1 Introduction to membrane filtration of liquids

Transcription

1 1 Introduction to membrane filtration of liquids 1.1 Introduction This book is largely concerned with solving process problems in the membrane filtration of liquids. In that sense, it is more a chemical engineering book than a membrane science book. There are many fine books available which provide much more information on membrane synthesis and structure, module design, transport processes in membranes and applications of membrane technologies in industrial and medical processes [1 4]. Nonetheless, a small amount of background on membrane separations is needed before the business of process modelling, analysis and design can begin. So, in this chapter, some general terminology is defined, membranes and modules are described and some of the main characteristics of the various membrane filtration techniques are outlined. A membrane is simply a physical barrier through which pure solvent can pass while other molecules or particles are retained. In the case of ultrafiltration (UF) and microfiltration (MF) this semi-permeability is largely a result of the relative sizes of the solute/particles and the membrane pores. Solute retention is a little different in reverse osmosis (RO) and is largely determined by charge effects. The RO membrane can be thought of as a matrix through which solvent and solute diffuse at different rates. Nanofiltration (NF) occupies a transition zone between UF and RO and solute retention is complex, involving both size and charge effects. MF is the filtration of suspensions containing particles that are generally less than 10 µm in diameter. For particles greater in size than this, the micro prefix is dropped but there is no significant difference, at least in terms of the basic concepts involved, between filtration and microfiltration. Throughout this book, the terms filtration and microfiltration are used at different times in the knowledge that any equations or models derived are applicable to either process. UF can be thought of as the filtration of solutions of high molecular weight molecules. In the NF region, we are usually dealing with molecules such as peptides, antibiotics and other compounds with molecular weights in the range Da. In RO, the solutions being filtered contain very small ionic species such as Na + and Cl - ions. It is worth mentioning in passing that there are situations where one might use a membrane in a filtration process for which it is not actually designed. For example, it would not be unheard of to use a UF membrane to filter a microbial suspension because such a choice may lead to better long-term performance.

2 2 Introduction to membrane filtration of liquids Whether this is UF or MF is a debatable point but it is probably more sensible to classify it as UF. Microfiltration and ultrafiltration are used widely in industry. MF is used in areas like microbial and animal cell separation, often as part of the downstream processing of bioproducts. It is also used in clarification of beverages such as wine, beer and fruit juice, and in dairy processing and wastewater treatment. UF and the related technique of diafiltration (DF) are used for concentration and purification of solutions of macromolecules of all kinds, especially proteins. All of the above examples of MF and UF are bio in nature and that is the emphasis throughout this book. The main application of RO is in the desalination of water or the production of ultra-pure water for electronics industries [5]. NF, which is an emerging technology, lies somewhere between RO and UF and has potential for use in wastewater treatment, food processing and textile dye removal. Further details are given in the work of Schafer et al. [6], probably the only book available at present which gives a broad and detailed introduction to this relatively new, and complex, area. All of the techniques mentioned above can be termed pressure-driven processes. Table 1.1 shows the approximate pressure ranges required for the various membrane filtration processes. Table 1.1 Approximate pressure ranges for membrane filtration processes. Technique Pressure range (bar) MF UF 1 10 NF 7 40 RO The variety of pressures required reflects the nature of the suspension/solution in each case and the type of membrane employed. In MF and UF, the membranes have distinct pores and fluid flow theory tells us that smaller pores lead to greater pressure drops across the membrane. In the case of UF, the need for greater pressures, resulting from the smaller pores, is enhanced by the osmotic pressure of the solution adjacent to the membrane. This creates an osmotic back pressure that opposes the applied pressure. In RO, the permeability of the solvent through the dense, pore-free, membrane is low while the osmotic back pressure created by the low molecular weight solutes is much higher than it is for the high molecular weight solutes that arise in UF. The pressures employed in NF fall between those used in UF and RO, reflecting the fact that NF has both UF and RO characteristics. To understand membrane filtration processes, some basic knowledge of fluid mechanics is required. For MF, one needs to understand some complex ideas relating to the compression of highly concentrated suspensions, or cakes, as well as issues concerning the various electrostatic and adhesive interactions that can occur between particles. However, for modelling purposes, much of the complexity of MF is lumped into a small number of experimental parameters, a common chemical engineering strategy when

3 1.2 Definitions and terminology 3 confronted by highly complex problems. Therefore, the equations derived in Chapters 2 and 3 do not involve any intricate theory, but remain at a level that should be accessible to most. UF is the best understood of the membrane filtration processes. As a general rule, predicting the behaviour of solutions is easier than predicting the behaviour of suspensions. If you are from a chemical engineering background, you only have to think of a subject like distillation where accurate models of the vapour liquid equilibrium of multi-component systems have been developed, allowing for accurate design of distillation columns. In contrast, describing an apparently simple operation like the settling of solid particles is a very complicated task that often defies a purely theoretical approach. To understand UF, the key engineering subjects required, in addition to fluid flow, are physical chemistry and mass transfer. The latter is a subject that is probably unique to chemical engineering and can involve some quite tricky ideas. However, in the context of UF, it is usually enough to know the basic concepts without getting bogged down in some of the notoriously difficult ideas that are scattered throughout mass transfer. The rigour and detail of Treybal s famous work comes to mind [7]. Physical chemistry, in the form of chemical thermodynamics, is important in UF, not only because it is an inherent part of mass transfer, but because a key thermodynamic concept, osmotic pressure, appears explicitly in the equations. One aspect of UF that makes it simpler than NF and RO is the fact that the solute rejection process, i.e., the process by which the solute is rejected by, or retained by, the membrane, is essentially a sieving process. This means that the transmission of a solute through the membrane largely depends on the relative size of the solute and the pores. Matters are not so simple in NF and RO. While the behaviour of the fluid adjacent to the membrane can be described with a similar theoretical approach to that used in UF, the transport of solvent and solute through the membrane is more complex, being driven by chemical potential differences rather than simple hydraulic pressure differences. In RO, the rejection of ions is normally complete and determined by charge. In NF, there is a range of rejection mechanisms, from RO-like behaviour at the lower end of the pore size spectrum to UF-like behaviour at the upper end. In between, the mechanism of rejection is a complex mixture of charge and size effects. 1.2 Definitions and terminology Dead-end and crossflow configurations Figure 1.1 below illustrates the difference between the dead-end and crossflow configurations using MF as an example. In the dead-end configuration, the feed flows normal (perpendicular) to the membrane while in crossflow it flows parallel, or tangential, to the membrane. The dead-end terminology tends to be used mainly in the bioprocessing field, and most chemical engineers would drop the dead-end term and refer to this configuration simply as filtration or microfiltration. However, given the importance of the crossflow mode of operation, the dead-end prefix has become important for clarity.

4 4 Introduction to membrane filtration of liquids Feed Feed Filtrate Filtrate Figure 1.1 Dead-end and crossflow modes of operation illustrating the accumulation of a solid cake on the membrane. When MF is done in crossflow mode, it is termed crossflow microfiltration (CFMF), at least in this book. In the literature you will sometimes see it referred to as tangential flow filtration (TFF), crossflow microfiltration or crossflow filtration. As UF, NF and RO are normally done in crossflow mode, the CF prefix is generally not used for these operations. The rationale for using a crossflow mode of operation in CFMF is that it reduces cake formation, i.e., the accumulation of a layer of solids on the membrane. Cake formation leads to a reduction in flow through the membrane so it is always desirable that it be minimised. It is important to note that cake formation does not really occur in UF, NF and RO. These processes exhibit a phenomenon referred to as concentration polarisation. This is the formation of a gradient in solute concentration whereby its concentration is highest adjacent to the membrane and lowest in the bulk flow. It is somewhat akin to boundary layer formation in fluid mechanics and is explained in Chapter 4. An additional and attractive feature of crossflow operation, which follows from the fact that cake formation/concentration polarisation is limited, is that steady state operation is possible, at least in principle. This is in contrast to most dead-end operations, which tend to be batch in nature. In dead-end filtration, cake continues to accumulate until the capacity of the filter is reached. At this point the equipment must be dismantled, the cake recovered or discarded, the system cleaned and a new batch commenced. There are exceptions, of course, such as when a rotary vacuum filter is used for continuous filtration of suspensions. This is explained in Chapter 2. It is commonplace, particularly when perusing the research literature, to read statements to the effect that crossflow techniques offer significant advantages over dead-end techniques. There is clearly some truth in this but the two techniques are not always comparable. In dead-end filtration, for example, the product of the process can be a cake of particles, whereas in CFMF the product is typically a concentrated suspension. Of course, both techniques can be used to recover a solute in the filtrate, also known as permeate. This is the liquid that has passed through the membrane. The word, filtrate, is generally used for MF processes, whereas permeate tends to be used for UF, NF and RO. Both terms are used throughout this book and the reader should take them as being identical.

5 1.2 Definitions and terminology 5 It is important to mention some terminology that is used throughout the research literature and other textbooks, but which is not used so much in this book. This is gel formation, which is the idea, commonly put forward in UF work, that when the solute reaches a very high concentration at the membrane, a gel forms. A gel is a solid, jellylike substance formed by the action of intermolecular forces, or cross-links, between the solute molecules. Evidence for gel formation in UF is essentially circumstantial as it is based purely on flux measurements. However, UF can be explained without having to invoke this phenomenon. We return to this issue in Chapter 4. Figures 1.2 to 1.4 show the range of process configurations that are available for crossflow membrane filtration techniques. They can be divided into continuous, batch and fed-batch systems and all are important in practice. The continuous configurations are shown below. In the feed-and-bleed mode, a portion of the retentate is recirculated as shown above and this requires an additional pump. The presence of the recirculation loop increases the pressure of the fluid entering the module as well as the volumetric flowrate through the module. Both of these changes have the effect of increasing the filtrate flowrate. Recirculation can also be used in batch systems as shown in Fig Filtrate Filtrate Feed Feed Figure 1.2 Process configurations for single pass and feed-and-bleed continuous membrane filtration. Figure 1.3 Process configurations for batch membrane filtration processes showing simple and recirculation options.

6 6 Introduction to membrane filtration of liquids Feed / Water Figure 1.4 Process configurations for fed-batch membrane filtration and diafiltration. Fed-batch systems are used when the retentate tank has insufficient capacity. In Fig. 1.4, fed-batch operation with a recirculation loop is shown. The volume of solution in the retentate tank is typically kept constant while the feed tank empties. When the feed to the retentate tank is pure water this operation is referred to as constant volume diafiltration (CVD) Key process parameters in membrane filtration All the membrane processes described in this book are pressure driven processes. This means that the flowrate of filtrate / permeate is controlled by the trans-membrane pressure in the same way that a heat flow is controlled by a temperature gradient in heat transfer processes. In the dead-end configuration, this is simply given by P TM = P feed P filtrate, (1.1) where P feed is the feed pressure and P filtrate is the filtrate pressure. In the crossflow configuration, one has to take account of the pressure loss of the feed as it passes along the membrane. Thus, the trans-membrane pressure is typically computed as the difference between the average pressure of the feed and the filtrate pressure, i.e., P TM = P feed + P retentate 2 P filtrate. (1.2) The retentate is technically the concentrated fluid that emerges from the membrane module. It is at a lower pressure and a higher concentration than the feed. Only in continuous, single pass configurations is there a significant difference between the particle/solute concentrations at the inlet and exit of the membrane. In all other configurations, there is very little change in concentration in a single pass through the membrane and the exit concentration is essentially the same as the concentration at any point in the flow channel. Rather than dealing with permeate flowrates, one normally talks about the permeate flux. This is simply the permeate flowrate divided by the membrane area and is usually denoted by J, with units of m/s. An important concept that arises in membrane filtration

7 1.2 Definitions and terminology 7 processes is the idea of a resistance to flow. If pure water of viscosity, µ, passes through a membrane with a flux, J, the flux is related to the trans-membrane pressure, by the expression J = P TM μr m, (1.3) where R m is termed the membrane resistance and has units of m 1. This parameter typically depends on the properties of the membrane, including pore size, porosity, charge, hydrophobicity, etc. The origins of Eq. 1.3 are explained in Chapter 2. The advantage of the resistance idea is that it is easily extended to account for multiple resistances in series. Thus, for an MF process, it is shown in Chapter 2 that J = P TM μ(r m + R c ), (1.4) where R c is the resistance of the cake. As is also seen in Chapter 2, the resistance of a filter cake is a complex function of particle and liquid properties and, usually, the trans-membrane pressure, a phenomenon referred to as cake compressibility. In this book, this resistance-in-series approach is not used for the molecular filtration process of UF, NF and RO. Many researchers and practitioners do indeed adopt this strategy but in Chapter 4, an entirely different language, based on the idea of osmotic pressure, is used to describe molecular filtration processes. The tangential or crossflow velocity is a key parameter in all membrane processes. This parameter is simply the average velocity of the solution or suspension as it flows tangentially to the membrane. Suppose, for example, the feed flowrate is Q m 3 /s and the membrane module contains N individual tubular membranes of diameter d t.the crossflow velocity, u, is then simply given by u = Q/N πd 2 t /4. (1.5) Membrane fouling is a very important phenomenon in determining fluxes and it is important to be clear as to how it is defined. Many people use the term loosely to describe the reduction of flux with time that occurs almost universally with membrane processes. However, in this book we take the term fouling to mean an increase in membrane resistance. Thus, concentration polarisation and cake formation are not considered as fouling even though they contribute significantly to the reduction in flux from its clean membrane value Solute and particle rejection The whole purpose of membrane filtration is that a solute or particle is retained on the membrane. This phenomenon is usually referred to as rejection. In many MF applications, membranes with a pore size of 0.2 µm are employed, and even allowing for the fact that real membranes have a distribution of pore sizes, one can be almost certain that no particles, even bacteria, will pass into the filtrate. In that case, there is complete

8 8 Introduction to membrane filtration of liquids rejection. However, complete rejection does not always occur in molecular filtration processes and the degree of rejection depends on the properties of the solute, the properties of the membrane and the precise operating conditions. Rejection is quantified with a parameter known as the rejection coefficient, σ, defined by σ = 1 c p c, (1.6) where c p is the solute concentration in the permeate and c is its concentration in the retentate. Thus, complete rejection implies σ = 1. If a membrane presents no barrier to a solute, σ = 0. In reality, there are some subtleties that must be dealt with when defining rejection coefficients and these are explained in Chapters 7, 8 and 9. Even in UF, where rejection is a result of the difference between the size of the solute molecules and the pore size, rejection is a relatively complex phenomenon. One reason for this is that a molecule, such as a protein, does not have a definite size, regardless of the solvent characteristics. The precise conformation and size of the molecule usually depend on liquid properties such as ph and ionic strength. Furthermore, molecules are not spheres and whether they pass through a pore, or not, may well depend on their orientation relative to the membrane as they approach the pore entrance. In addition, the pores of a typical membrane have a distribution of diameters. As a consequence, the rejection of macromolecules by UF membranes is something of a probabilistic phenomenon and rather than talking about the size of the molecule, it is generally better to deal with its molecular weight. UF membranes are then defined, not in terms of their pore size, but in terms of their molecular weight cut-off (MWCO). The exact definition of this parameter varies depending on the manufacturer but it is typically defined as the molecular weight of a standard test molecule for which the rejection coefficient has some value between 0.95 and 1.0. Often the standard test molecule used to determine the MWCO is a polysaccharide such as dextran. Characterisation of the membrane involves measuring σ in a stirred cell apparatus under well-defined conditions for a range of molecular weights of this molecule. A typical plot of σ versus solute molecular weight will look something like Fig σ Molecular weight (kda) Figure 1.5 Rejection curve for a UF membrane.

9 1.3 Membranes and their properties 9 It is clear that rejection is not a simple process and, in practice, it is made more complicated by the fact that the solute undergoing UF is usually not the same as the test solute used to characterise the membrane. Users will generally choose a MWCO that is significantly lower than the molecular weight of the solute they are filtering. 1.3 Membranes and their properties Ideally a membrane should provide as low a resistance to solvent transport as possible. This resistance is determined, at least in the case of MF and UF, by the size of the pores within the membrane and the membrane porosity, i.e., the fraction of the total membrane area taken up by the pores. From basic fluid mechanics, we know that the smaller the pores, the more resistant is the membrane, but the choice of pore size is usually determined by the process specification. For example, UF of a low molecular weight protein, of necessity, requires a membrane with much smaller pores than UF of a high molecular weight protein. In the case of RO, it is better not to think of a membrane as being a barrier with identifiable pores, but a three-dimensional mesh through which molecules diffuse at different rates, if at all. Charge rather than size is the key determinant of membrane transport in RO. NF is a complex process lying between RO and UF and has characteristics of both of these techniques. Finally, it is very desirable that a membrane be as resistant to fouling as possible. Many manufacturers produce membranes that are low fouling. Typically this means that solutes, especially proteins, have a low tendency to bind to the membrane Membrane materials Nowadays, most membranes are made from synthetic polymers. For UF and MF applications, polysulphones are often used. Other materials that are used commercially include polyvinylidenefluoride (PVDF), polyacrylonitrile (PAN) and polypropylene (PP). In recent times, interest in the use of ceramic materials has increased. These tend to be very robust and can withstand harsh operating and cleaning conditions. RO membranes are often made of polyamide while NF membranes are constructed from a variety of polymers, including polysulphone and polyimide, and ceramic materials Membrane morphology MF membranes are typically symmetric with a sponge-like structure. This means that the pore size and porosity are the same throughout the depth of the membrane. Ultrafiltration membranes are typically asymmetric i.e., they have a thin region ( skin ) of well-defined molecular weight cut-off, supported by a thicker, highly porous region. It is important to note that this asymmetry is an inherent part of the membrane and arises naturally from the chemistry of the membrane synthesis process. It is not two separate components stuck together. The less dense part of the membrane provides very little resistance to

10 10 Introduction to membrane filtration of liquids flow and gives the skin mechanical strength. All the flow and separation characteristics are determined by the membrane skin. RO and NF membranes are typically classified as thin film composites (TFCs). In contrast to asymmetric UF membranes, TFCs do involve separate polymeric layers bonded together, the top layer often being made of polyamide or polyimide and the support layer typically made from polysulphone. The polysulphone layer itself is often supported by a fabric backing, especially in RO membranes. Further details on membranes and their structure are to be found in the references at the end of this chapter, although there is some further discussion of NF and RO membrane characteristics in Chapter Membrane modules This section describes the types of equipment used in crossflow membrane filtration processes. Discussion of dead-end equipment is left for Chapter 2. The housing in which a membrane is placed is termed a module, or in some instances, a cartridge. Thesame basic module types can be used for all membrane processes ranging from CFMF to RO, although specific types of module tend to be preferred for different types of filtration Flat sheet and spiral wound modules Flat sheet modules, shown schematically in Fig. 1.6, use multiple flat sheet membranes in a sandwich arrangement consisting of membrane sheets, attached at their edges to a backing support. Spacer screens, or meshes, provide the channel for the feed flow and are typically mm in height. The space between the membrane support and the membrane itself acts as the channel for permeate flow. Several of these membranes and spacers are stacked alternately and held tightly together to form a cassette. Although flat sheet modules are prone to having their flow channels clogged, they have still been used successfully Spacer mesh Feed Permeate Membrane sheet Backing support Figure 1.6 Expanded view of a portion of a flat sheet cartridge showing examples of feed splitting into retentate and permeate.