Theory of Transport in Membranes
Theory of Transport in Membranes
Theory of Transport in Membranes
A membrane process is a separation process that covers a broad range of problems from
particles to molecules and a wide variety of membranes are available to design a process.
Although the membranes may vary in material (organic vs. inorganic) and structure (porous
vs. nonporous) the basic principle of membrane separation is the same. The separation
process and mass transport through the membranes is a function of the membrane being used
and the constituents being separated, and subsequently the theory used to describe the process
and mechanisms. However, common for all systems is the principle illustrated in Figure 1
where a membrane is considered a permselective barrier, or interface between two phases, and
the separation process takes place due to a specific driving force transporting a compound
through the membrane from the one phase to the other.
membrane
Phase 1:
Phase 2:
Permeate
Feed
Driving force; C, P, T, E
Figure 1. Schematic of membrane separation process with different driving forces that
are present
The membrane separation process is defined by which compound is more readily transported
(selectivity) through the membrane and by the flow of the specific compound (flux). Though
this may occur by various mechanisms the performance and efficiency of the process is
described by these parameters. The selectivity of a membrane is generally expressed as a
retention factor (R) or by a separation factor (). The definition of the retention factor is given
by equation (0);
R=
C feed C permeate
C feed
= 1
C permeate
C feed
(0)
A/ B =
xA
yB
(2)
xB
The selectivity is chosen such that its value is greater than unity and it is expressed by which
component passes through the membrane, i.e. for A/B the permeation rate of compound A is
greater than compound B. If A/B = B/A = 1, no separation is achieved.
Given a membrane selectivity, the flow through a membrane, or flux, is defined by the driving
force that governs the transport mechanism. When a membrane is the permselective barrier
between two phases the transport of a molecule or particle across from one phase to the other
will depend on a force that acts on the molecule or particle. The amount of force acting on the
constituents will depend on the gradient in potential, or the difference in potential, across the
membrane. The chemical potential difference () and the electrical potential difference are
the main potential differences in membrane processes. [Mulder, 1997] The driving force is
generally determined by the difference in chemical potential and is a result of either pressure
differences across the membrane, concentration gradients between the two phases or from
temperature differences. The electro potential difference is mainly important in electrodialysis
and similar processes.
The definition of the chemical potential relates back to the thermodynamic principles that
stem from Gibbs free energy equations. [Mulder, 1997; Baker et.al., 1995; Sirkar/Ho (eds)
et.al., 1992] For isothermal conditions the chemical potential is determined by pressure and
concentration, and the chemical potential is expressed as in equation (3)
i = io + RT ln ai + Vi P
(3)
Where io is a constant expressing the chemical potential of the pure compound while ai is the
concentration of the compound in terms of activity, ai = i xi and determined by the activity
coefficient and mole fraction. Vi is the volume of compound i and P the pressure, R is the
universal gas constant and T temperature. The difference in chemical potential is therefore
determined by a concentration term and pressure. This potential difference can be expressed
as a differential as in equation (4);
i = RT ln ai + Vi P or d i = RTd ln ai + Vi dP
(4)
The chemical potential across a membrane is defined as the chemical potential on the
permeate side minus the chemical potential on the feed side. From thermodynamics, diffusion
(i.e. mass transport of a compound) is defined as spontaneous only if the change in chemical
potential of the diffusing species is negative. A driving force from the feed side of a
membrane to the permeate side can therefore be generated by decreasing the pressure on the
permeate side or pressurizing the feed side.
2.2
A schematic of basic pore geometries is shown in Figure 2. Two basic geometries are shown
that possibly represent the two extremes, though in reality pore geometries are not as clearly
defined and have many variations in between these structures. The simplest geometry is a
series of regular cylindrical pores that are perpendicular to the membrane surface. In an
idealized membrane of cylindrical pores with the same radius, the volume flux through the
membrane may be described with the Hagen-Poiseuille equation.
rp2 P
8 x
where:
(5)
rp
P
x
For cylindrical pores the tortuosity is equal to unity. As the shape of the pore changes
determining the tortuosity factor of a membrane is difficult. From equation (5) it is apparent
that the morphology and characteristics of a membrane structure are important and influence
the transport across the membrane. The equation gives a good description of well-defined
parallel pores, however, such structures are not common in reality and the challenge is
defining the parameters describing the membrane characteristics.
The opposite of this symmetrical shape is the porous structure that is formed when spherical
shapes are packed in a bed to form a porous media. This structure is typically found in
sintered and some phase inversion membranes. Several researchers have investigated flow
through a porous media, particularly flow in sand-beds for filtration and groundwater flow.
An expression for flow through a porous media of packed spheres has been developed by
Carmen-Kozeny and is shown in equation (6);
J=
3
K S (1 )
2
where;
K
S
P
x
P
x
(6)
In most practical cases equation (6) is not easy to use either as the equation parameters are
very much dependent on the geometry of the pores. This is a function of the media grain size,
shape of the grains (sphericity), packing density etc. that define the membrane characteristics
and subsequently it is difficult to evaluate these equation parameters.
Further investigation of the flow through pores show that the transport of molecules through
porous membranes can be described by the mechanisms of convective flow, Knudsen
diffusion and pore surface diffusion. When the pores are much larger than the molecule,
transport will be by convective flow. However, if the pore size is smaller than the mean free
path the convective flow is replaced by Knudsen diffusion. The mean free path () is defined
as the average distance traversed by a molecule between collisions. Knudsen diffusion is
primarily a phenomenon that occurs in gas transport in porous media. In liquids molecules are
Convective
flow
Knudsen
diffusion
Molecular
sieving
k T
d2 P 2
where; T
P
d
k
(7)
is the temperature
is the pressure within the membrane pore
is the diameter of the gas molecule
is the Planck constant
Under atmospheric conditions and for pore diameters around 0.01 m Knudsen flow was
found to be dominant in gas flow through microporous membranes. [Callahan, 1988]
Considering the Hagen-Poiseuille and Carmen-Kozeny models presented above, the equation
parameters immediately before the P/x term can be viewed as expressions for the hydraulic
permeability expressed in terms of the membrane characteristics. The transport models for
porous membranes can thus be modified to account for specific membrane structures or
mechanisms such as Knudsen flow. When Knudsen flow is dominant, a transport model
defining the flux as in equation (8) may be used. [Mulder, 1997]
n rp2 DK P
J=
R T
x
where; rp
n
DK
R
T
P
x
(8)
Nonporous membranes are primarily made from organic polymers and the presentation in this
section will therefore only refer to synthetic organic membranes. Nonporous membranes are
defined as such that the membrane has a homogenous structure without any defined pores.
However, on a molecular level the polymer matrix and structure of the molecular chains does
result in a morphology that can be defined as molecular pores. This distinction will be
referred to when discussing transport through different types of nonporous membranes. In
simple terms, transport through nonporous membranes can be described as a solutiondiffusion mechanism. Nonporous membranes are sometimes referred to as solution-diffusion
membranes because the transport occurs when molecules dissolve into the membrane and
diffuse across it. In principle the mechanism is the same whether the separation process is for
liquids or gases. However, the solubility of liquids and gases into the membrane polymer is
quite different, with liquids in general showing a higher affinity for polymers than gases, and
the actual transport of a liquid or gas through the membrane is therefore different.
The transport of a gas, vapor or liquid through a membrane can be described by the
permeability of a membrane, which is a function of the solubility and diffusivity of the
compound in the membrane material. Permeability can be expressed as;
Permeability (P) = Solubility (S) x Diffusivity (D)
(9)
The solubility is a thermodynamic parameter and will depend on whether it is a gas or liquid
that is dissolving into the membrane. In general, the solubility of gases in a polymer is low
and can be described by Henrys Law but does not apply to liquids. The diffusivity term on
the other hand is a kinetic parameter that defines how fast a dissolved component is
transported through the membrane. The diffusivity is dependent on the molecule being
transported and the geometry of the polymer membrane. Different transport models have been
developed for nonporous membranes and the solution-diffusion model and the free-volume
diffusion model are probably the most widely accepted. [Baker et.al., 1995]
The underlying assumptions of mass transport in a solution diffusion processes are the
following three steps: [Nijhuis, 1990; Wijmans et.al., 1995]
1. Selective uptake (sorption) of one of the components on the feed side of the
membrane,
2. Selective transport (diffusion) through the membrane,
3. Desorption (evaporation) on the permeate side of the membrane in the vapor phase.
In the solution-diffusion model the starting point to describe the kinetics of the diffusion
process is based on thermodynamics where the driving force is related to the potential
difference. The flux of a component can then be described by the following equation;
J = Li
d i
dx
(10)
i = io + RT ln ai + Vi ( P Pi o )
(11)
For compressible phases (gas), the molar volume changes as a function of pressure. The ideal
gas laws need to be applied when integrating equation (4) and gives;
i = io + RT ln ai + Vi ln
P
Pi o
(12)
dC
dx
J
D
C
(13)
is the flux
is the diffusivity
is the concentration gradient over distance x.
Integrating this expression over the membrane thickness (l) we get the following expression,
where Cf and Cp are the component concentrations at the membrane interface; [Boddeker,
1995]
D ( Cf - Cp )
l
(14)
This model was first proposed in the early 1960s and has since been developed by several
studies. [Sirkar/Ho (eds) et.al., 1992] Several investigators have attempted to develop a
simple and useful version of the theory, however, providing a precise definition for the freevolume parameters has been difficult. The theory states that the movement of molecules
within a matrix depends on the free volume available as well as the energy to overcome the
polymer-polymer interactions. The definition of free volume is therefore dependent on
whether a polymer is in the glassy or rubbery state. The state of a polymer is important in
defining the free volume in this theory.
The state is defined as the phase in which the polymer appears. The phase of a polymer can be
distinguished with two regions defined as rubbery or glassy polymers. For non-crystalline or
amorphous polymers there is a transition temperature where the polymer changes from a
glassy to a rubbery state. These states are determined by structural factors relating to polymer
chain flexibility, chain interaction and molecular weight. A rubbery polymer is an amorphous
polymeric material that is above its softening or glass transition temperature while a glassy
polymer is below this temperature. [Billmeyer, 1971]
The transport models presented in sections 2.2.1 - 2.2.4 have been discussed based on the
definition of porous and nonporous membranes. These transport models can be described as
using a phenomenological approach or a mechanistic approach to describe a membrane
separation process. The first group does not give any information on how the actual
separation occurs while the second tries to relate the process to specific structure related
parameters of the membrane. As discussed in previous sections, membranes are manufactured
for specific applications and have a large variety in structure and composition such that a
simple definition porous / nonporous is not always valid. In many cases, membranes will have
characteristics and properties that can be related to both a porous and nonporous definition.
Ultimately the need for a more general transport model is apparent which can be used for
various applications.
Phase 1
Membrane
Phase 2
R1
RM
R2
R1 resistance in phase 1
RM resistance in membrane
R2 resistance in phase 2
Driving force
The principle for a resistance in series model is illustrated in Figure 4 where the overall
resistance to transport is the sum of a series of resistances in each phase. The resistance in
phase 1, which is normally the feed, can be caused by the formation of a cake layer on the
surface of the membrane or by liquid-film boundary layer effects such as concentration
polarization or gel formation. The membrane resistance will inevitably be a function of the
membrane structure (porous vs. nonporous) and various fouling phenomena. In phase 2, the
resistance will relate to how efficient the permeating component is removed from the
membrane. The overall resistance to mass transport in a membrane process will therefore be a
function of the separation application (liquid vs. gas) and the specific properties of the
membrane used in the process. For all cases, the reciprocal of the overall resistance will be
equal to the sum of the reciprocal resistances defined for a system as expressed by equation
(15);
1
Roverall
1
1
1
+
+
R1 RM R2
where: R1
RM
R2
(15)
The resistance-in-series model can be applied to both porous and nonporous membranes. The
definition of the membrane resistance will depend on the membrane properties and needs to
be defined for each application. For a nonporous membrane, the membrane resistance can be
expressed by the permeability and the fiber thickness. In porous membranes, the pore size and
geometry will determine if the transport is by convective flow, Knudsen diffusion or
The resistance-in-series model has been chosen to describe the mass transport in a degassing
application using a membrane contactor. When a vacuum is applied to one side of a gas
permeable membrane that is immersed in water, a concentration gradient is created across the
membrane. Dissolved gases in the water diffuse in the direction of decreasing concentration,
through the membrane and into the vacuum. With hydrophilic microporous membranes the
pores are wetted by the water and are filled by the solute. In the case of hydrophobic
microporous membranes the pores are generally gas-filled and provide the interface between
the vacuum and liquid phase. For hydrophobic membranes the pores will remain dry and gasfilled as long as the penetration pressure is not exceeded. The critical pressure at which
wetting of the pores occurs is referred to as the breakthrough pressure. [Kiani et.al., 1984;
Kim et.al., 1987; Nirmalakhandan et.al., 1987; Gabelman et.al., 1999] The breakthrough
pressure for a liquid/liquid phase or gas/liquid phase has been reported as in equation (16).
P =
2 cos
rp
(liquid/liquid), P =
2 cos
rp
(gas/liquid)
(16)
Here is the interfacial tension and the surface tension, while is the contact angle and rp is
the pore radius. Relating to a gas/liquid phase, this pressure is therefore a function of the
surface tension, the contact angle and the pore diameter. The contact angle is defined as the
Liquid
film
Gas phase
Gas
film
Liquid phase
CL
Bulk
Membrane
Liquid
film
Gas phase
Gas
film
Bulk
Ci
Pi
A. Hydrophobic membranes
P or C*
Ci
Pi
P or C*
B. Hydrophilic membranes
Figure 5. Partial pressure and concentration profiles for a gas/liquid phase with: A. gasfilled pores, and B. liquid-filled pores
A membrane contactor as shown in Figure 5 is essentially used as a phase barrier between the
gas and liquid phases, where a hydrophobic membrane prevents the liquid phase from
entering the pores. Under a vacuum the pores remain dry and gas filled while under pressure
the barrier may prevent bubbles from entering the liquid. As long as the liquid pressure is
equal to or higher than the gas pressure bubbles will not form. In effect, the membrane serves
to create a phase interface between the gas/liquid interface and an equilibrium separation
process of either gas absorption or gas stripping will take place depending on the direction of
the driving force.
The kinetics of gas transfer modeled using the resistance-in-series model, is therefore like the
two-film theory for mass transport across and gas/liquid interface with the addition of a
membrane resistance. [Letterman (ed), 1999] The mass transfer of gases across a membrane is
therefore determined by three resistances in series as shown in Figure 6. These consist of a
resistance across the liquid-film layer created by the aqueous solution around the fibers, the
resistance of the membrane, and finally the gas-film layer created by the gases within the
fiber. The sum of these resistances defines the overall resistance to gas transfer with a
microporous membrane system. The reciprocal of the overall resistance is the overall mass
transfer coefficient of the system.
Membrane
Liquid
film
Gas phase
Gas
film
Resistance:
Bulk
Ci
kL
P or C*
Pi
kM
1 1
1
1
= +
+
K kL H k M H kG
kG
where: C
H
(17)
is the interface concentration in the water and gas phase
is partition coefficient expressed by Henrys law constant
In vacuum degassing, the driving force for mass transfer is from the liquid phase to the gas
phase. The over all mass transfer coefficient and individual mass transfer coefficients for a
hydrophobic membrane, based on the aqueous phase concentration are then expressed as;
[Sirkar/Ho (eds) et.al., 1992]
1
1
1
1
= +
+
K k L H k M H kG
where: K
kL
kM
kG
H
(18)
The terms on the right hand side of equation (18) represent the resistance to mass transfer in
the two boundary layers and the membrane. Development of this equation also assumes the
following; a steady state system, equilibrium exists at the interface, the pores are uniform
through out the membrane, no transport occurs through the nonporous parts of the membrane,
and mass transfer can be described by simple film-type mass transfer coefficients.
N = K(C - C*)
where: K
C
C*
(19)
The over all mass transfer efficient, K, is determined by the type of membrane used, and the
appropriate expression representing the respective mass transfer resistances that dominate the
overall resistance to mass transfer. From an evaluation of equation (18) one can see that
resistance to mass transfer may be controlled by one or several stages, depending on the value
of the respective mass transfer coefficients and the partition coefficient. For example, if the
mass transfer coefficients in each phase (liquid-membrane-gas) are in the same order of
magnitude, the value of the partition coefficient will be important. In gas transfer, the value of
Henrys law constant will determine the significance of the membrane resistance and gas-film
boundary layer. For soluble or sparingly soluble gases the value of Henrys law constant is
high, and thus the second and third term in equation (18) become negligible, and the
resistance in the liquid-film boundary becomes dominant.
The mass transfer characteristics and individual mass transfer coefficients for microporous
gas-permeable hollow-fibers have been studied and the results indicate that the mass transfer
coefficient is always controlled by the resistance in the liquid phase [Yang et.al., 1986]. This
is due to the nature of the hydrophobic microporous membranes. Gas-phase diffusion is
several orders of magnitude greater than liquid-phase diffusion, 104 times faster than liquid
phase diffusion [Cussler, 1997]. The resistance in the gas-phase boundary layer can therefore
be neglected. Gas transfer through hydrophobic membranes is by gas-phase diffusion and the
membrane resistance can be negligible. As with aeration systems, the overall mass transfer
can thus be estimated by evaluating the liquid-film transfer coefficient, kL. Several
investigators have developed mass transfer correlations for aeration using hollow fiber
membranes [Sirkar/Ho (eds) et.al., 1992; Noble/Stern (eds) et.al., 1995; Perry, 1997]. Since
degassing is simply the reverse process of aeration, the same correlations can be applied to the
vacuum degassing system. The overall mass transfer coefficient may therefore be determined
by calculating the Sherwood number (Sh) based on these correlations.
An alternative method to determine the membrane resistance for a system as well as getting
some insight to the effect of the fluid velocity on the individual mass transfer coefficients is
the Wilson plot method. [Wislon, 1915; Prasad et.al., 1988] The assumption is that the mass
transfer coefficient is proportional to the fluid velocity, v, where the exponential is an
empirical constant for a given system. The Wilson plot gives a straight line when the mass
transfer coefficient is plotted against the fluid velocity as K-1 vs. v-. For hydrophobic
membranes the resistance from the gas phase boundary layer can be neglected and the Wilson
plot is made with data collected at high fluid velocities. The intercept represents the
membrane resistance from which the membrane mass transfer coefficient can be determined.
The Wilson plot is illustrated in Figure 7. The data used in the plot is from this study to
determine the membrane resistance for the composite membranes tested; reference chapter
Error! Reference source not found., section Error! Reference source not found. for
details.
4,0E+02
y = 7143,5x + 71,885
1/K
3,0E+02
R2 = 0,9888
2,0E+02
1,0E+02
0,0E+00
0,0E+00
1,0E-02
2,0E-02
3,0E-02
4,0E-02
-0,83
Describing the mass transfer coefficients of any degassing system or membrane contactor
process is important for the design of a unit. Correlations describing the mass transfer
coefficients of different types of mass transfer equipment have been developed, including
various membrane module configurations. In general, the mass transfer coefficient is
described using equation (20) and can be used to predict the performance of a system.
[Noble/Stern (eds) et.al., 1995; Perry, 1997]
Sh a Re b Sc
(20)
where: Sh
is the Sherwood number
Re
is the Reynolds number
Sc
is the Schmidt number
a, b, , are functions of geometry
The functions in equation (20) are system dependent for membrane processes and the
respective values are specific for a module configuration and mode of operation, i.e. feed flow
outside or inside tubular modules, co-current/counter-current/cross-flow modes etc. By
definition the dimensionless numbers (Sh, Re and Sc) are expressed in equations (21) (23).
Sh =
Re =
Sc =
kL de
D
v L de
(21)
(22)
(23)
where: kL
de
D
vL
By convention the hydraulic diameter for hollow fiber membranes is expressed as,
de =
(24)
Several investigators have reported correlations based on equation (20) to predict the mass
transfer coefficient for a given unit design. For the sake of comparison of correlations the
membrane processes have been grouped as follows; tube side flow, shell side flow parallel to
the fibers and shell side cross-flow. The summary is shown in Table 1.
The Schmidt number (Sc) is in fact a correction factor for temperature such that the different
mass transfer coefficients may be compared at a normalized temperature. [Geankoplis, 1972]
As can be seen from the summary of selected correlations reported in the literature, listed in
Table 1, the correction factor is Sc raised to the 0.33 power.
Comments
Reference
Sh=1.62(d2v/LD)1/2
Shell side parallel flow:
Lvque, 1928
Sh=[de(1-)/L]Re0.6Sc0.33
Sh=1.25(Re de/L)0.93Sc0.33
Sh=8(Re de/L)Sc0.33
Sh=0.019Gz
Sh=(0.53-0.56)Re0.53Sc0.33
Sh=0.18Re0.81Sc0.33
Prasad et.al..1988
Yang et.al..1986
Dahuron et.al..1988
Wickramasinghe, 1992
Costello et.al..1993
Letterman.1999
Wickramasinghe, 1992
Wickramasinghe, 1992
Wickramasinghe, 1992
Wickramasinghe, 1992
Prasad et.al..1988
Prasad et.al..1988
Ct et.al..1989
Ahmed et.al..1996
Johnson et.al..1997
Weiss et.al..1996
Crespo et.al..1994
Note: is the packing fraction. WeEu = gauge feed pressure x pore diameter/water surface tension.
(25)
Although this term represents a fairly good description of the mass transfer it was not found
suitable to predict experimental data. [Wickramasinghe et.al., 1992] The discrepancy was
attributed to effects of module geometry that are not included in equation (25). Variations of
the equation have been developed by several investigators to include geometrical factors and
operating conditions that take into account the module configuration, packing densities, flow
regimes and so forth. [Yang et.al., 1986; Dahuron et.al., 1988; Prasad et.al., 1988;
Wickramasinghe et.al., 1992; Costello et.al., 1993; Letterman (ed), 1999] One of the
problems related to shell side flow parallel to the fibers is channeling effects around the fibers
and fluid flow profiles along the fibers. The definition of the fluid velocity is important as the
mass transfer coefficient is a function of the local fluid velocity along the fibers.