Ultrafiltration uses semi-permeable membranes to separate solution
components based on selective molecular weight and structure. In
ultrafiltration, the solution is contacted with the membrane under
an applied pressure (Figure 1). The applied pressure forces the
solvent and the smaller molecules through the membrane. The membrane
retains larger molecules. The solution passed through the membrane is
the permeate. The solution containing the retained molecules is the
retentate (Mulder, 1991).
Figure 2 illustrates the basic structure of a spiral-wound membrane
module. A permeate spacer is placed between two membranes. The two
membranes and permeate spacer are glued together along three edges
to form what is called a leaf. The unglued edge is connected to a
perforated central collection tube. The membrane leaf along with a
mesh feed spacer is spirally wound around the central collection
tube. To make the leaf length shorter, several membrane leaves are
wound simultaneously around the central tube. Multi-leaf designs
are used to minimize the pressure drop experienced by the permeate
fluid (Matsuura, 1994). The wound module is placed inside a tubular
pressure vessel. Figure 3 illustrates a multi-leaf, spiral-wound module
and the pressure vessel in which it is placed.
Figures 4 and 5 illustrate the fluid flow of the spiral-wound membrane
module. The feed flows parallel to the central collection tube along
the mesh feed spacer (Matsuura, 1994). The feed spacer acts as a
turbulent generator to reduce concentration polarization. As the feed
makes its way down the module, it flows across (normal to) the membrane
surface (Matsuura, 1994). The crossflow of the solution to the membrane
surface facilitates diffusion of the solute from the membrane surface back
to the bulk flow (Long, 1999). Crossflow counteracts concentration
polarization. A portion of the feed permeates the membrane. The
permeate fluid flows along the permeate spacer. The direction of flow
of the permeate is spiral and roughly perpendicular to the feed flow
direction. The permeate is collected in the central tube.
Fouling refers to any phenomena that result in reduced transmembrane
flux over time. Fouling is primarily caused by concentration polarization.
Concentration polarization is the accumulation of molecules at the membrane
surface (Bungay et al., 1986). Three types of fouling are cake layer
formation, pore blockage, and internal pore fouling. Cake layer formation
and pore blockage refer to fouling on the surface of the membrane.
Deposited molecules build up on the membrane surface during cake layer
formation (Figures 6) while, rejected molecules block pore openings
during pore blockage (Figures 7). Internal pore fouling refers to fouling
within the pores of the membrane (Figure 8). Internal pore fouling occurs
when molecules are deposited within the pores of the membrane, which reduces
the average membrane pore size. All three forms of fouling resist the
transport of solvent and small solute molecules through the membrane
pores (Mulhern, 1995). Membrane fouling models for pore blocking, cake
formation and internal pore plugging are discussed in detail by
Jaffrin et al. (1996).
Work by Meireles et al. (1992) shows the effects of fouling on
ultrafiltration performance. Table 1 shows the permeabilities of
various membranes before and after either adsorbtion or
ultrafiltration with a bovine serum albumin solution. Permeabilities were
determined before and after either adsorbtion or ultrafiltration to
estimate the role of adsorption in fouling. In the adsorbtion process,
a clean membrane was brought into contact (non pressurized) with an
albumin solution for 15 hrs to reach adsorbtion equilibrium. In absence
of denaturation, the permeability of each membrane was reduced by the
same amount by fouling and by adsorption regardless of ultrafiltration
conditions. When protein molecules are denatured, fouling was found
to be more severe. Protein denaturation is sensitive to operating conditions.
A possible explanation for the more severe fouling is that the denatured
proteins were deposited on the membrane surface. These observations
suggest that membrane fouling is primarily due to adsorption of proteins,
and possibly to the deposition of denatured molecules on the membrane surface.
Meireles et al. (1992) also measured ultrafiltration fluxes and
retention coefficients at the initial stage of the run and at steady
state (Table 2). Since the concentrations, velocity, and transmembrane
pressure remained constant in these tests, the observed decrease of
permeate flux with time was due to membrane fouling. The initial retention
coefficient was 100% for the 10 kDa membrane and did not change while
the flux decreased. The fluxes and retention coefficients of the 40 kDa
and 100 kDa membranes both varied during the run. The initial retention
coefficient was 85% for the 40 kDa membrane and 20% for the 100 kDa membrane.
At steady state the retention coefficients of the 40 kDa and 100 kDa
membranes were 100% and 98% respectively. This increase was a result of
membrane fouling. The results of Meireles et al. (1992) illustrated how
fouling can induce changes, not only in the solvent flux but also in the
separation potential. They also illustrated that fouling is a function
of membrane properties. The least significant change occurs for smallest
MWCO membrane and the most significant change occurs for the largest MWCO membrane.
In general, it is observed that high cross flow velocity and low operating
pressure provide the most economical solute-solvent separations (Figure 9).
A high cross flow velocity will limit concentration polarization by
facilitating transport of the solute from the membrane surface back to
the bulk flow. At low operating pressures, the permeate flux increases
almost linearly with the applied pressure. As the pressure rises, the
permeate flux increases at a steadily slower rate (i.e. the relationship
is no longer linear). A pressure is reached beyond which no further
increase in flux is observed. At this pressure, the limiting flux has
been reached. A possible explanation for limiting flux is that higher
pressure causes more severe concentration polarization and fouling.
Higher crossflow velocities can reduce concentration polarization and
fouling. It can be seen from Figure 9 that the higher the crossflow
velocity, the higher the limiting flux. A balance of crossflow velocity
and pressure must be achieved for optimal separation. The balance varies
by membrane type and by feed solution characteristics. In our proposed
study our experimental setup prevents us from varying the crossflow velocity.
The optimum operating pressure will be determined as the pressure just below
the limiting flux.
As the protein concentration in the feed solution increases, the maximum
achievable flux decreases (Figure 10). At low pressures, the flux
increases with applied pressure. As the pressure increases, the concentration
of the protein at the membrane surface increases. The flux reaches a maximum
value when the concentration of the protein at the membrane surface is so high
that it forms a cake layer. The cake layer acts like a secondary membrane.
Further increases in pressure increase the thickness of the cake layer but
not the flux. The transmembrane flux is then determined by the rate of the
diffusion of protein from the membrane surface. The more concentrated the solution,
the lower the flux at which the cake layer forms. In our proposed work
we will be working with dilute protein solutions 1 mg/ml to 4 mg/ml.