M.J Lewis It can be determined experimentally for each and every component in the feed, by sampling the feed and permeate at the same time and analysing the component in ques tion. It is a very important property of a membrane, as it will influence the extent (quality )of the separation that can be achieved Rejection values normally range between 0 and 1; sometimes they are expressed as percentages(0-100%) when c E 0: nt is retained in the feed when c =c R=0; the component is freely permeating An ideal RO membrane would give a rejection value for all components of 1, whil an ideal UF membrane, being used to concentrate a high molecular weight component or remove a low molecular weight component would give respective rejection values of 1 and 0. If the concentration factor and rejection value are known, the yield of any component, which is defined as the fraction of that component present in the feed, which is recovered in the concentrate, can be estimated. Obviously for reverse osmosis, the yield for an ideal membrane is 1.0. Rejection data for membranes and their effects on eld and separation performance will be discussed in greater detail in Chapter 4 3.4 MEMBRANE CHARACTERISTICS The membrane itself is crucial to the process. The first commercial membranes were made of cellulose acetate and these are termed first-generation membranes. For food processing applications, they had some limitations, with temperatures below 30C and PH range of 3-6. These were followed in the mid-1970s by other polymeric membranes cond-generation membranes), with polyamides and, in particular, polysulphones being widely used for foods. The resulting improvements in cleaning and hygiene are covered in Section 3.8. It is estimated that over 150 organic polymers have now been investigated for membrane applications, Inorganic membranes based on sintered and ceramic materials are also now available. The physical structure of these membranes is complex and as most of them are used for microfiltration. their structure is described in more detail in Chapter 5 The main terms used to describe membranes are microporous or asymmetric. Microporous membranes have a uniform porous structure throughout, although the pore size may not be uniform across the thickness of the membrane. They are usually charac terised by a nominal pore size and no particle larger than this will pass through the nembrane. In contrast to this. most membranes used for ultrafiltration are of a type, having a dense active layer or skin of 0.5-1 um in thickness, and a further support layer which is much more porous and of greater thickness(Fig. 3.3). Overall the porosity of these membranes is high, although the surface porosity may be low, with quoted alues in the range 0.3-15%(Fane and Fell, 1987). Often the porous path may be quite tortuous, the distance covered by the solvent or solute being much greater than the thickness of the membrane; the term tortuosity has been used as a measure of this property. The pores are not of a uniform size, as can be seen when viewed under the electron microscope, and are best characterised by a pore size distribution. This
70 M. J.Lewis It can be determined experimentally for each and every component in the feed, by sampling the feed and permeate at the same time and analysing the component in question. It is a very important property of a membrane, as it will influence the extent (quality) of the separation that can be achieved. Rejection values normally range between 0 and 1; sometimes they are expressed as percentages (0-1 00%). when cp = 0; when cp = CF R = 1; all the component is retained in the feed R = 0; the component is freely permeating. An ideal RO membrane would give a rejection value for all components of 1, whilst an ideal UF membrane, being used to concentrate a high molecular weight component or remove a low molecular weight component would give respective rejection values of 1 and 0. If the concentration factor and rejection value are known, the yield of any component, which is defined as the fraction of that component present in the feed, which is recovered in the concentrate, can be estimated. Obviously for reverse osmosis, the yield for an ideal membrane is 1.0. Rejection data for membranes and their effects on yield and separation performance will be discussed in greater detail in Chapter 4. 3.4 MEMBRANE CHARACTERISTICS The membrane itself is crucial to the process. The first commercial membranes were made of cellulose acetate and these are termed first-generation membranes. For foodprocessing applications, they had some limitations, with temperatures below 30°C and a pH range of 3-6. These were followed in the mid-1970s by other polymeric membranes (second-generation membranes), with polyamides and, in particular, polysulphones being widely used for foods. The resulting improvements in cleaning and hygiene are covered in Section 3.8. It is estimated that over 150 organic polymers have now been investigated for membrane applications. Inorganic membranes based on sintered and ceramic materials are also now available. The physical structure of these membranes is complex, and as most of them are used for microfiltration, their structure is described in more detail in Chapter 5. The main terms used to describe membranes are microporous or asymmetric. Microporous membranes have a uniform porous structure throughout, although the pore size may not be uniform across the thickness of the membrane. They are usually characterised by a nominal pore size and no particle larger than this will pass through the membrane. In contrast to this, most membranes used for ultrafiltration are of asymmetric type, having a dense active layer or skin of 0.5-1 pm in thickness, and a further support layer which is much more porous and of greater thickness (Fig. 3.3). Overall the porosity of these membranes is high, although the surface porosity may be low, with quoted values in the range 0.3-15% (Fane and Fell, 1987). Often the porous path may be quite tortuous, the distance covered by the solvent or solute being much greater than the thickness of the membrane; the term tortuosity has been used as a measure of this property. The pores are not of a uniform size, as can be seen when viewed under the electron microscope, and are best characterised by a pore size distribution. This
Pressure-activated membrane processes 71 ctive layer Microporous Fig. 3.3. Closer examination of the membrane structure distribution can be measured by electron microscope techniques, or by combined bubble size and solvent permeability methods( Munari et al., 1985). It is claimed that thi method is capable of measuring the pore-size distribution in the thin skin. Another technique mentioned by Fane and Fell is the capillary condensation/permeability method Pore sizes may range from 1 to 100 nm. This distribution of pore size is one of the main factors preventing a sharp separation of components of almost similar size, e.g. mono and di-saccharides(see Rejection). An indirect measurement of pore size can be made by measuring the permeability of solutes, such as dextrans with a range of molecular It can be seen that in physical terms alone, there are a number of membrane structures available. The membrane also has a chemical nature, and many materials have been evaluated. It may be hydrophilic or hydrophobic in nature. Fane and Fell(1987) stated that the hydrophobic nature can be characterised by measuring its contact angle( e). The higher the contact angle the more hydrophobic is the surface. Polysulphones are generally much more hydrophobic than cellulosic membranes. There was shown to be a good correlation between flux decline and hydrophobicity, with the least hydrophobic mem brane showing the least flux loss over a period of 150 min. The surface may also be charged. All these factors will give rise to interactions between components in the feed and influence the components passing through the membrane, as well as the fouling of the membrane The physical chemistry of membranes has been described in more detail by Cheryan (1986), Gutman(1987)and Tsujita(1992) 3.5 PERMEATE RATE Two other important processing parameters are the flux or permeate rate and the power The flux is usually expressed in terms of volume per unit time per unit area (I m"). Expressed in this way it permits a ready comparison of different membrane configurations of different surface areas. It can also be expressed as a permeate velocity If energy is to be taken into account, it may be relevant to measure and maximise the flux to energy consumption ratio. Flux values may range from higher than 500 1 m h less than 51 m-2 h te: Imperial units are still sometimes used, where l gal ft-2d-=2.036- Factors affecting the flux rate are the applied pressure: the flow rate and viscosity, both of which affect turbulence, and the processing temperature. Increasing th
72 M.. Lewis temperature and inducing more turbulence increases the flux. However, the flux is only ffected by the applied pressure in the pressure-dependent region. These factors are discussed in more detail in Chapter 4 The power utilisation(W) is related to the pressure(head) developed and the mass flow rate as follows: where m= mass flow rate(kg s"), h=head developed(m)and g= acceleration du (981ms-2) ture rise. Cooling may be necessary if a constant processing temperature is required oera- This energy is largely dissipated within the fluid as heat and will result in a tempera- The membrane offers a resistance to the transfer of bo th solvent (normally water) and The permeate flux is a measure of the flow rate of solvent through the membrane, whereas the rejection describes the amount of solute which passes through(see eq. (3. 8)). From a process engineering standpoint, it is highly desirable to be able to predict the flux and rejection from the physical properties of the solution, the membrane characteristics and the hydrodynamics of the flow situation, in order to optimise the performance of the system. Membrane operations have been subject to a number of modelling processes, in order to achieve these objectives. However, before these models are discussed in more detail, it is important to consider the phenomena of concentration polarisation and uling 3.6 TRANSPORT PHENOMENA AND CONCENTRATION POLARISATION A very important consideration for pressure-driven membrane processes is that the sepa ration takes place not in the bulk of solution, but in a very small region close to the membrane, known as the boundary layer, as well as over the membrane itself. This gives rise to the phenomenon of concentration polarisation over the boundary layer. (Note that streamline flow the whole of the fluid will behave as a boundary layer. ) It is mani fested by a quick and significant reduction(2-10 fold)in flux when water is replaced by the feed solution, for example in a dynamic start oncentration polarisation occurs whenever a component is rejected by the membrane. As a result, there is an increase in the concentration of that component at the membran rface oncentration gradient over the boundary layer. Eventually dynamic equilibrium is established, where the convective flow of the component to the membrane surface equals the flow of material away from the surface, either in the permeate or back into the bulk of the solution by diffusion, due to the concentration gradient established. This increase in concentration, especially of large molecular weight components, offers a very significant additional resistance. It may also give rise to the formation of a gelled or fouling layer on the surface of the membrane(see Fig. 3. 4) Whether this occurs will depend upon the initial concentration of the component and the physical properties of the solution; it could be very important as it may affect the subsequent separation performance. Concentration polarisation itself is a reversible phenomenon; thus if the solution is then replaced by water, the original water flux should be restored. However, this rarely occurs in practice due to the occurrence of fouling
72 M.J.Lewis temperature and inducing more turbulence increases the flux. However, the flux is only affected by the applied pressure in the pressure-dependent region. These factors are discussed in more detail in Chapter 4. The power utilisation (W) is related to the pressure (head) developed and the mass flow rate as follows: W = m'hg (3.9) where m' = mass flow rate (kg s-'), h = head developed (nz) and g = acceleration due to gravity (9.81 m s-*). This energy is largely dissipated within the fluid as heat and will result in a temperature rise. Cooling may be necessary if a constant processing temperature is required. The membrane offers a resistance to the transfer of both solvent (normally water) and solute. The permeate flux is a measure of the flow rate of solvent through the membrane, whereas the rejection describes the amount of solute which passes through (see eq. (3.8)). From a process engineering standpoint, it is highly desirable to be able to predict the flux and rejection from the physical properties of the solution, the membrane characteristics and the hydrodynamics of the flow situation, in order to optimise the performance of the system. Membrane operations have been subject to a number of modelling processes, in order to achieve these objectives. However, before these models are discussed in more detail, it is important to consider the phenomena of concentration polarisation and fouling. 3.6 TRANSPORT PHENOMENA AND CONCENTRATION POLARISATION A very important consideration for pressure-driven membrane processes is that the separation takes place not in the bulk of solution, but in a very small region close to the membrane, known as the boundary layer, as well as over the membrane itself. This gives rise to the phenomenon of concentration polarisation over the boundary layer. (Note that in streamline flow the whole of the fluid will behave as a boundary layer.) It is manifested by a quick and significant reduction (2-10 fold) in flux when water is replaced by the feed solution, for example in a dynamic start. Concentration polarisation occurs whenever a component is rejected by the membrane. As a result, there is an increase in the concentration of that component at the membrane surface, together with a concentration gradient over the boundary layer. Eventually a dynamic equilibrium is established, where the convective flow of the component to the membrane surface equals the flow of material away from the surface, either in the permeate or back into the bulk of the solution by diffusion, due to the concentration gradient established. This increase in concentration, especially of large molecular weight components, offers a very significant additional resistance. It may also give rise to the formation of a gelled or fouling layer on the surface of the membrane (see Fig. 3.4). Whether this occurs will depend upon the initial concentration of the component and the physical properties of the solution; it could be very important as it may affect the subsequent separation performance. Concentration polarisation itself is a reversible phenomenon; thus if the solution is then replaced by water, the original water flux should be restored. However, this rarely occurs in practice due to the occurrence of fouling
Pressure-activated membrane processes 73 C Boundary Bulk Fig 3.4.(a)Concentra etected by a decline of flux rate at constant composition. Fouling is caused by the deposition of material on the surface of the membrane or within the pores of the membrane. Fouling is irreversible and the flux needs to be restored by cleaning. There- fore, during a concentration process, flux declines due to a combination of these two phenomena A number of mechanisms have been proposed to explain the transport of solvent and and Nichols(1992) The simplest mechanism to visualise conceptually is a simple sieving mechanism, based on size. However, for reverse osmosis, this does not explain the high rejection of salt and the permeation of water, as the molecules are about the same size. Other physice hemical factors concerned with the structure of the membrane and interaction of solvent and solutes with the membrane influence the performance. It is believed that for reverse osmosis, the phenomena are more complex than those occurring with ultrafiltration which is generally regarded as a sieving process. Therefore the models proposed for reverse osmosis and ultrafiltration are different in nature One of the most common models used for reverse osmosis is the solute diffusion model, in which the solvent flux (Jw) is influenced in the main by the pressure driving force and the solute flux (s) by diffusion. All the resistance to mass transfer takes place in the active layer, which is between 0.5 and 1.0 um in thickness. In this case the pressure driving force is the difference between the applied pressure and the osmotic pressure. The transport of solvent and solute are not connected. It is assumed that the solute dissolves the skin instantaneously and then passes through the pores of membrane by diffusion, whereas the flow of solvent(which also dissolves) is influenced by the pressure differential. The relationship between concentration dissolved in the membrane at its surface and that in the solution depends upon the partition coefficient for that component
II II II II II 11 I1 I1
74 M.J. Lewis Thus the diffusivity and membrane thickness and its partition coefficient all influence the transport of solute( Convective flow of solute is ignored at high solute rejections. According to this model, increasing the pressure will have a preferential effect on solvent flow, thereby reducing solute concentration in the permeate and rejection. It also predicts that increasing the solute concentration in the feed preferentially increases solute transport and increases the concentration in the permeate, thereby decreasing rejection. Increasing the temperature increases both solvent flux and solute flux by about the same amounts, thereby leaving the rejection unchanged. A 1C change in temperature changes the flux(solute and solvent) by approximately 3% Although many of these trends are observed with simple solutions, the theory does not account for all the observed facts for multi-component solutions, whereby the presence of one component increases the permeability of other components. Glover(1985) pointed out that this theory does not describe the behaviour of complex systems precisely, but it gives a background of understanding. In multicomponent systems, differences in the partition coefficients could explain the difference in permeabilities between different A second proposed model is the preferential adsorption, capillary flow model, which predicts that the component concentrated preferentially in the permeate will be that component which is adsorbed most strongly on the membrane surface. The component that is preferentially adsorbed onto the membrane surface provides a thin layer of tha component adjacent to the surface. This thin layer then moves through the pores of the membrane by capillary flow, under the influence of a pressure gradient, and in this way thus preferentially transported through the membrane, For membranes of a hydrophilic nature, the component preferentially absorbed and transported is water. Thus it is postulated that there is a thin film of water adjacent to the membrane surface. This theory also explains the low rejections and sometimes negative rejections for highly polar organic solutes, found with cellulose acetate membranes, due to their preferential adsorption on the surface. Other situations, where rejection of organic solutes decreases with increasing pressure, are explained by their adsorption onto the more hydrophobic regions of the membrane A third model is based upon the wetted surface mechanism, whereby water adsorbs onto the surface of the membrane by hydrogen bonding. It is postulated that these clusters f water prevent solute entering the pores and that the water passes through the mem- brane from one adsorbed site to the next. The energy requirements for water migration are much less than salt migration, thereby promoting separation of the salt and water. All these models are based on knowing the transport mechanisms involved. The physical hemistry of a wide variety of membrane materials, including permeability data, diffusion data and sorption data which are required for the models described earlier, been reviewed by Tsujita(1992). Also reviewed are transport properties related equilibrium thermodynamics with uncharged and charged membranes Other models are based upon irreversible thermodynamics, where the driving force for transport olvent and solute is expressed in terms of differences in their chemica potential over the membrane. However, the fluxes for solvent and solute are coupled and the flux for each component is influenced by the chemical potential difference for both components. With such models, the exact mechanisms are not known, but the
74 M.J.Lewis Thus the diffusivity and membrane thickness and its partition coefficient all influence the transport of solute. (Convective flow of solute is ignored at high solute rejections.) According to this model, increasing the pressure will have a preferential effect on solvent flow, thereby reducing solute concentration in the permeate and increasing rejection. It also predicts that increasing the solute concentration in the feed preferentially increases solute transport and increases the concentration in the permeate, thereby decreasing rejection. Increasing the temperature increases both solvent flux and solute flux by about the same amounts, thereby leaving the rejection unchanged. A 1°C change in temperature changes the flux (solute and solvent) by approximately 3%. Although many of these trends are observed with simple solutions, the theory does not account for all the observed facts for multi-component solutions, whereby the presence of one component increases the permeability of other components. Glover (1985) pointed out that this theory does not describe the behaviour of complex systems precisely, but it gives a background of understanding. In multicomponent systems, differences in the partition coefficients could explain the difference in permeabilities between different components. A second proposed model is the preferential adsorption, capillary flow model, which predicts that the component concentrated preferentially in the permeate will be that component which is adsorbed most strongly on the membrane surface. The component that is preferentially adsorbed onto the membrane surface provides a thin layer of that component adjacent to the surface. This thin layer then moves through the pores of the membrane by capillary flow, under the influence of a pressure gradient, and in this way is thus preferentially transported through the membrane. For membranes of a hydrophilic nature, the component preferentially absorbed and transported is water. Thus it is postulated that there is a thin film of water adjacent to the membrane surface. This theory also explains the low rejections and sometimes negative rejections for highly polar organic solutes, found with cellulose acetate membranes, due to their preferential adsorption on the surface. Other situations, where rejection of organic solutes decreases with increasing pressure, are explained by their adsorption onto the more hydrophobic regions of the membrane. A third model is based upon the wetted surface mechanism, whereby water adsorbs onto the surface of the membrane by hydrogen bonding. It is postulated that these clusters of water prevent solute entering the pores and that the water passes through the membrane from one adsorbed site to the next. The energy requirements for water migration are much less than salt migration, thereby promoting separation of the salt and water. All these models are based on knowing the transport mechanisms involved. The physical chemistry of a wide variety of membrane materials, including permeability data, diffusion data and sorption data which are required for the models described earlier, have been reviewed by Tsujita (1992). Also reviewed are transport properties related to nonequilibrium thermodynamics with uncharged and charged membranes. Other models are based upon irreversible thermodynamics, where the driving force for transport of solvent and solute is expressed in terms of differences in their chemical potential over the membrane. However, the fluxes for solvent and solute are coupled and the flux for each component is influenced by the chemical potential difference for both components. With such models, the exact mechanisms are not known, but the