13.11. ULTRAFILTRATION MEMBRANE PROCESSES

13.11A. Introduction

Ultrafiltration is a membrane process that is quite similar to reverse osmosis. It is a pressure-driven process where the solvent and, when present, small solute molecules pass through the membrane and are collected as a permeate. Larger solute molecules do not pass through the membrane and are recovered in a concentrated solution. The solutes or molecules to be separated generally have molecular weights greater than 500 and up to 1 000 000 or more, such as macromolecules of proteins, polymers, and starches as well as colloidal dispersions of clays, latex particles, and microorganisms.

Unlike reverse osmosis, ultrafiltration membranes are too porous to be used for desalting. The rejection R, often called retention, is also given by Eq. (13.9-7), which is defined for reverse osmosis. Ultrafiltration is also used to separate a mixture of different-molecular-weight proteins. The molecular-weight cut-off of the membrane is defined as the molecular weight of globular proteins, which are 90% retained by the membrane. A rule of thumb is that the molecular mass must differ by a factor of 10 for a good separation (P7).

Ultrafiltration is used in many different processes at the present time. Some of these are separation of oil-water emulsions, concentration of latex particles, processing of blood and plasma, fractionation or separation of proteins, recovery of whey proteins in cheese manufacturing, removal of bacteria and other particles to sterilize wine, and clarification of fruit juices.

Membranes for ultrafiltration are in general similar to those for reverse osmosis and are commonly asymmetric and more porous. The membrane consists of a very thin, dense skin supported by a relatively porous layer for strength. Membranes are made from aromatic polyamides, cellulose acetate, cellulose nitrate, polycarbonate, polyimides, polysulfone, and so forth (M2, P6, R1).

13.11B. Types of Equipment for Ultrafiltration

The equipment for ultrafiltration is similar to that used for reverse osmosis and gas separation processes, described in Sections 13.3C and 13.10D. The tubular-type unit is less prone to fouling and more easily cleaned than any of the other three types; however, this type is relatively costly.

Flat sheet membranes in a plate-and-frame unit offer the greatest versatility but at the highest capital cost (P6). Membranes can easily be cleaned or replaced by disassembly of the unit. Spiral-wound modules provide relatively low costs per unit membrane area. These units are more prone to fouling than tubular units but are more resistant to fouling than hollow-fiber units. Hollow-fiber modules are the least resistant to fouling as compared to the other three types. However, the hollow-fiber configuration has the highest ratio of membrane area per unit volume.

Cross-flow filtration is the most common type of model used (P7). Spiral-wound flat sheets are used most of all, followed by hollow-fiber units. Batch processes are also quite common. Stirred tanks with a membrane are used which approximate cross-flow operation (R1).

13.11C. Flux Equations for Ultrafiltration

The flux equation for diffusion of solvent through the membrane is the same as Eq. (13.9-2) for reverse osmosis:

Equation 13.9-2


In ultrafiltration the membrane does not allow passage of the solute, which is generally a macromolecule. The concentration in moles/liter of the large solute molecules is usually small. Hence, the osmotic pressure is very low and can be neglected. Then Eq. (13.9-2) becomes

Equation 13.11-1


Ultrafiltration units operate at about 5100 psi pressure drop, compared to 400-2000 for reverse osmosis. For low-pressure drops of, say, 510 psi and dilute solutions of up to 1 wt % or so, Eq. (13.11-1) predicts the performance reasonably well for well-stirred systems.

Since the solute is rejected by the membrane, it accumulates and starts to build up at the surface of the membrane. As pressure drop is increased and/or concentration of the solute is increased, concentration polarization occurs, which is much more severe than in reverse osmosis. This is shown in Fig. 13.11-1a, where c1 is the concentration of the solute in the bulk solution, kg solute/m3, cs is the concentration of the solute at the surface of the membrane, and cp is the concentration in the permeate.

Figure 13.11-1. Concentration polarization in ultrafiltration: (a) concentration profile before gel formation, (b) concentration profile with a gel layer formed at membrane surface.


As the pressure drop increases, this increases the solvent flux Nw to and through the membrane. This results in a higher convective transport of the solute to the membrane, that is, the solvent carries with it more solute. The concentration cs increases and gives a larger back molecular diffusion of solute from the membrane to the bulk solution. At steady state the convective flux equals the diffusion flux:

Equation 13.11-2


where Nwc/ρ = [kg solvent/(s · m2)](kg solute/m3)/(kg solvent/m3) = kg solute/s · m2; DAB is diffusivity of solute in solvent, m2/s; and x is distance, m. Integrating this equation between the limits of x = 0 and c = cs and x = δ and c = c1,

Equation 13.11-3


where kc is the mass-transfer coefficient, m/s. Further increases in pressure drop increase the value of cs to a limiting concentration, at which the accumulated solute forms a semisolid gel where cs = cg, as shown in Fig. 13.11-1b. For the usual case of almost-complete solute retention (P7), cp = 0 and Eq. (13.11-3) becomes

Equation 13.11-4


Still further increases in pressure drop do not change cg and the membrane is said to be "gel polarized." Then Eq. (13.11-3) becomes (P1, P6, R1)

Equation 13.11-5


With increases in pressure drop, the gel layer increases in thickness, causing the solvent flux to decrease because of the added gel-layer resistance. Finally, the net flux of solute by convective transfer becomes equal to the back diffusion of solute into the bulk solution because of the polarized concentration gradient, as given by Eq. (13.11-5).

The added gel-layer resistance next to the membrane causes an increased resistance to solvent flux, as given by

Equation 13.11-6


where 1/Aw is the membrane resistance and Rg is the variable gel-layer resistance, (s · m2 · atm)/kg solvent. The solvent flux in this gel-polarized regime is independent of pressure difference and is determined by Eq. (13.11-5) for back diffusion. Experimental data confirm the use of Eq. (13.11-5) for a large number of macromolecular solutions, such as proteins and so forth as well as colloidal suspensions, such as latex particles and so forth (P1, P6).

13.11D. Effects of Processing Variables in Ultrafiltration

A plot of typical experimental data for flux versus pressure difference is shown in Fig. 13.11-2 (H1, P6). At low pressure differences and/or low solute concentrations, the data typically follow Eq. (13.11-1). For a given bulk concentration, c1, the flux approaches a constant value at high pressure differences, as shown in Eq. (13.11-5). Also, more-dilute protein concentrations give higher flux rates, as expected from Eq. (13.11-5). Most commercial applications are flux-limited by concentration polarization and operate in the region where the flux is approximately independent of pressure difference (R1).

Figure 13.11-2. Effect of pressure difference on solvent flux.


Using experimental data, a plot of Nw/p versus In c1 is a straight line with a negative slope of kc, the mass-transfer coefficient, as shown by Eq. (13.11-5). These plots also give the value of cg, the gel concentration. Data (P1) show that the gel concentration for many macromolecular solutions is about 25 wt %, with a range of 5 to 50%. For colloidal dispersions it is about 65 wt %, with a range of 50 to 75%.

The concentration-polarization effects for hollow fibers are often quite small, due to the low solvent flux. Hence, Eq. (13.11-1) describes the flux. In order to increase the ultrafiltration solvent flux, cross-flow of fluid past the membrane can be used to sweep away part of the polarized layer, thereby increasing kc in Eq. (13.11-5). Higher velocities and other methods are used to increase turbulence and hence kc. In most cases the solvent flux is too small to operate in a single-pass mode. It is necessary to recirculate the feed past the membrane, with recirculation rates of 10/1 to 100/1 often used.

Methods for predicting the mass-transfer coefficient kc in Eq. (13.11-5) are given by others (P1, P6) for known geometries such as channels and so forth. Predictions of flux in known geometries using these methods and experimental values of cg in Eq. (13.11-5) in the gel polarization regime compare with experimental values for macromolecular solutions within about 25–30%. However, for colloidal dispersions the experimental flux is higher than the theoretical by factors of 20-30 for laminar flow and 8-10 for turbulent flow. Hence, Eq. (13.11-5) is not useful for predicting the solvent flux accurately. Generally, for design of commercial units it is necessary to obtain experimental data on single modules.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
34.201.8.144