Untitled

Download as pdf or txt
Download as pdf or txt
You are on page 1of 117

POLYMER PHASE INVERSION MEMBRANES: EFFECT OF BINARY

INTERACTION PARAMETERS ON MEMBRANE MORPHOLOGY

by
VIDYA MAHANKALI BURRA, B.Tech., M.Tech.

A DISSERTATION
IN

CHEMICAL ENGINEERING
Submitted to the Graduate Faculty
of Texas Tech University in
Partial Fulfillment of
the Requirements for
the Degree of
DOCTOR OF PHILOSOPHY
Approved

May, 1993
r'
ACKNOWLEDGEMENTS 7/. - •'

I am deeply indebted to many people who have made important contributions to

this work. In particular, I am grateful to Dr. Raghu S. Narayan for extending to me

complete freedom in defining and directing my research. I also appreciate his insightful

and constnictive critique and his emphasis on writing with clarity and precision. I would

like to thank Dr. Richard Wm. Tock, Dr. Raymond E. Desrosiers and Dr. Harry W.

Parker for their valuable advice and suggestions during the course of this research. I also

thank Dr. Richard A. Bartsch for serving on my doctoral committee.

I also owe thanks to Mr. Robert Spruill for helping me in the fabrication of the

apparatus required for the experimental work, and to the Department of Chemical

Engineering for providing me with financial support throughout this research.

A very special "thank you" to my husband, Gangadhar, without whom none of

this would have been possible, and to my family back home in India. My grandmother,

parents, brothers and sister have encouraged me in aU my endeavors, and will always be

a source of inspiration to me.

11
TABLE OF CONTENTS

ACKNOWLEDGEMENTS ii

LIST OF TABLES v

LIST OF FIGURES vi

LIST OF SYMBOLS vii

CHAPTER

1 INTRODUCTION 1

2 LITERATURE SURVEY 11

2.1 Empirical models 11

2.2 Mathematical models 14

2.2.1 Mass transfer models 16

2.2.2 Thermodynamic models 19

2.3 Summary of previous studies 26

2.4 Aim of this study 28

3 DEVELOPMENT OF MODEL 29

3.1 Theory 29

3.1.1 Flory-Huggins binary interaction parameters 29

3.1.2 Binodal curve (cloud point curve) and


and coagulation values 30

3.1.3 Effect of binary interaction parameters


on system miscibility 32

3.2 Plan of work 35

111
4 EXPERIMENTS 37

4.1 Choice of polymers 37

4.2 Initial experiments 39

4.3 Materials 40

4.4 Preparation and examination of membranes 40

4.5 Determination of coagulation values 41

4.6 Determination of nitrogen transport

rates of membranes 42

5 COMPUTATIONS 46

5.1 Xi3 values for PSf-water and CA-water pairs 46

5.2 Development of equations for estimating Xj^ and X23 46

5.3 Determination of concentration-dependent


Xj2 and X23 values 49

6 RESULTS AND DISCUSSION 52

7 CONCLUSIONS 79

8 RECOMMENDATIONS 81

REFERENCES 83

APPENDICES

A LISTING OF PROGRAM FOR UNIFAC CALCULATIONS 90

B LISTING OF PROGRAM FOR UNIFAP CALCULATIONS 98

C DATA FILE OF UNIFAC PARAMETERS 105

D AX CALCULATIONS FOR PSf-(NMP-DMSO) SYSTEM 109

E NITROGEN FLUX CALCULATIONS


FOR DENSE CA MEMBRANES 110
iv
LIST OF TABLES

6.1 Coagulation values, AX values and nitrogen


fluxes for PSf systems, at 25 °C 67

6.2 Coagulation values, AX values and nitrogen


fluxes for CA systems, at 25 °C 68
LIST OF FIGURES

1.1 Cross-sectional view of a dense polymer


membrane (magnification 1500) 5

1.2 Cross-sectional view of a porous polymer


membrane (magnification 1500) 6

1.3 Cross-sectional view of an asymmetric integrally-


skinned polymer membrane (magnification 1500) 7

2.1 Temary phase diagram for nonsolvent-solvent-


polymer membrane-forming system 15

3.1 Effect of binary interaction parameters


on location of miscibility gap 33

4.1 Schematic representation of apparatus used

in gas transport experiments 43

6.1 Scanning electron micrographs of various PSf membranes 55

6.2 Scanning electron micrographs of various CA membranes 63

6.3 Coagulation value versus AX for PSf systems, at 25 °C 69

6.4 Coagulation value versus AX for CA systems, at 25 °C 70

6.5 Nitrogen flux versus AX for PSf systems, at 25 °C 72

6.6 Nitrogen flux versus AX for CA systems, at 25 °C 73

6.7 Combined coagulation value versus AX data for the PSf


and CA systems examined in this study, at 25 °C 76
6.8 Combined nitrogen flux versus AX data for the PSf
and CA systems examined in this study, at 25 °C 77

VI
LIST OF SYMBOLS

a^ activity of species i

d^ combinatorial component of a^

aj^ free volume component of a^

a^*^ residual component of a^

Gjj^ excess molar Gibbs free energy for binary solution of species i and j

AGjj molar Gibbs free energy change of mixing for binary solution of species
i and j

^Q_^M4d molar ideal Gibbs firee energy change of mixing for binary solution of
species i and j

nij molar volume ratio of component i (V/Vi)

n total number of moles

Ui number of moles of species i

T temperature

Vi molar volume of species i

Xj mole fraction of component i

AXjj binary interaction parameter for species i and j

YJ activity coefficient of component i

|Xj chemical potential of component i in mixture

|Xj® chemical potential of pure species i in standard state (25 °C, 1 atm)

(j). volume fraction of component i

(().' segment fraction of component i

vii
CHAPTER 1

INTRODUCTION

A membrane is defined as a thin barrier or film that separates two fluid streams

and restricts or permits the selective transport of a chemical species across it.

The permeability of a membrane to a chemical species (permeant) is defined as

the absolute rate at which the permeant is transported across the membrane. The rate at

which two different species permeate through a membrane relative to one another is

referred to as the selectivity of the membrane. Membranes, therefore, can be successfully

used to design separation processes if adequate differences between the transport rates of

the chemical species involved can be achieved.

There are several naturally occurring membrane processes, such as the uptake of

water and nutrients from soil by plant cells, the transfer of oxygen into blood cells inside

the lung, the removal of solutes such as urea from the blood stream by the kidney, etc.

The barriers or films that are encountered in these processes are caUed biological

membranes or biomembranes. These are essential to the very existence of life.

Most commercial and industrial membrane separation processes are based on

synthetic membranes. Conceptually, these processes are similar to the naturally occurring

processes mentioned above. The transport properties of these synthetic membranes can

be tailored to meet the needs of specific applications.

Membrane processes are normally classified according to the driving force for the

transport of a chemical species across the membrane. For example, in a pervaporation


process, the driving force is the activity gradient of the individual permeants across the

membrane, while in a gas separation process, the driving force is the gradient of the

partial pressures of the gases involved. The driving force in a reverse osmosis process is

the pressure gradient across the membrane, and in dialysis applications, is the concentra-

tion gradient of the permeating species.

Synthetic membranes can be fabricated from a wide variety of materials such as

polymers, metals and ceramics. However, the vast majority of the membranes in

commercial use are based on polymeric material. The permeability and selectivity

characteristics of polymeric membranes, such as those examined in this study, are

influenced not only by the molecular structure of the polymer, but also by the physical

structure, or morphology, of the membrane.

The molecular structure of the polymer includes factors such as the physical state

of the polymer, i.e., whether it is in a glassy or rubbery state. The molecules of an

amorphous polymer are constantly in motion. However, below a certain temperature Tg,

the segments of the polymer chains do not possess sufficient energy to move past one

another, and are trapped in a chaotic, disordered state. A polymer in such a state is called

a glass, and the temperature Tg below which molecular motion is restricted is referred to

as the glass transition temperature of the polymer. An amorphous polymer which is at a

temperature above its Tg is called a rubbery polymer. The glassy or rubbery state of the

polymer, the nature of the chemical group comprising the polymer chains, the size and

nature of side groups attached to the main polymer chains, etc, determine the size of the
spaces between the polymer chains and the chemical affinity of the polymer to a chemical

species. These affect the inherent permeability and selectivity properties of the polymer.

The physical structure, or morphology, of these polymer membranes can range

from a completely dense film to completely open, skinless, sponge-like structure, as

shown in Figures 1.1 and 1.2. The morphology of a polymer membrane, determines the

mode of transport of a chemical species across the membrane, and therefore the eflective

transport rates of the chemical species.

The transport of a chemical species across a dense polymer film, for example,

takes place by a solution-diffusion mechanism under a concentration or pressure gradient

The chemical species dissolves in the polymer membrane matrix and diffuses across it.

The permeability of the chemical species in a dense polymer film, therefore, is

phenomenologicaUy the product of its diffusivity and solubility in the membrane polymer.

Dense films of polymers exhibit the inherent permeability and selectivity characteristics

of the polymer material. Dense films of rubbery polymers generally exhibit high

permeabilities but low selectivities for real gases, whereas glassy polymers exhibit high

selectivities at lower transport rates. For example, the pure gas permeabilities of CO^ and

CH4 in s icone rubber at a temperature of 35 °C and a pressure of 1 MPa [80] are

approximately 5700 barrers* and 1800 barrers respectively, which results in an intrinsic

(CO2/CH4) selectivity of 3.16 for this polymer. The large diffusivities, and therefore the

large permeabilities, and low selectivities of rubbery polymers are consistent with the

large scale segmental motion in rubbery material relative to the sizes of the penetrant

• 1 barrer = 10"'" [cc(STP).cm]/[cmls.cm Hg]


molecules. The permeabilities of CO^ and CH4 in polysulfone, a glassy polymer, under

identical conditions, on the other hand, are 4.6 and 0.18 barrers respectively [79]. The

intrinsic selectivity in this case would be approximately 25.6. For glassy materials, the

range of segmental motions, and hence the size of the gaps, are much smaller and more

narrowly distributed, thereby providing a much more restrictive environment for different

penetrants. Films of glassy polymers are therefore used in barrier applications, such as

protective coatings and packaging material where low permeabilities are desirable, but are

not viable in separation applications where high permeabilities are required.

Porous polymer membranes with pore sizes in the range of 5-10 nm use the

"sieving" mechanism to separate large molecules from smaller ones and find use in

ultrafiltration applications in biotechnology and food processing industries. Membranes

with pore sizes in the range of 0.1-10 |im are employed in microfiltration applications,

such as particulate removal in the pharmaceutical and beverage industries.

Completely dense or porous membranes with a uniform pore size are referred to

as symmetric or isotropic, since membrane structure does not vary across the thickness

of the membrane. Prior to 1960, the only known polymer membranes were isotropic or

slightly anisotropic in nature.

The first asymmetric, "integrally skinned" polymeric membrane was invented by

Loeb and Sourirajan in 1960 [1,2]. Skinning in a polymer membrane refers to the

formation of a thin dense layer called the "skin" (typically 0.1-1.0 ^im in thickness) which

is bonded to a thick, porous substructure (100-200 |im in thickness). The skin and the

substructure are composed of the same material and are formed at the same time during
s
S
9

Figure 1.1 Cross-sectional view of a dense polymer membrane


(magnification 1500)
Figure 1.2 Cross-sectional view of a porous polymer membrane
without a skin (magnification 1500)
^-•>'-v^-<.

•.JLW

Figure 1.3 Cross-sectional view of an asymmetric integrally skinned


polymer membrane (magnification 1500)

7
the membrane preparation process - hence the term integrally-skinned. The thickness and

integrity of the skin determine the transport characteristics of the membrane. The porous

sublayer functions primarily as a physical support for the thin and fragile skin and has

little or no effect on the transport characteristics of the membrane. The stnicture of a

typical asymmetric, integrally skinned polymeric membrane, is shown in Figure 1.3.

Membrane separation processes based on asymmetric polymer membranes have

attracted considerable commercial interest [3] since the late 1960s. Asymmetric polymer

membranes are used on a large scale in reverse osmosis [4], ultrafiltration [5,6] and, more

recently, in gas separation [7,8] and medical applications [9,10].

Most of these asymmetric polymer membranes are manufactured on a commercial

scale by the phase inversion technique [11]. Phase inversion refers to the process in which

a homogenous solution of a polymer in a solvent (or solvent mixture) inverts firom a

single phase into a two-phase system. The two-phase system consists of a solid (polymer-

rich) phase which forms the membrane structure and a liquid (polymer-poor) phase which

forms the pores in the final membrane. Phase inversion membranes can be prepared from

any polymer-solvent mixture which forms a homogenous solution under certain conditions

of temperature and composition, but separates into two phases when these conditions are

changed. For example, phase inversion can be induced by the evaporation of a volatUe

solvent from a homogenous polymer solution, or by cooling a casting solution which is

homogenous only at elevated temperatures. Phase separation can also be brought about

by the introduction of a third component (nonsolvent) into a homogenous polymer

8
solution. This particular variation is referred to as the wet phase inversion technique. The

general procedure for the preparation of a wet phase inversion membrane is as foUows.

(a) An amorphous polymer is dissolved in an appropriate solvent to form a solution

containing 10 to 30 weight% polymer.

(b) The homogenous polymer solution is cast into a thin film. The fihn is typically 100

to 200 (xm in thickness.

(c) The film is then quenched in a coagulation bath to precipitate the polymer membrane

out of the solution. The coagulation medium is typicaUy water, which acts as a strong

nonsolvent for most of the polymers used in the preparation of these membranes.

In some cases, the quenching step is preceded by a short evaporation period.

However, the principal structure-forming processes which influence the transport

properties of these membranes occur in the quenching step, during which nonsolvent-

solvent exchange takes place foUowed by various phase transformations in the polymer

film. Membranes with different morphologies and, therefore, different transport

characteristics can be obtained by the use of various combinations of polymer, solvent,

nonsolvent and processing conditions.

During the phase inversion process, the combination of steps leading to a given

membrane structure involves a complex interaction of thermodynamic and mass transfer

processes. However, in aU cases, the basic membrane formation mechanism is govemed

by simUar thermodynamic and kinetic parameters, such as the chemical potentials and

diffusivities of the individual components and the Gibbs free energy of mixing of the

system. An accurate description of the thermodynamic and kinetic phenomena occurring


during the phase separation process is vital to understanding the membrane formation

mechanism and, therefore, to optimizing membrane structure and transport properties.

The commercial importance of synthetic polymer membranes and the complexity

of the phase inversion process have led to several studies of the process conditions and

mechanisms leading to the formation of different membrane structures. However, even

though phase inversion is used extensively in the manufacture of asymmetric polymer

membranes, basic phenomenological understanding of the thermodynamic and kinetic

factors that influence the transport properties of these membranes is far from complete.

Tliis study, therefore, attempted to develop a simple model based on fundamental

thermodynamic factors that can correlate the process variables used in the membrane

preparation process to membrane morphology and transport characteristics.

10
CHAPTER2

LITERATURE SURVEY

The foUowing sections wiU discuss some of the empirical studies, kinetic and

thermodynamic models related to the formation of phase inversion membranes.

2.1 Empirical models

The earUest studies of membrane formation by the phase inversion technique were

largely empirical in nature. The emphasis of most of these studies was on detaUed recipe-

like descriptions of the membrane preparation techniques. Subsequent studies tried to

provide quaUtative descriptions which correlated various membrane preparation

parameters to membrane structure and transport properties.

Manjildan et al. [12,13] prepared various ceUulose acetate membranes using

acetone and formamide solvents. These studies attempted to obtain the "optimum"

conditions for the formation of ceUulose acetate (CA) reverse osmosis membranes.

The composition of the casting solution and operating parameters such as

evaporation time, gelation bath composition and temperature, and relative humidity of the

surrounding environment, affect the porosity of phase inversion membranes. Kesting et

al. [14,15] examined the effects of these variables on the water content and, therefore, the

porosity of the swoUen membranes obtained at the end of the coagulation step. SimUar

studies were performed by Leymone et al. [16,17]. They also examined the effect of the

11
addition of inorganic salts to the casting solution of CA-acetone mixtures in order to

optimize the performance of these membranes.

Studies performed by Strathmann et al. [18,19] attempted to rationaUze the effect

of various preparation variables on the properties of Loeb-Sourirajan-type reverse osmosis

membranes. These studies revealed that low precipitation rates of the polymer from the

casting solution produced membranes with sponge-like structures, whUe high precipitation

rates yielded membranes with large finger-Uke macrovoids. The authors proposed a

mechanism based on the differences in the precipitation rates of the polymer to explain

the asymmetric nature of the membranes. Studies by Uragami et al. [20] concluded that

it is important to substitute the solvent in the polymer film with the nonsolvent in the

coagulation bath at high velocities in order to make structures with finger-like voids.

SimUarly Bottino et al. [21] performed experiments to estabUsh the relationship between

macrovoid formation and the ratio of solvent to nonsolvent flow between the casting

solution and the coagulation bath.

Strathmann et al. [22] also attempted to correlate the water content of the water-

swoUen membranes at the end of the coagulation step, and therefore the porosity, of

ceUulose acetate and polyacrylonitrUe phase inversion membranes to the solubiUty

parameters of the solvents used in the preparation of these membranes. Studies simUar

to these were also performed by Chawla and Chang [23], Cabasso et al. [24], and

Uragami et al. [25]. The studies by Chawla and Chang also examined the effect of

increasing the concentration of a weak nonsolvent poreformer, ethanol, in a casting

solution containing CA and acetone. The solubiUty parameters of the casting solution,

12
plotted on a two-dimensional solubiUty map, were correlated to the porosity and

permeabUity of the resulting membranes. Uragami and Cabasso observed that there were

simUar characteristics in certain solvent groups, which can l^e classified by their solubUity

parameters.

Kesting and Menefee [26] attempted to correlate the effect of increasing the

concentration of formamide in the casting solution and the viscosity of the casting

solution to the porosity, optical properties and permeabiUty characteristics of CA phase

inversion membranes. Cabasso and KUen [27] also showed tiiat high viscosities for the

spinning dope and relatively moderate quenching conditions in the bore fluid should be

used to get macrovoid-free polysulfone (PSf) hoUow fiber membranes. Empirical studies

of the conditions required for the formation of finger-like macrovoids in polysulfone

ultrafiltration membranes were also performed by Doi and Hamanaka [28]. They found

that a "mixabiUty" parameter, which was obtained from titration experiments, and the

viscosity of the coagulation medium could be used to predict membrane morphology. It

was also noted that finger-Uke void layers appeared under conditions where the coagulant

has both a low mixabiUty parameter and low viscosity. These results were appUed to

obtain pore size variations in PSf hoUow fiber membranes.

Badenhop and Bourgignon [29] examined processing conditions and casting

solution formulations to control pore formation in polysulfone (PSf) and polyamide (PA)

phase inversion membranes. More recently, Miyano et al. [30] studied the effect of

polymer molecular weight and solvent and casting solution composition on the pore sizes

and pore size distributions of polyethersulfone (PES) membranes. They attempted to

13
correlate the effective hydrodynamic radius of the polymer in the casting solution and the

average pore size on the membrane surface and proposed a mechanism for pore formation

based on these correlations.

2.2 Mathematical models

More recentíy, mathematical models based on fundamental kinetic and thermody-

namic concepts have attempted to provide a more definitive quantitative analysis of the

phase inversion process.

Kinetic models of the phase inversion process have examined the mass transfer

dynamics of the process. The focus of these models is the analysis of the mass transfer

paths of the solvent and nonsolvent inside the polymer film and their effects on

membrane structure formation. Several thermodynamic models have analyzed various

aspects of nonsolvent-solvent-polymer phase diagram behavior, such as spinodal and

binodal curves and tie-lines, and examined their relevance to the membrane formation

process.

The key aspects of the nonsolvent-solvent-polymer phase diagram wiU be

discussed in this section, foUowed by a review of some of the kinetic and thermodynamic

models.

The nonsolvent-solvent-polymer membrane-forming system can be represented by

a temary phase diagram as shown in Figure 2.1, where the comers of the triangle are the

14
Solvent
2

^Binodal curve (Cloud point curve)


Spinodal curve
A'
MiscibiUty gap

Nonsolvent

Figure 2.1 Temary phase diagram for nonsolvent-solvent-polymer


membrane-forming system

15
pure nonsolvent (1), solvent (2) and polymer (3). The polymer is assumed to be mono-

disperse, i.e., even though the polymer is a mixture of molecules of different sizes, aU the

polymer molecules are assumed to be of equal size and molecular weight.

The binodal curve, or cloud point curve, represents the limit of the miscibiUty of

the system. The area inside the binodal curve is caUed the miscibiUty gap. Any system

whose composition Ues inside this gap separates spontaneously into two phases which are

in equiUbrium with each other. The miscibiUty gap is again separated into metastable and

highly unstable regions by the spinodal curve. The area inside the spinodal curve is highly

unstable whUe that between the spinodal and binodal curves is metastable. A line which

connects two points representing two phases which are in equiUbrium with each other is

caUed a tie-line. For example, in Figure 2.1, if the points A and A' represent two phases

in equiUbrium with each other, the line AA' represents a tie-Une across these two points.

Any system whose composition corresponds to a point on the tie-Une AA' demixes

spontaneously into two phases with compositions A and A'.

2.2.1 Mass transfer models

Early models [31,32,33] of the mass transfer processes occurring during the

membrane formation process focused mostly on the optional evaporation step. These

models assumed tíiat skin formation and, therefore, tiie determination of the membrane

characteristics occur during the evaporation step. The basic stmcture of aU the evaporation

models is simUar. The solvent volume fraction in the film was determined by the solution

of a one-dimensional, unsteady-state, diffusion equation, after specifying the initial and

16
boundary conditions in tíie film. Ray [34] presented a model which claimed to t>e an

improvement on the earUer approaches due to the incorporation of volume changes in the

film and local thinning effects into the model.

However, otiier studies [35,36] have concluded tíiat asymmetric, skinned

membranes can be obtained in the absence of an evaporation step. In addition,

experimental observations [14,15,37,38,39] have showntfiatmembraneskin characteristics

are strongly affected by changes in the quench batíi conditions, such as bath composition

and temperature. Most of the subsequent efforts, therefore, used the evaporation models

simply in the form of an initial condition for the modelUng of the quenching step.

Cohen et al. [40] developed one of the earUest models for the diffusion-controUed

formation of porous membranes obtained by the phase inversion process. Their theory was

based on the diffusion-induced phase separation at the spinodal of the temary nonsolvent-

solvent-polymer system. By assuming that phase separation begins at compositions lying

at the limit of metastabiUty, i.e., at the spinodal, and by treating the flow of the solvent

and nonsolvent in the film as a one-dimensional diffusion process, they formulated criteria

for the formation of the membrane stmctures. These criteria are expressed in terms of the

chemical potentials of the components in the stable and metastable regions of the temary

phase diagrams.

The above model of Cohen et al. neglects mass transfer in the coagulation bath

and assumes that surface concentrations of the nonsolvent in the bath is constant

throughout the coagulation process. This, however, is contradictory to the observations

17
of Cabasso [41]. In addition, other studies [14,15,18,42] have concludedtíiatmass transfer

on the bath side can have a strong influence on the stmcture of the membranes.

Studies by Yilmaz and McHugh [43,44] were based on the premise that the

principal stmcture-forming processes for both the skin and the sublayer occur during the

quenching step. They simplified a temary, three-dimensional diffusion equation into a

one-dimensional, pseudo-binary diffusion model by neglecting the net macroscopic

movement of the polymer molecules. Their model accounted for the effects of the surface

boundary conditions on asymmetric membrane formation. By use of suitable initial and

boundary conditions, they were able to plot the concentration profiles of the solvent and

nonsolvent in the film on the temary phase diagram to determine the conditions under

which the profile enters the unstable region of the miscibiUty gap. These temary

concentration plots were used to differentiate formation conditions for completely dense

morphologies from those of asymmetric and porous stmctures. However, since this model

treats the flux ratio of the nonsolvent to solvent as a parameter that must be determined

experimentaUy, its predictive abiUty is limited.

Tsay and McHugh [45,46] developed a simUar model to study the mass transfer

processes associated with the quench bath period of the phase inversion process. In

addition to the diffusion equations formulated by Yilmaz and McHugh, the model of Tsay

and McHugh included separate diffusion equations to account for the dynamics of the

bath side mass transfer process. In this model, it was considered that phase separation

occurs whenever the mass transfer path crosses the binodal curve and reaches the

metastable region. DetaUed calculations were carried out for the water-acetone-cellulose

18
acetate system. By imposing the mass transfer path on the temary phase diagram, the

model calculated the precipitation time required for these membranes and attempted to

correlate this time to membrane morphology.

Tsay and McHugh [47] also analyzed the effects of the combined evaporation and

quench steps of the phase inversion technique. The primary focus of this study was the

effects associated with the evaporation step variables such as evaporation time, initial

cast-film thickness, and vapor phase composition.

Young and Chen [48] investigated the mass transfer mechanism for the diffusion-

controUed formation of ethylene vinyl alcohol (EVAL) phase inversion membranes. They

proposed a two-step mechanism model to describe the mass transfer processes occurring

during the formation of these membranes. The authors hypothesized that the ratio of the

nonsolvent inflow to solvent outflow is important in determining the stmcture of the top

layer or skin of the membrane. More specificaUy, the compactness of the skin depends

on the local composition at the time of phase transition. The solvent and nonsolvent flow

across the lower layers of the membrane and, therefore, the stmcture of the lower layers,

depend on the compactness of the skin. The morphology of various EVAL membranes,

obtained by changing the solvent content of the coagulation bath, was explained on the

basis of this model.

2.2.2 Thermodynamic models

One of the earUest systematic studies of the thermodynamics of the phase

inversion process was conducted by Koenhen et al. [49]. They proposed a general

19
two-step mechanism for the formation of these membranes based on the phase equiUbria

of the polymer solutions. According to their hypothesis, the membrane skin is formed by

the direct desolvation of the top layer of the polymer solution, foUowed by gelation of

the polymer chains. The stmcture of the porous sublayer below the skin layer is

dependent on the liquid-Uquid separation in the polymer casting solution. In most cases,

the porous substructure is formed by the nucleation and growth of the dUute polymer

phase. The concentrated polymer phase forms the continuous matrix that constitutes the

fmal membrane after complete desolvation. The authors also noted that when the

difference in chemical potential for the solvent between the polymer solution and the

nonsolvent is large, finger-like macrovoids are present in the membrane stmcture.

A simUar study by Broens et al. [50] attempted to describe the membrane

formation mechanism based on theoretical and experimental knowledge of phase

separation phenomena in concentrated polymer solutions of CA, PSf, polyacrylonitrUe

(PAN) and polydimethylphenyleneoxide (PDPO). They demonstrated that different types

of phase separation phenomena, such as Uquid-Uquid demixing and gelation of the

concentrated polymer phase, are responsible for the buUdup of the dense skin layer and

the supporting porous sublayer in asymmetric membranes. From the temary phase

diagrams of CA-dioxane-water and PSf-dimethylformamide (DMF)-water systems, they

concluded that the factors which favor tíie formation of a dense skin, i.e., tíie formation

of a reverse osmosis type of membrane are: (a) a higher initial polymer concentration in

the casting solution, which increases supersaturation of the polymer in the top layer of

the membrane during the coagulation step; (b) a lower tendency of the nonsolvent to

20
penetrate the top layer of the cast fUm; and (c) a lower coagulation bath temperature

which, once again, increases the supersaturation of the polymer in the top layer. On the

other hand, factors favoring the formation of ultrafUtration type of membranes are lower

polymer concentrations in the initial casting solution and addition of nonsolvent to the

polymer casting solutíon.

The authors also observed that membrane-forming systems which require large

nonsolvent concentrations for precipitation to occur result in reverse osmosis membranes,

whUe systems requiring low nonsolvent concentrations favored the formation of

ultrafiltration membranes.

Other studies by Altena and Smolders [51], Wijmans et al. [52], and Gaides and

McHugh [53] examined phase separation phenomena in membrane-forming systems of

CA and PSf using differential scanning calorimetry (DSC). CA-dioxane-water systems at

low polymer concentrations showed Uquid-Uquid phase separation. At higher polymer

concentrations, these solutions showed characteristic endothermic peaks upon cooling,

demonstrating that gelation/crystallization occurs in tíiese systems. Calorimetric

measurements of PSf-dimethylacetamide (DMAc)-water systems at very high polymer

concentrations, on the other hand, suggested that gelation in these systems is amorphous.

Altena and Smolders also discussed the characteristic properties of some of the PSf-DMac

membranes obtained in their study. A PSf-DMAc solution with a polymer concentration

of 15 weight% yielded a typical porous ultrafUtration membrane, whUe a 35 weight%

polymer solution resulted in a pervaporation-type membrane. The difference in the

stmcture of the membrane skin obtained in these two cases was explained as foUows in

21
terms of the "composition path" foUowed by these systems during the coagulation step.

The different composition changes for the top layer (skin) and the sublayer as visuaUzed

in the temary phase diagram are referred to as tíie composition path foUowed by the

system. The autíiors hypotíiesized tíiat tíie 35 weight% polymer solution yields a

membrane with a dense skin because the composition path foUowed by the system during

the coagulation step does not crosstíiebinodal curve. Lowering the polymer concentration

in the casting solution to 15 weight % increases the possibiUty of the coagulation path

crossing the binodal curve and, therefore, of liquid-Uquid demixing occurring in the skin

layer. From this analysis, it was concluded that the location of the binodal curve in the

temary phase diagram is important in determining the nature of the membrane skin.

Burghardt et al. [54] also examined the glass transition, crystaUization and

thermoreversible gelation in temary polyphenyleneoxide (PPO) membrane-forming

systems. The phase equiUbrium t)ehavior of PPO-trichloroethylene solutions in the

presence of ethanol, methanol and octanol nonsolvents was examined. The effect of the

solvent-nonsolvent mixtures on the glass transition temperature depression of the polymer

was studied using DSC. CrystaUization and gelation behavior was examined by DSC and

the falling baU method [55]. Gelation in these systems was found to be entirely

amorphous. The authors hypothesized that knowledge of the glass transition behavior of

membrane-forming systems is essential to understanding the mechanism of skin formation

in these membranes. Using theoretical equations for the glass transition temperature, they

generated glass transition curves in the temary phase diagrams of these systems. For the

systems under consideration in this study, it was observed that the branch of the binodal

22
curve corresponding to the concentrated polymer phase Ues in the glassy region of the

temary phase diagram in aU cases. Based on these results, it was concluded that as the

solvent diffuses out of the polymer film during the quenching step, the concentrated

polymer phase eventuaUy becomes a glass. The authors thus provided a simple

mechanism for the formation of the membrane skin in these systems.

Reuvers et al. [56] examined the demixing and gelation behavior of temary CA-

solvent-water systems for polymer concentrations of up to 40 weight% in dioxane,

acetone and tetrahydrofuran solvents. They measured cloud points as a function of the

rates at which the solutions were cooled and attempted to distinguish the process of

polymer aggregate formation from the more rapid process of Uquid-Uquid demixing using

scanning electron microscopy (SEM). It was observed that the appearance of polymer

aggregates depended strongly on the type of solvent used in the casting solution. The

gelation behavior of these systems was characterized by the falling baU method in order

to determine the position of the sol-gel transition curve in the temary phase diagrams.

Mulder et al. [57] concluded that stmcture formation of asymmetric membranes

reflects different time regimes required for composition changes to occur at different

positions in the coagulating film. They hypotíiesized that for the formation of the skin,

there are two possibiUties: (a) tíie coagulation path crosses the gel region before

intersecting the binodal demixing gap or (b) the coagulation path intersects the binodal

curve prior to entering the gel region. The location of the binodal curve is therefore

important in the formation of the dense top layer. It was observed that as the proximity

of the binodal curve to the polymer-nonsolvent axis increased, the greater was the

23
lUceUhood that the gel region is reached before the binodal curve is crossed by the

coagulation path and,tíierefore,tíiatthe membrane skin is dense.

Altena and Smolders [58] attempted to predict tíie tíieoretical binodal curves for

CA-solvent-water and PSf-solvent-water systems used in the preparation of asymmetric

ultrafiltration and reverse osmosis membranes. The effect of the concentration-dependent

Flory-Huggins binary interaction parameters on tiie location of tíie miscibUity gaps in tíie

temary phase diagrams of these systems was examined. Once again, it was observed that

systems which require only smaU amounts of nonsolvent for Uquid-Uquid phase separation

to occur resulted in membranes with ultrafiltration properties, whUe those requiring large

amounts of nonsolvent yielded reverse osmosis type of membranes.

Yilmaz and McHugh [59] performed calculations based on the Flory-Huggins

theory to analyze the behavior of the temary nonsolvent-solvent-polymer phase diagram

of the water-acetone-CA membrane-forming system. This study took into consideration

the behavior of the spinodal as weU as the binodal curves, tie-Une slopes and critical

points as functions of the Flory-Huggins binary interaction parameters. These calculations

demonstrated the importance of the interaction parameters in controlUng phase diagram

behavior and showed particularly the importance of the concentration-dependence of the

solvent-polymer and solvent-nonsolvent interaction parameters in affecting the nature of

the miscibiUty gap. Two different values of water-acetone interaction parameters obtained

from Uterature sources [40,58] were used to calculate the binodal and spinodal curves and

the tie-Unes for this system. It was observed that by use of varied values of the

nonsolvent-solvent interaction parameters there is a considerable difference not only in

24
the locatíon of the curves, but also in the slopes of the tie-lines. Based on these results,

the autíiors hypothesized that during the quenching step, the point of entry of the

coagulation path into the region of immiscibUity wUl have a strong influence on the

porosity of tíie precipitated layers. Therefore, tíie calculation of the sequential buUdup of

porosity in the membrane would require knowledge of the positions and slopes oftiietie-

Unes, reiterating the importance of the concentration-dependence of the interaction

parameters. The concentration-dependent nonsolvent-solvent interaction parameters used

in this study were obtained from experimentaUy determined excess Gibbs free energy data

for the binary water-acetone system. The concentration-dependent solvent-polymer

interaction parameters were obtained from tabulated values in Uterature for the acetone-

CA system [60].

A simUar study was performed by Shuguang et al. [61] who investigated solution

thermodynamics for the water-DMAc-PSf system. This study assumed constant values for

the water-PSf and DMAc-PSf interaction parameters, whUe the water-DMAc interaction

parameter was determined exactíy as a concentration-dependentparameter. The theoretical

cloud point curve calculated from these parameters was in reasonably good agreement

with the experimentaUy determined cloud point data.

Zeman and Tkacik [62] performed a thermodynamic analysis of the water-N-

methylpyrroUdinone(NMP)-PSf membrane-forming system. The concentration-dependent

water-NMP interaction parameter was obtained from the excess Gibbs free energy data

for this system. The concentration-dependent solvent-polymer interaction parameter was

obtained from light-scattering and refractive index measurements of PES-NMP solutions.

25
The water-polymer interaction parameter was obtained from polymer swelUng data. The

experimental values of the cloud point curves agree very weU with the theoreticaUy

determined temary phase diagram for this system

Swinyard and Barrie [63] studied phase separation in nonsolvent-dimethylformam-

ide(DMF)-PES and nonsolvent-DMF-PES systems. The nonsolvents used in this study

were water, methanol, ethanol, 1-propanol, acetone and carbon tetrachloride. The authors

determined the cloud point curves for these systems and attempted to correlate the amount

of nonsolvent required for the polymer to precipitate out of the solution to the binary

interaction parameters of the membrane-forming systems.

A similar study by Lau et al. [64] examined phase separation in water-solvent-PSf

and water-solvent-PES systems. The solvents used in this study were NMP, DMAc,

dimethylsulfoxide (DMSO), tetramethylurea (TMU) and dimethylpropyleneurea (DMPU).

The authors obtained tie-Une data and cloud point curve data for these systems, and

correlated the size of the miscibiUty gap in each case to the solvating power of the

solvent for the polymer and the degree of tolerance of the polymer to the nonsolvent,

which in this case was water.

2.3 Summarv of previous studies

The earUest empirical studies were too recipe-oriented [12,13,18,19], Other

empirical studies were largely limited to experimental observations of membrane

morphology and transport properties and qualitatively correlating these to various

membrane preparation parameters. Also, the results of some of these studies were too

26
specific to be appUed to other systems in general. For example, Kesting et al. [15]

observed that increasing coagulation bath temperature resulted in increased membrane

porosity and permeabUity in CA membranes. This phenomenon was quaUtatively

explained on the basis of the gelation rate of the casting solution in the batíi. SimUarly,

the increasing permeabiUty of CA membranes with increasing concentrations of

formamide in the CA-acetone casing solutions [26] was attributed to the abUity of

formamide to form hydrogen-bonds with the coagulation medium, water.

Some of the empirical studies also required extensive experimentation. For

example, Chawla and Chang [23] developed a two-dimensional solubiUty map for CA

using nearly tíiirty solvents and swelUng agents and obtained a "solubiUty envelope".

They hypothesized that any solvent lying inside the envelope was a "good" solvent and

therefore yielded dense membranes, whUe the others were either sweUing agents or

nonsolvents and yielded porous membranes. WhUe this concept is useful in the selection

of a solvent for a particular polymer, it did not in any way account for the coagulation

bath conditions or any of the other membrane preparation parameters. But this approach

was used very often in the membrane manufacture industry and solubiUty maps were

developed for various polymers. Such a process can therefore be time-consuming and

cumbersome.

The approaches developed by the kinetic and thermodynamic models have

contributed significantíy to the understanding of the phase inversion process at a

molecular level, they too have some drawbacks. In most cases, these models are highly

complex and cannot be easUy simpUfied. For example, in model developed by Tsay and

27
McHugh [45,46], the system and model parameters that had to be specified in order to

perform calculations for the quench step are: (a) temperature and pressure; (b) molecular

weights and partial specific volumes of the components; (c) thermodynamic interaction

parameters and their composition dependence; (d) scaling factors for tíie thickness of the

membrane and for the component diffusivities; and (e)frictioncoefficients describing the

binary interactions between the components.

The biggest drawback of aU the above models is that for a given set of membrane

formation conditions, none of them can actuaUy make any predictions about membrane

morphology or transport rates.

2.4 Aim of this studv

The aim of this study, therefore, was to develop a general model based on

fundamental thermodynamic considerations, so that it would (a) not be restricted to any

particular membrane-forming system; (b) require data that can be easUy obtained

experimentaUy or from Uterature sources; and (c) most importantíy, be able to predict

membrane morphology and transport properties for a given set of membrane-formation

conditions.

28
CHAPTER3

DEVELOPMENT OF MODEL

3.1 Theorv

The theoretical basis of the model that was developed as part of this study is

explained in the foUowing sections.

3.1.1 Flory-Huggins binary interaction parameters

The thermodynamic stabiUty of a nonsolvent-solvent-polymer system with limited

miscibiUty, such as the one represented by Figure 2.1, can be described in terms of the

Gibbs free energy change of mixing. For a stable state to exist, the Gibbs free energy of

an equiUbrium state must be a minimum with respect to aU possible changes in

composition, at a given temperature T and pressure P. The criterion for the equiUbrium

of a single phase yields the condition

AG^ < 0 {constant T, F). ^^'^^

When a homogenous single phase system separates spontaneously into two phases, the

Gibbs free energy change of mixing for the unstable state is

AG^ > 0 (constant T, F). ^^*^^

29
According to the extended Flory-Huggins theory for polymer solutions [59,65], the

Gibbs free energy change of mixing for a temary nonsolvent-solvent-polymer system is

given by

kí.

—^^=/iiln<t)i+/i2ln<|)2+n3ln<|)3+Xi2ni<t)2-HXi3ni<|)3+X23/i2<|)3 , ^^'^^

where n^ and <t)i are the number of moles and volume fraction of component i, and n is

the total number of moles in the system. X^j is the concentration-dependent binary interac-

tion parameter for a solution of species i and j . The nonsolvent, solvent and polymer are

denoted by the subscripts 1, 2 and 3 respectively.

In the simplest approximation, the binary interaction parameter X^j is a phenomen-

ological dimensionless quantity characterizing the nonideaUty of the Gibbs free energy

change of mixing of a binary solution of species i and j . It is thus a measure of the

miscibiUty of i and j within each other. A larger value of X^j indicates a smaUer decrease

in the free energy of mixing, so that when X^j b)ecomes sufficientíy large, a homogeneous

solutíon of species i and j separates into two phases, which are in equiUbrium with each

other. Xij is dependent on state variables such as composition, temperature and pressure.

A more rigorous thermodynamic definition of X^j can be found in other sources [60,66].

3.1.2 Binodal curve (cloud point curve)


and coagulation values

The binodal curve or cloud point curve, as shown in Figure 2.1, represents the

limit of the miscibiUty of the nonsolvent-solvent-polymer membrane-forming system. It

30
can be experimentally obtained by a simpletitrationmethod [67]. Binary solutions of the

polymer in the solvent (polymer dopes) at various concentrations are titrated with the

nonsolvent (water, m tfiis study) untU phase separation occurs in the polymer solution.

This is indicated by tíie formation of a permanent precipitate, i.e., persistent turbidity is

observed in the solution. The solution is held at a constant temperature. The curve formed

by joining these turbidity points is referred to as the binodal curve or cloud point curve.

The area inside this curve, as explained in Chapter 2, is the miscibiUty gap. Any system

whose composition Ues within the miscibiUty gap is unstable and separates spontaneously

into two phases which are in equiUbrium with each other.

The coagulation value (C.V.) of a binary polymer dope is a measure of the

miscibiUty of the nonsolvent-solvent-polymer membrane-forming system. It is defined

[67] as the amount of nonsolvent required to cause phase separation and yield the first

permanent precipitate (or persistent turbidity) in a homogenous polymer solution.

Therefore, when a polymer dope has a large coagulation value, it indicates that the

components of the temary nonsolvent-solvent-polymer system have greater miscibiUty in

each other; a smaU coagulation value impUes that phase separation or demixing is favored

by the system.

It is possible to determine the amount of nonsolvent required to cause phase

separation in a dope at any polymer concentration. But at very high polymer concentra-

tions, it is difficult to maintain homogeneity in the solution, since coagulation occurs

locaUy as the nonsolvent is added to the polymer dope. For this reason, it is desirable to

perform these experiments at a low polymer concentration. The coagulation value of a

31
polymer dope is therefore defined arbitrarUy as the weight of nonsolvent required to cause

phase separation in a solution of 2 g of polymer dissolved in 98 g of solvent at 25 °C.

By definition, the coagulation pomt Ues on the cloud point curve of a system at 25 °C.

3.1.3 Effect of binary interaction parameters


on system miscibiUty

The effects of changing the nonsolvent-solvent, solvent -polymer and nonsolvent-

polymer binary inleraction parameters Xj^, X^^ and X13 on tíie location of the miscibiUty

gap [58,61,63] are shown in Figure 3.1.

First, consider the effect of X12, assuming that the other two binary interaction

parameters remain unchanged. As can be observed from Figure 3.2, increasing Xj^

decreases the miscibiUty gap. QuaUtatively, this impUes that a lower tendency of mixing

between the nonsolvent and solvent favors the mixing of the system.

Next, consider the effect of increasing solvating power of the solvent for the

polymer, or decreasing X23, assuming that the other two binary interaction parameters

remain the same. As can be observed from Figure 3.2, this too decreases the miscibiUty

gap. Therefore, greater solvent-polymer affinity favors increased miscibiUty of the system.

FinaUy decreasing Xj^ , or increasing the nonsolvent-polymer affinity, assuming

that the other two parameters remain constant, decreases the miscibiUty gap. This impUes

that increasing nonsolvent-polymer affinity increases the miscibUity of the system.

Considering the combined effects of aU the parameters on the miscibiUty gap, it

can be observed from Figure 3.1 that the effect of increasing X^^ is opposite to that of

32
Increasing X

Decreasing X

Decreasing X

Nonsolvent

Figure 3.1 Effect of binary interaction parameters on


location of miscibiUty gap

33
increasing X23 and Xj^. The miscibiUty gap decreases as (a) Xj^ increases and (b) X23 and

Xi3 decrease. AII three parameters can therefore be combined to quantify the miscibiUty

of the system by a new term AX, where AX is defined as

AX=-Xi2+X23+Xi3 . (3.4)

The negative sign on X,^ takes into account its opposing effect on the miscibiUty

gap, as compared to the effects of X23 and Xj^, as discussed above. AIso, the additive

form of the expression on the right-hand side of the equation can be explained on the

basis of the fundamental assumption of additivity in the group contribution models which

were used to estimate these parameters, as explained in Chapter 5. No weighting factors

were used for the binary interaction parameters because the group contribution models

also take into account the concentration-dependence of these parameters and, therefore,

the composition of the system. The expression on the right-hand side of eqn. 3.4,

therefore, consists of the simplest and most logical choice of parameters that can be used

to define AX.

From its definition in eqn. 3.4, a smaller value of AX should indicate greater

miscibiUty of the components of system, whUe a larger AX value should favor system

demixing. Since the coagulation value of a polymer solution is also a measure of the

miscibiUty of the nonsolvent-solvent-polymer system, it would be reasonable to expect

that the coagulation value is related to the AX value of a system if the above hypothesis

is valid. Membrane-forming systems with greater miscibiUty between the components, i.e..

34
with smaU AX values should have large coagulation values. Conversely, systems with

lower miscibiUty, i.e., with large AX values, should have smaU coagulation values.

Previous studies of PES, CA and PA membrane-forming systems [50,58,67,68]

have reported the existence of a relationship between their coagulation values and the

morphology and transport properties of tiiese membranes. If AX is a vaUd measure of the

system miscibiUty, it can also be expected that it is related to tíie transport properties of

the membranes obtained from various nonsolvent-solvent-polymer systems. AX can then

be a simple parameter that can correlate the thermodynamic properties of the nonsolvent-

solvent-polymer system to the morphology and transport properties of these membranes.

3.2 Plan of work

The experiments and computations that were performed as part of this study were

therefore aimed towards achieving two objectives, which were: (a) to verify the existence

of any relationship between the AX and coagulation values of phase inversion membrane-

forming systems and examine the vaUdity of AX as a measure of the miscibiUty of the

membrane-forming system; and (b) more importantiy, to study the effect of AX, or the

miscibiUty of the nonsolvent-solvent-polymer systems, on the morphology and transport

properties of the corresponding phase inversion membranes.

The sequence of experiments that were performed involved: (a) preparation and

examination of various phase inversion membranes; (b) determination of the coagulation

values of the polymer dopes; and (c) determination of the gas transport rates of these

membranes.

35
The computations involved determination of the concentration-dependent binary

interaction parameters for the nonsolvent-solvent-polymer systems that were used in the

preparation of these membranes.

The detaUed descriptions of the experiments and computations foUow in Chapters

4 and 5.

36
CHAPTER4

EXPERIMENTS

4.1 Choice of polvmers

The polymers used in this study were Bisphenol A polysulfone (PSf) and cellulose

acetate (CA), two of the most commonly used polymers in tíie membrane manufacture

industry. The molecular stmctures of these polymers are shown in Schemes 4.1 and 4.2.

The choice of these two polymers was based on their completely differing chemical and

mechanical properties, as weU as different degrees of tolerance for water.

CH, o
II
-o- \ / s- \ / Jn
CH3 II
o

Scheme 4.1. Polysulfone (PSf)

CH2OCOCH3 OCOCH3
-0

OCOCH3 CH.OCOCH3

Scheme 4.2 Cellulose acetate (CA)

37
Polysulfone is a rigid, strong, engineered thermoplastic. The polymer backbone

made up of benzene groups with ether linkages. These groups cause polysulfone to be

strongly hydrophobic in nature. Among the properties which qualify PSf as good polymer

for membrane preparation are a glass transition temperature (Tg) of 195 °C, an amorphous

glassy state, thermal and oxidative stability, resistance to extremes of pH, and low creep

even at elevated temperatures.

CeUuIose acetate is prepared by the partial acetylation of ceUuIose. The buUding

unit of the polymer chain consists of cyclic anhydroglucose stmctures to which hydroxyl

and acetyl groups are attached. Cellulose acetate,tíierefore, has both basic (acetyl) and

acidic (hydroxyl) character. The hydroxyl groups are hydrophobic sites which form

hydrogen bonds with water, whUe the acetyl groups are hydrophobic in nature. But

overall, CA is a hydrophiUc polymer. It is used extensively in the manufacture of reverse

osmosis membranes due to its high permeabiUty rates for water. CA possesses outstanding

tractabiUty and exhibits reasonable resistance to oxidation by chlorine. However, it has

a relatively low Tg of 68.6 °C, rendering it inviable in high temperature and high pressure

membrane appUcations. It also undergoes hydrolysis in aUcaline media.

The choice of two extremely diverse polymers, PSf and CA, in this study, was

based on tíie rationale that if the model developed in Chapter 3 was vaUd for tíiese

membrane-forming systems, it is possible that this model can easily be extended to

membrane-forming systems based on other polymers.

38
4.2 Initial experiments

Initially, PSf membranes were prepared using N-methyI-2-pyrroUdinone (NMP),

N,N-Dimethylacetamide (DMAc), dimethylsulfoxide (DMSO) and N,N-DimetiiyIpropy-

leneurea (DMPU) solvents, and CA membranes were prepared using acetone, dioxane and

DMSO solvents, and examined by scanning electron micrography, as áescnbeá in Section

4.4. These solvents are the most commonly used ones reported in literature for PSf and

CA. The coagulation values of the binary polymer dopes at 25 °C and the nitrogen

transport rates of these membranes were determined as described in Section 4.5 and 4.6.

The binary interaction parameters for all these systems were estimated by group-

contribution models, as described in Chapter 5.

The results of these initial experiments, coupled with the estimated AX values for

these systems, indicated that there is a relationship between their AX and coagulation

values. Also, the visual inspection of these membranes showed that the morphologies of

these membranes varied with the AX of these systems. The morphologies of tiie PSf

membranes ranged from a dense, skinned structure supported by a sublayer with large

macrovoids for the PSf-DMPU system, to a very porous stmcture with a very thm skin

for the PSf-DMSO system; for the CA systems, tiie morphologies ranged from a

completely dense membrane for tiie CA-acetone system to a porous stmcture for the CA-

DMSO system,

Further experiments were therefore performed using N-formyl piperidine (NFP)

and dimethylformamide (DMF) solvents for PSf, and pyridine, methyl acetate (MA) and

39
tetrahydrofuran (THF) solvents for CA. Similar experiments were performed with PSf

using a (1:1) molar mixture of NMP and DMSO solvents.

The foUowing sections deal with the materials used in the preparation of the

polymer membranes and the procedures used in the membrane preparation, coagulation

value determination, and nitrogen transport experiments.

4.3 Materials

The UDEL P-3500 polysulfone and ceUuIose acetate (DS=2.3) were obtamed

from Amoco Performance Products, Inc. and Aldrich Chemical Company, respectively.

The solvents NMP, NFP, DMF, MA, acetone, pyridine and dioxane were obtained from

Aldrich Chemical Company. The solvents DMAc, DMSO and THF were obtained from

Fisher Scientific. The solvent DMPU was obtained from Sigma Chemical Company. The

99.5% purity nitrogen used in the gas transport experiments was obtained from Big Three

Industries, Inc. The polymers, solvents and nitrogen gas were used as suppUed.

4.4 Preparation and examination of membranes

The membranes were prepared from solutions of 30.00 g polymer in 100.00 g

solvent (or solvent mixture). In most cases, the solvent (or solvent mixture) was preheated

to about 60 °C before adding the polymer to it. The solvent-polymer mixture was further

heated (for up to about 4 h in some cases) and stirred occasionaUy with a glass rod, untU

it formed a homogenous solution completely free of air bubbles. The polymer solution

was cast as a thin film (approximately 200 fxm in thickness) on a glass plate (12.5 cm X

40
20 cm) and aUowed to stand for 1 minute before quenching it in a water bath at 25 °C.

The PSf membranes were washed in mnning water for 24 h and dried in the controUed

environment of a Precision Scientific air dryer for 24 h. The CA membranes were washed

in mnning water for 24 h, t)efore being subjected to an isopropanol-n-hexane solvent-

exchange process [69]. The water-swoUen CA membranes were immersed successively

in isopropanol and n-hexane baths for approximately 6 h each. The CA membranes were

then dried in the air dryer for 24 h.

For each polymer-solvent system, three membranes were cast using the procedure

described above.

Samples of the PSf and CA membranes (approximately 1.5 mm X 1.5 mm) were

mounted on a steel plate and coated with platinum in a Seevac Hummer V sputter coater.

Scanning electron micrographs (SEMs) membranes were then obtained using a Hitachi

S570 scanning electron microscope in the Department of Biology, Texas Tech University.

4.5 Determination of coagulation values

The coagulation values for 2% dopes of PSf m DMPU, NMP, NFP, DMAc, DMF

and the (1:1) NMP-DMSO mixture, as weU as coagulation values for 2% dopes of CA

in acetone, dioxane, pyridme, MA, DMSO and THF, were obtained experimentaUy at

25 °C as described below. The coagulation value for the PSf-DMSO dope was taken as

zero, since PSf does not dissolve in DMSO at 25 °C.

In each case, 2.00 g of polymer was dissolved in 98.00 g of solvent (or solvent

mixture). In some cases, the mixture had to be heated to a temperature sUghtiy above

41
25 °C and stirred to facUitate complete dissolution of the polymer in the solvent, and then

cooled to 25 °C. DistiUed water was slowly added to the homogenous polymer dope,

stirring it constantiy, until the solution yielded the first permanent precipitate (or

persistent turbidity). The temperature of the mixture was held constant at 25 °C. The

coagulation value in each case was the weight of water required to obtain the first

permanent precipitate in the polymer solution.

4.6 Determination of nitrogen transport rates of membranes

The gas transport experiments in this study were performed with the objective of

obtaining a measure of the openness of the membranes obtained from various solvent-

polymer systems, relative to each other. The aim of these experiments was not to

determine the absolute permeabUities of the membranes; the experimental conditions were

performed were therefore chosen such that any appreciable differences in the gas transport

rates of these membranes could be easUy detected.

A schematic representation of the apparatus used to perform these experiments is

shown in Figure 4.1. The permeabiUty ceU of virgin grade extmded Teflon^ consisted of

two soUd cylindrical blocks, each of diameter 10 cm and height 2 cm, which formed the

upstream and downstream components of the ceU. Each of these blocks was provided with

a circular opening at the center. The openings, which acted as the gas inlet and outíet on

the upstream and downstream sides respectively, were fitted to Swageloc male connec-

tions (1/2 " NPT X 1/4 " tube). The upstream side of the ceU was connected by 1/4 "

42
Nitrogen gas supplv

Permeability cell
Pressure guage

On-off
valve
Membrane
Gasket

-^To flowmeter
Permeate

Figure 4.1 Schematic representation of apparatus


used in gas transport experiments

43
PVC piping to a nitrogen gas cylinder. The downstream side was connected to a soap-

bubble meter to measure theflowrate of the permeate. The upstream side of the permeab-

iUty cell was also connected to a pressure gauge (0-300 psig) and a simple on-off valve.

The PSf and CA membranes, which were prepared as describ>ed previously, were

cut in the form of circular discs of 7.5 cm diameter. For the nitrogen transport

experiments, the upstream and downstream pressures in tiie ceU were arbitrarUy chosen

as 65 psig and atmospheric pressure, respectively. All these experiments were performed

at room temperature.

The foUowing procedure was repeated with each oftíiethree membranes cast from

a particular polymer-solvent system.

The membrane whose nitrogen flux was to be determined was placed between the

two blocks of the ceU, with the skinned surface of the membrane facing the upstream

side. A rubber gasket of 5.5 cm diameter was placed between the blocks, which were then

bolted together. The nitrogen gas inlet flow rate was adjusted so that the upstream

pressure in the ceU, as indicated by the pressure gauge, was 65 psig. The permeate flow

rate through the membrane was continuously monitored by the soap-bubble meter until

it reached a steady state. The nitrogen supply was then switched off. The pressure on the

upstream side was relieved by opening the shut-off valve, before removing the membrane

from the ceU.

The next set of experiments was based on the resistance model approach, proposed

by Henis and Tripodi [78]. This model was developed to explain the behavior of

composite gas separation membranes, which were made up of a porous asymmetric

44
substrate of one polymer, and a coating of another polymer. The authors examined the

permeation propertíes of a polysulfone hollow fiber membrane and a siUcone-coated

polysulfone composite fiber for a hydrogen/carbon monoxide gas mixture. They obtained

polysulfone hoUow fiber membranes with a very thin skin. But smaU defects, or pinholes

on the surface of tiie membranes prevented the fuU separation properties of the polymer

form being achieved. By using an appropriate coating such as siUcone, which offers very

low resistance to tiie permeation of gases, the authors were able to effectively plug the

defects or pores on the surface of the substrate, whUe not adversely affectmg its

separation properties in anyway.

In this study, therefore, a quaUtative measure of the relative porosity of the

membranes, obtained from various polymer-solvent systems, was obtained by performing

transport experiments with siUcone-coated membranes.

The siUcone solution was prepared by dissolving 2.00 g of siUcone in 100.00 g

of n-hexane. In each case, the membrane was dipped into the siUcone solution, removed

and dried in the air dryer for approximately 1 h. The nitrogen fluxes of these siUcone-

coated membranes were also determined under experimental conditions identical to those

used for the uncoated membranes, as described previously.

The results of these experiments are discussed in Chapter 6.

45
CHAPTER5

COMPUTATIONS

In the foUowing sections, tiie choice of X^^ values for the PSf-water and CA-water

systems wiU be discussed. This wUI be foUowed the development of the theoretical

equations for estimating the binary concentration-dependent interaction parameters X,^ and

X23.

5.1 Xi, values for PSf-water and CA-water pairs

In this study, the water-polymer interaction parameters were assumed to be

independent of the composition of the solution. This is a vaUd assumption based on

previous studies [58,61,62].

There are three reported values of Xj^, 5.9, 3.7 and 3.8, for the PSf-water system

[52,58,63]. Shuguang et al. [61] concluded in their study that the value of 5.9 for this pair

is not a reasonable one. However, as there is good agreement between the other two

values, the PSf-water interaction parameters was assumed to be 3.7 in this study.

The Uterature value of X13 for the CA-water paU is 1.4 [58].

5.2 Development of equations for estimating Xi^ and Xo,

The importance of the concentration-dependence of the nonsolvent-solvent and

solvent-polymer pairs has been demonstrated previously [58,59,61,62]. In this study,

therefore, the concentration-dependence of these parameters was taken into consideration.

46
These interaction parameters can be determined experimentaUy by several methods

[60,62,72], or can be estimated by the UNIFAC (UNIQUAC Functional-group Activity

Coefficient) and UNIFAP (UNIQUAC Functional-group Activity Coefficient for Polymer

solutions) methods. These are well-established group-contribution models for the

thermodynamic behavior of Uquid and polymer solutions [70,73].

Experimentally determined values of Xj^ and X23 reported in Uterature for some

of the water-solvent and solvent-polymer systems are contradictory [40,58,60,61,63,66,75].

In this study, therefore, aU the Xj^ and X23 values were estimated by the UNIFAC and

UNIFAP methods.

The UNIFAC approach is used to calculate component activities in Uquid mixtures

based on Prausnitz's group contribution model [71]. The excess molar Gibbs free energy

G,2^ for a binary solution of nonsolvent (1) and solvent (2) is defined as

G^^-RT [^jlnYi+x^lnYj ^^^^

=RT [x^ifna^-lnx^^+x^ilna^-lnx^] ,

where x^, a^ and YÍ are the mole fraction, activity and activity coefficient of component

i. The molar Gibbs free energy change of mixmg AG,^^ is then given by

AGI?=GI2+AGS

=RT [x^(lnai-lnx{)+xJilna^-lnx2)]+RT [x^lnx^+xjnx^] ^^-^^

=RT [x^lna^+x^lna^] .

47
From eqns. 3.3 and 5.2, it follows tíiat

_ [(^G,^/RT)-(x,]ni»,+x^lRÍ,^)]
-^12 T
(53)
_ [Xiln(a^/4)^)+jt^ln(fl^4)^)]

^1*2

where ^., is tíie volume fraction of component i in tiie binary solution and is estimated by

x^m^+x^m^

The molar volume ratio mj = (Vj / V,), where V, is tiie molar volume of component i,

reflects the difference in the molecular sizes of the components i and j . Here, the

nonsolvent molar volume V, is chosen as the reference molar volume.

Similarly, the solvent activity in a polymer solution can be estimated by UNIFAP

[73], a modified UNIFAC method. For polymer solutions, the mole fraction is an

awkward unit of concentration because the molecular weight of the polymer is so much

larger than that of the solvent that mole fractions are always close to 1.0 for the whole

range of polymer concentrations. Therefore, the UNIFAP estimation of solvent activities

is based on weight fractions. This modified method also takes into account the free-

volume effects in polymer solutions. The complete theoretical equation for solvent activity

in a polymer solution contains a free-volume term in addition to the residual and

combinatorial terms. Therefore,

48
Ina^^hia^'^+lna^^+lna/^ , (5.5)

where tiie superscripts R, C and FV refer to the residual, combinatorial and free-volume

components of the solvent activity a^. The binary solvent-polymer mteraction parameter

X23 is related to tíie solvent activity in the solution [74] by

ln.,=i!^
RT (5.6)
=ln(|)/+<|)/-.X^(<t)/)^ ,

where fi^ and 1X2*^ refer to the chemical potentials of the solvent in solution and pure

solvent at 25 °C, respectively, and (j),' is the segment fraction, which is simUar to the

volume fraction, of component i. From eqn. 5.6 it foUows that

^ (4.3^

5.3 Determination of concentration-dependent


X,o and X-,, values

Two computer programs were developed to perform the calculations required to

estimate the binary nonsolvent-solvent and solvent-polymer interaction parameters based

on the UNIFAC and UNIFAP methods. The FORTRAN codes of these programs are

Usted in Appendices A and B. The UNIFAC groups that can be used by these programs

49
are listed in Appendix C. The group interaction parameters required by these programs

were obtained from the latest supplement to the UNIFAC program [76,77].

Some of the molecular groups in the PSf and CA repeat units, as shown in Section

4.1, were not available m tiie UNIFAC database (Appendix C). The combination of

molecular groups used to represent PSf and CA in the UNIFAP program, as shown below,

were therefore chosen in such a way that the estimated values for tiie water-PSf and

water-CA interaction parameters were close to the experimental values of 3.7 and 1.4

reported in Uterature [58,63].

PSf groups CA groups

CH3 2 CH2 2

C 1 CH 6

ACH 16 OH 2

AC 8 CH3COO 4

CHO 4

The X12 calculations were based on compositions corresponding to a mixture

containing water equivalent to the coagulation value of the polymer dope, and 98.00 g of

solvent. SimUarly, the X23 values were estimated based on compositions corresponding

to 2.00 g polymer dissolved m 98.00 g solvent For the PSf-(NMP-DMSO) system, the

various binary interaction parameters were estimated in a similar fashion (Appendix D).

50
The AX value had to also account for the interaction between the cosolvents NMP and

DMSO.

The free-volume correction term a^^ in eqn 5.5 jdelded a negative value in some

cases. In the other cases, it was not significant when compared to the residual and

combinatorial terms. It was therefore discarded from the UNIFAP program.

The results of these computations are tabulated in Chapter 6.

51
CHAPTER6

RESULTS AND DISCUSSION

The scanning electron micrographs (SEMs) of tfie cross-sections of the PSf and

CA membranes (magnification 200), which were prepared as described in Section 4.4, are

shown in Figures 6.1 and 6.2, respectively. Figure 6.1 also mcludes SEMs of tiie skins

of the membranes obtamed from the PSf-NMP (magnification 250) and PSf-DMSO (mag-

nificatíon 1000) systems. The coagulation values, X,^, X23 and AX values of tiie PSf and

CA systems are Usted in Tables 6.1 and 6.2, respectively. The average nitrogen fluxes of

the coated and uncoated membranes, along with the standard deviation for the three

membrane samples obtained for each polymer-solvent system, and the flux ratios are also

shown in Tables 6.1 and 6.2.

As expected, the results of the experiments performed with PSf polymer and NFP

and DMF solvents, and CA polymer with pyridine, THF and MA solvents, were

consistent with those obtained from the initial experiments (Section 4.2). The coagulation

values, as weU as the corresponding AX values, of the PSf-NFP and PSf-DMF systems

were found to lie between those for the PSf-DMPU and PSf-DMSO systems. The overaU

morphology of the PSf-NFP and PSf-DMF membranes also varied between the PSf-

DMPU membrane, which had a dense skin and a supporting substmcture with large

macrovoids, and the PSf-DMSO membrane, which had a very thin skin supported on a

sponge-like substmcture. SimUarly for the CA systems, the CA-MA system which had

a larger coagulation value than the CA-acetone system, had a correspondingly lower value

52
of AX, and yielded a dense membrane simUar to the one obtained from the CA-acetone

system. The CA-pyridine system, which had a coagulation value and the corresponding

AX value close to that of the CA-acetone system, also yielded a dense membrane. The

CA-THF system had a coagulation value and a AX value lying between tiiose of the CA-

acetone and CA-dioxane systems. In this case too,tiieresulting membrane was completely

dense. Therefore, aU tiie CA systems with AX lower than tiiat of the CA-THF system,

yielded completely dense membranes. CA systems with AX values greater than that of the

CA-THF system, on the other hand yielded skinned, asymmetric membranes.

Thus, by performing a few initial experiments and having a knowledge of the AX

values and the morphology of the membranes obtained from these selected polymer-sol-

vent systems, this model is capable of predicting, a priori, the morphology of membranes

obtained from other polymer-solvent systems once their AX values are known.

The coagulation values of the PSf and CA systems were plotted against the

corresponding AX values, as shown in Figures 6.3 and 6.4 . From these figures, it can be

observed that, for both the PSf and the CA systems, there is a strong correlation between

the coagulation values and the AX values. For both PSf and CA, membrane-forming

systems with low AX values have high coagulation values, and systems with high AX

values had low coagulation values. In both cases, the coagulation values decreased

systematicaUy with increasing values of AX, further consoUdating the results obtained in

the initial experiments (Section 4.2).

Among the various PSf systems, only PSf-NMP system did not foUow this general

trend (Figure 6.3). The high estimated value of AX (3.929) obtained for the PSf-NMP

53
system appears be the result of tiie low value of the binary interactíon parameter

estimated by UNIFAC for tíie water-NMP pair (X,^ = 0.178). The experimentaUy deter-

mmed value of X,^ for this system [62] is 1.20. In tíiis case, if the experimental value of

X,2 were used mstead of tíie estimated value, it would yield a AX^p^ value of 2.907. This

pomt is indicated by a (+) m Figure 6.3. This point appears to foUow the general trend

shown by the other PSf systems.

From Figures 6.3 and 6.4, it appears as if there is strong linear relationship

between coagulation value and AX for the PSf and CA systems examined in this study.

The coagulation value versus AX data were tiierefore fitted to straight lines. For the PSf

systems, the PSf-NMP point was considered an outUer due to the reason discussed above,

and was therefore discarded from the regression calculations for the PSf systems. The

correlation coefficients (R^) obtained for the PSf and CA systems were 94.9% and 85.9%

respectively. The high correlation coefficients obtained for both the PSf and CA systems

indicated that the relationship between coagulation value and AX is indeed linear.

These results therefore vaUdate the hypothesis developed in Chapter 3 that AX can

be used as a measure of the miscibiUty of the nonsolvent-solvent-polymer membrane-

forming systems.

For the PSf-(NMP-DMSO) system, the AX value includes an additional term to

account for the binary interaction between the cosolvents NMP and DMSO, as discussed

in Section 5.3. The coagulation value and AX for this system Ues between those obtained

54
Figure 6.1. Scanning electron micrographs of various PSf membranes
(a) PSf-DMPU system (magnification 200)

55
Figure 6.1. (Continued): (b) PSf-NMP system (magnification 200)

56
Figure 6.1. (Continued): (c) PSf-DMAc system (magnification 200)

57
Figure 6.1. (Continued): (d) PSf-DMF system (magnification 200)

58
Figure 6.1. (Continued): (e) PSf-DMSO system (magnification 200)

59
Figure 6.1. (Continued): (f) PSf-(NMP-DMSO) system (magnification 200)

60
Figure 6.1. (Continued): (g) Skin of PSf-NMP membrane
(magnification 250)

61
' ^ . •' ^ •ii^'^>.''W '•*;»-«•

Figure 6.1. (Continued): (h) Skin of PSf-DMSO membrane


(magnificatíon 1000)

62
Figure 6.2. Scanning electron micrographs of various CA membranes
(a) MA-CA system (magnification 200)

63
Figure 6.2. (Continued): (b) THF-CA system (magnification 200)

64
Figure 6.2. (Continued): (c) Dioxane-CA system (magnification 200)

65
m- - ^.-v

Figure 6.2. (Continued): (d) DMSO-CA system (magnification 200)

66
o

>

t—«
O

H
a

""•
T3

cd

vo
cd
<o

cS
o

X
o
00

3
cd CN

o cu
^

U qii
^

>
X
00

«
System

S "

ro
Flux
ratio
Uncoated Coated
PSf-DMPU 9.5 1.350 0.062 2.412 3.478 0.945 0.27
± 0.0770 ±0.0105
PSf-NMP 8.2 0.178 0.407 3.929 3.758 0.965 0.26
±0.2119 ± 0.0215
PSf-NFP 6.9 0.570 0.030 3.160 4.294 1.525 0.36

67
± 0.2582 ± 0.0243
PSf-DMAc 5.2 0.465 0.330 3.565 4.625 1.618 0.35
±0.1365 ± 0.0429
009*0
PSf-DMF 3.5 0.553 3.653 4.675 2.391 0.51
± 0.2349 ± 0.0730
PSf-DMSO 0.0 -0.65 0.903 5.203 4.923 3.006 0.61
±0.1331 ± 0.0757

-.
PSf-(NMP/ -0.04 1.328 5.108 4.838 2.806 0.58
DMSO 1:1 ±0.1783 ± 0.0793
mixture)**
o

C
C
C

>
T3

(A

<rS

C/3
cd

'cd
cd
o^
'B ^I
Q

B c

c
o
B c
t-c
T3

B ^

T3 T3
B <^

^
2 B

t^

•o
2O

CU
C!

2 >
...
(X


a> cx,
cx,
O
X

.1—<
c

a
c
(/3

Cj
13

<N
o

CA
u

u
cO
o
'cd
dS

(U

VO
C/5
Ci)

<L> o

<
CN
•«-»

(D
•«->

(J

C/5

C/3

u
»o

s
cS

>%
^

>
U
X

X
Z X

System Flux
ratio

00
Uncoated Coated
1

CA-MA 94.9 1.694 0.275 -0.020 0.114* • 1


1

1
CA-Acet- 67.8 1.470 0.241 0.171

68
one
1
1

1
CA-Pyri- 66.9 1.140 -0.22 0.040
dine
1
1

CA-THF 53.6 0.891 0.469 0.978 0.103*


CA-Diox- 45.4 0.813 0.710 1.297 (8.58 ± (2.07 ± 0.24
ane 0.794)** 0.108)**
CA-DMSO 20.5 -0.62 -0.17 1.850 (35.07 ± (17.76 0.51
1.247)**
C 0.695)**
C

O
^

^
c

0)

cO

00
-o

<ií

JD
(L)

eo
. .-
P.

c/.

u
^
C

.^Î3
(D
w B

^
cd

J^

C
C/3

^
4_>

s
C/3

cs o
-^ S
•-^ "2

^
C nD
<a

D
"L)

ii
< .2

(U
C o
O C

cg |_
u >

-^

^* c
Í;

g CX,
T vo
<D
o
Ob

- «o
P H íd<3>

CÍH

'i^ o
o
\pMA

Q
00
o 3
\oNFP

00

less

O §;Q
oo

ens
cn •3S
-f-
co ?i
:^ 1—1
• r-^
o 6
Q + B 13
^ -a
<^ S^
O <u X I—H CA

c« 3
W ) «3
>^
co >> <I
Oi CQ
cd c ^
1—1 O 00
co (N U CU
Oi
1 II rS
vo
II c<
D
?3i
cn
CLH
s
ûi)
;-! • ^
UH
O

ts,
H
«>
X
II
+
oo VO CN

J9'^'BM jo 3 '^npA uoi'^'BjnS'eoQ

69
CO

»o
oi

- <N
O
Cj-I

^
00
13
L ^ CO co
S u

ion
> °»o
0)
co
?5 c5
CU >
B. c g"
, ^TZ
^
.2
a <D6
Ctí t i
X .— c/^
3 >>
Dû t^
<1
O <
U U
Tt
»o vd
o (L>
i-i
3
OJ
UH

- o

»o
o

jg^-eM jo ã '9npA uoi^ietnS'BOQ

70
for the PSf-NMP and PSf-DMSO systems. The results obtained for this system foUows

the general trend exhibited by other PSf systems, as shown in Figure 6.3.

The effect of AX, or the miscibility, of the PSf systems on the nitrogen fluxes of

these membranes can be observed firom Figure 6.5. The average nitrogen fluxes of the

uncoated and the silicone-coated PSf membranes, obtained from three samples for each

PSf system, increased systematically as the AX values increased from 2.412 to 5.203. The

nitrogen flux versus AX data for the uncoated and coated PSf membranes were fitted to

straight lines. The correlation coefficients obtained for the uncoated and coated PSf

membranes were 74.6% and 90.5% respectively. The slopes of both these lines are

positive, but the low correlation coefficient for the uncoated membranes indicates that the

nitrogen flux versus AX data may not be linear in nature. Therefore, straight lines fitted

to these data points have no other significance other than to show that the flux increases

with AX of the PSf systems.

The flux ratio of the coated to the uncoated membranes also increased from about

27% for the PSf-DMPU membrane to about 58% for the PSf-(NMP-DMSO) membrane

and 61% for the PSf-DMSO membrane, as can be observed in Table 6.1. The effect of

the silicone-coating therefore appears to be more pronounced on membranes obtained

from systems with lower AX values. This behavior of the PSf membranes can be qualita-

tively explained by observing the SEMs of the membrane skins obtained from the PSf-

NMP and PSf-DMSO systems (Figure 6.1). The uncoated PSf-NMP membrane has a

dense skin which has defects through which convective gas flow can occur. It appears that

these defects were effectively sealed by the silicone-coating since the flux of the coated

71
1\ r vo
\
\
\
\
\
\ lO
\
CO

\+ »o

\ ^ o.
- '^
\ ^ II
\ II cs
+ \ \V
\ + + co
^
00
\ \ c«

'c ^ u
1-1 o
Y O
> '^
cn

<N
CO • 1—1

oo
cs ><
+ \+ C "^
f-î ^ o c
oo CO
T—1
-^ <N p tí
^ t- tl >»
•^ C/5
o II . oo
«o Cu
11 c^ v6
Si û:í
w^
W)

c3 a
o o
OU
D

—• O
<N O
VO »o co

09S UID/DD 'gOI X ^ ^ n U9S0J'^tU 9S^J9AV

72
•o
<N

- + CN

o \ +
.

~ — «o

O + ^
/

co 00
co
3 U
289

'c
.2 ^ »o
co 1> 'co X ^
G
1 O
^ C/3

»o c c
T-H o S
oq oi O c/5
t—
co
1-H II
II
-
3i
— »o v6 ^
o v6
<o
í-i

o+ tu
JQ
Uncoated
Coated


1>

1 1 1 1 « 1 »o
o o
o »o «o o «o »o
co
co <N CN

D9S^mD/DD '^OT X ^ ^ 5 U9S0J'^IU 9S^J9AY

73
membrane is only 27% of the flux of the uncoated membrane. On the other hand, the skin

of the uncoated PSf-DMSO membrane is extremely thm, to the extent that the porosity

of the supporting substructure is visible through the sldn and appears to destroy the

integrity of the skin. The silicone-coating does not appear to be very effective in plugging

the defects in this case, as indicated by the high ratio of the fluxes for the coated to

uncoated DMSO membranes. These results are consistent with the observations of Henis

and Tripodi [78], who used silicone rubber to coat PSf hollow fiber membranes.

The nitrogen fluxes of the CA membranes prepared by using MA, acetone,

pyridine and THF solvents were extremely small under the experimental conditions used

in this study and could not be measured. This can be qualitatively explained by observing

the SEMs of these membranes. The membranes formed from the CA-MA, -acetone, -

pyridine and -THF systems were completely dense, barrier-like films. The SEMs of the

CA-MA and -THF membrane cross-sections are shown in Figure 6.2 . The permeabilities

of these membranes are therefore probably close to the intrinsic permeability of the

polymer. The nitrogen fluxes of the CA-MA and CA-THF membranes were estimated

using the intrinsic permeability of nitrogen in CA [79]. The effective thickness of the

membrane in each case was obtained from the SEMs. The detailed calculations are shown

in Appendix E.

The CA-dioxane and CA-DMSO systems on the other hand yielded skinned

membranes with porous substructures. The nitrogen fluxes of the uncoated and sUicone-

coated CA-dioxane and CA-DMSO membranes are shown in Table 6.2. The flux ratio

74
increased from about 24% for the CA-dioxane system to about 51% for the CA-DMSO

system. The silicone coating therefore appears to be more effective in the case of the CA-

dioxane membrane than the CA-DMSO membrane, which has a higher AX value than the

former.

The calculated and experimentally determined nitrogen fluxes of various CA

membranes were plotted against the corresponding AX values, as shown in Figure 6.6.

The nitrogen flux of the CA membranes increased as AX increased from -0.02 to 1.85.

However, even though the CA-MA system had a lower AX value than the CA-THF

system, it had a marginally higher calculated nitrogen flux than the latter. The SEMs of

these membranes show that the CA-THF membrane is thicker than the CA-ma membrane.

This can probably be explained on the basis of the boiling points of these two solvents.

Methyl acetate and tetrahydrofuran are both extremely volatile solvents, with boiling

points of 57 °C and 67 °C respectively. The lower boiling point of MA causes more of

it to evaporate from the CA-MA casting solution during the 1 minute evaporation step

(prior to the quenching step), than the corresponding solvent evaporation from the CA-

THF solution. This results in a CA-MA membrane which is thinner than the CA-THF

membrane, and correspondingly greater calculated value of nitrogen flux than the latter.

The calculated and experimentally determined nitrogen flux versus AX data for the

uncoated CA membranes were fitted to a straight line. The low correlation coefficient (R^

= 79.7%) again indicates that a linear regression model may not be the correct model for

this set of data. But the positive slope of the line shows that the nitrogen flux increases

with increasing values of AX for the CA systems examined in this study.

75
VO

o+
00 C/5
B B
on 00
»o
>^ >%
00 00
< ^ ^r* o
J=í
cuU 4_>

Vi
O U
<4-l
ccJ o
»o
irs r^
T3 4 - >
<>
o ^ >i
C/5 -n
p
3 4—>
O 00
>-i
C>0

co co
1) C/3
> *P
c« -4—>
O
'c 3 c
.2 ctí
> -o
'cn
C c>
(U
C c
o
4->
<>» — cd cn

gul
X
+
<l CO
o
00

e
o
4—>
-a C/3
+ c >^
00
• ^

omb
+ u
U T3
C
t^ d
Vt-I
VO
+ o O^
oo
+ t-l

- + 3

tJU

o o o o o o o o o o
o o\ oo vo »o co <N

J9'^'BAV JO S ' ^ n p A UO '^'B{nST30Q

76
1 1 1 ll I • l l • 1 l l "T—
1' ' •1— ll 1 • ll • vo

O+ • X
n

PSf uncoated
CA uncoated
CA coated
PSfcoated
— lO
<
U
-o
c
cc3
«*-!
00
fX.
o + ^
o
Æ

$++ v-< oU
*^
o
C4-1 »o
CCÍ < ^

o + ca -^
•ti cd

^ co
-o
»s> >»
m
(D
9-§ 4->
'c C/5 00

sio
o + c
s.-

trogeri iux ver


exami: d in th
o

D X •^ Q-i <D
_. C
X
<3
• X
• ^H
C 00
• ^
-o
0 e
<D
syst
mbin

0
U
— • — o CXJ
vd
<D
>-i
3
CiO

u.
• 1 . 1. . . 1. . . 1 . 1 1 ii . • 1. . . 1. . 1

o <N CO »0 vo
o 0 0 0 0 0 0 0
+ <D <D <D 0 <D <D <D
0 0 0 0 0 0 0
O

D9S ^mo/Do 'xng u^Soj'ji^

77
The combined coagulation value versus AX data and niu-ogen flux versus AX data

for all the systems examined in this study are shown in Figures 6.7 and 6.8 . From Figure

6.7, it can be observed that there is a sharp change in the curvature of the coagulation

value versus AX data, approximately between AX = 1.5 and AX = 3.5 . The nitrogen flux

versus AX curve also showed a corresponding change in curvature for the same range of

AX values, as shown in Figure 6.8. Systems which had AX values lower than about 1.5

yielded completely dense membranes, while systems which had AX values greater than

3.5 yielded porous membranes. CA and PSf systems with AX values in this transition

range yielded asymmetric membranes with dense skins, with the nitrogen fluxes

progressively increasing with AX. The results obtained for the binary solvent-PSf systems

examined in this study also seem to apply to the PSf-CNMP-DMSO) system.

These results prove the validity of AX as a measure of the miscibility of

nonsolvent-solvent-polymer membrane-forming systems. More importantly, AX could

successfully correlate the miscibility of these systems to the morphology of the

membranes. Systems with low AX values yielded completely dense membranes, while

those with high AX values jdelded porous membranes. Systems with intermediate values

of AX yielded asymmetric membrane structures.

78
CHAPTER7

CONCLUSIONS

The results of this research can be summarized as follows. A simple thermody-

namic parameter AX was introduced to quantify the miscibility of PSf and CA

nonsolvent-solvent-polymer membrane-forming systems. PSf and CA systems with low

AX values had high coagulation values and those with high AX values had low

coagulation values. These results confirmed the validity of AX as a measure of the

miscibility of the membrane-forming systems.

More importantly, AX could successfully correlate the miscibility of the

nonsolvent-solvent-polymer systems to the morphology to the transport properties of the

corresponding PSf and CA phase inversion membranes. PSf systems with lower AX

values resulted in membranes with denser skins and lower fluxes, while those with higher

AX values produced membranes with more porous skins and higher fluxes. The results

obtained for the PSf-(NMP-DMSO) system appear to indicate that this approach can also

possibly be extended to systems with solvent mixtures. CA systems with lower AX values

resulted in completely dense membranes, the nitrogen fluxes of which could not be

measured under the experimental conditions that were used in this study. CA systems with

higher AX yielded asymmetric, skinned membranes with higher nitrogen fluxes. AX is

therefore a useful parameter that can be used to differentiate formation conditions for

completely dense morphologies from those of either asymmetric, skinned, or completely

porous structures.

79
CHAPTER8

RECOMMENDATIONS

The simplicity and generality of the model that was developed in this study

indicates that further investigation of the potential for its predictive abilities is warranted.

This study utilized PSf and CA, two polymers which are very different in terms

of their physical and chemical properties. However, the scope of this model can be

significantly enhanced by examining membrane-forming systems using a variety of other

polymers such as polycarbonates, polyimides, etc.

Previous studies [63] have examined phase separation in polyethersulfone and

polysulfone membrane-forming systems using nonsolvents such as methanol, ethanol, 1-

propanol, acetone and carbon tetrachloride. The effect of different nonsolvents on

membrane transport properties could therefore be another area of future investigation.

The model also permits the examination of the effects of numerous other variables

such as, the coagulation bath temperature, initial concentration of solvent in the

coagulation bath, etc., on membrane morphology. Experimental observations [14,15,37,38,

39] have shown that membrane skin characteristics are strongly affected by changes in

the quench bath conditions, such as bath composition and temperature. It would be

interesting, therefore, to examine the trends in the coagulation value versus AX data and

gas transport rates versus AX data with varying coagulation bath compositions and

temperatures, to understand the effect of these variables in determining membrane

morphology.

81
Often, in commercial membrane manufacture processes, a second solvent

(cosolvent) is used in the polymer casting solution. The cosolvent usually acts as a

poreformer during the membrane formation process. The PSf-(NMP-DMSO) mixture that

was examined in this study is an example of a casting solution where DMSO acts as the

cosolvent poreformer. Other poreformers such as ethylene glycol, glycerol, etc, are

commonly used in the commercial preparation of polysulfone membranes. The applicab-

ility of the model can therefore be extended further by performing experiments with

membrane-forming systems which use solvent mixtures in the casting solutions.

All the transport experiments in this study were performed using nitrogen gas with

the objective of obtaining a measure of the openness of the membrane structures obtained

from various polymer-solvent systems. Transport experiments can also be performed with

gas mixtures. For example, gas transport experiments can be performed using a simple

gas mixture such as air and obtaining not only the flow rate of the permeate but also

analyzing its composition. This would give a better qualitative and quantitative measure

of the thickness and integrity of the membrane skin.

82
REFERENCES

[I] S. Loeb and S. Sourirajan, US Patent 3,133,132 (1964).

[2] S. Loeb and S. Sourirajan, Seawater demineralization by means of an osmotic


membrane, Adv. Chem. Ser., 38 (1963) 117.

[3] M.C. Porter, Handbook of Industrial Membrane Technology, Noyes Publications,


New Jersey, 1990.

[4] S.Sourirajan, Reverse Osmosis and Synthetic Membranes, National Research


Council of Canada, Ottawa, 1977.

[5] Munir Cheryan, Ultrafiltration Handbook, Technomic Publishing Company,


Lancaster, Pennsylvania, 1986, Chap.8 ,pp. 231-340.

[6] John R. Dorgan, Polymer membranes for bioseparations, in:Davis S.Soane(Ed.),


Polymer Applications for Biotechnology, Prentice Hall, Englewood Cliffs, New
Jersey, 1992, Chap. 3,pp. 64-97.

[7] A.E. Fonda, J.D. Hazlett,T. Matsuura and J. Johnson (Eds.), Membrane Separations
in Chemical Engineering, AIChE Symp. Ser., No.272, Vol.85, Published by
AIChE, New York, New York, 1989.

[8] R.W. Spillman, Economics of gas separation membranes, Chem. Eng. Prog., 85
(1989) 41.

[9] R.A. Ward, P.W. Feldhoff and E. Klein, Membrane materials for therapeutic
applications in medicine, in: Materials Science of Synthetic Membranes, Douglas
R. Lloyd (Ed.), ACS Symposium Series No.269, American Chemical Society,
Washington, D.C., 1985, Chap.5, pp. 99-118.

[10] R.E. Kesting, Synthetic Polymeric Membranes, McGraw-Hill, New York, 1971,
Chap.3, pp. 90-98.

[II] R.E. Kesting, Synthetic Polymeric Membranes, McGraw-Hill, New York, 1971,
Chap.7, pp. 251-261.

[12] S. Manjikian, S. Loeb and J.W. McCutchan, Proceedings of the First Intemational
Desalination Symposium, Washington, D.C., Vol 2 (1965), 159.

83
[13] S. Manjikian, Desalination membranes from organic casting solutions, Ind. Eng.
Chem. Prod. Res. Dev., 6 (1967) 23.

[14] R.E. Kesting, Semipermeable membranes of cellulose acetate for desalination in


the process of reverse osmosis. I. Lyotropic swelling of secondary cellulose
acetate, J. Appl. Polym. Sci., 9 (1965) 663.

[15] R.E. Kesting, M. Barsh and A. Vincent, Semipermeable membranes of cellulose


acetate for desalination in the process of reverse osmosis. H. Parameters
affecting membrane gel strucmre, J. Appl. Polym. Sci., 9 (1965) 1873.

[16] C. Lemoyne, M. Guillotin, C. Noel and L. Monnerie, Physicochemical processes


occurring during the formation of cellulose diacetate membranes. Research of
criteria for optimizing membrane performance. IV. CeUulose diacetate-acetone-
organic additive casting solutions, Desalination, 21 (1977) 165.

[17] C. Lemoyne, C. Friedrich, J.L. Halary, C. Noel and L. Monnerie, Physicochemical


processes occurring during the formation of cellulose diacetate membranes.
Research of criteria for optimizing membrane performance. V. Cellulose
diacetate-acetone-water-inorganic salt casting solutions, J. Appl. Polym. Sci., 25
(1980) 1883.

[18] H. Strathmann, P. Scheible and R.W. Baker, A rationale for the preparation of
Loeb-Sourirajan type cellulose acetate membranes, J. Appl. Polym. Sci., 15 (1971)
811.

[19] H. Strathmann, K. Kock, P. Amar and R.W. Baker, The formation mechanism of
asymmetric membranes, Desalination, 16 (1975) 179.

[20] T. Uragami, Y. Ohsumi and M. Sugihara, Studies on syntheses and permeabilities


of special polymer membranes : 40. Formation conditions of finger-like cavities
of cellulose nitrate membranes, Desalination, 37 (1981) 293.

[21] A. Bottino, G. Capnnelli and S. Munari, Effect of coagulation medium on


properties of sulfonated polyvinylidene membranes, J. Appl. Polym. Sci., 30
(1985) 3009.

[22] H. Strathmann, M.T. So, F.R. Eirich and R.W. Baker, Preparation of asymmetric
Loeb-Sourirajan membranes, J. Polym. Sci., Polym. Lett., 11 (1973) 201.

[23] A. Chawla and T. Chang, Use of solubility parameters for the preparation of
hemodialysis membranes, J. Appl. Polym. Sci., 19 (1975) 1723.

84
[24] I. Cabasso, E. Klien and J.K. Smith, Polysulfone hoUow fibers. I. Spinning and
properties, J. Appl. Polym. Sci., 20 (1976) 2377.

[25] T. Uragami, Y. Ohsumi and M. Sugihara, Studies on syntheses and permeabUities


of special polymer membranes : 38. Formation mechanism of finger-Uke cavities
in membranes from ceUulose nitrate and single solvent, Polymer, 23 (1982) 999.

[26] R.E. Kesting and A. Menefee, Role of formamide in preparation of ceUulose


acetate membranes, KoUoid Z. Z. Polym., 230(2) (1969) 341.

[27] I. Cabasso, E. Klien and J.K. Smith, Polysulfone hoUow fibers. n. Morphology,
J. Appl. Polym. Sci., 21 (1977) 165.

[28] S. Doi and K. Hamanaka, Pore size control techniques in the spinning of
polysulfone hoUow fiber ultrafiltration membranes, Desalination, 80 (1991) 167.

[29] C.T. Badenhop and A.L. Bourgignon, The characterization of polymer membranes
as produced by phase inversion, in: E. DrioU and M. Nagagaki (Eds.), Membrane
and Membrane Processes, Plenum Press, 1986, pp. 429-445.

[30] T. Miyano, T. Matsuura and S. Sourirajan, Effect of polymer molecular weight,


solvent and casting solution composition on the pore size and pore size distribution
of polyethersulfone (Victrex) membranes, Chem. Eng. Comm., 95 (1990) 11.

[31] J.E. Anderson and R. Ulmann, Mathematical analysis of factors influencing the
skin thickness of asymmetric reverse osmosis membranes, J. Appl. Phys., 44
(1973) 4303.

[32] C. CasteUari and S. Ottani, Preparation of reverse osmosis membranes : A


numerical analysis of asymmetric membrane formation by solvent evaporation
from ceUulose acetate casting solutions, J. Memb. Sci., 9 (1981) 20.

[33] M. Ataka and K. Sasald, Gravimetric analysis of membrane casting. I. CeUulose


acetate-acetone binary casting solutions, J. Memb. Sci., 11 (1982) 11.

[34] R.J. Ray, Interfacial instabiUties arising from excess intermolecular potential
gradients: AppUcation to asymmetric membrane formation, PhD thesis, University
of Colorado, 1983.

[35] M.N. Sarbolould, Preparation of sldnned membranes without the evaporation step,
J. Polym. Sci., Polym. Lett., 11 (1973) 753.

85
[36] K.K. Sirkar, N.K. Agarwal and G.P. Rangaiah, The effect of short air exposure
periods on the performance of ceUulose acetate membranes from casting solutions
with high ceUulose acetate content, J. Appl. Polym. Sci., 22 (1978) 1919.

[37] T.A. Tweddle and S. Sourirajan, Effect of ethanol-water mixture as gelation


medium during formation of ceUulose acetate reverse osmosis membranes, J. Appl.
Polym. Sci., 22 (1978) 2265.

[38] M.A. Frommer, R. Matz and U. Rosenthal, Mechanism of formation of reverse


osmosis membranes, Ind. Eng. Chem., Prod. Res. Dev., 10 (1971) 193.

[39] J.G. Wijmans, J.P.B. Baaij and C.A. Smolders, The mechanism of formation of
microporous or skinned membranes produced by immersion precipitation, J.
Memb. Sci., 14 (1983) 263.

[40] C.Cohen, G.B. Tanny and S. Pragen, Diffusion-controUed formation of porous


structures in temary polymer systems, J. Polym. Sci., Polym. Phy. Ed., 17 (1979)
477.

[41] I. Cabasso, Ultrastmcture of asymmetric and composite membranes, in: A.F.


Turbak (Ed.), Synthetic membranes : DesaUnation, ACS Symp. Ser., Vol. 153,
American Chemical Society, Washington, D.C., 1981, p. 267.

[42] H. Strathmann and K. Kock, The formation mechanism of phase inversion


membranes, DesaUnation, 21 (1977) 241.

[43] L. Yilmaz and A.J. McHugh, ModeUing of asymmetric membrane formation. I.


Critique of evaporation models and development of a diffusion equation formaUsm
for the quench period, J. Memb. Sci., 28 (1986) 287.

[44] L. Yilmaz and A.J. McHugh, Modelling of asymmetric membrane formation. II.
The effects of surface boundary conditions, J. Appl. Polym. Sci., 35 (1988) 1967.

[45] C.S. Tsay and A.J. McHugh, Mass transfer modeUing of asymmetric membrane
formation by phase inversion, J. Polym. Sci., Polym. Phy., 28 (1990) 1327.

[46] C.S. Tsay and A.J. McHugh, An improved numerical algorithm for temary
difftision with a moving interface, Chem. Engg. Sci., 46(4) (1991) 1179.

[47] C.S. Tsay and A.J. McHugh, The combined effects of evaporation and quench
steps on asymmetric membrane formation by phase inversion, J. Polym. Sci.,
Polym. Phy., 29 (1991) 1.

86
[48] T. Young and L. Chen, A two-step mechanism of diffusion-controUed ethylene
vinyl alcohol membrane formation, J. Memb. Sci., 57 (1991) 69.

[49] D.M. Koenhen, M.H.V. Mulder and C.A. Smolders, Phase separation phenomena
during formation of asymmetric membranes, J. Appl. Polym. Sci., 21 (1977) 199.

[50] L. Broens, F.W. Altena, C.A. Smolders and D.M. Koenhen, Asymmetric
membrane stmctures as a result of phase separation phenomena, Desalination, 32
(1980) 33.

[51] F.W. Altena and C.A. Smolders, Phase separation phenomena in solutions of
ceUuIose acetate in mixtures of dioxane and water, J. Polym. Sci.: Polym. Symp.,
69 (1981) 1.

[52] J.G. Wijmans, J. Kant, M.H.V. Mulder and C.A. Smolders, Phase separation
phenomena in solution of polysulfone in mixtures of a solvent and a nonsolvent:
relationship with membrane formation, Polymer, 26 (1985) 1539.

[53] G.E. Gaides and A.J. McHugh, Gelation in an amorphous polymer : a discussion
of its relation to membrane formation, Polymer, 30 (1989) 2118.

[54] W.R. Burghardt, L. Yilmaz and A.J. McHugh, Glass transition, crystaUization and
thermoreversible gelation in temary polyphenyleneoxide solutions : relationship
to asymmetric membrane formation, Polymer, 28 (1987) 2085.

[55] D.R. Paul, Reversible gelation of acrylonitrUe-vinyl acetate copolymer solutions,


L Appl. Polym. ScL, 11 (1967) 439.

[56] A.J. Reuvers, F.W. Altena and C.A. Smolders, Demixing and gelation behavior
of temary ceUulose acetate solutions, J. Polym. Sci., Part B: Polym. Phys., 24
(1986) 793.

[57] M.H.V. Mulder, J.O. Hendrikman, J.G. Wijmans and C.A. Smolders, A rationale
for the preparation of asymmetric pervaporation membranes, J. Appl. Polym. Sci.,
21 (1977) 199.

[58] F.W. Altena and C.A. Smolders, Calculation of Uquid-Uquid phase separation in
a temary system of a polymer in a mixture of a solvent and a nonsolvent,
Macromolecules, 15 (1982) 1491.

[59] L. Yilmaz and A.J. McHugh, Analysis of nonsolvent-solvent-polymer phase


diagrams and their relevance to membrane formation modeUing, J. Appl. Polym.
ScL, 31 (1986) 997.

87
[60] R.A. OrwoU, The polymer-solvent interaction parameter X, Rubber Chem.
TechnoL, 50 (1977) 451.

[61] L. Shuguang, J.Chengzhang and Z.Yuanqi, The investigation of solution


thermodynamics for the polysulfone-DMAc-water system, DesaUnation, 62 (1987)
79.

[62] L. Zeman and G. Tkacik, Thermodynamic analysis of a membrane-forming system


water/N-methyI-2-pyrroUdinone/poIyethersulfone, J. Memb. Sci, 36 (1988) 119.

[63] B.T. Swinyard and J.A. Barrie, Phase separation in nonsolvent/dimethylformamide-


/polyethersulfone and nonsolvent/dimethylformamide/polysulfone systems, Brit.
Polym. J., 20 (1988) 317.

[64] W.W.Y. Lau, M.D. Guiver and T. Matsuura, Phase separation in polysulfone-
solvent-water and polyethersulfone-solvent-water systems, J. Memb. Sci., 59
(1991) 219.

[65] P.J. Flory, Thermodynamics of polymer solutions , Discuss. Farad. Soc, 49


(1970) 7.

[66] A.F. Barton, Handbook of Polymer-Liquid Interaction Parameters and SolubUity


Parameters, CRC Press, Boston, 1990, pp. 3-22.

[67] M. Kai, K. Ishu, H. Tsugaya and T. Miyano, Development of polyethersuUone


ultrafUtration membranes, in: T.Matsuura and S.Sourirajan (Eds.), Reverse Osmosis
and Ultrafiltration, ACS Symposium Series No. 281, American Chemical Society,
Washington, D.C., 1985, Chap.2, pp. 21-33.

[68] A.G. Fane, C.J.D. FeU, C.W. Yao and R.P. Burford, Effect of coagulation
conditions on stmcture and properties of membranes from aliphatic polyamides,
J. Memb. ScL, 38 (1988) 113.

[69] R.E. Kesting, Synthetic Polymeric Membranes, McGraw-HiU, New York, 1971,
Chap.4, pp. 119-120.

[70] A. Fredenslund, J. Gmehling, M.L. Michelsen, P. Rasmussen and J.M. Prausnitz,


Computerized design of multicomponent distiUation columns using the UNIFAC
group contribution method for calculation of activity coefficients, Ind. Eng.
Chem. Proc Des. Dev., 16 (1977) 450.

[71] D.S. Abrams and J.M. Prausnitz, Statistical thermodynamics of Uquid mixtures:
A new expression for the excess Gibbs energy of partly or completely miscible
systems, AIChE J., 21 (1975) 116.

88
[72] P. Munk, P. Hattam, D. Qiangguo and A.A.A. Abdel Azim, Determination of
polymer-solvent interaction coefficients by inverse gas chromatography, J. AppL
Polym. ScL, Appl. Polym. Symp. Ed., 45 (1990) 289.

[73] T. Oishi and J.M. Prausnitz, Estimation of solvent activities in polymer solutions
using a group contribution method, Ind. Eng. Chem. Proc Des. Dev., 17(3) (1978)
333.

[74] M. GottUeb and M. Herskowitz, Estimation of the X parameter for poly (dimethyl-
sUoxane) solutions by the UNIFAC group contribution method, Macromolecules,
14 (1981) 1468.

[75] A.F. Barton, Handbook of Polymer-Liquid Interaction Parameters and SolubUity


Parameters, CRC Press, Boston, 1990, pp. 88-95.

[76] A. Fredenslund, R.M. Jones and J.M. Prausnitz, Group-contribution estimation of


activity coefficients in nonideal Uquid mixtures, AIChE J, 21 (1975) 1086.
Supplement to program UNIFAC, Fifth revision: J.P. Baker, Jan. 1992.

[77] H.K. Hansen, P. Rasmussen, A. Fredenslund, M. SchiUer and J. Gmehling, Vapor-


Uquid equUibria by UNIFAC group contribution. 5. Revision and extension, Ind.
Eng. Chem. Res., 30 (1991) 2352.

[78] J.M.S. Henis and M.K. Tripodi, Composite hoUow fiber membranes for gas
separation: The resistance model approach, J. Memb. ScL, 8 (1981) 233.

[79] W.J. Koros, G.K. Fleming, S.M. Jordan, T.H. Kim and H.H. Hoehn, Polymeric
membrane materials for solution-diffusion based permeation separations, Prog.
Polym. ScL, 13 (1988) 339.

[80] W.J. Koros and M.W. HeUums, Transport Properties, in: Encyclopedia of Polymer
Science and Engineering, Supplement Volume, Second Edition, John Wiley and
Sons Inc, 1989, pp. 724-802.

89
APPENDDC A

LISTING OF PROGRAM FOR UNIFAC CALCULATIONS

This appendix Usts the FORTRAN code for the UNIFAC program used in the

calculation of the binary nonsolvent-solvent interaction parameter Xj^.

90
C THIS PROGRAM CALCULATES THE INTERACTION PARAMETER FOR A
C BINARY LIQUID SYSTEM. ALL CALCULATIONS AND NON-INTEGER TERMS
C ARE DOUBLE-PRECISIONED.
Q **********************************:|e*j(:**:(c****************************

DOUBLE PRECISION R(89),Q(89),MW(89),A(48,48),TC,TK,DENS(2),W(2),DW


DOUBLE PRECISION SPVOL(2),MOLARVOL(2),M(2),Rl(2),Ql(2),MWl(2)
DOUBLE PRECISION TMOL,TVOL,MOLFR(2),VOLFR(2),R2,Q2,SIGMAXL
DOUBLE PRECISION PH(2),TH(2),L(2),LNGAMC(2), AMN(48,48),SHI(48,48)
DOUBLE PRECISION SUMXQ,XQ(2,10),THETA(2,10),SUM1,SUM2,SUM3
DOUBLE PRECISION LNTAUK(2,10),LNTAUK1(2,10),SUMXQ1,XQ1(2,10)
DOUBLE PRECISION THETA1(2,10),LNGAMR(2),LNGAM(2),LNA(2),CHI
INTEGER AN(2,10)
DIMENSIONN1(89),N2(89),L\R(2,10),MGNO(2,10),ILS(2)
CHARACTERN1$(89)*10,N2$(89)*10,C$(2)*6,CM$(2,10)*10,CS$(2,10)*10
C
C THIS PART OF THE PROGRAM READS R AND Q DATA FROM DATAFILE
C UNFRQ48.DTA
C
OPEN(UNIT=4,FILE='A:UNFRQ48.DTA',STATUS='OLD')
PRINT*,'READING A:UNFRQ48.DTA'
DO 10 1=1,89
READ(4,*)Nl$(I),Nia),N2$(I),N2(I),R(I),Qa),MWa)
10 CONTINUE
CLOSE (UNIT=4)
C
C THIS PART OF THE PROGRAM READS INTERACTION PARAMETER DATA
C FROM DATAFILE UNFA48.DTA . THIS DATA WAS OBTAINED FROM THE
C FIFTH REVISED VERSION OF PRAUSNTTZ'S UNIFAC PARAMETERS
C
OPENaJNIT=5,FILE='A:UNFA48.DTA',STATUS='OLD')
PRINT*,'READING A:UNFA48.DTA'
DO 20 1=1,48
DO 30 J=l,45,4
30 READ(5,*)A(IJ),Aa,J+l),A(I,J+2),A(I,J+3)
20 CONTINUE
CLOSEaJ]SÍIT=5)
C
C ALL SUBSEQUENT CALCULATIONS ARE FOR A BINARY NONSOLVENT-
C SOLVENT SYSTEM.
C
IMX=2
C
C

91
C INmALIZING THE VALUES OF THE SECONDARY GROUP NOS. AND
C NUMBER OF GROUPS OF EACH TYPE FOR BOTH COMPONENTS
35 DO 40 I=1,IMX
DO 50 J=l,10
IARa,J)=0
50 ANaj)=0
40 CONTINUE
C
C THIS PART OF THE PROGRAM IS FOR ENTERING COMPONENT DATA
C
PRINT*,'ENTER COMPONENT DATA IN THE REQUIRED FORMAT'
PRINT*,'ENTER CHARACTER DATA IN APOSTROPHES'
PRINT*,'ENTER FIRST COMPONENT AS NONSOLVENT'
DO 60 1=1,IMX
PRINT*,'ENTER NAME OJPTO 6 CHAR) OF COMPONENT ',1
READ*,C$a)
PRINT*,'INPUT NO.OF SECONDARY GROUP TYPES IN ',C$(D
READ*,ILSa)
PRINT*,'INPUT MAINGR. NAME, SECGR. NAME (UPTO 10 CHAR)'
PRINT*,'MAIN GR. NO.,SEC.GR. NO. AND NO.OF EACH TYPE'
DO 70 J=1,ILS(I)
70 READ*,CM$a,J),CS$a,J),MGNO(IJ),L\Ra,J),ANa,J)
60 CONTINUE
C
C ENTERING COMPONENT INFORMATION INTO OUTPUT FILE
0PEN(UNIT=7,FILE='A:CHINSS.DTA',STATUS='NEW')
WRITE(7,4900)C$(2),C$(1)
4900 FORMAT(20X,'CALCULATION OF INTERACTION PARAMETER FOR'y
1 28X,A,'-',A,' SYSTEM')
WRrrE(7,*)' ',
1 ' '
WRITE(7,*)
WRITE(7,*)
WRITE(7,*)' COMPONENT LIST'
WRITE(7,*)' '
WRrrE(7,*)
DO 80 I=1,IMX
WRITE(7,5000)I,C$a)
5000 FORMAT(30X,I2,5X,A)
80 CONTESrUE
WRITE(7,*)
DO 90 I=1,IMX
WRITE(7,5001)C$a)
5001 FORMAT(23X,'TYPE AND NO.OF GROUPS IN ',A)

92
DO 100 J=1,ILS(I)
100 WRITE(7,5005)CS$a,J),ANa,J)
5005 FORMAT(28X,A,5X,I2)
WRITE(7,*)
90 CONTINUE
C
C ENTERING TEMPERATURE AND DENSITIES OF COMPONENTS
C
PRINT*,'INPUT TEMPERATURE ,DEG C'
READ*,TC
TK=TC+273.15D-FO0
DO 110I=1,IMX
PRINT*,'ENTER DENSITY (gm/cc) OF COMPONENT ',1
READ*,DENS(I)
110 CONTINUE
C
C ENTERING TEMP AND WEIGHT FRACTION DATA ESfTO OUTPUT FILE
C
WRITE(7,5010)TC
5010 F 0 R M A T ( 2 3 X , ' T E M P E R A T U R E = ',D10.4,' C')
WRITE(7,*)
WRrrE(7,*)' DENSITY OF COMPONENTS'
WRITE(7,*)' '
WRITE(7,*)
DO 115I=1,IMX
WRITE(7,5020)I,DENS(I)
5020 F0RMAT(23X,'DENSITY(',I1,') = ',D8.3,' gm/cc')
115 CONTINUE
WRITE(7,*)
C
^:|c4c3ic*******:ic:|c************************************************************

C THIS PART OF THE PROGRAM PERFORMS THE UNIFAC CALCULATIONS


C FOR DETERMINING COMPONENT ACnVITIES.

C SOME PRELIMINARY CALCULATIONS TO OBTAIN Rl AND Ql VALUES


C DSfmALIZING ALL VALUES
DO 120 I=1,IMX
Ria)=0.0D+00
Q1(I)=0.0D-K)0
MWia)=0.0D+00
SPVOLa)=0.0D+00
MOLARVOLa)=0.0D+00
Ma)=0.0D+00
DO 122 J=1,ILS(I)

93
R1 a)=R 1 (i)+RaARa,j))* ANa, j)
Q1 (i)=Q 1 a)+QaARa, J)) * A N ^ , J)
122 MWl(I)=MWia)+MW(IARa,J))*AN(I,J)
SPVOLa)=1.0D+00/DENS(I)
MOLARVOLa)=MWl a)*SPVOLa)
Ma)=MOL ARVOL(I)/MOLAR VOL( 1)
PRINT*,'M(',I,') =',M(I)
120 CONTINUE
C
C CALCULATING SHI VALUES
C
IFLAG=0
DO 150 I=1,IMX
DO 160 J=1,ILS(I)
MM=MGNO(I,J)
DO 170 II=1,IMX
DO 180 JJ=1,ILS(II)
NN=MGNO(n,JJ)
AMN(MM,NN)=A(MM,NN)
SHI(MM,NN)=DEXP(-AMN(MM,NN)/TK)
IF (AMN(MM,NN).EQ.O.O) THEN
IF (MM.NE.NN) THEN
WRITE(7,5021)CS$a,J),CS$(n,JJ)
5021 FORMAT(lX,'INTERACTION VALUES NOT AVAILABLE FOR ',A,'-',A,'
1 PAIR')
IFLAG=1
ENDIF
ENDIF
180 CONTINUE
170 CONTINUE
160 CONTINUE
150 CONTDSfUE
C
C
IF aFLAG.EQ.O) THEN
WRITE(7,*)'ALL INTERACTION PARAMETERS ARE AVAILABLE'
ENDIF
WRITE(7,*)
WRITE(7,*)
WRrrE(7,*)' WEIGHT VOLUME MOLE',
CHI'
WRrrE(7,*)' FR. OF FR. OF FR. OF'
WRrrE(7,*)' H20 H20 H20'
WRITEÍ7.*)'

94
C ENTER STEP SIZE FOR WEIGHT FRACTION, INITL^LIZING
C THE WEIGHT FRACTIONS
DW=0.01
W(l)=0.01
124 W(2)=1.0-W(l)
C
C CALCULATING MOLE AND VOLUME FRACTIONS
C
TM0L=(W(1)/MW1(1))+(W(2)/MW1(2))
TVOL=MOLFR( 1 )*M( 1 )+MOLFR(2)*M(2)
DO 125 I=1,IMX
MOLFRa)=(W(I)/MWl a))/rMOL
VOLFR(I)=(MOLFR(I)*Ma))/TVOL
125 CONTINUE
R2=0.0
Q2=0.0
DO 130 I=1,IMX
R2=R2+MOLFRa)*Rl(I)
Q2=Q2+MOLFR(I) *Q 1 (I)
130 CONTINUE
SIGMAXL=0.0
DO 140 I=1,IMX
PH(I)=Ria)*MOLFRa)/R2
TH(I)=Ql(I)*MOLFR(I)/Q2
L(I)=5.0*(R1(I)-Q1(I))-(R1(I)-1.0)
SIGMAXL=SIGMAXL+La)*MOLFR(I)
140 CONT NUE
C
C CALCULATION OF COMBINATORL^L PART GnGAMC) OF ACTIVITY
C COEFFICEENT
DO 145 I=1,IMX
145 LNGAMC(I)=DLOG(PHa)/MOLFRa))+ 5.0*Qia)*
1 DLOG(TH(I)/PHa))+ La)- (PHa)/MOLFRa))*SIGMAXL
C
C CALCULATION OF RESIDUAL PART (hiGAMR) OF ACTIVITY COEFFICIENT,
C CALCULATION OF THETA VALUES FOR EACH GROUP IN EACH COMPO-
C NENT
C
SUMXQ=0.0D+00
DO 200 1=1,IMX
DO 210 J=l,ILSa)
XQaj)=MOLFRa)*ANa,J)*Q(L\RaJ))
210 SUMXQ=SUMXQ+XQ(I,J)
200 CONTINUE

95
DO 220 1=1,IMX
DO 230 J=1,ILS(I)
THETAa,J)=XQ(I,J)/SUMXQ
230 CONTINUE
220 CONTINUE
C
C CALCULATION OF Ui(Tauk) FOR ALL GROUPS IN BOTH COMPONENTS
C
DO 235 1=1,IMX
DO 240 J=l JLS(I)
SUMl=0.0D+00
DO 250 n=l,IMX
DO 260 JJ=l,ILSai)
260 SUMl=SUMl+THETA(njJ)*SHia^GNOaiJJ),MGNOaj))
250 CONTINUE
SUM3=0.0D+00
DO 270 n=l,IMX
DO 280 JJ=l,ILSai)
SUM2=0.0D+00
DO 290 in=l,IMX
DO 300 JJJ=1,ILS(III)
300 SUM2=SUM2+THETAan,JJJ)*SHia^GNOan,JJJ),MGNO(n,JJ))
290 CONTINUE
280 SUM3=SUM3+THETAaiJJ)*SHI(MGNOaJ),MGNOai,JJ))/SUM2
270 CONTINUE
LNTAUKa,J)=Q(L\Ra,J))*(l-DLOG(SUMl)-SUM3)
240 CONTINUE
235 CONTINUE
C
C CALCULATION OF UiTaukl VALUES WITHIN EACH MOLECULE
C
DO 350 1=1,IMX
IF(ILS(I).EQ.1)THEN
LNTAUKl a, l)=0.0D+00
GO TO 345
ENDIF
C CALCULATION OF THETAl VALUES FOR EACH GROUP WTTHIN EACH
C MOLECULE
SUMXQ1=0.0
DO 301 J=1,ILS(I)
xQia,j)=ANa,j)*QaARa,J))
SUMXQl=SUMXQl+XQia,J)
301 CONTESfUE
DO 302 J=1,ILS(I)

96
THETAl(I,J)=XQia,J)/SUMXQl
302 CONTINUE
DO 310 J=1,ILS(I)
SUMl=O.OD+00
DO 320 JJ=l,ILSa)
320 SUMl=SUMl+THETAl(I,JJ)*SHI(MGNOa,JJ),MGNO(I,J))
SUM3=0.0D+00
DO 330 JJ=l,ILSa)
SUM2=0.0D+00
DO 340 JJJ=1,ILS(I)
340 SUM2=SUM2+THETAl(I,JJJ)*SHia4GNO(I,JJJ),MGNO(I,JJ))
330 SUM3=SUM3+THETAiajJ)*SHI(MGNOa,J),MGNO(I,JJ))/SUM2
LNTAUKia,J)=Q(L\Ra,J))*(l-DL0G(SUMl)-SUM3)
310 CONTINUE
345 CONTDSrUE
350 CONTINUE
C
DO 360 I=1,IMX
LNGAMR(I)=O.OD+00
DO 365 J=1,ILS(I)
365 LNGAMR(I)=LNGAMRa)+ANa,J)*a-NTAUKa,J)-LNTAUKl (I, J))
360 CONTDSfUE
C
C CALCULATDSfG LN(ACnVITY COEFFICIENT) AND LN(ACTIVITY) FOR
C EACH COMPONENT
C
DO 370 I=1,IMX
LNGAMa)=LNGAMCa)+LNGAMRa)
LNA(I)=LNGAM(I)+DLOG(MOLFR(I))
370 CONTINUE
C
C CALCULATING BINARY INTERACTION PARAMETER Xj^
C
CHI=( MOLFR(l)*a'NA(l) -DL0G(V0LFR(1))) + MOLFR(2)*(LNA(2)-
1 DLOG(VOLFR(2))) ) / (M0LFR(1)*V0LFR(2))
WRITE(7,6100)W(1),VOLFR(1),MOLFR(1),CHI
6100 FORMAT(4X,4(D12.6,6X))
C
C INCREMElSÎTESfG WEIGHT FRACTION
W(1)=W(1)+DW
IF (W(1).LT.1.0) GO TO 124
CLOSEaJNIT=7)
PRINT*,' JUST A REMEsfDER TO CHANGE OUTPUT FILE NAME!!'
END

97
APPENDDC B

LISTESfG OF PROGRAM FOR UNIFAP CALCULATIONS

This appendix Usts the FORTRAN code for the UNIFAP program used in the

calculation of the binary solvent-polymer interaction parameter X23.

98
C THIS PROGRAM CALCULATES THE BESÍARY ESfTERACTION PARAMETER
C FOR A SOLVENT-POLYMER SOLUTION. ALL CALCULATIONS AND NON-
C ESJTEGER TERMS ARE DOUBLE-PRECISIONED.
Q * * * * * * * * * * * * : | c : J c :(£*:(: : j c * : ( : : j c * * * * * * * * * : J c * * * * * * * : | c * * * : f . : ( : : ^ ; j c * * * * * * * * * * * * * * * * * * * * *

DOUBLE PRECISION R(89),Q(89),MW(89),A(48,48),TC,TK,DW,W(2),R2,Q2


DOUBLE PRECISION Rl(2),Ql(2),MWl(2),RlPRD4E(2),QlPRn4E(2)
DOUBLE PRECISION LNC0MB1,PHPREV[E(2),THPREVIE(2),AMN(48,48)
DOUBLE PRECISION SHI(48,48),SUMQPRW,QPRW(2,10),THETAPR(2,10)
DOUBLE PRECISION SUM1,SUM2,SUM3,LNTAUK(1,10),LNTAUK1(1,10)
DOUBLE PRECISION SUMQPRW1,QPRW1(1,10),THETAPR1(1,10),LNRES1
DOUBLE PRECISION LNA1,CHIPAR
INTEGER AN(2,10)
DD4ENSIONN1(89),N2(89),IAR(2,10),MGNO(2,10),ILS(2)
CHARACTER Nl$(89)* 10,N2$(89)* 10,C$(2)*6,CM$(2,10)* 10,CS$(2,10)* 10
C
C THIS PART OF THE PROGRAM READS R AND Q DATA FROM DATAFELE
C UFNRQ48.DTA
C
OPENaJNIT=4,FILE='A:UNFRQ48.DTA',STATUS='OLD')
PRD^*,'READING A:UNFRQ48.DTA'
DO 10 1=1,89
READ(4,*)Nl$(I),Nia),N2$(I),N2(I),Ra),Q(I),MW(I)
10 CONTBSfUE
CLOSE aJNIT=4)
C
C THIS PART OF THE PROGRAM READS EsfTERACTION PARAMETER DATA
C FROM DATAFELE UNFA48.DTA .THIS DATA WAS OBTAINED FROM THE
C FIFTH REVISED VERSION OF PRAUSNITZ'S UNIFAC PARAMETERS
OPEN(UNIT=5,FILE='A:UNFA48.DTA',STATUS='OLD')
PRINT*,'READING A:UNFA48.DTA'
DO 20 1=1,48
DO 30 J= 1,45,4
30 READ(5,*)A(IJ),Aa,J+l),Aa,J+2),Aa,J+3)
20 COlSfTINUE
CL0SEaJ]SfIT=5)
C
C ALL SUBSEQUENT CALCULATIONS ARE FOR A BE^ARY POLYMER-SOLV
C ENTSYSTEM
DvlX=2
C
C DsfmALIZDSfG THE VALUES OF THE SECONDARY GROUP NOS. AND
C NUMBER OF GROUPS OF EACH TYPE FOR BOTH COMPONENTS
C

99
35 DO 40 I=1,DVDC
DO 50 J=l,10
L\R(U)=0
50 ANaj)=0
40 CONTDSfUE
C
C THIS PART OF THE PROGRAM IS FOR ENTERDSfG COMPONENT DATA
C
PRD^*,'ENTER COMPONENT DATA DSf THE REQUIRED FORMAT'
PRDSfT*,'ENTER CHARACTER DATA JN APOSTROPHES'
PRDSÍT*,'ENTER FIRST COMPONENT AS SOLVENT'
DO 60 I=1,IMX
PRDVÍT*,'ENTER NAME OJPTO 6 CHAR) OF COMPONENT ',1
READ*,C$(I)
PRDSfT*,'DSfPUT NO.OF SECONDARY GROUP TYPES Dvf ',C$(I)
READ*,ILSa)
PRDSÍT*,'DSfPUT MADSf GR.NAME,SEC GR.NAME (UPTO 10 CHAR)'
PRINT*,'MAIN GROUP NO.,SEC.GROUP NO. AND NO.OF EACH
1 TYPE'
DO 70 J=1,ILS(I)
70 READ*,CM$a,J),CS$(I,J),MGNOa,J),IARa,J),ANa,J)
60 CONTINUE
C
C ENTERD^G COMPONENT EsfFORMATION E^O OUTPUT FILE
C
0PEN(U]SÍIT=7,Fn.E='A:S0LP0LY.DTA',STATUS='NEW')
WRITE(7,4900)C$(2),C$( 1)
4900 FORMAT(20X,'CALCULATION OF DSfTERACTION PARAMETER FOR'y
1 25X,A,'-',A,' SYSTEM')
WRITE(7,*)
WRITE(7,*)' COMPONENT LIST'
DO 8 0 1=1,IMX
WRITE(7,5000)I,C$a)
5000 FORMAT(30X,I2,5X,A)
80 CONTDWE
WRITE(7,*)
DO 90 I=1,DV1X
WRITE(7,5001)C$a)
5001 FORMAT(23X,'TYPE AND NO.OF GROUPS IN ',A)
DO 100 J=1,ILS(I)
100 WRITE(7,5005)CS$(U),AN(I,J)
5005 FORMAT(28X,A,5X,I2)
WRrrE(7,*)
90 CONTDSfUE

100
105 PRINT*,'DSfPUT TEMPERATURE ,DEG C'
READ*,TC
TK=TC+273.15D+00
WRITE(7,5010)TC
5010 FORMAT(23X,'TEMPERATURE = ',D10.4,' C')
WRITE(7,*)
WRITE(7,*)
^ * * * * * * * * * * * : i c * * * : i c * * * * * * : i c 4::): 4: : ( : * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

C THIS PART OF THE PROGRAM PERFORMS THE UNIFAP CALCULATIONS


C FOR DETERMDSÍDSfG SOLVENT ACnVITY
/-<:ic***********************:ic:j::ic:ic:ic:|c:ic:ic******************************** * * * * * * * *

C SOME PRELIMD^ARY CALCULATIONS TO OBTAIN RlPRDvlE AND Q1PRD4E


C VALUES
C
DO 120 I=1,IMX
Ria)=0.0D+00
Qia)=0.0D+O0
MWl(I)=O.OD+00
RlPRIME(I)=O.OD+00
Q1PRDV1E(I)=O.OD+00
DO 130 J=1,ILS(I)
Ria)=Ri(i)+RaARa,j))*AN(i,j)
Qia)=Qia)+Q(iARa,j))*AN(u)
130 MW1 (I)=MW 1 a)+MW(L\R(I,J))* ANa, J)
RlPRIME(I)=Rl(I)/MWia)
QlPRIME(I)=Ql(I)/MWia)
120 CONTDSfUE
C
C CALCULATD^G BmARY D^RACTION VALUES A(M,N) BETWEEN GROUPS
IFLAG=0
DO 150 I=1,IMX
DO 160 J=1,ILS(I)
MM=MGNOa,J)
DO 170 n=l,IMX
DO 180 jj=i,n.sai)
NN=MGNOai,JJ)
AMN(MM,NN)=A(MM,NN)
SHiavlM,NN)=DEXP(-A]VD^(MM,NN)A'K)
IF (AMNavlM,NN).EQ.O.O) THEN
IF (MM.NE.NN) THEN
WRITE(7,5021)CS$a,J),CS$aiJJ)
5021 FORMAT(15X,' AOvlJSO VALUE NOT AVAILABLE FOR ',A,'-',A,' PAIR')
IFLAG=1
ENDIF

101
ENDff
180 CONTTNUE
170 CONTINUE
160 CONTDSfUE
150 CONTDsfUE
IF aFLAG.EQ.O) THEN
WRITE(7,*)'ALL INTERACTION PARAMETERS ARE AVAILABLE'
EISÍDIF
WRITE(7,*)
WRITE(7,*)
C
WRITE(7,*)' WEIGHT WEIGHT',
1' CHI'
WRITE(7,*)' FR. OF FR. OF'
WRITE(7,*)' LIQUro POLYMER'
C
C DSfmALIZD^G WEIGHT FRACTIONS
C
DW=0.01
W(l)=0.01
110 W(2)=1.0-W(l)
c
C SOME PRELIMDSfARY CALCULATIONS
C
R2=0.0D-K)0
Q2=0.0D+00
DO 135 I=1,EVDC
R2=R2+Wa)*RlPRIMEa)
Q2=Q2+W(I)*Q1PRIME(I)
135 CONTDWE
DO 140 I=1,IMX
PHPRDVlE(I)=RlPiaME(I)*W(I)/R2
THPRIMEa)=QlPRIMEa)*W(I)/Q2
140 CONTINUE
C
C CALCULATION OF COMBDSfATORIAL PART (bialc) OF SOLVENT ACTIVITY
C
LNCOMB1=DLOG(PHPRIME(1))+(1-PHPRIME(1))+5.0D+00*Q1(1)*
1 (DLOG(THPRDV1E(1)/PHPRDV[E(1))+(PHPRIME(1)/THPRIME(1))-1.0D+00)
C
C CALCULATION OF RESDDUAL PART Onalr) OF SOLVENT ACnVITY
C CALCULATION OF THETAPRDVIE VALUES FOR EACH GROUP DSf EACH
C COMPONENT
C

102
SUMQPRW=O.OD+00
DO 200 I=1,IMX
DO 210 J=l JLS(I)
QPRWa,J)=W(I)*AN(I,J)*Q(L\R(I,J))/MWl(I)
210 SUMQPRW=SUMQPRW+QPRWa,J)
200 CONTDSfUE
DO 220 I=1,DÆX
DO 230 J=1,ILS(I)
THETAPRa,J)=QPRWa,J)/SUMQPRW
230 CONTINUE
220 CONTDSfUE
C
C CALCULATION OF bi(Tauk) FOR ALL GROUPS IN SOLVENT SPECIES
C
1=1
DO 240 J=1,ILS(1)
SUMl=0.0D+00
DO 250 n=l,D4X
DO 260 jj=i,n.sai)
260 SUMl=SUMl+THETAPR(n,JJ)*SHI(MGNO(n,JJ),MGNO(l,J))
250 COlSfTDSfUE
SUM3=0.0D+00
DO 270 n=l,IMX
DO 280 JJ=l,ILSai)
SUM2=0.0D+00
DO 290 ni=l,IMX
DO 300 JJJ=l,ILS(ni)
300 SUM2=SUM2+THETAPR(m,JJJ)
1 *SHiavlGNO(m,JJJ),MGNOai,JJ))
290 CONTDSÍUE
280 SUM3=SUM3+THETAPR(n,JJ)*SHI(MGNO(l,J),MGNOai,JJ))/SUM2
270 CONTINUE
LNTAUK(l,J)=QaAR(l,J))*(l-DL0G(SUMl)-SUM3)
240 CONTDWE
C
C CALCULATION OF UiTaukl VALUES WITHm SOLVENT MOLECULE GROUPS
C
IF aLS(l).EQ.l) THEN
LNTAUKl(l,l)=0.0D+00
GO TO 350
ENDIF
C
C CALCULATION OF THETAPRIMEl VALUES FOR EACH GROUP D^ SOLVENT
C MOLECULE

103
SUMQPRWl=O.OD+00
DO 301 J=1,ILS(1)
QPRWl(l,J)=AN(l,J)*QaAR(l,J))/MWl(l)
SUMQPRW1=SUMQPRW1+QPRW1(1,J)
301 CONTDWE
DO 302 J=1,ILS(1)
THETAPRl (1 ,J)=QPRW1 (1 ,J)/SUMQPRW1
302 CONTDSÍUE
C
DO 310 J=1,ILS(1)
SUMl=O.OD+00
DO 320 JJ=l,n.S(l)
320 SUMl=SUMl+THETAPRl(l,JJ)*SHia4GNO(l,JJ),MGNO(l,J))
SUM3=0.0D+00
DO 330 JJ=1,ELS(1)
SUM2=0.0D+00
DO 340 JJJ=1,ILS(1)
340 SUM2=SUM2+THETAPRl(l,JJJ)*SHI(MGNO(l,JJJ),MGNO(l,JJ))
330 SUM3=SUM3+THETAPR1 (1 ,JJ)*SHI(MGNO( 1 ,J),MGNO( 1, JJ))/SUM2
LNTAUKl(l,J)=Q(IAR(l,J))*(l-DLOG(SUMl)-SUM3)
310 CONTDsfUE
C
350 LNRESl=O.OD+00
DO 360 J=1,ILS(1)
LNRES 1=LNRES l+AN^l^J^^O-NTAUK^l^J^-LNTAUKUl^J))
360 CONTINUE
C
C CALCULATDSfG SOLVENT ACTIVITY
C
LNAl=LNCOMB 1+LNRES1
C
C CALCULATDsfG INTERACTION PARAMETER
CHIPAR=(LNA1-DLOG(PHPRDV1E(1))-PHPRDV1E(2))/((PHPRDV[E(2))**2.0)
WRITE(7,6100)W(1),W(2),CHIPAR
6100 FORMAT(2(13X,D8.3),13X,D12.5)
C
C INCREMENTDSfG WEIGHT FRACTION W(l)
W(1)=W(1)+DW
IF (W(1).LE.0.99) GO TO 110
CLOSEaJNIT=7)
PRINT*,' JUST A REMDSfDER TO CHANGE OUTPUT FILE NAME!!'
END

104
APPENDDC C

DATA FILE OF UNIFAC PARAMETERS

This datafile (UNFRQ48.DTA) contains the parameters required by the UNIFAC

and UNIFAP programs Usted in Appendices A and B. The UNIFAC interaction para-

meters between the 48 molecular groups Usted here were obtained from the Supple-

ment to Program UNIFAC, Fifth Revision : January 1992 by John P. Baker (Refs. 76,77).

105
P ***********4c***********************************************************

C THIS DATAFn.E PROVIDES THE PARAMETERS REQUIRED BY THE UNIFAC


C AND UNIFAP PROGRAMS. FOR EACH MOLECULAR GROUP, THEY ARE
C LISTED DSf THE ORDER MADSÍ GROUP, MAIN GROUP NUMBER, SECONDARY
C GROUP, SECONDARY GROUP NUMBER, GROUP VOLUME AND SURFACE
C AREA PARAMETERS R AND Q, AND GROUP MOLECULAR WEIGHT MW.
^^% ^^ ^^ ^^ ^L ^L ^u ^u^u ^u ^u ^u ^u ^u^u ^u^u ^ ^u ^u ^u ^u ^u ^u ^u ^u ^u ^u ^u ^u^u ^^^^ ^ ^u ^u^u ^u^u ^u ^ ^u ^ ^ ^ ^u^u ^u ^u ^u^u ^ ^ ^ ^ ^ ^u ^u ^ ^u ^u ^ ^u ^ ^ ^ ^^ ^^ ^ ^ ^ slc slc
• ' • n ^ ^ " ^ * ^ ^ * ^ " ^ " ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ * ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^'t* ^ ^ ^ ^ ^ •!* 'F" T* *I*'T*

"CH2",1,"CH3M,.9011,.848,15.034
"CH2",1,"CH2",2,.6744,.54,14.026
"CH2",1,"CH",3,.4469,.228,13.018
"CH2",1,"C",4,.2195,0,12.010
"C=C",2,"CH2=CH",5,1.3454,1.176,27.044
"C=C",2,"CH=CH",6,1.1167,.867,26.036
"C=C",2,"CH2=C",7,1.1173,.988,26.036
"C=C",2,"CH=C",8,.8886,.676,25.028
"C=C",2,"C=C",9,.6605,.485,24.020
"ACH",3,"ACH",10,.5313,.4,13.018
"ACH",3,"AC",11,.3652,.12,12.01
"ACCH2",4,"ACCH3",12,1.2663,.968,27.044
"ACCH2",4,"ACCH2",13,1.0396,.66,26.036
"ACCH2",4,"ACCH",14,.8121,.348,25.028
"OH",5,"OH",15,1,1.2,17.008
"CH3OH",6,"CH3OH",16,1.4311,1.432,32.042
"H2O",7,"H2O",17,.92,1.4,18.016
"ACOH",8,"ACOH",18,.8952,.68,29.018
"CH2CO",9,"CH3CO",19,1.6724,1.488,43.044
"CH2CO",9,"CH2CO",20,1.4457,1.18,42.036
"CHO",10,"CHO",21,.998,.948,29.018
"CCOO",11,"CH3COO",22,1.9031,1.728,59.044
"CCOO",11,"CH2COO",23,1.6764,1.42,58.036
"HCOO",12,"HCOO",24,1.242,1.188,45.018
"CH2O",13,"CH3O",25,1.145,1.088,31.034
"CH2O",13,"CH2O",26,.9183,.78,30.026
"CH2O",13,"CH-O",27,.6908,.468,29.018
"CH2O",13,"FCH2O",28,.9183,l.l,30.026
"CNH2",14,"CH3NH2",29,1.5959,1.544,31.058
"CNH2",14,"CH2NH2",30,1.3692,1.236,30.50
"CNH2",14,"CHNH2",31,1.1417,.924,29.042
"CNH",15,"CH3NH",32,1.4337,1.244,30.50
"CNH",15,"CH2NH",33,1.207,.936,29.042
"CNH",15,"CHNH",34,.9795,.624,28.034
"(C)3N",16,"CH3N",35,1.1865,.94,29.042
"(C)3N",16,"CH2N",36,.9597,.632,28.034
"ACNH2",17,"ACNH2",37,1.06,.816,28.034

106
"pyridine",18,"C5H5N",38,2.9993,2.113,79.098
"pyridine",18,"C5H4N",39,2.8332,1.833,78.090
"pyridine",18,"C5H3N",40,2.667,1.553,77.082
"CCN",19,"CH3CN",41,1.8701,1.724,41.052
"CCN",19,"CH2CN",42,1.6434,1.416,40.044
"COOH",20,"COOH",43,1.3013,1.224,45.018
"COOH",20,"HCOOH",44,1.528,1.532,46.026
"CCr,21,"CH2Cr,45,1.4654,1.264,49.483
"CCI",21,"CHCI",46,1.238,.952,48.475
"CCr,21,"CCl",47,1.006,.724,47.467
"CC12",22,"CH2C12",48,2.2564,1.988,84.940
"CCI2",22,"CHC12",49,2.0606,1.684,83.932
"CC12",22,"CC12",50,1.8016,1.448,82.924
"CC13",23,"CHC13",51,2.87,2.41,119.389
"CC13",23,"CC13",52,2.6401,2.184,118.381
"CC14",24,"CC14",53,3.39,2.91,153.838
"ACC1",25,"ACC1",54,1.1562,.844,47.467
"CNO2",26,"CH3NO2",55,2.0086,1.868,61.042
"CNO2",26,"CH2NO2",56,1.7818,1.56,60.034
"CNO2",26,"CHNO2",57,1.5544,1.248,59.026
"ACNO2",27,"ACNO2",58,1.4199,1.104,58.018
"CS2",28,"CS2",59,2.057,1.65,76.142
"CH3SH",29,"CH3SH" ,60,1.877,1.676,48.108
"CH3SH",29,"CH2SH",61,1.651,1.368,47.100
"furfural",30,"furfural",62,3.168,2.481,96.090
"DOH",31,"(CH2OH)2",63,2.4088,2.248,60.052
"r,32,"I",64,1.264,.992,126.92
"Br",33,"Br",65,.9492,.832,79.916
"C-C",34,"CH'-C",66,1.292,1.088,25.028
"C'-C",34,"C~C",67,1.0613,.784,24.020
"Me2SO",35,"Me2SO",68,2.8266,2.472,78.131
"ACRY",36,"ACRY",69,2.3144,2.052,53.06
"aCC",37,"Cl(C=C)",70,.791,.724,35.45
" ACF",38," ACF",71 ,.6948,.524,31.01
"DMF",39,"DMF-1 ",72,3.0856,2.736,73.09
"DMF",39,"DMF-2",73,2.6322,2.12,71.09
"CF2",40,"CF3",74,1.406,1.38,69.01
"CF2",40,"CF2",75,1.0105,.92,50.01
"CF2",40,"CF",76,.615,.46,31.01
"COO",41,"COO",77,1.38,1.2,44.010
"SiH2",42,"SiH3",78,1.6035,1.2632,31.110
"SiH2",42,"SiH2",79,1.4443,1.0063,30.102
"SiH2",42,"SiH",80,1.2853,.7494,29.094
"SiH2",42,"Si",81,1.047,.4099,28.086

107
"SiO",43,"SiH20" ,82,1.4838,1.0621,46.101
"SiO",43,"SiHO",83,1.303,.7639,45.093
"SiO",43,"SiO",84,1.1044,.4657,44.085
"NMP",44,"NMP",85,3.981,3.2,99.13
"Tert-N",45,"-N=" ,86,0.2854,0.0920,14.008
"N,N-Amide",46,"CON(CH3)2",87,2.859,2.428,72.086
"Amide",47,"CONH2",88,1.466,1.336,44.034
"OCCOH",48,"C2H5O2",89,2.1226,1.904,57

108
APPENDK D

AX CALCULATIONS FOR PSf-(NMP-DMSO) SYSTEM

The AX value for the PSf-(NMP-DMSO) system was calculated as foUows. The

PSf-solvent mixture system consisted of 2 g of the polymer and 98 g of the (1:1) solvent

mixture ie., 54.78 g of NMP and 43.22 g of DMSO. The coagulation value of the system

was 1.1 g water. Based on this overaU composition, the binary concentration-dependent

interaction parameters between aU the components were estimated by UNCFAC and

UNIFAP. The values of Xj^ and X^j in this case are then given by

X i 2 = X^j^.NMP + ^water-DMSO (Dl)

-^23 = -^NMP-PSf + -^DMSO-FSf \^^)

In addition to these interactions, the binary interaction between NMP and DMSO

must also be accounted for by Xf^.D^so- Therefore, the AX value in this case is given

by

AX = -Xi2 + X23 + Xsi^2 + Xi3 (D3)

where Sl and S2 represent the components of the solvent mixture, NMP and DMSO.

From the UNIFAC and UNIFAP calculations, the AX value in this case was estimated to

be

AX = -(0.491 - 0.530) +(0.409 + 0.919) +0.0411 +3.7

= 5.108

109
APPENDDC E

NITROGEN FLUX CALCULATIONS FOR DENSE CA MEMBRANES

From the fundamental defmition of permeabiUty coefficient, the flux (N) of a

gaseous species across a polymer membrane with effective thickness (1) is given by

N = P.(Ap/l) (El)

where P is the permeability coefficient of the species in the polymer material, and Ap is

the partial pressure driving force the gas across the membrane interphase.

The nitrogen fluxes of the dense membranes obtained from the CA-MA and CA-

THF systems were, therefore, calculated as foUows, based on the permeabUity coefficient

of CA, as reported in Uterature [79] under approximately the same experimental

conditions as those used in this study, and the thickness of the membranes, which were

measuredfi^omthe SEMs of these membranes.

PermeabUity coefficient of N^ in CA, (P) = 0.23 barrer

Upstream pressure of N^ gas = 65 psig

Downstream pressure of N^ gas = 0 psig

Partial pressure driving force for N^ transport, Ap = 65 psi

= 336.05 cm Hg

(a) For CA-MA membrane:

Thickness of membrane, 1 '^ 67.5 |im

= 67.5 X 10-^ cm

Therefore, from eqn. El,

110
flux across CA-MA membrane, (N) = [0.23 X 10"^° (cc(STP).cm)/(cmls.cm Hg)]

X [336.05 cm Hg/(67.5 X lO^ cm)]

= 1.145 X 10-' cc(STP)/(cm^s)

(b) For CA-THF membrane:

Thickness of membrane, 1 - 7 5 |im

= 75 X lO^ cm

SimUarly, from eqn. El,

flux across CA-THF membrane, (N) = [0.23 X lO'^" (cc(STP).cm)/(cmls.cm Hg)]

X [336.05 cm Hg/(75 X 10"^ cm)]

= 1.0305 X 10-^ cm^ (STP)/(cmls)

111

You might also like