You are on page 1of 10

ENG3185: Separation Processes 2

Bioseparations

Section 3 - Ultra-filtration Theory - Students

3.1 Introduction
In Section 2 we saw the general principles of a membrane process and now we focus on the
design calculations. The key design calculations are the trans-membrane pressure and the
membrane area.

3.2 Theory
The general case is where we have a porous membrane and the pores are large compared to the
size of the solvent molecules. In this case there is flow in the pores not just diffusion. If we
assume that smaller solute molecules can also pass through the pores, then there is the possibility
that solute molecules will be entrained in the solvent flow. The extent to which this happens will
depend on the size of the solute molecules.

We assume that the flux of solvent (often water) is linearly proportional to the difference
between the applied pressure and the osmotic pressure noting one acts against the other.

(3.1)

where Jw = volumetric flux (m3 m-2 s-1)


P = trans-membrane pressure (kPa)
= the reflection coefficient
= osmotic pressure (kPa)
LW (the proportionality constant) = volume based membrane permeability for the
solvent (often water) (m3 m-2 s-1 kPa-1)

The reflection coefficient () is the fraction of the pores that are too small to allow the passage of
solute and from which the solute is reflected and therefore lies in the range (01). For most
UF membranes the pores are quite large compared to the size of the solutes molecules which are
responsible for generating the osmotic pressure and therefore the reflection coefficient is small.
This means that the solvent flux is dominated by the trans-membrane pressure difference which
is large compared to the (osmotic pressure x reflection coefficient) i.e. the second term in the
brackets in equation 3.1 is very small compared to the first.

At low concentration, the osmotic pressure may be related to the concentrations on either side of
the membrane.

(3.2)

Page 1
where M = molarity (concentration)
R = universal gas constant
T = absolute temperature
i = vant Hoff factor

From Physical Chemistry, the van't Hoff factor (i), is the number of moles (ions if the solute
ionizes) of solute actually in solution per mole of solid solute that was added to make the
solution. So, for sodium chloride NaCl which gives one mole of Na+ ion and one mole of Cl- ion
per mole of NaCl dissolved, i = 2, and for calcium chloride CaCl2 (one Ca+ ion and two Cl- ions
per mole of the salt added) i = 3. Notice that the actual charges on the individual ions do not
affect the osmotic pressure directly.

For a solute we normally assume two mechanisms are causing it to cross through the membrane.
The first is the direct diffusion of solute through the membrane and second is the solute being
dragged through the membrane with the solvent flow. The first mechanism is characterised by
a conductivity coefficient (membrane permeability) and a concentration driving force and the
second by a solvent transfer rate (volumetric flux), the fraction of the pore capable of passing
solute (1 - reflection coefficient) and the mean solute concentration on either side of the
membrane.

(3.3)

where Js = flux of solute (kg s-1 m-2) [note the unit difference between JW and JS]
Cs = concentration driving force of solute across the membrane
= mean of the concentrations of solute on either side of the membrane
Ls = membrane permeability for the solute (m s-1).

Note that the units of the solvent and solute fluxes are different with one being mass based
(solute) and the other being volume based (solvent).

Now consider the mass balance on the membrane unit as illustrated in Figure 3.1

Retentate (R, CSF)

Feed
Membrane P
(F, CSF)
Separation
Unit

Permeate (P, CSP)


Figure 3.1 General Arrangement
Assuming steady state conditions we can write the overall volume balance assuming there is no
Page 2
change of density and a solute balance based on volumetric flows and concentrations.

Overall balance

(3.4)

Solute balance

(3.5)

Finally, if the area of membrane is A, then, assuming that the value of JW (solvent flux) varies
with position over the area A, the Permeate rate is the integral of the solvent flux (JW) across the
area.

(3.6)

where P = volumetric permeate rate


JW = local solvent flux
A' = dummy variable of the integration

Similarly, the solute flow in the permeate is given by

(3.7)

Now in Lecture 2 we saw the definition of the rejection coefficient:

(3.8)

this parameter has to be determined experimentally and may, unfortunately, depend on the trans-
membrane pressure and the feed concentration.

In practice, predicting the variation in the solvent and solute fluxes is impossible unless we can
predict the variation in concentrations across the membrane surface. Measuring the local
concentrations and hence the local fluxes is possible but predicting them for design in other than
very simple flow regimes is very difficult. For cross flow devices we can carry out a steady state
differential analysis across the membrane to give a concentration vs. position relationship and
then integrate the resulting differential equations to get a value of the mean transfer fluxes.

We are usually only concerned with the overall operation of the membrane so we can simplify

Page 3
the above expressions greatly if we assume that either there are no gradients in concentration
across the area of the membrane surface or the variation may be characterised by a mean
concentration hence giving a constant value of both the solvent and solute fluxes everywhere
across the area.

Then Equations (3.6) and (3.7) can be integrated simply as

(3.9)
and

(3.10)

Now, by dividing Equation (3.10) by (3.9), we get the concentration of solute in the permeate
flow (normally the maximum concentration) which is what we wish to determine.

(3.11)

We are left with one problem namely what concentration to use as being representative of the
upstream surface of the membrane remembering that for diffusion from the bulk to the
membrane there must be a concentration gradient. If the feed compartment is well mixed, then

(3.12)

where Ci,U = concentration of component i at upstream surface of membrane


Ci,R = concentration of component i in the retentate

If the liquid on the retentate side is in plug flow, then approximately

(3.13)

3.2.1 Theory for Solution Membranes

Here the solute and solvent both dissolve in the membrane and are transported through the
membrane at different rates by molecular diffusion and emerge on the permeate side of the
membrane.

In the absence of any mass transfer resistance in the fluid on either side of the membrane we
postulate that the solute or solvent dissolves in the membrane.

Page 4
Thus we can express the concentration of component i just inside each face of the membrane in
terms of the concentration in the fluid at equilibrium on the exterior of the membrane

Figure 3.2 Simple Solution across the Membrane

Note that the concentration of component i just inside the upstream surface of the membrane is
greater than the concentration in the fluid in contact with surface because the only component i
transfers from the solution to the membrane leaving diluent material behind hence exhibiting a
greater concentration just inside the membrane surface.

(3.14)

where ki = partition coefficient of component i between the membrane and the adjacent
solution
Ci,1 = concentration of component i inside the upstream membrane surface
Ci,2 = concentration of component i inside the downstream membrane surface

Using Ficks Law of diffusion for the diffusion in the membrane we can define the flux of
component i passing through the membrane.

(3.15)

where Ji = flux of component i through the membrane


DE,I = the diffusivity of component i through the membrane
M = thickness of the membrane

Using the partition coefficient from equation 3.12 we can eliminate the internal membrane
concentrations in equation 3.13 we get

(3.16)

Page 5
The above expression will replace Equations (3.1) and (3.3) when a solution/diffusion membrane
in use.

3.2.2 Concentration Polarisation


If there are differential rates at which species pass through the membrane and there are finite
mass transfer resistances in the vicinity of the surface of the membrane, then Figure 3.2 will be
wrong and the picture will be more as shown in Figure 3.3

Figure 3.3 Concentration Polarisation across the Membrane

Because more solvent than solute passes through the membrane, solute is concentrated upstream
of the membrane because the solvent moves more readily than the solute. If the solute is a
macromolecule or a protein this layer can develop into a gelatinous mass on the upstream surface
of the membrane and this is called gel polarisation.

The concentration layer increases the driving force for the solute and reduces the driving force
for the solvent, so it acts to reduce the selectivity and the permeability of the membrane, both of
which are bad news since it encourages the solute to pass through the membrane and reduced the
driving force for solvent to pass through the membrane.

Suppose the concentration on the upstream face of the membrane is low and the concentration on
the downstream face is negligible, then the osmotic pressure is given by

(3.17)

Page 6
where = equilibrium concentration of solute at the upstream surface of the membrane
under to gel layer

Figure 3.4 Mass Balance on plane within the Gel film

If we assume a steady state and Ficks First Law controls the diffusion, we can perform a solute
balance across any plane within the upstream film and the flux of solute to the left and right must
be equal.

(3.18)

where z = the position inside the film measured from the upstream surface of the film
= the concentration of solute in the film at position z

We can write the general solution of this ODE as

(3.19)

where B = integration constant

Now we can substitute for from Equation (3.11) as CSP giving

(3.20)

Now we know the following boundary condition for the system namely

Page 7
(3.21)

hence substituting z = 0 into 3.20 and setting CS to CS,U we get




Hence

which rearranges to give

If we now apply this across the entire film noting that CS = CS* at z =


(3.23)

This analysis is also applicable if the solute flux is zero and hence the downstream concentration
is zero which is the case when the membrane passes only solvent.

(3.24)

We note that the effective diffusivity divided by the film thickness is an effective mass transfer
coefficient (Year 2 Mass Transfer notes) which must be experimentally determined. The
analysis also ignores the resistance to flow of solvent through the gel layer.

It is usual to define the polarisation modulus, PM as

(3.25)

There are major complications if the concentration of the solute at the membrane exceeds the
solubility of the solute because we get precipitation of solute on the membrane surface.

3.3 The Design Problem Absence of Concentration Polarisation

Suppose we are required to treat a known flow rate of feed, F, at a known concentration of the

Page 8
solute of interest CSF and we may have one of two product specifications namely,

a) A specified minimum concentration of solute in the retanate,


or,
b) A specified maximum concentration of solute in the permeate.

The problem is to calculate the flow rates of the two product streams, the area of the membrane
required and the trans-membrane pressure to be used.

Example 1
0.1 m3 s-1 of a protein solution at a feed concentration of 10 mg dm-3 is to be concentrated by
removal of the water through an ultra-filtration membrane, the protein solution is required at a
concentration of 25 mg dm-3. Assume that the membrane is impermeable to protein and the
maximum trans-membrane pressure of 5 bar is to be used. The membrane permeability for
solvent is 1x10-4 m3 s-1 m-2 bar-1 and the reflection coefficient is zero.

Example 2
Suppose we have the same example as above but that the membrane has a finite permeability to
the protein such that Ls = 10-5 m s-1 and the reflection coefficient = 0.999

3.3 The Design Problem Absence of Concentration Polarisation

Example 3 Concentration Polarisation


With concentration polarisation we have an extra variable (the concentration of solute at the
surface of the membrane) and an extra equation (Equation (3.24)) provided we know, or can
estimate, the effective mass transfer coefficient we can solve the equation set if we make sensible
assumptions.

If we can ignore the resistance to flow for the solvent that the concentration layer represents,
then we can still calculate the solvent flux and hence calculate the concentration at the surface of
the membrane from Equation (3.23) or (3.24). Otherwise the calculation is identical to Example
2 above. We expect that the area of membrane required will increase significantly.

3.4 Concluding Comments


We have developed and used a simple theory of ultra-filtration. In very simple circumstances it
is possible to obtain direct analytical solutions to the design problem. For any other case, an
iterative procedure is required and the usual procedure is to guess the permeate concentration.

This can then be calculated and direct substitution provides rapid and stable convergence.
We have also ignored the effects of the concentration polarisation on the flux of solvent through
the membrane.

However, an important lesson has been learnt: fouling of membranes leads to loss of
permeability and loss of selectivity and is therefore very important to avoid.

Page 9
We have largely ignored the complications due to osmotic effects.

Page 10

You might also like