to as bulk, sometimes with the attribute “well
mixed”. Several ways of expressing practical selectivity are in use
(i, j = liquid mixture components):
• The first compares directly with binary distillation, accounting for
the separation effect in terms of a separation factor αij as depicted
by a McCabe-Thiele diagram of vapor composition versus liquid
feed composition (i = faster moving component).

(ci / c j )′′ ci′′(1 − ci′ ) ( pi / p j )′′

α ij = = = (5.5)
(ci / c j )′ ci′(1 − ci′′) (ci / c j )′

As a fitting example, Fig. 5.2 compares evaporation (at reduced pres-

sure) and pervaporation (through a hydrophilic CA membrane) of
water-ethanol using water fractions as coordinates. It is observed
that hydrophilic pervaporation favors the higher boiling water, en-
richment being particularly effective near the azeotropic compositi-
on of the feed at about 4 w-% H2O.
86 5 Pervaporation versus Evaporation

Fig. 5.2. Hydrophilic pervaporation (CA membrane) versus low pressure

evaporation of water-ethanol in the presentation of a McCabe-Thiele dia-
gram. Thin lines are curves of constant separation factor relative to the
diagonal at which αij = 1 (no separation).

• A practical measure of selectivity takes heed of the fact that, as

a rule, pervaporation is applied to separate (enrich) minority
components out of dilute feed solutions. If that component is the
wanted species, its enrichment is of interest, expressed as ratio of
concentrations (in weight or volume or molar units) in permeate
over feed: The enrichment factor βi = ci′′/ ci′ > 1 , which relates to
the solute rejection R of reverse osmosis (Eq. 3.7), – enrichment
seen as negative rejection (i = minority species):
 5.3 The capability of pervaporation 87

α ij =
βi
=
( ci′′ ci′) and βi = 1 − R (5.6)
βj ( )
c ′′j c ′j

With some arithmetic ado, separation factor and enrichment factor

are interconvertible; at very low feed concentration (at the far ends
of the composition range) they come close, xi → 0 rendering
α ij → βi .
• Enrichment of high boiling/low solubility organics from water by
organophilic pervaporation easily extends beyond the miscibility
limit of the components, leading to phase separation (“demi-
xing”) of the permeate once it becomes liquefied. The permeate
now consists of two coexisting liquid phases of different composi-
tion and density: One an aqueous solution saturated with (little)
organic, the other an organic solution saturated with (little) water.
“Natural enrichment” under conditions of phase separation is the
proportion of the organic component in the organic-rich phase. It
is temperature dependent, just as the mutual solubility of the
components is (Sect. 5.5 has examples).
• If the minority feed component is unwanted, to be removed in
refining a wanted product, loss of product into the permeate is
a realistic indication of selectivity. A relevant example is residual
alcohol in the aqueous permeate when dehydrating alcohol-water
mixtures (Sect. 5.4.1).

5.3 The capability of pervaporation

Figure 5.2 is an example of barrier interference. What a McCabe-
Thiele diagram does not show is rates: Pervaporation is “slow” in
comparison to “instantaneous” evaporation, raising the question
under which conditions the separation capability of pervaporation
may be used to advantage. With a view at aqueous-organic liquid
separations, these are the facets to be considered:
• Economy of affinity. Prevailing modes of liquid-polymer interac-
tion dictate two directions of process design: Selective water per-
meation through hydrophilic barriers, or selective permeation of
organics using organophilic barriers. By the same argument, using
88 5 Pervaporation versus Evaporation

membranes to separate equimolar (“even”) liquid aqueous-

organic mixtures is an unlikely proposition.
• Economy of mass transport. Extending the argument, mass trans-
port economics suggests applying pervaporation to dilute feed so-
lutions, treating the respective minority component to be the
preferentially permeating one.
• Nonequilibrium separations. Preferential sorption is the key to
uncommon separation effects: Selective transport of water by
glassy barriers to separate constant boiling mixtures (azeotropes);
enrichment, aiming at recovery, of high boiling organics by rub-
bery polymers (“high boiler pervaporation”).
• Bioseparations. Membranes allow for gentle (low temperature)
and chemically noninterfering liquid separations such as are de-
sirable in bioprocessing; direct coupling to “life” fermenters is
a distinct possibility (membrane bioreactor, Sect. 4.5).
In the following, the two principal directions of aqeous-organic
pervaporative separation – hydrophilic and organophilic – are pre-
sented and illustrated by case studies.

5.4 Hydrophilic pervaporation

5.4.1 General observations
The aim is to dehydrate organic solvents with particular attention to
dewatering azeotropes (constant boiling mixtures). Process design is
dictated by the predicament of all barriers: They are “leaky”. Even
the most water-selective membrane will not completely block the
passage of organics, turning the nonvalue permeate (water) into
a problem waste. Dense hydrophilic membranes, insensitive to or-
ganic attack, are thus called for. To offset the limited permeability
(low flux) inherent to dense membranes, thin (asymmetrically struc-
tured) membranes are employed, preferably in combination with
elevated temperatures. As a semi-standard in hydrophilic pervapo-
ration have emerged composite membranes of crosslinked poly-
vinylalcohol (PVAL) on a microporous support, the thickness of the
active layer being in the micrometer range.
 5.4 Hydrophilic pervaporation 89

The conceptual answer to limited water permeability is vapor

permeation through microporous inorganic barriers, raising the feed
temperature to above boiling, thereby enhancing the rate of mass
transport and eliminating the phase change associated with liquid
pervaporation. The situation is illustrated in Fig. 5.3, comparing
pervaporative flux as function of water content of isopropanol-water
mixtures of a polymer membrane (PVAL) and a zeolite barrier
(NaA) at the temperatures indicated (permeate pressure
p ′′ ≈ 20 mbar).
This is the information conveyed.
Performance. – Pervaporation is “slow” by comparison. For per-
spective, even at a temperature of 90°C the flux of the PVAL com-
posite membrane at azeotropic feed composition (approximately

Fig. 5.3. Dehydration of isopropanol by pervaporation and vapor permeation:

Flux of a polymer membrane (PVAL composite) and a zeolite barrier (NaA)
as function of feed composition at the temperatures indicated [1]. Prevailing
curvature reflects Langmuir sorption isotherms.
90 5 Pervaporation versus Evaporation

10 w-% H2O) is one tenth of what is considered to be an economic

lower limit of water flux in reverse osmosis: 40 against 400 L/d m2.
The superior performance of the zeolite membrane is apparent,
especially at the highest temperature shown (120°C), which is likely
beyond the long-term reach of most polymeric membranes. Selectiv-
ity is judged by the amount of carried-over isopropanol in the aque-
ous permeate (reported at below 1% in the example shown).
Transport modeling. – As is the design purpose, the process oper-
ates against a diminishing water content of the feed (reading the
curves from right to left). This brings to focus the concentration
dependence of the flux of the target species, designed to be the pref-
erentially permeating minority component of the feed solution. That
dependence is related to shape and slope of the sorption isotherm of
the permeant-membrane system considered, identifying sorption to
be a key transport parameter in pervaporation ( cim in Eq. 5.4). At
sufficiently low sorption, that is to say, as long as the sorption iso-
therms are linear or of the Henry type, flux varies linearly with feed
concentration. The system H2O-PVAL in Fig. 5.3 shows such a linear
dependence throughout, indicating “well behaved” sorption in ac-
cordance with a regular (Nernst type) distribution of water between
isopropanol and membrane.
Permeance. – A linear dependence of flux on feed concentration
conceptually corresponds to a permeance describing the proportion-
ality between flux and driving force at given membrane and given
operating conditions (Sect. 1.6). Deviations from linearity – now
seen as deviation from well-behaved permeance – occur in both di-
rections, as documented in Figs. 5.3 (for hydrophilic) and 5.6 (for
organophilic pervaporation). The flux curves for the zeolite barrier,
when likened to sorption isotherms, resemble the dual sorption iso-
therms known in gas sorption, suggesting pervaporation respective-
ly vapor permeation across non-swelling barriers to be variants of
gas permeation. By contrast, pervaporation of organic solutes
through high-swelling (“rubbery”) membranes shows a higher than
linear dependence of organic flux on feed concentration akin to
Flory-Huggins isotherms, ascribed to a plasticizing effect which si-
multaneously enhances sorption capacity (polymer swelling) and
overall permeant mobility (Sect. 2.3.1).
 5.4 Hydrophilic pervaporation 91

5.4.2 Pervaporation versus reverse osmosis

Hydrophilic pervaporation and reverse osmosis are related mem-
brane processes in that both are designed to preferentially transport
water using basically similar polymeric membranes, both interpret-
ing water transport by a solution-diffusion mechanism (taking ex-
ception to inorganic barriers). Conceptual differences notwithstan-
ding, it is instructive to compare the driving forces and resultant
fluxes for the two processes.
For a model comparison, pure water transport is considered,
identifying the driving force as gradient of water activity between
feed and permeate interfaces inside the membrane [ ai′ (m ) and
ai′′(m ) = boundary activities within the membrane]. The argument
is based on the swelling state of the membrane under process condi-
tions as it influences permeant activity. Viewing water as incom-
pressible, the activity of the liquid feed is unity for both processes,
a′i = 1.
In pervaporation there is equilibrium swelling at the feed side of
the membrane [ a′i = a′i (m)], declining to a state of near-zero swell-
ing (“dryness”) at the permeate side [ a′′ i
(m) < a′i (m)]. The gradient
inbetween is pictured naively as linear, as provides for Fickian diffu-
sion. (It is noted that the swelling profile in organophilic pervapora-
tion is decidedly nonlinear, Fig. 2.2). The permeate is water vapor, its
activity defined as ratio of downstream pressure over saturation
(pure component) vapor pressure of water, a′′ i
= p ′′ / pio , ranging
from 1 → 0 as the downstream pressure is lowered. At equilibrium
[ a′′i
(m) = a′′
i
] and the relevant activity gradient is that between liq-
uid feed and gaseous permeate, namely [1 − a′′ i
(m)].
In reverse osmosis the membrane is considered uniformly swollen
throughout (isotropic swelling), implying level water activity from
feed side to permeate side within the membrane, [ a′i (m) = a′′ i
(m)].
Pressurizing the feed and, in turn, the membrane will lower the wa-
ter activity inside the membrane by compressing the polymer ma-
trix; water is literally “squeezed out”, flowing off freely. The relevant
activity gradient, therefore, is that between liquid feed and com-
pacted membrane, once again [1 − a′′ i
(m)].
Minimizing the downstream boundary activity of the permeant,
a′′
i
(m), is seen to be the common handle to enhance the driving
potential. Eqs. 5.7 relate the boundary activity to the forces actually
92 5 Pervaporation versus Evaporation

employed: Lowering the pressure of the gaseous permeate in per-

vaporation ( p′′ (v) → min); raising the hydraulic feed pressure in
reverse osmosis ( p′ (l) → max).
p ′′
Pervaporation ai′′(m ) =
pio
(5.7)
⎛ V ⎞
Reverse osmosis ai′′( m ) = exp ⎜ − i Δ p ⎟
⎝ RT ⎠
When comparing the driving forces to effect equal water flux,
a surprisingly large advantage in favor of pervaporation is suggested.
For example: Theoretically, a reverse osmosis feed pressure ( p′ ) of
400 bar is needed to produce the same water flux as a pervaporation
permeate pressure ( p′′ ) of 5 mbar (one fifth of the saturation vapor
pressure of water, p ′′ /p°) will. Actually, observed pervaporation
fluxes consistently are lower than reverse osmosis fluxes. Analysis of
this seeming anomaly is a lesson in barrier interference. Three in-
fluence factors are held responsible:
• Trivia first: Membrane thickness. In form of the active layer of
flat-sheet composite membranes, the thickness of current hydro-
philic pervaporation membranes (PVAL) is higher by an order of
magnitude than that of composite reverse osmosis membranes
(PA), – 2 μm as against 0.2 μm for illustration; water permeance is
consequently lower.
• Membrane swelling in pervaporation declines from “fully swol-
len” at the feed side to “virtually dry” at the permeate side, the
swelling (= sorption) profile inbetween being a function of the in-
tensity of molecular interaction between permeant(s) and poly-
mer. Irrespective of the shape of the sorption profile, overall wa-
ter sorption (cim) is lowered and water mobility hindered by
regressive sorption, resulting in a transport resistance higher than
would be encountered under level swelling conditions, as in re-
verse osmosis.
• Structural pressure loss. Asymmetric (thin) membranes are me-
chanically stabilized by a microporous support, onto which the
membrane is applied skin-like (composite membrane structure).
Evaporation of the permeants occurs into the supporting sub-
 5.5 Organophilic pervaporation 93

structure, the associated volume increase, being inversely pro-

portional to pressure, reducing the available driving potential
(peff > p").

5.5 Organophilic pervaporation

5.5.1 General observations
While water is the single target permeant in hydrophilic pervapora-
tion, organophilic pervaporation is as diverse as there are volatile
organic solutes (VOC’s), – to be recovered (as product) or removed
(as contaminant) from aqueous solutions. If biosynthesis is to be
a guide, an upper limit to organic concentration encountered is
of the order of 10%, as witnessed by the prevalent ethanol concen-
tration in wine; lowest concentration is in the ppm range, found,
for example, with aroma compounds or else trace industrial water
pollutants.
The membranes of choice are rubbery (elastomeric) polymers
disposed to swelling when exposed to organic solvents. The proto-
type of a rubbery hydrophobic polymer is silicone rubber (poly-
dimethylsiloxane, PDMS), available in sheet form since 1951, first
discovered by Kammermeyer for its extraordinary permeability to
gases (oxygen). Other noteworthy elastomers are polyether-
blockamide (PEBA) and polyurethane (PUR), both “segment-
elastomeric” polymers (Sect. 6.2).
Performance. – All polymer films, even those dubbed “hydropho-
bic”, are permeable to water to a degree, still more so when swollen.
As a consequence, while hydrophilic pervaporation may strife for
exclusive water selectivity, organophilic pervaporation enriches the
organic solute against an unavoidable undercurrent of water. Con-
comitant water flux, in turn, is an indicatior for the swelling state of
a rubbery membrane: Enhancement of permeability due to swelling
is immanent when, as function of organic feed concentration, water
flux starts to increase (exemplified by the shaded area in Fig. 5.6).
The separation effect under these conditions is determined by com-
paring the organic target concentration in two aqueous process solu-
tions: One the permeate (following condensation), the other the bulk
feed; the ratio of the two is the organic enrichment ( βi = ci′′ci′ > 1 ).
94 5 Pervaporation versus Evaporation

Activity coefficients. – From the statement of driving force

(Eq. 5.3) organic enrichment in terms of mol fractions is obtained as
yi γ po
βi = ≤ i i (5.8)
xi p ′′

Solute enrichment may thus be estimated from the thermodynamic

condition of the feed (Henry coefficient, γ i pio ) and from process
conditions (downstream pressure, p′′ ), without prior knowledge of
membrane properties. The key parameter for further assessment is
the activity coefficient of the organic solute, both in the aqueous feed
solution and in the membrane (polymer) phase. Being concentration
dependent (see Fig. 2.1), activity coefficients for consistency are re-
corded at “infinite dilution”, that is, at their numerical highest. Ex-
cepting the few instances of negative deviation from Raoult’s law
(notably carboxylic acids), the solutes under consideration form
positively nonideal solutions with water, activity coefficients ranging
from γ ≈ 2 for methanol (the species “closest” to water) to several 104
for nonpolar solutes. The sheer magnitude of this range is notewor-
thy; it is paralleled by a vast range of nominal selectivities as exem-
plified by the data of Table 5.1 (observed with an uncommonly thick

Table 5.1. Organophilic pervaporation of aliphatic alcohols: Correlation be-

tween aqueous solution activity coefficient and separation factor. Beginning
with
n-butanol miscibility with water is limited. Derived from [2].

Alcohol Boiling point Activity coefficient Separation

(°C) factor
(infinite dilution) (αij)
methanol 064.5 000002 0009
ethanol 078.3 000005 0017
propanol 097.2 000015 0067
butanol 117.5 000050 0074
pentanol 138 000200 0265
hexanol 157 001000 1050
heptanol 176 003000 1600
PDMS membrane of 200 μm thickness; 25°C.
Feed concentration: 1 vol-%; hexanol 0.5 vol-%; heptanol 0.1 vol-%
 5.5 Organophilic pervaporation 95

membrane). It is recalled that solubility and activity correlate in-

versely; for sparingly soluble organics in water the activity coeffi-
cient is the inverse of molar solubility, and vice versa (Eq. 2.2).
Activity coefficients for the system ethanol-water (completely
miscible) are presented in Appendix A; Table 5.1 has activity coeffi-
cients for the homologous series of aliphatic alcohols (going far
beyond the realm of miscibility); Table 5.2 has activity coefficients
for the four isomeric butanols (near the limit of aqueous miscibility).
When contemplating the influence of activity coefficients on sol-
ute relocation (transfer) in liquid separations, three practical situa-
tions merit attention:
• Relocation into the vapor phase: Evaporation is dictated (and
predictable) by VLE; the significance of the Henry coefficient
( γ i pio ) is to describe solute volatility relative to the vapor pres-
sure of the pure component.
• A variant of evaporation is steam distillation: Certain high boiling
organics are “volatile with steam”, meaning that they are volatil-
ized by steam blown into the feed mixture. Given a low level of
molecular interaction (as witnessed by low water solubility), the
partial pressures of water and organic in the extracted vapor add
up to the sum of the pure component vapor pressures. Organic
mol fraction in the vapor phase under these conditions is deter-
mined by the vapor pressure of the organic species according to
xorg ≈ porgo o
/ pwater (Eq. 2.3). Condensation of the vapor immediately
results in phase separation, the coexisting phases being the two
components as nearly pure as their limited mutual miscibility will
allow them to be. The organic phase, saturated with water, repre-
sents the “natural enrichment” of steam distillation.
• Relocation into a polymeric phase (and thence to vapor): Perva-
poration, dictated by the distribution of organic solute between
aqueous feed and polymeric membrane (Eq. 2.18). Distribution in
favor of the polymer phase has two consequences, which are
counteracting: (a) Organic enrichment is apt to be higher than
suggested by volatility: High boiler pervaporation (Sect. 5.5.2); (b)
a boundary layer depleted of organic solute will develop, reducing
solute permeance: Concentration polarization.
96 5 Pervaporation versus Evaporation

5.5.2 High boiler pervaporation

What is intriguing about organophilic pervaporation: It enables
aqueous-organic separations which seemingly defy the limitation of
volatility of the organic species, even at gentle temperatures. High
activity coefficients (low solubility) of the organic solutes in water,
combined with preferential sorption by the membrane polymer, re-
sult in high (occasionally extreme) enrichment of low volatile organ-
ics from aqueous solution. The case studies below illustrate some
aspects of high boiler pervaporation.
Mostly water, after all. – High enrichment notwithstanding, most
of the permeate still will be water. The following example is taken
from the repertoire of microbially accessible (and thereby “natural”)
aroma compounds. γ-Decalactone (MW 170.2; bp. 281°C) is an
aroma compound of “fruity” fragrance (peach), forming a highly
nonideal solution with water (activity coefficient ~ 14000). Vacuum
distillation (VLE) of the aqueous solution has but little effect on
composition, a high activity coefficient just about offsetting the low
vapor pressure ( γ i pio ≈ 1). Pervaporation, on the other hand, enri-
ches γ-decalactone from a feed concentration of 100 ppm (0.01%) to
a permeate concentration of nearly 3% (PEBA; 40°C). The necessary
information is that even at a 300-fold enrichment of the organic tar-
get species, 97% of the permeate still is water. Organic flux, for per-
spective, is low: 0.8 g/h m2 of γ-decalactone in the example presented.
Phase separation. – High boiler enrichment, as a rule, leads to
phase separation (demixing) of the condensed permeate in accor-
dance with the phase diagram (diagram of miscibility versus tem-
perature) of the aqueous-organic system under consideration. Phase
separation may be employed as a means to improve the overall se-
lectivity of high boiler pervaporation.
Phenol (MW 94.1; bp. 182°C) is a rewarding study object, first of
all for its eminent industrial relevance and water polluting prowess,
but also on account of its relatively high water solubility thanks to
the weakly acidic function of the OH-group. At a temperature of
30°C phase separation commences at a phenol concentration of
about 10 w-%, yielding two coexisting liquid phases,
• “phenol in water”, approximately 10% phenol; and
• “water in phenol”, approximately 70% phenol,
 5.5 Organophilic pervaporation 97

Fig. 5.4. Two-stage pervaporation/phase separation process scheme for

phenol recovery at the enrichment level of “water in phenol” [3].

the phenol-rich phase precipitating due to higher density. A process

scheme combining pervaporation and phase separation to effect
phenol enrichment to the level of “water in phenol” is presented in
Fig. 5.4. The permeate of the first pervaporation stage, enriched to
a phenol concentration of > 10%, undergoes phase separation to
yield the two coexisting phenolic solutions indicated. Of these, the
fraction “water in phenol” is withdrawn as target product while the
fraction “phenol in water” is subjected to a second pervaporation/
phase separation stage, once more producing “water in phenol” to
be recovered, and “phenol in water” being recycled as shown (a feed
and bleed situation). Nominal phenol enrichment is the ratio of
concentration of the phenol-rich phase (fixed by the phase diagram)
to that of the prevailing feed solution.
Salting-out. – Upon addition of salt to aqueous solutions the ac-
tivity of water is lowered (colligative properties, Sect. 2.1.1), whereas
the activity of dissolved nonelectrolytes increases, both effects hav-
ing the same origin: a decrease in “availability” of water. In keeping
with the influence of activity coefficients on pervaporation perfor-
mance, salting-out is expected to enhance both organic flux and
organic enrichment. Confirmation is presented in Fig. 5.5, again
using phenol as sample solute.
The effect is clearly confined to nonelectrolytes, as again the per-
vaporation behavior of phenol demonstrates. With increasing pH
value phenol (a weak acid) gradually converts into ionic phenolate:
C6H5OH + NaOH ⇔ [C6H5O−] Na + H2O. Pervaporative phenol en-
+
98 5 Pervaporation versus Evaporation

Fig. 5.5. Salting-out to enhance pervaporation: Effect of electrolyte concen-

tration on flux and enrichment of phenol. Feed 400 ppm; PEBA membrane;
50°C [4].

richment as function of pH diminishes precisely as the concentra-

tion of undissociated phenol in equilibrium with phenolate does.
Conversely, phenolate is rejected by reverse osmosis whereas phenol
is not. The cross-over between phenol enrichment and phenolate
rejection occurs at a pH value equal to the dissociation constant,
pH = pKa = 10.4.
Concentration polarization. – Concentration polarization is
a phenomenon of process dynamics. Preferential sorption under
conditions of mass transfer will lead to accelerated depletion of or-
ganic solute near the feed-membrane interface. The causality of high
solute activity coefficient in the feed leading to high solute sorption
by the membrane (Eq. 2.18), in turn leading to high solute enrich-
ment on pervaporation suggests a correlation between organic en-
richment and concentration polarization.
Estimation of the significance of concentration polarization is ba-
sed on a comparison of “actual” (experimental) solute enrichment
data with the “intrinsic” enrichment of a given membrane. Actual
enrichment is the ratio of solute concentrations in permeate and
bulk feed, and is affected by boundary layer depletion; intrinsic en-
 5.5 Organophilic pervaporation 99

richment is the highest achievable by the membrane in the absence

of boundary layer effects. Evaluation of the polarization equation
(Eq. 4.4) to emphasize the concentration polarization modulus
cw /cb ≤ 1 yields the following expression [5]
cw exp ( J v δ / D )
= (5.9)
cb 1 + βo ⎡⎣exp ( J v δ / D ) − 1⎤⎦
showing concentration polarization to be dependent on intrinsic
enrichment (βo) and process conditions through the ratio of convec-
tive and diffusive mass transport in the boundary layer (Jv δ / D). Per-
vaporation being “slow”, it is the mass transfer coefficient (Eq. 4.5)
and the intrinsic enrichment which are of concern. Surveying avail-
able data it appears that concentration polarization in organophilic
pervaporation requires attention at solute enrichment higher than
β ≈ 100. It is noted that this proviso necessarily limits the feed con-
centration of concern to below 1%. Primarily affected by concentra-
tion polarization, therefore, are sparsely soluble organic solutes such
as aroma compounds and higher alcohols (Table 5.1), which may
exhibit extreme nominal enrichment.
On the other hand, low MW commodity chemicals of biosynthetic
promise, including lower alcohols, are far from the critical limit,
pervaporative recovery from the respective fermentation broths
proceeding at more modest enrichment: Ethanol by a factor of about
5; n-butanol by a factor of about 30 (PDMS membranes).

5.5.3 Butanol, a glimpse at bioseparations

When Weizmann introduced ABE (acetone-butanol-ethanol) fer-
mentation in 1912, the incentive was acetone, – badly needed by the
military for the manufacture of smokeless gunpowder according to
one of Nobel’s countless patents. Ultrafiltration using asbestos fiber
devices to clarify fermenter broths was known and practiced at the
time, however, recovery of volatile metabolites was strictly by distil-
lation from batch fermentation. Butanol (n-butanol, Table 5.2), al-
ready then recognized as a precursor to synthetic rubber, today
would be the product of value.
Today, Weizmann would have considered a membrane bioreactor
(MBR; Sect. 4.3), product recovery by pervaporation inclusive.
A sampling of butanol pervaporation, laboratory scale, follows.
Table 5.2 Physical constants of the structural isomers of butanol (C4H9OH; MW 74.12).
100

Boiling point Vapor pressure Solubility in water Activity Water content

at 20°C at 20°C coefficient in azeotrope
at 20°C
(°C) (mbar) (w-%) (w-%) (°C)

n-butanol •-•-•-•-OH 117.5 05.7 07.7 41.1 42.5 (93)

iso-butanol •-•-•-OH 108 12 08.5 44.4 32 (90)

•-•
5 Pervaporation versus Evaporation

sec-butanol •-•-•-• 099.5 16 12.5 20.8 30 (88)

•-•-OH
•-•
tert-butanol •-•-OH 082.5 41 miscible 11.4 11.8 (80)

•-•
 5.5 Organophilic pervaporation 101

There are four structural isomers of butyl alcohol (C4H9OH; MW

74.1); Table 5.2 summarizes the physical constants. Two of the alco-
hols are seen to be “high boilers”; water miscibility and aqueous ac-
tivity correlate inversely, only tert-butanol, closest to water in activ-
ity, being completely soluble; all butanols form positive azeotropes
with water. Why microbial action exhibits a preference for n-butanol
(soon to develop “product inhibition”) is a matter of speculation.
Limited miscibility leads to phase separation. The coexisting pha-
ses at room temperature in case of n-butanol are
• “butanol in water”, 7.7% butanol;
• “water in butanol”, exceeding 60% butanol,
the latter representing the “natural limit” of enrichment.

Fig. 5.6. Organophilic pervaporation of the butanol isomers: Organic flux

(curves) and realm of concurrent water flux (shaded area) as function of
organic feed concentration (PEBA membrane; 50°C) [6]. Prevailing curva-
ture reflects Flory-Huggins sorption isotherms.
102 5 Pervaporation versus Evaporation

Figure 5.6 summarizes the organophilic pervaporation of the bu-

tanol isomers from aqueous solution as flux versus feed concentra-
tion (permeance). [Fig. 5.3 is an analogous presentation for hydro-
philic pervaporation]. What is observed is an illustration of per-
vaporation thermodynamics:
• The order of fluxes is counter to the order of both boiling points
and solubility, the highest boiling and least soluble n-butanol
showing highest permeance.
• Up to an organic feed concentration of ~ 1 w-% (the presumed
limit to biotoxicity of n-butanol in ABE fermentation) fluxes in-
crease linearly, the slopes reflecting the order of Henry coeffi-
cients; thereafter, flux increase is stronger than linear.
• Water flux, high to begin with, is practically independent of or-
ganic feed concentration up to the 1% limit, increasing thereafter
in compliance with polymer plasticization; true to the general pat-
tern, the effect is least pronounced with tert-butanol.
Table 5.3 compares the performance of three “classical” elastom-
ers, a polymer blend incorporating silicalite (a hydrophobic mole-
cular sieve), and a supported liquid membrane (oleyl alcohol,
C18H35OH) with equilibrium evaporation in n-butanol enrichment.
While organic flux (= flow density) would be a key parameter in
establishing the membrane area of a fictitious membrane bioreactor,

Table 5.3. Pervaporation of n-butanol through organophilic membranes and

a liquid membrane, comparison with evaporation. Data in brackets esti-
mated from published accounts. Taken from [6] and [7].
2
Membrane Flux [g/h m ] Flux ratio Selectivity
BuOH H 2O H2O/BuOH αij βi
(i) (j)
PDMS (26 044 (1.7 (059 (37
PEBA (56 222 (4 (025 (20
PUR (10 078 (8 (013 (11
PDMS-SIL (90) (70) (0.8) (120) (55)
Oleyl alcohol (SLM) (50) (35) (0.7 (150 (60
Evaporation (VLE) - - - (025 (20
Feed 1 w-% n-BuOH; membrane thickness 50 μm; temperature 50°C.
SIL = silicalite molecular sieve in PDMS (~ 50%), composite on PEI
SLM = supported liquid membrane in microporous PP of 25 μm thickness
VLE = vapor-liquid equilibrium
 5.6 Pervaporation in perspective 103

it is the concurrent water flux which determines the enrichment

achieved. Accordingly, highest organic enrichment is observed with
the liquid membrane, conceptually having least interaction with
water.

5.6 Pervaporation in perspective

All things considered, pervaporation may well be the most versatile
and least adopted of barrier separations. A number of recent reviews
testify to the continuing fascination [8]–[14].
As an industrial separation process, pervaporation is far from
mature. Basically there are two problem areas, unrelated on first
sight only:
• Scaling up has not (yet) been achieved;
• Biogeneration of base chemicals is not (yet) attractive.
Scaling up faces the problem of handling (transporting and con-
densing) large volumes of low pressure vapor. Module design, by
adhering to high pressure reverse osmosis prototypes, does not meet
pervaporation needs adequately; alternative designs would have to
be modeled after the low pressure/high throughput membrane con-
figurations employed in cross flow microfiltration (pleated mem-
brane designs).
The generic “tree” branching into ever higher levels of carbonic
chemistry from roots nourished by either petroleum or biomass is
well known. As long as oil is “cheap”, prospects of seeing biotech-
nology developing into a serious contender of petrochemistry are
regrettably dim. This goes for the development of the supporting
membrane technology as well.

Bibliography
References
[1] T. Melin, R. Rautenbach, S. Sommer, U. Hömmerich, Einsatzpotential von
anorganischen Zeolithmembranen in der Pervaporation und Dampfper-
meation. Chemie Ingenieur Technik 70 (1998) 1101–1102.
[2] J. M. Watson, P. A. Payne, A study of organic compound pervaporation
through silicone rubber. J. Membrane Sci. 49 (1990) 171–205.
104 5 Pervaporation versus Evaporation

[3] K. W. Böddeker, Pervaporation of aqueous phenols. Proc. 2nd Int. Conf.

Pervaporation Processes (R. Bakish, ed.), San Antonio, Texas, 1987.
[4] K. W. Böddeker, G. Bengtson, E. Bode, Pervaporation of low volatility aro-
matics from water. J. Membrane Sci. 53 (1990) 143–158.
[5] R. W. Baker, J. G. Wijmans, A. L. Athayde, R. Daniels, J. H. Ly, M. Le, The
effect of concentration polarization on the separation of volatile organic
compounds from water by pervaporation. J. Membrane Sci. 137 (1997)
159–172.
[6] K. W. Böddeker, G. Bengtson, H. Pingel, Pervaporation of isomeric bu-
tanols. J. Membrane Sci. 54 (1990) 1–12.
[7] J. Huang, M. M. Meagher, Pervaporative recovery of n-butanol from aque-
ous solutions and ABE fermentation broth using thin-film silicalite-filled
silicone composite membranes. J. Membrane Sci. 192 (2001) 231–242.
M. Matsumura, H. Kataoka, M. Sueki, K. Araki, Energy saving effect of per-
vaporation using oleyl alcohol liquid membrane in butanol purification.
Bioprocess Engineering 3 (1988) 93–100.

Reviews
[8] X. Feng, R. Y. M. Huang, Liquid Separation by membrane pervaporation:
A review. Ind. Eng. Chem. Res. 36 (1997) 1048–1066.
[9] A. Baudot, M. Marin, Pervaporation of aroma compounds: Comparison
with vapour-liquid equilibria and engineering aspects of process improve-
ment. Trans. Inst. Chem. Eng. 75 (1997) 117–142.
[10] F. Lipnizki, R. W. Field, P-K. Ten, Pervaporation-based hybrid process:
A review of process design, applications and economics. J. Membrane Sci.
153 (1999) 183–210.
[11] T. C. Bowen, R. D. Noble, J. L. Falconer, Fundamentals and applications of
pervaporation through zeolite membranes. J. Membrane Sci. 245 (2004)
1–33.
[12] L. M. Vane, A review of pervaporation for product recovery from biomass
fermentation processes. J. Chem. Technol. Biotechnol. 80 (2005) 603–629.
[13] C. C. Pereira, C. P. Ribeiro Jr., R. Nobrega, C. P. Borges, Pervaporative re-
covery of volatile aroma compounds from fruit juices. J. Membrane Sci. 274
(2006) 1–23.
[14] P. Shao, R. Y. M. Huang, Polymeric membrane pervaporation. J. Membrane
Sci. 287 (2007) 162–179.
6 What Membranes are About

6.1 Prelude: Collodion membranes

Collodion, the viscous solution of cellulose nitrate in ether-alcohol,
is the prototype of “synthetic” film-forming materials and as such is
the parent of artificial membranes, likewise of photographic film.
Cellulose nitrate is also the parent of artificial fiber (rayon) and of
plastics (celluloid), not to mention explosives (in that capacity better
known as nitrocellulose). It was discovered by Schönbein (who also
discovered ozone) in 1846, who coined the name guncotton in hopes
of exploiting its hazardous nature. Cotton is cellulose at its natu-
ral purest; collodion uses partially nitrated cellulose also known as
pyroxylin.
Collodion membranes are prepared by allowing the solvents to
evaporate from the viscous solution (“dope”) spread unto a smooth
surface which, in the old days, sometimes was a pool of mercury.
Controlled evaporation (first ether, then alcohol) causes the cellulose
ester to precipitate into a cohesive film, evaporation conditions per-
mitting to influence the permeability characteristic of the resulting
membrane. Next to thickness and morphology, the all-important pa-
rameter determining water permeability is the water content of the
membrane, which is controlled by immersing the film in water be-
fore evaporation of the organic solvents is complete, whereupon the
remaining solvents (mostly alcohol at this stage) are being exchan-
ged against water. By this procedure, a membrane water content of
from 50 to 90% is attainable, most of which in the form of “pore
fluid”. This is to be contrasted with hydration following complete
solvent evaporation, which, by order of magnitude, amounts to only
10%, and is seen as “structural” (truly absorbed) water. – It is noted
that the water content of the dense (nonporous) skin layer of asym-
metric reverse osmosis membranes is of the same order (Sect. 1.3),
106 6 What Membranes are About

albeit at thickness of 0.2 μm as against 200 μm for the symmetrical

collodion film.
The process of membrane formation described, referred to as
phase inversion technique, in countless variations is still in use today
[1], as are microporous cellulose nitrate membranes for use in ul-
trafiltration.
First mention of collodion membranes is by Fick in his classic
studies on liquid diffusion (1855) [2]. His membranes were com-
pletely dry prior to exposure to water and, consequently, were
“tight”, showing low water permeability and practically no salt leak-
age in osmotic experiments. Comparison with the much more
“open” animal membranes (pig’s bladder) lead him to dismiss phy-
sical pores, vaguely speculating instead on “interstitial molecular
diffusion” as mechanism for water transport. The long-standing
controversy about “pores or no pores” in interpreting membrane
permeation has its origin here.

6.2 Membrane polymers –

polymer membranes
A list of commonly used (frequently quoted) membrane polymers is
presented in Appendix D, arbitrarily arranged in terms of increasing
glass transition temperature (Tg).
There is no single consistent system or figure of merit by which to
categorize membrane polymers. Instead, characterization is by
a number of practical criteria such as
• ways and means of polymer formation, chemical and structural
identity of polymeric materials;
• film forming properties, manufacturing conditions for polymeric
membranes, porous and nonporous;
• barrier properties, performance of polymeric films in fluid trans-
port and fluid separations.
The glass transition temperature used as guiding principle in this
survey reflects structure-relevant features of the polymeric materi-
als, foremost chain flexibility and chain interaction; it is not
a natural constant. It is based on the descriptive notion that, at high
enough temperature, all macromolecular organic structures are
 6.2 Membrane polymers – polymer membranes 107

somehow mobile and pliable, in short “rubbery”. On cooling the

random mobility freezes into a “glassy” state at the glass transition
temperature, discernible as change in slope of the temperature de-
pendent volume contraction (Fig. 6.1). Polymers with Tg below room
temperature (Appendix D), consequently, are considered elastomers;
all others are amorphous or semicrystalline with various degrees of
crystallinity appearing respectively disappearing at the glass transi-
tion temperature. High Tg values indicate high thermal and, implic-
itely, high chemical stability; the socalled “engineering plastics” are
high Tg polymers.
Chemical stability of organic polymers relates to solvent compa-
tibility as a film forming criterion: As a rule not without exception,
low Tg membranes are manufactured from suitable polymer solu-
tions (Sect. 6.3), whereas high Tg membranes, being insoluble in
common solvents, are formed either by in-situ interfacial polymeri-
zation (aromatic PA) or else from melt-extruded film by physical

Fig. 6.1. Glass transition: Schematic presentation of the temperature de-

pendence of the specific volume of polymeric materials. Glass transition
temperatures are reported from −120°C for PDMS to +300°C for PI (Ap-
pendix D).
108 6 What Membranes are About

methods like stretching (PTFE, PP, PE) or track-etching (PE, PC). As

a consequence, it is mainly from low Tg polymers that “homogene-
ous” (nonporous) membranes are accessible, all others variously
yielding microporous membrane structures (Sect. 6.4).
Solvent compatibility (chemical stability) has a different connota-
tion when it comes to membrane performance in terms of the two
prevailing mechanisms of mass transport, – solution/diffusion and
convection/diffusion. In membrane separation of aqueous solution
systems a key criterion is the hydrophilic versus hydrophobic charac-
ter of the membrane polymer or membrane surface. Hydrophilic
membranes are in demand; in Appendix D, hydrophilic polymers
are identified by italics. While water generally is a nonsolvent for
polymers (excepting water-soluble specialty polymers included as
“synthetic macromolecules” in Appendix C), water sorption is
a condition for osmotic as well as pervaporative water transport;
further, the only viable safeguard against protein fouling in (sub-
critical) ultrafiltration appears to be the hydrophilicity of the mem-
brane respectively membrane surface.
Chain propagation as mechanism of formation of high polymers
is not limited to homopolymers: Copolymerization and blending of
different monomers offer ways to “design” wanted membrane poly-
meric materials. Some design principles are typified as follows:
• Random copolymers of low Tg moieties, such as the synthetic rub-
bers listed in Appendix D.
• Block-copolymers, composed of soft (low Tg) and hard (high Tg)
segments in (more or less) stoichiometric order, the series of
polyetherblockamides (PEBA) as example.
• Polymer blends: high Tg polymers are hydrophilized by blending
with hydrophilic polymers, foremost polyvinylpyrrolidone (PVP);
Sect. 4.5 has an example.
• Filled membranes are polymer blends incorporating inorganic
fillers such as zeolites. In the extreme, the role of the polymer re-
duces to that of a binder; an example is silicalite in PDMS,
Sect. 5.5.3.
Cellulose and cellulose derivatives are a class of film forming
polymers by themselves, too well documented to be reiterated [3].
 6.3 Like dissolves like 109

6.3 Like dissolves like

To no small degree, polymer-solvent interaction is at the core of
liquid barrier separations, –
• as necessary condition for mass transfer according to the
solution-diffusion mechanism, – the sorption contribution
(Sect. 2.3.1);
• as dissolved polymer in preparation for membrane manufacture
by any of various solution casting procedures (Sect. 6.1).
As solution systems polymer-solvent solutions are dilute and
highly nonideal: Limited sorption of solvent (as permeant) in a swol-
len polymer phase in the first instance; limited dissolution of poly-
mers in organic solvent systems to form viscous casting solutions in
the second.
Rationalizing polymer-solvent compatibility is the same in both
instances, and is based on the independent premise that a corre-
lation exists between the cohesive energy (potential energy) of pure
substances and their mutual miscibility. The cohesive energy of
a pure substance (solid or liquid) is the sum total of inter-molecular
forces which define the condensed state, and which need to be over-
come on evaporation. In decreasing order the forces of concern
are hydrogen bonds (δh ~ 40 kJ/mol), polarity interaction (δp
~ 20 kJ/mol), and dispersion forces (δd ~ 2 kJ/mol). Genuine chemical
bonds (~ 400 kJ/mol) are not affected by evaporation.
The cohesive energy density (CED) is the heat of evaporation at
constant molar volume, by which the solubility parameter δi of com-
ponent i is formally defined (Hildebrand, 1916):
1
⎛ ΔE ⎞
δ total = (δ h2 + δ p2 + δ d2 )
2 1
δi = ⎜ ⎟ and
2
(6.1)
⎝V ⎠
In order for two volatile substances to be compatible their free en-
ergy of mixing (as heat of mixing, ΔH) is supposed to be small
(Sect. 2.1.2). The concept of solubility parameters seeks to predict
mixture compatibility by relating the heats of evaporation of the
mixture components to the heat of mixing: ΔH is small when the
difference in solubility parameters is small; highest compatibility
(miscibility) is therefore expected at δi ≈ δj . This is the statement of
“like dissolves like” in terms of solubility parameters.
110 6 What Membranes are About

For practical use, total and partial solubility parameters are tabu-
lated in units of (cal/cm3)1/2, the format apparently chosen to gain
tractable numbers [4]. Nonvolatile polymers are assigned solubility
parameters according to the like-dissolves-like principle by probing
their solubility in solvents of known solubility prowess. A practical
tool to illustrate polymer-solvent interaction is a solubility map,
a two dimensional graph which uses the prevalent interaction pa-
ramters δh and δp as coordinates (Fig. 6.2). On this graph every sol-
vent of interest is positioned according to its characteristic solubility
parameter listing (in convenient units between 0 and 16), ranging
from hexane (lower left: no hydrogen bonds, no polarity) to water
(upper right: maximal on both counts). Any given polymer, if at all
responsive to solvents, is inscribed into the solubility parameter map

Fig. 6.2. A solubility parameter map for ethyl cellulose. The coordinates re-
present the principal forces which determine the cohesive energy of the sol-
vents indicated: Hydrogen bonds (δ h) and polarity interaction (δ p). After [5].
 6.4 Microporous barriers 111

according to the criteria soluble, swelling, insoluble, graphically

forming a “compatibility island” with blurred contour.
Aside from examining individual polymer-solvent compatibility,
the real value of the solubility parameter analysis is that it allows to
compose mixed solvents from nonsolvents or solvents/nonsolvents.
The guiding principle is the condition that the tie line between non-
solvents on the map meets the “island” of polymer-solvent compati-
bility. As an example for the power of mixed solvents, it is noted that
cellulose nitrate is insoluble in either alcohol or ether alone, but is
soluble in a mixture of the two (collodion, Sect. 6.1). Experimenting
with the counteracting influences of solvents and nonsolvents pro-
duces phase-inversion membranes, the nonsolvent as a rule being
water.

6.4 Microporous barriers

Microporous membranes represent the filtration aspect of barrier
separation as opposed to the solution-diffusion behavior of “homo-
geneous” membranes. As such, microporous membranes are basi-
cally governed by size considerations, retaining their structural iden-
tity in the filtration operation, whereas homogeneous membranes
rely on specific polymer-permeant interactions which typically cause
the membrane to swell.
Within membrane separation technology, microporous structures
serve a variety of purposes determining material selection, target
structure, and method of preparation. Typical uses are:
• Their immediate use as barriers in ultra- and microfiltration in
one of the prevailing configurations of planar (spiral wound;
sandwich; pleated), cylindrical (tubular), or hollow fiber (micro-
tubular);
• their application as non-wettable porous barrier in membrane
distillation respectively osmotic distillation;
• their use as support for immobilized liquid membranes (SLM), or
else for enzymes (catalysts) in membrane bioreactors (MBR);
• their application as rigid porous barriers in gaseous diffusion and
aerosol filtration;
• their use as structural support in composite membrane constructs,
as by coating, interfacial polymerization, or dynamic precipitation.
112 6 What Membranes are About

A pictorial record of microporous structures along with some in-

dication of the techniques employed to create them is presented in
Appendix E, drawing on information provided by the manufacturers
identified, – expanding between the infinite variety of the silica re-
mains of ancient aquatic algae (kieselguhr, E.1) and the exacting
layers of the protein remains of processed archeo-bacteria (bacterial
S-layers, E.18). Clearly, electron microscopy is an indispensable tool
in elucidating membrane structure, limited only by the requirement
that specimen need to be “dry”; water containing polymer mem-
branes formed by solvent-nonsolvent interplay thus need to undergo
a solvent exchange treatment before being ready to reveal their
structure.
Membranes of uniform pore structure are termed isoporous.
Closest to isoporosity among the structures shown are the bacterial
S-layers (E.18) followed by track-etched porous films (E.9) and hon-
eycombed alumina (E.14). Woven fabric, polymeric or metallic, of
uniform mesh (not shown) is isoporous.
Where to from here? Life’s functioning is unthinkable without
membranes. Biomimicry, – learning how nature does –, is bound to
have an influence on future membrane science. Short of living mem-
branes, even life’s material science still needs elucidating [6]: How
do shells grow (biomineralization)? How to reproduce spider web
and silk (noncellulosic natural fibers)?

Bibliography
[1] R. E. Kesting, A. K. Fritzsche: Polymeric Gas Separation Membranes. Wiley-
Interscience, New York etc., 1993.
[2] A. Fick, loc. cit. Chap. 7, Ref. [8].
[3] A.F. Turback (ed.): Synthetic Membranes. Volume 1: Desalination; Volume
2: Hyper- and Ultrafiltration Uses. ACS Symposium Series 153 and 154,
Washington, 1981.
[4] A. F. M. Barton: Handbook of Solubility Parameters and Cohesion Parame-
ters. CRC Press, Boca Raton, Florida, 1983.
[5] E. Klein, J. K. Smith, Asymmetric membrane formation, solubility parame-
ter for solvent selection. Ind. Eng. Chem., Prod. Res. Dev. 11 (1972) 207–210.
[6] J.M. Benyus: Biomimicry, Innovation Inspired by Nature. HarperCollins
Publishers, New York, 1997.
7 Tracing Membrane Science,
an Historical Account

First published in Journal of Membrane Science, 100 (1995) 65–68.

Semipermeability of an animal membrane was discovered by chance,
if not by accident, by Nollet in 1748 [1]. This in short is the story: To
prove that it is dissolved air which causes ebullition of liquids under
reduced pressure, he intended to store a sample of deaerated alcohol
under conditions which would preclude any contact with air for
repeat experiments. To this end he closed the filled sample vial with
a piece of pig bladder, much as is done today with flexible sealing
film, and submerged it in water. The inevitable happened: water
being drawn into the alcohol, thereby straining the membrane. Fas-
cinated by the phenomenon, Nollet devised some clarifying experi-
ments which established the preferential permeability of his mem-
brane towards water. All this is recorded as an addendum to
a treatise on the ebullition of liquids, a topic quite unrelated to
membranes [1].
The force or “new power” manifested by the strained membrane
remained mysterious until Gibbs consolidated the free energy con-
cept in 1873. Nollet himself, in a later textbook on experimental
physics, relates the effect to the volume reduction observed when
alcohol and water are mixed as indicating a natural tendency for
interpenetration of miscible liquids [L’art des expériences, ou avis
aux amateurs de la physique, par M. l’Abbé Nollet, seconde édition,
tôme troisième. Durand, Paris 1770, p. 104].
To Dutrochet [2], who introduced the term osmosis to spontane-
ous liquid flow across permeable partitions, the cause is electricity,
“although I admit that I did not succeed in obtaining a reading on
the galvanometer, even after several attempts”. Noting that liquid
flow occurs both ways (exosmosis and endosmosis in his parlance),
he disproves an earlier capillary theory by Poisson: Capillary action
(another force to which considerable attention was paid at the time)
114 7 Tracing Membrane Science, an Historical Account

Fig. 7.1. Maurice Quentin de la Tour: Abbé Jean Antoine Nollet (1700–
1770). Munich, Alte Pinakothek. Reproduced by permission.

would predict flow exclusively in favor of that liquid which rises

highest in a vertical capillary. – As an aside it may be mentioned that
Dutrochet outpaced many a future membranologist in adopting the
metric system.
Graham is best remembered for his contributions on gas permea-
tion, one of his earliest communications being the “notice of the
singular inflation of a bladder” of 1829 [3]. A moist bladder partially
filled with air or methane (coal gas), when immersed into an atmos-
phere of carbon dioxide, becomes inflated. Once more, Nature’s
powerful urge to equilibrate, made visible through the interference
of a membrane. For it is not the carbon dioxide moving in, but air
barred from moving out, which causes the bladder to bulge. The real
significance of this little note, which also appeared in Schweigger-
 7 Tracing Membrane Science, an Historical Account 115

Seidel’s Jahrbuch der Chemie und Physik for that year [“Notiz über
das sonderbare Anschwellen einer Thierblase”, Band III (1829),
pp. 227–229], is that it contains the original statement of the solu-
tion-diffusion concept, reiterated more extensively in 1866 [7].
While animal membranes are microporous and hydrophilic, rub-
bery membranes of plant origin (gum elastic; caoutchouc) are ho-
mogeneous and hydrophobic (Mitchell, 1829 [4]). With only these
two types of membrane available to him, Mitchell finds the ratio of
permeation rates of various gases to be independent of the mem-
brane used, whereas liquid permeation depends on both the nature
of the liquids and the membrane. Selective withdrawal of oxygen
from air through a gum elastic membrane into water makes him
think of a method to obtain nitrogen gas (the reverse situation
would later be known as the artificial gill). Speculating on the power
of penetration, a resemblance is drawn between sorption affinity and
the ease of condensation of certain gases by charcoal. Sorption (e. g.,
of carbon dioxide by gum elastic) is envisioned as “interstitial infil-
tration”, leaving no room for conducting capillaries. Indeed,
Mitchell’s acidly polite dismissal of Graham’s notion of capillary
canals, besides making for amusing reading, foreshadows the con-
troversy between those advocating a solution-diffusion mechanism
and those insisting on pores when describing membrane permeabil-
ity, – irrelevant at last with the advent of molecular modeling.
A matter of real confusion is Graham’s law: Two “square-root
laws” of gas transport go by his name, which, in practically identical
terms (fluxes being in the inverse ratio of the square roots of their
molecular masses), describe two entirely different rate processes.
One is Graham’s Law of Diffusion, published in 1833 [5], applying
to the interdiffusion of gases at uniform pressure. Intermolecular
collisions are essential to this process, such as would result in hy-
drodynamic or viscous flow when proceding through pores. Unidi-
rectional viscous flow, of course, termed transpiration in the early
literature, does no longer obey Graham’s law but is described as
Poisseuille flow; it is obviously to no separative effect.
The other is Graham’s Law of Effusion, published in 1846 [Phil.
Trans. Royal Soc. (London) 4 (1846) 573], applying to the rates of
effusion of gases through small apertures into a vacuum. In true
effusion, intermolecular collisions are insignificant, the gas mole-
cules crossing the barrier independently of one another. Gas flow in
116 7 Tracing Membrane Science, an Historical Account

this limiting situation, originally termed atmolysis, is known as mo-

lecular flow or Knudsen flow, and yields separation effects as pre-
dicted by a square-root law.
In convenient generalization, any gas separation effected by
means of a porous barrier is nowadays considered a case of Gra-
ham’s law; the process itself is “gaseous diffusion”.
It was Knudsen who, much later, recognized the geometrical as-
pect of it all [6]. The Knudsen number (named after him, not by
him) relates the mean free path of the gas molecules (primarily
a function of pressure) to the dimensions of the duct (diameter and
length). It numerically identifies the flow regime between low (vis-
cous flow = transpiration) and high (molecular flow = atmolysis)
over three orders of magnitude. As to the interaction of gas mole-
cules with a solid wall, the essential feature of the Knudsen theory of
gas flow is that the direction into which an impacting molecule is
repelled is independent of the direction of impact, the analogy being
that of a glowing wall.
Graham’s paper “on the absorption and dialytic separation of
gases by colloid septa” of 1866 [7| is usually quoted to be the founda-
tion of the solution-diffusion model of membrane transport, exem-
plified at the fractionation of air through a rubbery membrane.
Whereas gaseous diffusion based on molecular flow through pores
would slightly favor the lighter nitrogen, the rubber membrane is
found to enrich oxygen. Drawing on the solubility of air in water
(a subject which occupied Nollet [1]), dissolved gases are considered
liquefied and thereby amenable to liquid diffusion. A correlation is
consequently expected between the penetration of rubber by differ-
ent gases and their ease of liquefaction, as was already noted by Mit-
chell [4], however, relative rates of solution-diffusion do not yield to
rationalization as “squarely” as do those of gaseous diffusion.
Back to liquids. It remained to Fick (1855) to interpret the liquid
analog to gaseous interdiffusion: The law by which a solute dissi-
pates in its own solvent [8]. The experimental data from which Fick
set out are again Graham’s. The idea, which presented itself “quite
naturally”, was to draw a parallel to the diffusion of heat (Fourier)
and that of electricity (Ohm) in their respective conductors, – no
small feat for a demonstrator of anatomy that he was at the time.
Fick’s attempts to model the diffusion of salt solutions through po-
rous partitions appear somewhat tedious, however, they produced
 7 Tracing Membrane Science, an Historical Account 117

the first mention of a collodion membrane, exhibiting an “endosmo-

tic equivalent” (ratio of water to salt diffusing) vastly higher than
that of animal membranes.
As to van’t Hoff ’s lasting contribution to solution theory [9], such
is the beauty of the apparent analogy between gas pressure and os-
motic pressure of dilute solutions that it persists although proven
wrong a long time ago. It has produced, nevertheless: The idea of the
semipermeable membrane; a reminder that molarity is the basis to
compare mass action in chemistry; an indication, privately hinted at
by Arrhenius, that if molarity does not work it might be due to elec-
trolytic dissociation; the concept of isotonic solutions having equal
vapor pressure of the same solvent; and the first Nobel prize in che-
mistry (1901). Actually, it was Pfeffer who lead van’t Hoff on the false
track with his measurements of osmotic pressure of sugar solutions
[W. Pfeffer: Osmotische Untersuchungen. Studien zur Zellmechanik.
Verlag W. Engelmann, Leipzig, 1877]. Indeed, the numerical corre-
spondence of the osmotic pressure of an aqueous sugar solution
with the ideal gas pressure on an equimolar footing must be regar-
ded as one of nature’s profound jokes.
As a preliminary study to an investigation of the state of soaps in
aqueous solution, Donnan examined solutions of Congo red, it being
known that this “colloidal” sodium salt will dissociate but not diffuse
through parchment paper. When such a solution is contacted, across
a parchment paper diaphragm, with a sodium chloride solution, all
ions present except the bulky anion of Congo red are free to move
about. The ensuing equilibrium distribution of sodium chloride, go-
verned by the condition of equal activity of any diffusible species on
both sides of the membrane, is unequal, tending to prevent chloride
(and thus sodium chloride) from entering the Congo red compart-
ment: Donnan equilibrium. Unequal electrolyte distribution, in turn,
gives rise to a potential difference: Donnan potential. Translation of
this situation into the exclusion principle of ion exchange mem-
branes is straightforward: The fixed charges of the membrane matrix
assume the role of the non-dialysable ions which act to prevent mo-
bile co-ions from entering the matrix: Donnan exclusion. Pores or
not, a fluid aqueous phase allowing dissolved ions to move within the
ion exchange matrix is a logical requirement for this analogy to hold.
Electrodialysis, of course, had to wait another thirty years (Meyer and
Strauss, 1940), physiology being the interest of the hour [10].
118 7 Tracing Membrane Science, an Historical Account

The only contribution in this essay devoted to membranes pro-

per, also the only multi-authored contribution, is a review by Bige-
low and Gemberling on collodion membranes [11]. First mentioned
by Fick in 1855 [8], collodion membranes were the first synthetic
membranes to compete with natural or processed animal skin like
pig’s bladder, goldbeater’s skin, and parchment paper. Collodion
(cellulose nitrate; pyroxylin) is the ancestor of the still thriving fam-
ily of cellulose ester membranes, which, it may be remembered, ge-
nerated the first membrane recognized to be asymmetrically struc-
tured (Loeb and Sourirajan, 1960).
Numerous researchers before him must have made the observa-
tion which prompted Kober to investigate pervaporation, viz., “that
a liquid in a collodion bag, which was suspended in the air, evapo-
rated, although the bag was tightly closed” [12]. Kober’s claim to
fame lies not so much in the profoundness of this investigation, but
in having named the effect. Ironically, it is his “vacuum perstilla-
tion” which today is addressed as pervaporation.
What is missing? Membranes and membrane separations remain-
ed laboratory tools until fairly recently, not in the least confined by
the fact that there was no polymer research to speak of during the
time period documented here. There was little intention of applying
membranes to industrial separations, unless the clarification of wine
by ultrafiltration using compacted asbestos, introduced towards the
end of the 19th century, is considered such. Even notions of medical
applications are conspicuously absent, which is all the more surpris-
ing considering that the majority of early membrane researchers are
biologists, botanists, physiologists or outright medical professionals.
The age of innocence for membrane science ended in 1942. At this
time, curiously coinciding, two totally unrelated membrane proces-
ses made their appearance, which have changed the world. One is the
separation of uranium isotopes by gaseous diffusion of UF6 (Man-
hattan Project), which, for better or worse, gave access to nuclear
energy. The other is hemodialysis (Kolff), marking the unsuspecting
beginning of the high-tech manipulation of life itself.
 Literature cited 119

Literature cited
[1] J. A. (Abbé) Nollet (1700–1770): Recherches sur les causes du Bouillonne-
ment des Liquides (Investigations on the Causes for the Ebullition of Liq-
uids). Histoire de l’Académie Royale des Sciences, Année MDCCXLVIII,
Paris, 1752, 57–104.
[2] R. J. H. Dutrochet (1776–1847): Nouvelles Observations sur l’Endosmose et
l’Exosmose, et sur la cause de ce double phénomène (New Observations on
Endosmosis and Exosmosis, and on the Cause of this Dual Phenomenon).
Annales de Chimie et de Physique, 35 (1827) 393–400.
[3] T. Graham (1805–1869): Notice of the Singular Inflation of a Bladder. Quar-
terly Journal of Science, No. II (1829) 88–89.
[4] J. K. Mitchell (1793–1858): On the Penetrativeness of Fluids. The Journal of
the Royal Institution of Great Britain, No. IV (1831) 101–118; No. V (1831)
307–321.
[5] T. Graham: On the Law of the Diffusion of Gases. The London and Edin-
burgh Philosophical Magazine and Journal of Science, Vol. II (1833) 175–
190; 269–276; 351–358.
[6] M. Knudsen (1871–1949): Die Gesetze der Molekularströmung und der
inneren Reibungsströmung der Gase durch Röhren (The Laws of Molecular
Flow and of Inner Friction Flow of Gases Through Tubes). Annalen der
Physik, 28 (1909) 75–130.
[7] T. Graham: On the Absorption and Dialytic Separation of Gases by Colloid
Septa. Part I. Action of a Septum of Caoutchouc. The London, Edinburgh,
and Dublin Philosophical Magazine and Journal of Science, Vol. XXXII
(1866) 401–420.
[8] A. Fick (1829–1901): Über Diffusion. Poggendorff’s Annalen der Physik und
Chemie, 94 (1855) 59–86. – Abstracted by the author as: On Liquid Diffu-
sion. The London, Edinburgh, and Dublin Philosophical Magazine and
Journal of Science, Vol. X (1855) 30–39.
[9] J. H. van’t Hoff (1852–1911): Die Rolle des osmotischen Druckes in der
Analogie zwischen Lösungen und Gasen. Zeitschrift für Physikalische Che-
mie, 1 (1887) 481–508. – Translation published as: The Role of Osmotic
Pressure in the Analogy Between Solutions and Gases. In: The Foundations
of the Theory of Dilute Solutions, Alembic Club Reprint No. 19, Oliver and
Boyd, Edinburgh, 1929, 5–42.
[10] F. G. Donnan (1870–1956): Theorie der Membrangleichgewichte und Mem-
branpotentiale bei Vorhandensein von nicht dialysierenden Elektrolyten.
Ein Beitrag zur physikalisch-chemischen Physiologie (Theory of Membrane
Equilibria and Membrane Potentials in the Presence of Non-Dialysing Elec-
trolytes. A Contribution to Physical-Chemical Physiology). Zeitschrift für
Elektrochemie und angewandte physikalische Chemie, 17 (1911) 572–581.
[11] S. L. Bigelow and A. Gemberling: Collodion Membranes. Journal of the
American Chemical Society, 29 (1907) 1576–1589.
[12] P. A. Kober: Pervaporation, Perstillation and Percrystallization. Journal of
the American Chemical Society, 39 (1917) 944–948.
Appendix

A Thermodynamic properties of aqueous solutions as

function of composition. Upper: NaCl-H2O (nonvolatile
solute to saturation); lower: EtOH-H2O (volatile solute,
completely miscible).
B Criteria of technical water quality in water desalination:
Hardness; alkalinity; corrosiveness.
C Membrane filtration: Molecular mass of frequently en-
countered nonelectrolytes and macromolecules (marker
molecules).
D A survey of commonly used membrane polymers, ar-
ranged in order of increasing approximate glass transi-
tion temperature.
E Microporous structures in barrier separation: A pic-
torial record by electron micrographs.
A Properties of Aqueous Solutions

Thermodynamic properties of aqueous solutions as function of com-

position. Upper: NaCl-H2O (nonvolatile solute to saturation); lower:
EtOH-H2O (volatile solute completely miscible).

Composition Osmotic Activity

Pressure Coefficient
m x1 x2 w% w% π1 π2 γ1 γ2
NaCl H2O NaCl H 2O NaCl H 2O NaCl H 2O NaCl
0.01 0.999+ 0.001− 99.9+ 0.058 000.48 – – 0.900
0.1 0.998 0.002 99.4 0.58 004.62 – – 0.767
0.2 0.996 0.004 98.8 1.16 009.16 – – 0.725
0.5 0.991 0.009 97.2 2.84 022.8 – – 0.674
1.0 0.982 0.018 94.5 5.52 046.4 – – 0.660
2.0 0.965 0.035 89.5 10. 5 097.5 – – 0.680
3.0 0.949 0.051 85.1 14. 9 155 – – 0.735
6.0 0.902 0.098 74 26 389 – – –
H 2O EtOH H 2O EtOH H 2O EtOH H 2O EtOH
– 0.99 0.01 97.5 2.5 0013.5 1487 1.000 3.028
– 0.95 0.05 88.1 11.9 0064.2 0863 1.004 2.629
– 0.90 0.10 77.9 22.1 0122 0632 1.017 2.264
– 0.50 0.50 28.1 71.9 0600 0210 1.291 1.221
– 0.10 0.90 4.2 95.8 2191 0041 2.024 1.009
– 0.05 0.95 2.0 98.0 3030 0021.0 2.196 1.002
– 0.01 0.99 0.4 99.6 5140 0004.3 2.360 1.000
m = molality of solute [mol/kg]
x = mol fraction (1, solvent; 2, solute)
π = osmotic pressure at 25 °C [bar]
γ2 for NaCl is the mean activity coefficient at 25 °C
124 A Properties of aqueous solutions

Data compiled and adapted from:

R. W. Stoughton, M. H. Lietzke, Calculation of some thermody-
namic properties of sea salt solutions at elevated temperatures from
data on NaCl solutions. J. Chem. Eng. Data 10 (1965) 254–260.
A. C. Schneider, C. Pasel, M. Luckas, K. G. Schmidt, J.-D. Herbell,
Bestimmung von Ionenaktivitätskoeffizienten in wässrigen Lösungen
mit Hilfe ionenselektiver Elektroden. Chem. Ing. Techn. 75 (2003)
244–249.
G. D. Mehta, Comparison of membrane processes with distillation
for alcohol-water separation. J. Membrane Sci. 12 (1982) 1–26.
B Criteria of Technical Water Quality

Hardness
Hardness is that part of the total salinity of a water (TDS) associated
with the alkaline-earth cations Ca++ and Mg++, total hardness com-
prising all divalent salt ions in solution, while carbonate hardness is
the fraction of total hardness which relates to dissolved CO2. Hard-
ness is expressed as meq/L of hardness-producing species, usually
CaCO3 or Ca++ (new degrees of hardness according to WHO/EU):
1 meq/L = 0.5 mmol/L = 50 mg/L CaCO3 = 20 mg/L Ca++

A rough classification of water in terms of hardness is

meq/L mg/L CaCO3
soft > 10−2 > 500−100
medium hard > 12−6 > 100−300
hard > 16−10 > 300−500
very hard > 10 > 500

The actual concentration of dissolved Ca++ (Mg++) in water de-

pends on the availability of CO2 which solubilizes CaCO3 (MgCO3)
according to

CaCO3 + CO2 + H2O ⇌ Ca(HCO3)2 (B.1)

Solubility of CaCO3 in pure water is little more than 10 mg/L, while

a concentration of CO2 corresponding to its ambient partial pressure
yields a carbonate hardness of about 100 mg/L (2 meq/L). Scale in-
hibitors like polymeric phosphates, which are widely used in water
desalination, will sustain a concentration of up to 250 mg/L CaCO3
(5 meq/L). WHO recommendation for drinking water is 2 meq/L as
highest desirable level, and a maximum permissible total hardness of
126 B Criteria of technical water quality

10 meq/L. The EU drinking water directive suggests a minimum

hardness for softened (demineralized) water of 3 meq/L.
Carbonate hardness may be reduced through addition of lime to
precipitate CaCO3 (lime softening), according to
Ca(HCO3)2 + Ca(OH)2 → 2 CaCO3 + 2 H2O (B.2)
Lime softening is a process of partial demineralization used exten-
sively in the treatment of brackish groundwaters. Overall reduction
of TDS is of the order of 10%, depending on the fraction of divalent
ions in solution. Conversely, hardness is conveyed to soft waters by
adding lime in combination with CO2,
Ca(OH)2 + 2 CO2 → Ca(HCO3)2 (B.3)
This is a method of posttreatment used to remineralize the product
water of reverse osmosis or thermal seawater desalination which, on
account of its negligible salinity, is extremely soft and therefore ag-
gressive (see below).

Alkalinity
The ratios of the carbonate-containing species in aqueous solution
are a function of pH value (and vice versa), as illustrated by the fol-
lowing figures:
CO2 / HCO-3 HCO-3 / CO=3
pH 6 2 20000
pH 7 0.2 02000
pH 8 0.02 00200

The standard acid consumption on titration of a water sample to

pH 4.3 is the total alkalinity (acid capacity) of the water, KA 4.3, re-
ported as mmol/L acid (HCl or HCO-3 ) or as mg/L CaCO3. At pH 4.3
the concentration of bicarbonate ( HCO-3 ) is down to 1% of the dis-
solved CO2, and neutralization of carbonate ( CO3= ) is essentially
complete. Alkalinity is thus related to carbonate hardness, specifi-
cally to HCO3− concentration, and is a measure of the capacity of the
water to resist changes in pH (buffering). A minimum alkalinity of
0.5 mmol/L HCO3– is suggested by the EU drinking water directive.
 B Criteria of technical water quality 127

Alkalinity (carbonate hardness) is destroyed by addition of acid,

with simultaneous generation of CO2,
Ca(HCO3)2 + 2 HCl → CaCl2 + H2O + 2 CO2 (B.4)
Acidification followed by aeration to remove excess CO2 is a com-
mon pretreatment practice in water desalination to reduce hardness
and to prevent carbonate scales from being deposited. In thermal
seawater desalination, scale formation proceeds by the reverse of
reaction B.1 at elevated temperature, again producing CO2.

Corrosiveness
Any excess of CO2 beyond that needed to keep Ca++ (Mg++) in solu-
tion according to Eq. B.1 is termed aggressive and must be removed
before the water is distributed. Excess CO2 is harmful in two ways:
Firstly, dissolved excess CO2 acts much like a mineral acid, particu-
larly in the presence of oxygen, and will corrode the ferrous ducts of
the distribution system; secondly, excess CO2 interferes with the cor-
rosion-protective mineral layer in ferrous ducts, either by preventing
its formation or by re-dissolving it. Formation of this protective
layer, which basically consists of crystalline CaCO3 and FeCO3, re-
quires a minimum alkalinity to exclude free CO2, a residual hardness
of at least 30 mg/L CaCO3, and dissolved oxygen.
Very soft waters, even after thorough aeration (deacidification),
always retain some excess CO2 (typically 5 mg/L) and thus are ag-
gressive on both above counts. Only with very hard waters a small
excess of CO2 may be tolerated, aeration then bearing the risk of
precipitation of CaCO3 according to Eq. B.1. – Mixing of waters of
different carbonate hardness always yields aggressive water requir-
ing deacidification.
Yet another kind of corrosion is due to sulfate ions ( SO=4 ) de-
composing cement materials (concrete). Again, soft waters are more
aggressive than hard water.
The above interrelations are especially significant for desalinated
water supplies. Water produced by desalination invariably is soft
and acidic, regardless of the process used: The product water of ther-
mal desalination is essentially devoid of all saline constituents, and
is likely to contain free CO2 originating from the acid pretreatment,
128 B Criteria of technical water quality

Eq. B.4. In reverse osmosis desalination there is a preferential rejec-

tion of the hardness-producing ions by the membranes, resulting in
an enrichment of CO2 (lowering of pH) and a relative increase of
monovalent (soft) ions in the permeate.
Corrosion control is through chemical treatment and appropriate
material selection. The European directive addressing posttreatment
of demineralized drinking water recommends removal of excess
CO2 , raising the pH value to 8, and remineralization to a compara-
tively high 3 meq/L (corresponding to 150 mg/L CaCO3). According
to an industrial standard, minimum remineralization for bulk trans-
port of desalinated seawater is 35 mg/L CaCO3.
C Marker Molecules

Membrane filtration: Molecular mass of frequently encountered

nonelectrolytes and macromolecules (marker molecules). This is
a teaching aid. Entries are compiled from various open literature
sources (including membrane manufacturer’s brochures), and are
not critically weighted. Molecular mass > 5000 usually is quoted as
“average”, individual assay depending on origin (biological matrix)
and method of mass determination.

D-Alanine 000089 amino acid

Creatinine 000113 (urine constituent)
Phenylalanine 000165 amino acid
Glucose 000180 blood sugar
Tryptophan 000204 amino acid
Sucrose 000342 cane/beet sugar
Lactose 000342 milk sugar
Raffinose 000504 tri-saccharide
Vitamin B12 001355 (in activated sludge)
Bacitracin 001400 antibiotic polypeptide
(globular)
Inulin 005200 polysaccharide
Insulin 005800 polypeptide hormone
β2-Microglobulin 011800 plasma protein
Cytochrome C 013000 respiratory proteid
(globular)
Lysozyme 014400 mucolytic enzyme
α-Lactalbumin 016000 cheese whey protein
Myoglobin 017500 respiratory proteid
ß-Lactoglobulin A 018700 milk protein
Trypsin 024000 proteolytic enzyme
Chymotrypsinogen A 024500 proteolytic enzyme
Carbonic anhydrase 031000 CO2 hydrating enzyme
130 C Marker molecules

Pepsin 034500 gastric enzyme (globular)

Ovalbumin 045000 egg white protein
Bovine serum albumin (BSA) 067000 plasma protein (globular)
Hemoglobin 068000 respiratory proteid
Human serum albumin 069000 plasma protein (globular)
Transferrin (s) 080000 iron-binding glycoprotein(s)
Phosphorylase B 094000 plasma enzyme
Aldolase 142000 plasma enzyme (lyase)
Immunoglobulin IgG 160000 antibody protein (globular)
Catalase 240000 anti-peroxide enzyme
Ferritin (apoferritin) 450000 iron storage protein
Myosin 500000 muscle protein
Thyroglobulin 680000 gluco protein (thyroid)
Immunoglobulin IgM 960000 antibody protein

Synthetic macromolecules
Polyethylene glycol (PEG) 00200 to 2 × 106
Polyvinylpyrrolidone (PVP) 02500 to 900000
Dextran (polysaccharide) 15000 to 50 × 106
D Membrane Polymers

A survey of organic membrane polymers, arranged in order of ap-

proximate glass transition temperature (°C). It is noted that by the
complex nature of the polymeric state of matter (chemical and phy-
sical structure) the temperature of glass transition does not have the
quality of a natural constant. Accordingly, published figures vary;
the data presented are meant to provide orientation. Hydrophilic
polymers are set in italics.

Natural polymeric materials

Natural rubber (polyisoprene) – 70
Cellulose (regenerated) CE
Cellulose derivatives (polycellobiose)
Ethylcellulose (ether) EC + 045
Cellulose nitrate (ester) CN + 060
Cellulose diacetate CA + 070
Cellulose triacetate CTA + 100

Synthetic polymers
Polydimethylsiloxane PDMS – 120
Polybutadiene PB – 080
Polyetherblockamide PEBA – 065
Polyethyleneoxide PEO – 050
Polyethylene (40 % cryst.) LDPE – 070
Polyethylene (70 % cryst.) HDPE – 020
Polyvinylidenefluoride PVDF – 040
Polypropylene PP – 015
Polyvinylacetate PVAC + 030
132 D Membrane polymers

Polyamide (Nylon) PA < 100

Polyvinylalcohol PVAL + 085
Polyvinylchloride PVC + 090
Polystyrene PS + 100
Polymethylmethacrylate PMMA + 110
Polyacrylonitrile PAN + 120
Polytetrafluoroethylene (Teflon) PTFE + 125
Polyetheretherketone PEEK + 140
Polycarbonate PC + 150
Polyphenyleneoxide PPO + 170
Polyvinylpyrrolidone (360 000) PVP + 180
Polysulfone PSU + 190
Polytrimethylsilylpropyne PTMSP > 200 (?)
Polyphenylsulfone PPS + 215
Polyetherimide PEI + 215
Polyethersulfone PES + 230
Polyamide (aromatic) PA + 270
Polyimide PI > 300 (?)
Polyethylene terephthalate PET (mp. ~ 240)

Synthetic copolymers (elastomers)

Acrylonitrile-butadiene NBR
Acrylonitrile-butadiene-styrene ABS
Styrene-butadiene SBR
Ethene-propene-diene EPDM
Ethylene-vinylalcohol EVAL
E Microporous Structures

A collection of electron micrographs (SEM = scanning electron mi-

croscopy), illustrating the scope and variety of organic and inorganic
porous materials, along with some indication of the imaginative
methods by which porous barriers are created.
134 E Microporous structures

Fig. E.1 Kieselguhr: Deposits of the silica shells of unicellular algae

(diatoms) from the Tertiary geologic period. When compacted into
cylinders: Berkefeld filter for drinking water disinfection. (When
employed as absorbant for nitroglycerin: Dynamite). – [x 1200;
Meyer-Breloh].
Fig. E.2 Glass fiber depth filter. Nonwoven (randomly compacted)
filters from fibers are ancient clarifying aids, their current rendition
being polymeric nanofiber devices. There is an anticipated relation-
ship between fiber diameter and retention capability: Asbestos,
20 nm (outlawed); glass fibers, thinness limited by health hazard
considerations; polymeric nanofibers (by electrospinning), 100 to
1000 nm (0.1 to 1 μm); stainless steel, 4 μm; human hair, 70 μm.
– [x 1000; FZK].
Fig. E.3 Cellulose acetate (CA): A member of the generic class of
cellulose derivatives, capable of forming homogeneous as well as
porous barriers. Shown is the porous substructure of an integral-
asymmetric (“skinned”) reverse osmosis membrane produced by
solvent-nonsolvent phase inversion according to Loeb-Sourirajan.
– [GKSS].
Fig. E.4 Polysulfone (PSU), the workhorse of ultrafiltration and
widely used as support for composite membranes. As generic class,
which includes sulfonated polysulfones as well as polyethersulfone
(PES), polysulfone is versatile: Itself hydrophobic, hydrophilicity is
imparted through sulfonation. Discernible is a structure of micros-
copic nodules and nodule aggregates. – [Rated 0.8 μm; Seitz].
Fig. E.5 Polyamide (aliphatic PA, Nylon 66). An interlaced structure
produced by a proprietary foaming process. Nylon is a trade name
for a series of related polymers containing the –CONH– bond in its
structure. It is also the “hard” component in segment-elastomeric
block-copolymers like PEBA. – [Rated 0.2 μm; x 3000; PALL].
Fig. E.6 Polyamide (aromatic PA). The original, and still going,
“FT-30” membrane for seawater desalination by reverse osmosis. In
view are surface and fractured edge of the composite membrane
produced by interfacial polymerization onto a polysulfone porous
support (Fig. E.4). Note the highly corrugated “skin” of this barrier.
– [FilmTec/Dow; Chapter 3, Ref. (4)].
E Microporous structures 135
136 E Microporous structures

Fig. E.7 Polypropylene (PP). A structure of interconnected (open)

vesicles produced by thermal inversion from a heated polymer solu-
tion. Porosity (vesicular size) is adjustable by controlling the rate of
cooling. The picture shown is a section out of a tubular membrane of
gradually narrowing (unisotropic) pore structure [Akzo].
In a related process, precipitation is from a polymer-diluent system
initially homogenized by heating to above the melting temperature
of the polymer; PP in mineral oil [TIPS = thermally induced phase
separation (Lloyd)].
Fig. E.8 Polymer blend: Polyethersulfon (PES)/polyamide (PA) blen-
ded with polyvinylpyrrolidone (PVP). An asymmetric (unisotropic)
globular structure with randomly distributed hydrophilic microdo-
mains in a hydrophobic matrix (for use in hemodialysis). – [Rated
6 nm; Gambro].
Fig. E.9 Track-etched pores: Surface of a capillary pore membrane
(complete with retained asbestos fibers). Uniform capillary pores are
created by irradiating melt-extruded polymer film followed by che-
mical etching of the nucleation tracks; applied to polymers which are
intractable by solution casting: Polycarbonate (PC) and polyester
(PE). – [Nuclepore].
Fig. E.10 Cross section of a track-etched pore structure: Capillary
pores in polyethylene terephthalate (PET), otherwise known as
a fiber forming polyester. – [Pore diameter 0.7 μm; GSI].
Fig. E.11 Pores by controlled stretching of semicrystalline polymer
film. Slit-shaped pores (ruptures) in melt-extruded polypropylene
(PP). – [Slit rating 0.02 by 0.2 μm; Celanese/Celgard].
Fig. E.12 A tortuous pore structure by expansion (biaxial stretching)
of polymer film, Teflon (PTFE): Nodules interconnected by fibrils
suggestive of the direction of stretching. – [Gore/Gore-Tex].
E Microporous structures 137
138 E Microporous structures

Fig. E.13 Alumina (aluminum oxide) barrier: An asymmetric inor-

ganic membrane composed of microporous layers on a macro-
porous mineral support, produced by slip-coating and sintering
reminiscent of ceramic techniques. – [Rated 0.2 μm; Alcoa/Ceraver].
Fig. E.14 Alumina (aluminum oxide): A straight channel structure
obtained by anodic oxidation of metallic aluminum, channel width
0.2 μm. By controlling the voltage the channels divide near the sur-
face to yield a 0.02 μm (20 nm) asymmetric pore structure as shown
[Alcan/Anotec].
Fig. E.15 Porous glass (CPG, controlled pore glass). Obtained by
a process of micro-dispersed phase separation followed by leaching
of the dispersed phase. A structure of interconnected nodules remi-
niscent of porous polysulfone. – [Rated 85 nm; Schott].
Fig. E.16 Pure metallic silver, formed from suspensions of amor-
phous silver into molecularly bonded (that is, not sintered) micro-
porous membranes of 50 μm thickness; antiseptic and reusable.
– [Graded 0.2 to 5 μm; Osmonics].
Fig. E.17 Asymmetric metallic barrier: A filter mat of randomly
compacted (nonwoven) stainless steel fibers (see E.2) sintered onto
a support of porous stainless steel made by powder metallurgy. Fiber
thickness, 4 μm; active layer, 200 μm; total thickness, 3 mm. – [Rated
1 μm; Krebsöge].
Fig. E.18 Bacterial S-layers. The crystalline protein envelopes of cer-
tain archeo-bacteria are isolated and deposited on a filter support to
form highly ordered two-dimensional grids with application poten-
tial as isoporous nanofilters. Uniform size of the underlying bacteria
makes for uniform pore size (2 to 6 nm depending on spcies) and
“sharp” molecular weight cutoff (near MW 50000), the “membranes”
having the solvent stability of the archeo-proteins. Shown is a com-
puter image reconstruction of an hexagonal grid. – [Nanosearch/
Biofil].
E Microporous structures 139
Subject Index

absolute rating (microfiltration) biotoxic metabolites; biotoxicity

77 79
ABE fermentation 99 biotechnology vs. petrochemistry
activity 20, 24, 30, 33, 39, 44, 85, 103
91, 97 boundary layer 12, 13
activity coefficient 20, 21, 22, 23, laminar boundary layer 53, 70, 71
94–97, 123 dynamic boundary layer 73
(see also mean activity coefficient) bulk; bulk phase 13, 14, 22, 53, 70,
activity vs. concentration 85, 93
(see mass action) 19, 20, 30 butanol 99, 101, 102
alembic teaching 17
alkalinity (water quality) 126 carrier gas pervaporation 81
aroma compounds 22, 45, 93, cellophane (regenerated cellulose)
96, 99 66, 79
artificial kidney (see hemodialysis) cells; cell debris 64, 74
asbestos 99, 118, 134 chain flexibility/chain interaction
aspect ratio 36, 66 106
asymmetric membrane 36, 50, charged membranes 7
66, 88 chemical potential (partial molar free
(see also composite membrane) energy) 8, 26, 27, 30, 47
azeotrope (constant boiling mixture) (see also Gibbs free energy)
20, 21, 23, 86 clearance (plasma clearance) 77
azeotrope splitting 1, 10, 45, 88 cohesive energy (density) 8, 109, 110
colligative properties 19, 30, 39,
back diffusion 53, 70 41, 97
Barrer unit 14 (see also osmotic pressure; vapor
barrier; barrier interference 3, 8, 14, pressure)
17, 114 collodion (cellulose nitrate) 66, 82,
barriers (see membranes) 105, 106, 118
batch; batch operation 14, 82, 85 colloids; colloidal solution 64, 73
biofouling (see fouling) compaction 72
biological membranes 39, 112 composite membrane 41, 50, 88,
biomimicry 112 92, 112
biotechnology concentration polarization 12, 14,
biological wastewater treatment 52, 53, 54, 70, 95, 98
78, 79 concentration-normalized flux 14
bioorganic synthesis 79, 93 (see also permeance)
bioseparation 1, 88, 103 convection (mass transfer) 7, 36, 75
142 Subject Index

conversion (see recovery) free energy (see Gibbs free energy)

copolymers 108 thermodynamic minimum
corrosiveness (water quality) 127 3, 54, 55
coupling 3, 12, 13, 46 pV energy (mechanical) 26, 42,
Crafoord prize 80 43, 54, 55, 60
critical flux 67, 72 πV energy (chemical) 42, 43
cross flow (tangential flow) 12, 54, enrichment 1, 2, 86, 93
60, 78 (see also selectivity)
entropy contribution 26
Dalton’s law 25, 84 extensive variables 26
Dead Sea 18, 61
dead end filtration 12, 77, 78 facilitated transport 7
dehydration; dewatering 9, 30, 45, feed flow rate; feed stream 51
71, 87, 88 feed pressure (operating pressure)
(see also volume reduction) 51, 54, 92
demixing 3, 8, 25, 96 feed-and-bleed operation 61, 97
(see also phase separation) filtration spectrum 6, 65
depletion dialysis 75, 77 Flory-Huggins isotherm 33, 34, 83,
Desalting Plants Inventory Report 90, 101
56 flow (flow rate) 51, 52
diafiltration 31, 75 flux (flow density) 11, 13, 32
dialysate (receiving phase) 75 critical flux 67, 72
dialysis 10, 19, 31, 48 pure water flux 67
difference approximation 28, 31 volume flux 13, 67, 68, 72
diffusion 7, 10, 31, 34, 36 flux equation 31, 32, 47, 84
diffusion coefficient 31, 35, 71 food processing 45
diffusion selectivity 32, 35 foot print (plant acreage) 79
diffusivity (mobility) 32, 35, 64, fouling/biofouling 7, 54, 58, 79
74, 83 fractionation (of proteins) 74
dilute solutions 2, 22, 30, 39, 42, 46, free enthalpy (see Gibbs free energy)
64, 88, 95, 99, 117 friction 3
direct osmosis (see osmosis) functionalized membranes 7
distribution; distribution coefficient
24, 47, 90, 95 gas separation (solution-diffusion)
Donnan potential/exclusion 7, 117 37
dynamic boundary layer 73 gaseous diffusion (porous barrier)
dynamic membrane (gel 6, 63, 111, 116, 118
polarization) 72, 112 gaseous dispersion (aerosol) 63
dynamic reverse osmosis 48, 51 Gay-Lussac’s law 41
(see also process dynamics) gel; gelation (solidified sol) 64, 71
gel polarization 14, 72, 74
economy of scale 56 Gibbs free energy (free enthalpy)
electrodialysis 7, 8, 10, 117 8, 25
electroneutrality 12, 24 Gibbs fundamental equation 26, 27
energy (of separation) 3 glass transition (temperature)
energy consumption (specific) 106, 107
54, 55 glassy polymers 6, 35, 67, 83, 107
energy recovery 51, 54, 56, 60 (see also membrane polymers)
 Subject Index 143

Graham’s laws 115 ion exchange membranes

gravity 36, 59, 60 7, 8, 117
ion hydration 18, 24, 59
Hagen-Poisseuille equation 67 isoporous; isoporous sieve 66, 112
hardness (water quality) 125 isotonic solution 43, 73, 75, 117
heat of mixing 25, 109
hemodialysis (artificial kidney) Journal of Membrane Science
1, 10, 44, 70, 75, 77–79, 118 32, 113
hemofiltration 79, 80
Henry coefficient (volatility) 22, 94, Knudsen diffusion 116
95, 102 (see also gaseous diffusion)
Henry’s law 32
Henry isotherm 33, 83, 90 laminar boundary layer 53, 70
high boilers (see volatile organics) Langmuir isotherm 33, 34, 89
high flux membrane leaky membrane 4, 46
(see asymmetric membrane) “like dissolves like” 34, 109, 110
homogeneous vs. porous barriers 6 liquid membranes 6, 84, 102,
(see also membranes) 103, 111
hydraulic low boilers (see volatile organics)
– conductor 36
– permeability 14, 67, 68, 77 macromolecules/macrosolutes
– resistance 6, 36, 64, 65 10, 64, 74
hydrodynamic lift force 74 (see also sol and gel)
hydrogen bond marker microorganisms
(see intermolecular forces) (microfiltration) 78
hydrophilic polymers 49, 131 marker molecules (ultrafiltration)
(see membrane polymers) 64, 76, 77, 129
hyperfiltration (reverse osmosis) mass action
9, 65 activity governed (“vigor”) 20
concentration governed
ideal mixture/solution 19, 20, 27, 29 (“number”) 20, 34
(see also Raoult’s law) mass transfer (mass transport)
ideal separation 11 2, 3, 10
ideally semipermeable 4, 42, 59 convection 7, 31, 36, 99, 108
infinite dilution 20, 24 diffusion 7, 31, 34, 36, 99, 108
intensive variables 8, 23, 25 plug flow 75
interfacial polymerization 41, 50, relocation 10, 95
107, 112 mass transfer coefficient 13, 14, 73,
interfacial precipitation 74, 99
(see fouling; scaling) 39 fluid dynamics (cross flow rate)
intermolecular forces (molecular 73
interaction) 17, 109 solute dynamics (diffusivity) 73
hydrogen bonds 8, 23, 109 (see also process dynamics)
polarity (charge-dipole; master flux equation
dipole-dipole) 8, 23, 24, 109 (see flux equation)
dispersion forces 109 Maxwell’s demon 8
intrinsic performance/properties 12, McCabe-Thiele diagram 85, 86, 87
14, 48, 50, 51, 53, 85, 98, 99 mean activity coefficient 24, 123
144 Subject Index

membrane – Office of Saline Water (OSW) 37

– bioreactor 78, 99, 103, 111 Ohmic analogy 11, 14,
– boundary 34 31, 116
– distillation 7, 9, 111 operating pressure
– filtration 5, 6, 63 (see feed pressure)
– stage (module; plant) 51, 72, 85 organophilic polymers 88
– thickness 32, 36, 49, 66, 83, 92 (see membrane polymers)
membrane polymers 131 osmosis 10, 39, 60, 113
glassy vs. rubbery 6, 35, 67, 83, osmotic –
84, 88, 93, 107, 115 – cell (Pfeffer) 17, 39, 40
hydrophilic vs. hydrophobic – distillation 9, 19, 45, 111
44, 45, 49, 80, 88, 108, 115 – equilibrium 4, 41
note: hydrophobic loosely taken – experiments 42, 44
as organophilic – power 60
membrane barriers – pressure 41, 43, 45, 46, 48, 59,
charged 7, 8 61, 117, 123
functionalized 7 – pump 59, 60
homogeneous (“dense”) 6, 31, osmotic pressure limitation
82, 108 45, 46, 48
liquid 6, 84 osmotically relevant (solute size)
porous (microporous) 5, 7, 31, 19, 44, 64
108, 134
membrane water content 7, 105 partial molar free energy
pore fluid (porosity) 105, 112 (see chemical potential)
sorption (structural water) partial molar volume 47
49, 105, 108 partial pressure (see pressure)
(see also swelling) partial solute rejection 74, 75
Merten model 46, 47, 48, 50, 70 pathogens 54, 66
microorganisms 6, 9, 64, 74, 78 perfect mixing 12
microsolutes 64, 74, 75, 77 permeability 3, 11, 12, 32
mixed solvents 111 permeance 11, 29, 36, 90
mixing/demixing 3, 8, 25, 96 permeate pressure 85
(see also phase separation) permeate recovery (see recovery)
molecular interactions 8, 17, 23 phase –
(see also intermolecular forces) – boundary 22, 24, 28
molecular modeling 11, 115 – inversion 106, 111
molecular weight cutoff – separation (demixing) 20, 22,
(ultrafiltration) 76 87, 96, 101
mol fraction/mol number 25, 26 – separation (enrichment) 95, 96,
97, 101
natural enrichment (phase phenol 96–98
separation) 87, 95, 101 phenol enrichment vs. phenolate
negative azeotrope 23 rejection 97, 98
negatively nonideal 23 pig’s bladder (animal membrane)
nominal rating (ultrafiltration) 76 4, 5, 106, 113
number effect vs. activity effect plasticization (see swelling)
19, 20, 34, 42 plasticizing parameter 35
(see also mass action) plug flow (see mass transfer) 75
 Subject Index 145

polarity interaction πV energy (chemical) 42, 43

(see intermolecular forces) (see also energy)
polarization equation 70
polarization model (stagnant film Raoult’s law 20, 23, 43, 84, 94
model) 68, 72 rating
polymer-solvent compatibility absolute (microfiltration) 77
87, 109, 111 nominal (ultrafiltration) 76
polymer swelling (see swelling) recovery; permeate recovery 14, 52,
pore blockage/pore constriction 54, 55, 72
74 (syn. recovery ratio; conversion)
pore fluid/pore space 7, 75 reflection coefficient 70
(see membrane water content) rejection vs. retention 76
pore size/pore size distribution relocation (see mass transfer)
6, 67, 68, 76, 80 retentate (reject) 51
“pores or no pores” 106 retention
porosity; membrane porosity 71 size retention 65
surface porosity 7, 66, 68 electrostatic retention 65
volume porosity 7, 8, 66 Ringer solution 43
porous (vs. homogeneous) barrier rubbery polymers 6, 35, 84, 90, 93,
positive azeotrope 20, 107, 116
positive nonideality 20, 34, 85, 94 (see also membrane polymers)
pressure (operating variable)
9, 59, 103 salinity 44
feed pressure 10, 51, 54 salt passage 46, 47, 48
hydraulic pressure 43, 59, 92 salt rejection (solute rejection)
permeate (downstream) pressure 48, 49, 70
85 salting-out 19, 97, 98
pressure loss 67, 92 saturation vapor pressure 83, 85,
pressure (state variable) 9, 26 91, 92
partial pressure 19, 20, 81, scaling (deposit) 54, 58
83, 84 scaling up 103
saturation vapor pressure 20 Schroeder’s paradox 34
vapor pressure 9, 10, 21, 30 seawater 31, 44, 50, 55, 59
(see also osmotic pressure) selectivity 3, 11, 49
pressure exchanger 55, 60 enrichment factor 86, 87, 93
pressure-normalized flux intrinsic selectivity 14
(permeance) 14, 29 natural selectivity
pressure retarded osmosis 61 (phase separation) 87
pretreatment/posttreatment ratio of rates 49
(of water) 54, 56 separation factor 76, 85
process dynamics 3, 10, 12, 48, 51, semipermeable membrane 4, 5, 19,
71, 73, 85, 98 39, 117
(see also mass transfer coefficient) (see also ideally semipermeable)
productivity (see permeance) separation factor (see selectivity)
proteins 43, 64, 65 separation categories (generic), see
protein fouling 7, 79, 80 enrichment
pV energy (mechanical) 42, 43, 54, depletion
55, 60 fractionation
146 Subject Index

phase separation surface skimming 42

precipitation swelling; polymer swelling 8, 33, 35,
volume reduction 36, 49, 83, 90, 91, 92, 93
dehydration/dewatering plasticization 33, 35, 90, 102
shear-induced diffusion 74 swelling profile 29, 36, 91, 92
Sherwood plot/diagram 2, 3, 30 (see also membrane water
SI units 13 content)
sieving 63, 66, 75
sieving coefficient 48, 76 tangential flow (see cross flow)
silt density index 74 “10 gfd criterion” 50, 67
skin, skinned membrane 50 thermodynamic activity (see activity)
(see asymmetric membrane) thin film composite membrane
sol (macromolecules in dispersion) (see asymmetric membrane)
64 tortuosity factor 68
solubility map 110
solubility parameter 34, 109, 110 vapor permeation 89
solute activity (see mass action) vapor pressure (see pressure) 30
solute rejection (salt rejection) vapor-liquid equilibrium (VLE)
48, 74, 87 4, 81
solute size 63, 64 “vigor” (solute activity),
effective 63, 64 see mass action 20, 24
geometric 63, 64 volatile organics
solution-diffusion 6, 14, 32, 46, 49, high boilers 22, 82, 87, 95, 101
65, 82, 91, 108, 109, 115 low boilers 22, 82
(see also transport modes) “volatile with steam” 95
solvent recovery (see recovery) volatility (see also Henry coefficient)
sorption (absorption) – 14, 32 22, 95, 96
– capacity 8, 49, 90 volume flux (see also flux)
– coefficient 34 convective 14, 67, 68, 72, 75
– isotherms 14, 24, 32, 33, 34, 83, diffusive 14
89, 90 volume reduction 1, 18, 30, 46, 53,
– profile 36, 83 55, 75
– selectivity 32, 33
(see also swelling) wall concentration 53, 71
sparingly soluble wastewater; wastewater treatment
(see dilute solutions) 78, 79
standard seawater 44, 46, 48, 59 water clustering 34, 49
state variables (intensive) 25 water content (see membrane water
temperature 9, 25 content) 7
pressure 9, 25 World Health Organization (WHO)
composition (mol numbers) 58, 125
9, 25
steam distillation 22, 95 zeolites (molecular sieves) 89, 90,
sterile filtration 78 102, 108