Introduction It is well understood today that our society has to face the challenge of modifying the traditional industr...
77 downloads
1740 Views
21MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Introduction It is well understood today that our society has to face the challenge of modifying the traditional industrial growth to a sustainable growth, if we want to keep developing for generations. In principle, adverse environmental impact can be notably reduced by optimizing the existing activities. Present design methods are effectively devoted, in most cases, to managing wastes better, to introducing methods for pollution abatement, and to realizing cleaner processes for cleaner products. Nevertheless, these positive effects are expected to be offset by the on going growth. Traditional environmental management and pollution prevention will not suffice in the long run; newer approaches, which are radically innovative and integrated, are needed. The chemical and engineering community is already paying significant attention to the request for technologies that would lead us to the goal of technological sustainability. A promising example with a lot of interest by process engineers is the strategy of process intensification. It consists of innovative equipment, design and process development methods that are expected to bring substantial improvements in chemical and any other manufacturing and processing, such as decreasing production costs, equipment size, energy consumption, waste generation, and improving remote control, information fluxes and process flexibility. How to implement this strategy is, however, not obvious. An interesting and important case is the continuous growth of modem membrane engineering whose basic aspects satisfy the requirements of process intensification. Membrane operations, with their intrinsic characteristics of efficiency and operational simplicity, high selectivity and permeability for the transport of specific components, compatibility between different membrane operations in integrated systems, low energetic requirement, good stability under operating conditions and environment compatibility, easy control and scale-up, and large operational flexibility, represent an interesting answer for the rationalization of chemical productions. Many membrane operations are practically based on the same hardware (materials), only differing in their software (methods). The traditional
2 Introduction membrane separation operations (reverse osmosis, micro-,ultra- and nanofiltration, electrodialysis, pervaporation etc.), already largely used in many different applications, are today conducted with new membrane systems such as catalytic membrane reactors and membrane contactors. At present, redesigning important industrial production cycles by combining various membrane operations suitable for separation and conversion units, thus realizing highly integrated membrane processes, is an attractive opportunity because of the synergic effects that can be attained. Interesting examples already exist in water desalination plants, in downstream processing of biological and biotechnological productions, etc. This strategy starts to penetrate also in new areas such as the petrochemical industry, the electronic industry. Limits however exist to the traditional membrane operations, as p.e. the level of feed concentrations which can be reached in a RO system or on the recovery factors in the same RO desalination units. New unit operations moreover might be invented and/or developed in same cases which better satisfy the requirement of the process intensification strategy. Among other new unit operations involving membranes, membrane contactors are expected to play a decisive role in this scenario. The key concept is to use a solid, microporous, hydrophobic (or hydrophilic) polymeric matrix in order to create an interface for mass transfer and/or reaction between two phases: large exchange area and independent fluid dynamics allow an easily controlled operation. These membrane systems, in the form generally of low cost hollow fibres, provide a high interfacial area significantly greater than most traditional absorbers between two phases to achieve high overall rates of mass transfer. In addition, whereas the design of the conventional devices is restricted by limitations in the relative flows of the fluid streams, membrane contactors give an active area, which is independent of the liquid fluid dynamics. Membrane crystallizers, membrane emulsifiers, membrane strippers and scrubbers, membrane distillation systems, membrane extractors, etc. can be designed and integrated in the production lines together with the other existing membranes operations for advanced
Introduction 3 molecular separation, and chemical transformations conducted using selective membranes and membrane reactors, overcoming existing limits of the more traditional membrane processes (for example the osmotic effect of concentration by reverse osmosis). It is amazing to note that, although the above mentioned systems are quite "young", the potentialities of membrane systems have been already discovered and suggested at the beginning of the XX Century [ 1]. In Table 1 are summarized the most traditional membrane contactors developped in these last years.
Table 1. Membrane contactors systems Membrane strippers Membrane scrubbers Membrane extractors Supported liquid membranes Membrane distillation Osmotic distillation Membrane emulsifiers Phase transfer catalysis
A first example might be considered the supported liquid membranes where the microporous hydrophobic membranes act as support to the liquid phase containing appropriate carriers for the selective transport of the species dissolved in the solutions facing the membrane; other most recent examples are membrane distillation contactors. In all the operations mentioned the role of the membranes is crucial; they not only serve as an ideal contactors between the two phases they separates, but contribute more to the efficiency of the overall processes.
4 Introduction
The relative simplicity of the hardware of these systems is combined with a certain complexity on the contrary of their software. A multidisciplinary background is certainly necessary for a deep basic knowledge of the membrane contactors properties in their various configurations and in their various applications. Transport phenomena in porous media, interphacial phenomena in liquid-.liquid, gasliquid, in gas-gas phases, basic properties of polymeric materials, as also of colloids and gels, are necessary and must be well integrated with a knowledge of fundamentals of chemistry as of the thermodynamics and kinetic aspects. In this book we will present the basic aspects of the various membrane contactors already existing, and their applications. The overall potentialities of these new technologies will be also temptatively discussed.
References
[1] P. A. Kober. Pervaporation, perstillation and percrystallization., Contribution read at the meeting of the Soc. Expt. Biol. Med., Feb. 21 (1917)
Chapter I. Basic principles of membrane contactors
1. Generalities on membrane contactors operations The term "membrane contactor" is used to identify membrane systems that are employed to "keep in contact" two phases. On the contrary of the more "traditional" idea of membranes as media for performing separations thanks to their selectivity, membrane contactors do not offer any selectivity for a particular species with respect to another, but simply act as a barrier between the phases involved, by allowing their contact in correspondence of a well defined interfacial area [ 1-9]. Being the two phases separate by the membrane, there is no mix of them and dispersion phenomena do not occur. The species are transferred from one phase to the other by only diffusion. The membranes are usually microporous and symmetric and can be both hydrophobic and hydrophilic. In the case of hydrophobic materials, the membrane can be wetted by non polar phases (e.g., non polar organics) or filled by gas, while the aqueous/polar phase can not penetrate into the pores (see Figure 1).
6 Chapter 1
Figure 1. Interface between a non polar/gas phase and a polar phase in a hydrophobic membrane.
In this way, it is possible to define the area of contact in correspondence of the pores mouths. In order to avoid the mixing of the two phases, it is important to carefully control the operating pressures. First of all, the pressure of the aqueous/polar phase has to be equal to or higher than the pressure of the wetting/filling phase. This permits to eliminate any possibility of dispersion as drops of one phase into the other phase. Moreover, the interfacial area can be established at the pore mouth only if the penetration of the aqueous/polar phase into the membrane pores is prevented. The hydrophobicity of the material is not, in fact, a warranty for keeping the pores aqueous/polar phase-free. If a critical value of pressure, called generally
breakthrough pressure, is exceed, the membrane loses its hydrophobic character and the aqueous/polar phase starts to wet it [10-12]. For a particular material the breakthrough
Basic Principles of Membrane Contactors 7 pressure depends on the pore radius, surface/interfacial tension, contact angle between the
membrane and the fluid, and can be calculated by using the Laplace's equation (see Chapter 2). In figure 1, as well as in all the other figures, for simplicity, straight pores are considered for symmetric membranes. In practice, membrane pores have an un-defined shape, mainly related to the tortuosity of the membrane along its thickness. With asymmetric membranes in which the pore size reduces along the thickness, it is possible to keep in non-dispersive contact the two phases also by working, at the bigger pores side, at pressures higher than the breakthrough value. In fact, being the breakthrough pressure inversely dependent on the pore size, there is a partial wetting of the membrane for the bigger pores, whereas the smaller pores continues to be aqueous/polar phase free. The interfacial area is now established within the pores (see Figure 2).
Figure 2. Interface between a non polar/gas phase and a polar phase in a partially wetted asymmetric membrane.
8 Chapter 1
The hydrophobicity of the membrane can also vary because of the interactions with the phases involved that lead to changes in the membrane structure and morphology. This last aspect can be minimized by using composite membranes with a non-porous thin layer coated on the microporous surface that prevents the penetration of the aqueous/polar phase (Figure 3) [13-17].
Figure 3. Composite membrane with a dense thin layer coated on the microporous surface. The non-porous thin layer allows also to enlarge the range of the operating pressures, but, in order to do not increase too much the resistance to the mass transport, it has to be highly permeable for the trasferred species. The membrane wetting can be partial or complete; in the first case the two phases are in contact somewhere in the membrane pores, whereas for complete wetting the two phases are mixed and the membrane contactor loses its function.
Basic Principles of Membrane Contactors 9
When hydrophilic materials are used, the aqueous/polar phase wets the membrane pores while the non polar/gas phase is blocked at the pore mouth. In this configuration the interface is established at the pore mouth at the non polar/gas phase side and the dispersion as drops between the phases is avoided by working with pressures of the non polar/gas phase equal to or higher than the wetting phase pressure (Figure 4).
Figure 4. Interface between a non polar/gas phase and a polar phase in a hydrophilic membrane.
As for the hydrophobic membranes, the interface is kept at the pore mouth until the breakthrough pressure is not exceed. As reported by Sirkar [10], two liquid phases can be in contact also by means of a composite hydrophobic-hydrophilic membrane where the polar phase wets the hydrophilic
10 Chapter 1 part and the non polar phase enters the hydrophobic one (Figure 5). The interface is now located at the hydrophobic-hydrophilic interface and can be well defined by operating with one of the two phases at higher pressure, taking care in not exceeding the critical pressure value.
Figure 5. Interface between a non polar/gas phase and a polar phase in a composite hydrophilic hydrophobic membrane. Until now, we did not consider any reaction between the phases involved. When the species present into the two phases react, an interface where the reaction occurs can be formed and it can correspond with the phase interface or can be located into one phase.
Basic Principles of Membrane Contactors 11
Table 1 summarizes the main characteristics of the membranes used in membrane contactors. A more detailed analysis on the membrane materials is reported in Chapter 2.
Table 1. Membranes used in membrane contactors i
Microporous membranes Hydrophobic Hydrophilic Symmetric Asymmetric Composite (hydrophilic-hydrophobic or dense-microporous)
All operations that are based on the mass transport between two contacting phases can be in principle carried out by membrane contactors. For example, liquid-liquid extraction, the removal of gases/volatiles dissolved in a liquid phase by stripping with a gaseous stream or the addition of a gas/volatile contained in a gaseous stream into a liquid. In the following, the different types of membrane contactors that can be used depending on the specific application are described. Table 2 reports about them in terms of phases involved and driving force.
12 Chapter 1 Table 2. Membrane contactors systems Membrane Supported strippers/scrubbers/ liquid extractors
Membrane Osmotic Membrane Phase transfer distillation distillation emulsifiers catalysis
membranes
Phase 1 Gas/Liquid
Gas/Liquid
Liquid
Liquid
Liquid
Liquid
Phase 2 Liquid
Gas/Liquid
Liquid
Liquid
Liquid
Liquid
Driving Concentration force gradient
Partial Partial pressure/conc, pressure ~radient ~radient
Partial pressure gradient
Pressure gradient
Concentration gradient
In all different types of membrane contactors the species to be transferred encounters several resistances during its passage from one phase to another. In general, these resistances are offered by the phases and the membrane. Depending on the particular system, the mass transfer can be controlled by the resistance offered by the phase/phases, by the membrane or by both. Although a more detailed analysis of the equations that regulate the mass transfer will be furnished in next Chapters, a discussion on the resistances involved and general expressions for calculating the mass flux are also shortly reported in the following.
1.1. Membrane strippers/scrubbers and membrane extractors
In both membrane strippers and scrubbers a liquid is in contact with a gas, the difference between the two systems being the direction in which the species are transferred: from the liquid to the gas and viceversa, respectively. These systems are used for the transport of
Basic Principles of Membrane Contactors 13
volatile species contained in the phases. A generic species i moves from a phase to the other due to a partial pressure gradient. In the case of streams containing different volatile species, a simultaneous transfer can be achieved. For example, dissolved oxygen can be removed from water by stripping with a CO2 stream while, due to the partial pressure gradient, the C02 diffuses into the water. The membranes are usually hydrophobic and gas-filled, because the volatile species have higher effective diffusion in gas than in liquid and, thus, the resistance offered by the membrane is strongly reduced, with a consequent improvement of the mass transport. These systems can be considered as alternative to traditional packed and bubble columns.
Figure 6. Hydrophobic membrane contactors as strippers.
Figure 7. Hydrophobic membrane contactors as scrubbers.
14 Chapter 1
Membrane extractors can be used for carrying out liquid-liquid extractions, usually conducted in columns, mixer-settler or centrifugal devices. The driving force is due to a difference of concentration and the membranes can be both hydrophobic and hydrophilic, depending on the affinity of the species to be transferred with the streams involved. The choice is dictate by the need to reduce the membrane resistance. For example, if the species has higher affinity with the polar phase, then the membrane will be hydrophilic with the pores filled with the polar stream. If there is higher affinity with the non-polar phase, the membrane will be hydrophobic. The possibility to simultaneously transfer different solutes is valid also for these systems. The figure below refers to a concentration of the species i higher in phase
Figure 8. Transfer of the species i from the phase 1 towards the phase 2.
Basic Principles of Membrane Contactors 15
In membrane strippers/scrubbers/extractors, a generic species contained in phase 1 that moves towards phase 2 encounters a first resistance in the phase 1-self close to the membrane surface, then the resistance of the membrane and, finally, the resistance in phase 2 close to the other membrane side. The presence of these resistances leads to a concentration profile for the species, as depicted in Figure 9, that determines the driving force available for the transport.
Figure 9. Concentration profile for a species that moves from the phase 1 towards the phase 2.
A general expression used to calculate the flux of the species is the following [5]:
J = K'(CI-Ce)
(1)
with
X =f(kl, km, k2)
(2)
where:
J, flux; C1,C2, concentrations in the two phases;
16 Chapter 1
K, overall mass transfer coefficient," kl, k2, phases mass transfer coefficients," kin, membrane mass transfer coefficient.
1.2. Supported liquid membranes In supported liquid membranes the micropores of the membrane are usually filled by an organic phase and the membrane is located between two aqueous phases. One of the aqueous phase is the feed to be treated, the other representing the stripping phase. The removal of the species from the feed to the stripping phase occurs by diffusion through the organic phase and the stripping one, the concentration difference being the driving force (Figure 10).
Figure 10. Supported liquid membrane with aqueous feed and strip and organic phase into the micropores.
The effectiveness of the process is mainly depending on the affinity between the species and the organic phase. In order to increase the mass transport rate, a facilitated transport can be achieved by introducing a carrier in the organic phase. The carrier reversibly complexes
Basic Principles of Membrane Contactors 17
with the species and the carrier-species complex moves from the feed side to the strip side. Once at the strip side, being the reaction reversible, the carrier releases the species that is removed (Figure 11) [ 18-20].
Figure 11. Transfer of the species i by means of a carrier. In this way, the species leaves the feed stream both as uncomplexed, by permeating through the organic layer, and as a complex, by means of the carrier (Figure 12).
Figure 12. Permeation of the species i both as free and as a complex.
18 Chapter 1
The transport of the species by means of the carrier is faster than the simple diffusion of the species into the organic phase. The transport rate is, thus, enhanced and, if the carrier is high specific for the species of interest, high selectivities can be reached. For this configuration the membranes used are hydrophobic and the interfacial areas are established at the pore mouth of the membrane (on both sides) by properly acting on the aqueous pressures. In order to keep the membrane pores organic-filled, it is essential that the organic phase/carrier is immiscible with the aqueous streams. The properties of the immobilized solution (volatility, viscosity, degree of miscibility with the feed/strip phase) and of the carrier (stability, selectivity) are, in fact, at the basis of the performance of these systems. The membrane micropores can be also filled by an aqueous phase in which the carrier is dissolved; in this case, the membrane is hydrophilic and separates two organic phases immiscible with the aqueous one (Figure 13).
Figure 13. Supported liquid membrane with organic feed and strip and aqueous phase into the micropores.
Basic Principles of Membrane Contactors 19
Although most of the applications of supported liquid membranes refer to liquid phases [21-25], gaseous phases can be also treated by this type of membrane contactor [26-27].
In supported liquid membranes the membrane micropores are usually liquid-filled and the mass transfer resistance offered by the membrane mainly matches with the mass transfer resistance offered by the liquid. The two phases also contribute to the overall resistance to the mass transport and the general expression for the mass flux is the same reported above (eqs. 1 and 2). The mass flux through the membrane will be now dependent on the diffusion coefficient of the species in the liquid and, for a liquid containing a carrier, on the diffusion coefficient of the complex species-carrier in the liquid.
A typical expression describing this flux is [18]:
J = km'AC + kmcomplex "Afcomplex
(3)
where:
J, flux through the membrane; kmcomp+ex,membrane mass transfer coefficient for the complex; AC, difference of concentration of the species across the membrane; ACcomplex,difference of concentration of the complex across the membrane.
1.3. Membrane distillation
Membrane distillation is the only example of membrane contactor where the driving force is related to a temperature gradient across the membrane. The membranes used are
20 Chapter 1
hydrophobic and the feed streams are aqueous solutions. The stripping can be performed by using an aqueous stream at the permeate side (direct-contact membrane distillation) or by applying vacuum or by sending a strip gas. The first type of stripping has been the mostly applied. In this case, the hydrophobic membrane separates the two aqueous solutions (feed and strip). By imposing a temperature difference across the membrane (the feed solution is heated and the strip solution is cooled), a partial pressure gradient is created from the hot to the cold side. Due to this gradient, the water molecules evaporated at the warm side of the membrane migrate through the membrane micropores and, then, condensate at the permeate side (Figure 14) [28-29]. Membrane distillation can be effectively used for producing ultrapure water or for concentrating aqueous solutions and can be view as an alternative process to traditional distillation columns.
Figure 14. Scheme of the membrane distillation.
Basic Principles of Membrane Contactors 21
In membrane distillation the mass transport is strictly related to the difference of temperature imposed across the membrane thickness. The resistances offered by the phases and the membrane create now a temperature profile (see Figure 15) that determines the partial pressure gradient available for the transport. The values of the partial pressures at the membrane interfaces are, in fact, dependent on the temperatures values at the interfaces.
Figure 15. Temperature profile in membrane distillation. The equations that describe the membrane distillation operations are based both on mass and energy balances. The water vapour mass flux through the micropores is calculated by:
J = km'(Pz-P2)
(4)
where:
J, flux through the membrane; PI,P2, water vapour partial pressures at the membrane interfaces.
22 Chapter 1
The membrane mass transfer coefficient for flux of vapour/gas molecules through micropores is usually derived as function of the Knudsen and molecular mass transfer coefficients.
Referring to the heat flux, at steady-state it can be written as [30]:
Q = Hv(Th-Tc)
(5)
where:
Q, heat flux; Th, To, temperatures at the hot and cold side," H, membrane heat transfer coefficient," r,, temperature polarization coefficient
The temperature polarization coefficient is due to the resistances offered by the boundary layers adjacent to the membrane surfaces and is defined as:
r = (Thm-Tcm)/(Th-Tc)
(6)
where:
Thin,, Tcm, temperatures at the membrane interfaces. Usually, iterative procedures are implemented for solving the above equations.
1.4. Osmotic distillation
Osmotic distillation performs the same work of the membrane distillation but uses a different method for creating the partial pressure gradient. In this case, the operation is carried out at ambient temperature and the gradient is achieved by sending at the strip side an
Basic Principles of Membrane Contactors 23
aqueous solution containing non-volatiles compounds (usually salts, as CaC12). The difference in solute concentrations between the solution to be treated and the strip side leads to a vapor pressure difference which causes the transport of the water vapor molecules (Figure 16). The membranes used are hydrophobic. The possibility to concentrate a solution at ambient temperature is quite important for streams containing labile or easily denaturated compounds [31].
Figure 16. Scheme of the osmotic distillation. Working at ambient temperature, no heat flux is usually considered and the water vapour mass flux through the micropores can be calculated by the same equations derived for membrane distillation (equ. 3 and 4).
Osmotic distillation can suffer from concentration
polarization phenomena that consist in the increase of the concentration of the species contained in the aqueous solution at the membrane surface with respect to their bulk
24 Chapter 1
concentration. The phenomenon is usually described by means of a concentration polarization coefficient, CPC, defined as the ratio between the concentration of the species at the membrane surface and its concentration in the bulk:
CPC
= C,,,/Cb
(7)
This phenomenon also occurs in membrane distillation, but its effect on the water vapour flux through the membrane can be neglected, being the driving force directly dependent on the difference of temperature. Osmotic distillation is also applied for the removal of volatile compounds (e.g., alcohols) from water streams. In this case, the aqueous strip can be pure water [31]. The resistances offered by the two phases and the membrane lead to a concentration profile that determines the driving force for the transport.
1.5. Membrane crystallizers
Membrane crystallizers represent a particular application of membrane and osmotic distillation. These systems, in fact, are based on the same principles that regulate the above operations but are specifically mentioned here because the feed solutions they treat are close to the saturation values and usually are the results of previous treatments. The aim of membrane crystallizers is to perform the crystallization of the solutes of interest by removing water from the almost saturated feeds. An important task of the process is to avoid the
Basic Principles of Membrane Contactors 25
formation and precipitation of crystals on the membrane surface that could cause pore blocking. This type of membrane contactor is altemative to conventional methods used for producing crystals, such as evaporation.
1.6. Membrane emulsifiers
Membrane emulsifiers employ both hydrophobic and hydrophilic membranes for creating microemulsions. These systems are not used to keep in contact the two phases, but to force one phase into the other. We report here about them as a type of membrane contactors because the membrane properties required for carrying out this operation are similar to those needed in membrane contactors processes. In membrane emulsifiers, one side of the membrane is in contact with the liquid phase emulsified ("dispersed phase") while the other side is in contact with the liquid phase that contains the emulsified phase ("continuous phase"). The dispersion phase is forced, by applying a pressure, to permeate through membrane into the continuous phase where it is emulsified (Figure 17).
26 Chapter 1
Figure 17. Emulsion formation by means of a microporous membrane. The driving force is, thus, related to the difference of pressure between the two phases. During the process it is important that the membrane surface is not wetted by the dispersed phase and the choice of the membrane strongly depends on this aspect. For example, for oil/water emulsions, the membrane used is hydrophilic, whereas for water/oil emulsions is hydrophobic [32-34]. In membrane emulsifiers the flux is directly proportional to the difference of pressure between the two phases and mainly depends on the membrane resistance and the resistance offered by the continuous phase. A generic expression is:
J = K'(P1-P2)
(8)
with K= f(km, k2)
(9)
where:
PI, P2, pressures of the dispersed and continuous phase.
Basic Principles of Membrane Contactors 27 1.7. Phase transfer catalysis Membrane contactors can be also used to carry out catalytic reactions. In this case, the membrane, that can be both hydrophilic and hydrophobic, is catalytically active (e.g. enzymes are immobilized into its micropores). When two liquid phases (aqueous/organic) are kept in contact, a compound of one phase can diffuse to the catalytic sites where reacts and the formed products can be stripped in the other phase, without mixing of the two streams (Figure 18). This type of system is an example of the so-called "phase transfer catalysis" [35]. The process is regulated by a difference of concentration, for both reactants and products.
Figure 18. Schematic representation of the phase transfer catalysis. The concept can be applied also to systems in which both streams contain reactants. Now both reactants have to diffuse towards the catalytic sites and products can move towards both streams, the degree of affinity between products and streams controlling their distribution
28 Chapter 1 (Figure 19). These types of membrane contactors can be, thus, effective also for three-phase reaction where a gas and a liquid come in contact on the catalytic membrane (solid).
Figure 19. Separate feed of reactants and products diffusion towards the two phases. Phase transfer catalysis couples the transport of the species with the reaction. The flux of reactants towards the catalytic sites as well as the flux of products from the reaction zone towards the phases is always depending on the resistances offered by the phases and the membrane. For a product that is formed in themembrane pores on the catalytic sites and that moves towards one of the phases (e.g., phase 1), the flux can be described as:
Jp = K (Cpr - C m )
(10)
w i t h X = f(kmc, kl)
(11)
where:
Basic Principles of Membrane Contactors 29
Jp, flux of the product P; Cec, Cm, concentrations of P in the catalytic membrane pores and in the phase I," kmc, catalytic membrane mass transfer coefficient.
For the case of a phase 1 containing the reactant that moves towards a membrane with a catalytic surface and a phase 2 where the formed products are recovered (see Figure 20), the equations that describe the fluxes are:
JR = K'(CR1-CRm)
(12)
with K = f(kl)
(13)
where:
Jl~ flux of the reactant R; CRI, GRin, concentrations of R in the phase 1 and at the catalytic membrane surface.
Jp = K'(Cpm-Cp2)
(14)
with K = f(km, k2)
(15)
where:
Jp, flux of the product P; CPm,CP2, concentrations of P at the catalytic membrane surface and in the phase 2;
30 Chapter 1
Figure 20. Concentration profiles of the reactant contained in the phase 1and of the products for a membrane with a catalytic surface and a phase 2 with high affinity for products. The different examples of membrane contactors described, with the exception of membrane emulsifiers, can be further grouped into three main classes: -
Carrier-free, that include all membrane contactors working without any carrier;
-
Carrier-charged, that include membrane contactors where carriers are used to facilitate the transport ;
-
Reactors, that include membrane contactors where a reaction occurs within the membrane pores.
Table 3 shows the general equations describing the mass flux through the membrane for the different classes.
Basic Principles of Membrane Contactors 31 Table 3. General equations describing the mass transport through the membrane for the different classes of membrane contactors Membrane contactors
Equation for the mass flux
Carrier-free
J = km AC or J = km AP
Carrier-charged
J = km AC + km complex ACcomplex
Reactors
J = kmc AC
2. Advantages and disadvantages of membrane eontactors Membrane
contactors
have different interesting properties that make them more
advantageous with respect to traditional operations. For example, it is possible to work with a well defined and constant interfacial area. This means that the exchange area is known and all the device works with the same efficiency. The constance of the interfacial area with changes in the operating conditions or fluid properties leads also to a higher efficiency with respect to conventional units. Moreover, a higher interfacial area can be provided in a small volume, that corresponds to higher compactness, and, thus, to reduced size and weight. The typical interfacial area per unit of volume of membrane contactors varies between 1500-3000 m2/m 3, whereas for conventional contactors this ratio is in the range of 100- 800 m2/m 3 [36]. It is important to point out that the higher interfacial area is the major responsible of the enhanced efficiency in membrane contactors with respect to traditional devices. As a matter of fact, the mass transfer coefficients reachable in membrane contactor are usually the same or sligthly lower than those of conventional systems. Ding et al. [37] compared the ka (with a representing the interfacial area) achievable in membrane contactors with those related to a
32 Chapter 1 high-efficiency rotating column and a conventional extractor. From the above comparison it resulted a ka value of 0.053 s -1 for membrane contactors versus 0.0007 and 0.00005 s-1 for the rotating column and the conventional extractors, respectively. Another positive aspect is that there is no dispersion between the two phases and, thus, no need to separate the two phases downstream the process and no need to work with fluids of different densities. Furthermore, being the two phases separate by the membrane, phenomena as flooding, loading, foaming are avoided, leading to a higher flexibility in changing the operating flowrates that can be varied, also independently, in a wider range of values. In gas-liquid transfer, the size of the gas bubbles introduced into the liquid bulk is depending on the micropores size. By ensuring a minimal distance between adjacent pores, any possible coalescence is avoided. This implies that very small bubbles of gas reach the liquid and, then, a better dispersion is achieved. The same concept is valid for the microemulsions production. In membrane distillation, ultrapure water and high recovery factors up to crystals production can be obtained at relatively low temperatures with respect to the classical distillation (the temperature of the feed stream is usually of the order of 35~ and the temperature of the strip phase is in the range of 15-25~
Moreover, azeotropic mixtures, hardly separated by
distillation column, can now be treated. Solutions containing compounds that can deteriorates with temperatures (pharmaceutical compounds, vitamins, aromes) can be processed by osmotic distillation. By carrying out a reaction with membrane contactors, it is possible to reduce the mass transport resistances of the reactants towards the catalyst sites (the phases are in direct contact with the catalytic zone and the reactants do not have to diffuse through the
Basic Principles of Membrane Contactors 33
other phase before reaching the catalyst, as usually happens in multiphase reaction systems). The system can be also used to simultaneously separate the products. In this way, the conversion of reversible reactions can be increased and the further reactions of the desired products are avoided. As all membrane operations, membrane contactors are flexible, easy in the scale-up and control, modular in design, do not present any moving part and are generally characterized by low pressure drops. Unfortunately, these systems offer some disadvantages too! First of all, the presence of the membrane is cause of a further resistance to the mass transport. However, this resistance can significantly be reduced by operating properly. This aspect will be treated in more details in next Chapters. Other drawbacks related to the membrane are its limited life-time and the risk of fouling, that sometimes implies pre-treatments before the process. The limited operating pressures allowed, based on the breakthrough value, is another weak point of these systems. Specifically for the supported liquid membranes, the stability of the solvent and the lifetime and selectivity of the carrier, represent hard problems to solve. Finally, as it will be discussed in next Chapters, sometimes during the operations channeling and bypassing can not be completely avoided, with a consequent reduction of the mass transport efficiency. In Table 4, for each type of membrane contactor is reported the corresponding conventional unit operation.
34 Chapter 1
Table 4. Membrane contactors systems and corresponding conventional operations Membrane contactors
Conventional operations
Membrane strippers/scrubbers
Packed and bubble columns
Membrane extractors
Packed columns, mixer-settler, centrifugal devices
Supported liquid membranes
Packed and bubble columns, mixer-settler, centrifugal devices
Membrane distillation and osmotic distillation Distillation columns Membrane crystallizers
Evaporators
Membrane emulsifiers
High pressure homogenizers
Phase transfer catalysis
Chemical reactors
Table 5 summarizes the main advantages and disadvantages of membrane contactors.
Basic Principles of Membrane Contactors 35
Table 5. Positive and negative aspects of membrane contactors Positive
Negative
Well defined and constant interfacial area
Further resistance offered by the membrane
High interfacial area in small volumes
Membrane limited life-time
Reduced size and weight
Membrane fouling
No dispersion between phases
Pre-treatments before the process
No need of phase separation downstream
Limited operating pressures, based on the breakthrough value
No need to work with fluids of different densities
Channeling and bypassing of fluids
No flooding, loading, foaming
Limited stability of the solvent and of the lifetime and selectivity of the carrier in supported liquid membranes
Wide range of operating flow-rates Flow-rates can be varied independently No coalescence phenomena Controlled and very small size of the bubbles and the emulsions produced Lower operating temperatures with respect to distillation processes Azeotropic mixtures can be easier treated than in conventional units Reaction and separation carry out simultaneously Flexible, easy in scale-up, control and automatization Modular design and no moving parts
36 Chapter 1 References [ 1] Z. Qi and E.L. Cussler. Microporous hollow fibers for gas absorption. I. Mass transfer in the liquid, J. Membrane Sci., 23 (1985) 321-332 [2] Z. Qi and E.L. Cussler. Microporous hollow fibers for gas absorption. II. Mass transfer across the membrane, J. Membrane Sci., 23 (1985) 333-345 [3] E.L. Cussler. Hollow fiber contactors, in J.G. Crespo and K.W. Boddeker (Eds.), Membrane Processes in Separation and Purification, Kluwer Academic Publishers, The Netherlands (1994) 375-394 [4] A. Kiani, R.R. Bhave and K.K. Sirkar. Solvent extraction with immobilized interfaces in a microporous hydrophobic membrane. J. Membrane Sci., 20 (1984) 125-145 [5] A. Gabelman and S.T. Hwang. Hollow fiber membrane contactors. J. Membrane Sci., 159 (1999) 61-106 [6] B.W. Reed, M.J. Semmens and E.L. Cussler. Membrane Contactors, in: R.D. Noble and S.A. Stern (Eds.), Membrane Separation Technology: Principles and Applications, Elsevier, Amsterdam (1995) 467 [7] E. Drioli and A. Criscuoli. Microporous inorganic and polymeric membranes as catalytic reactors and membrane contactors, in: Nick Kanellopoulos (Ed.), Membrane Science and Technology Series, 6 entitled: "Recent advances in gas separation by microporous membranes", Elsevier, Amsterdam (2000) 497-510 [8] A. Criscuoli, E. Curcio and E. Drioli, Polymeric membrane contactors, in: S.G. Pandalai (Ed.), Recent research developments in applied polymer science, Transworld Research Network Publication by Research Signpost, ISBN: 81-7895-102-9, Kerala, 37/66 (2), 7 (2003) 1-21
Basic Principles of Membrane Contactors 37 [9] E. Drioli, A. Criscuoli and E. Curcio. Membrane contactors and catalytic membrane reactors in process intensification. Chem. Eng. Technol., Vol. 26 N. 9 (2003) 975-981 [10]R. Prasad and K.K. Sirkar. Membrane-based solvent extraction, in: W.S.W. Ho and K.K. Sirkar (Eds.), Membrane Handbook, Chapman and Hall, New York (1992) 727-763 [11]H. Kreulen, C.A. Smolders, G.F. Versteeg and W.P.M. van Swaaij. Determination of mass transfer rates in wetted and non-wetted microporous membranes. Chem. Eng. Sci., 48 (1993) 20932102 [12]A. Malek, K. Li and W.K. Teo. Modeling of microporous hollow fiber membrane modules operated under partially wetted conditions. Ind. Eng. Chem. Res., 36 (1996) 784-793 [13]H. Kreulen, C.A. Smolders, G.F. Versteeg and W.P.M. van Swaaij. Microporous hollow fibre membrane module as gas-liquid contactors. Part 2. Mass transfer with chemical reaction. J. Membrane Sci., 78 (1993) 217-238 [14]J.S. Cha, V. Malik, D. Bhaumik, R. Li and K.K. Sirkar. Removal of VOCs from waste gas streams by permeation in a hollow fiber permeator. J. Membrane Sci., 128 (1997) 195-211 [15] K. Li, D. Wang, C.C. Koe and W.K. Teo. Use of asymmetric hollow fibre modules for elimination of H2S from gas streams via a membrane absorption method. Chem. Eng. Sci., 53 N. 6 (1998) 1111-1119 [ 16]D. Bhaumik, S. Majumdar and K.K. Sirkar. Pilot-plant and laboratory studies on vapor permeation removal of VOCs from waste gas using silicone-coated hollow fibers. J. Membrane Sci., 167 (2000) 107-122 [17]S. Majumdar, D. Bhaumik and K.K. Sirkar. Performance of commercial-size plasmapolymerized PDMS-coated hollow fiber modules in removing VOCs from N2/air. J. Membrane Sci., 214 (2003) 323-330
38 Chapter 1 [18]M. H.V. Mulder. Basic Principle of Membrane Technology., second edition, Kluwer Academic Publishers, The Netherlands (1996) 339-357 [19]A.J.B. Kemperman, D. Bergeman, Th. Van den Boomgaard and H. Strathmann. The stability of supported liquid membranes: A state of the art literature review. Sep. Sci. Technol., 31 (1996) 2733-2762 [20]R.W. Baker. Membrane Technology and Applications, McGraw-Hill, New York (2000) 405-442 [21]D.L. Bryant, R.D. Noble and C.A. Koval. Facilitated transport separation of benzene and cyclohexane with poly(vinyl alcohol)-AgNO3 membranes. J. Membrane Sci., 127 (1997) 161-170 [22]W.S.W. Ho and T.K. Poddar. New membrane technology for removal and recovery of metals from waste waters and process streams. Proc. of the AIChE Spring National Meeting, Atlanta, March 5-9 2000, 38-43 [23]X.J. Yang, A.G. Fane, J. Bi and H.J. Griesser. Stabilization of supported liquid membranes by plasma polymerization surface coating. J. Membrane Sci., 168 (2000) 29-37 [24]S.H. Lin and R.S. Juang. Mass.transfer in hollow fiber modules for extraction and back-extraction of copper(II) with LIX64N carriers. J. Membrane Sci., 188 (2001) 251-262 [25]A. Gherrou, H. Kerdjoudj, R. Molinari and E. Drioli. Facilitated co-transport of Ag(I), Cu(II) and Zn(II) ions by using a crown ether as carrier: influence of the SLM preparation methos on ions flux. Sep. Sci. Technol., 37 N. 10 (2002) 2317-2336 [26]A. Figoli, W.F.C. Sager and M.H.V. Mulder. Facilitated oxygen transport in liqid membranes: review and new concepts. J. Membrane Sci., 181 (2001) 97-110 [27]J.D. Way and R.D. Noble. Facilitated transport, in: W.S.W. Ho and K.K. Sirkar (Eds.), Membrane Handbook, Chapman and Hall, New York (1992) 833-866
Basic Principles of Membrane Contactors 39 [28]R.W. Shofield, A.G. Fane and C.J.D. Fell. Gas and vapor transport through microporous membranes. II. Mebrane distillation. J. Membrane Sci., 53 N.1 &2 (1990) 173-185 [29]K.W. Lawson and D.R. Lloyd. Membrane distillation. J. Membrane Sci. 124 (1997) 25 [30]M. Gryta and M. Tomaszewska. Heat transport in the membrane distillaton process. J. Membrane Sci., 144 N. 1&2 (1998) 211-222 [31]P.A. Hogan, R.P. Canning, P.A. Peterson, R.A. Johnson and A.S. Michaels. A new option: osmotic distillation. Chem. Eng. Prog., (1998) 49-61 [32]V. Schroder, O. Behrend and H. Schubert. Effect of dynamic interfacial tension on the emulsification process using microporous, ceramic membrane. J. Colloid and Interf. Sci., 202 (1998) 334-340 [33]R.A. Williams, S.J. Peng, D.A. Wheeler, N.C. Morley, D. Taylor, M. Whalley and D.W. Houldsworth. Controlled production of emulsions using a crossflow membrane. Part II: Industrial scale manufacture. Trans IchemE, 76 part A (1998) 902-910 [34] V. Schroder and H. Schubert. Production of emulsions using microporous, ceramic membranes. Colloid and Surf. A: Physochemical and Eng. Aspects 152 (1999) 103-109 [35]S.J. Taverner and J.H. Clark. Recent highlights in phase transfer catalysis. Chem. Ind., (1997) 2227 [36]P.S. Kumar, J.A. Hogendoorn, P.H.M. Feron and G.F. Versteeg. New absorption liquids for the removal of CO2 from dilute gas streams using membrane contactors. Chem. Eng. Sci., 57 (2002) 1639-1651 [37]H.B. Ding, P.W. Carr and E.L. Cussler. Racemic leucine separation by hollow-fiber extraction. AIChE J., 38 n.10 (1992) 1493-1498
Chapter 2. Membrane materials
I. Introduction
The membrane itself represents the core of any membrane process. A large variety of membranes exists, depending on their structure, transport properties and separation mechanism; all those different characteristics are generally originated by dissimilar raw materials or preparation methods. The class of synthetic membranes includes organic (polymeric) and inorganic membranes. Due to the possibility to modulate their intrinsic properties (thermal, mechanical and chemical stability, selectivity and permeability etc.), polymeric membranes have attracted much more interest. A large part of membranes in use for membrane contactors applications are polymeric; the most significant exception probably concerns the use of ceramic membranes in the emulsification process. The microstructure of a membrane is also a critical subject, and strictly depends on the preparation procedures: commonly, one can discriminate between symmetric and asymmetric membranes. Symmetric membranes may be dense or have straight or sponge-like pores: such a kind of microporous structures are widely employed in membrane distillation and related operations, in membrane absorption, stripping and extraction processes, as support for liquid membranes, in membrane emulsification technology. Asymmetric membranes show a thin dense skin layer with or without pores on the top of a high porous sublayer: the thickness of the selective skin offers the advantage of a low resistance to the transport through the membrane. In phase transfer catalysis, if pores in the dense layer are small enough to retain the catalyst- but large enough to freely pass substrates and products - asymmetric membranes provide an interesting support for its immobilization.
Membrane Materials 41
In the next paragraphs, a survey on some polymeric and inorganic materials and on the preparation and characterization techniques for membranes used as contactors is presented. It is beyond the scope of this book to give details on this extremely complex matter, and readers are referred to specific handbooks in this field. Information on commercial modules used in membrane contactors applications are furnished in Chapter 3.
2. Membrane polymers
When producing porous membranes, the selection of the material is mainly driven by the necessity to achieve a high chemical and thermal stability. Microporous polymeric membranes are prepared by various techniques: sintering, stretching, track-etching, phase inversion. The processing requirements and related characteristics of the resulting membrane also determine and limit the choice of the polymeric materials. Typology and main characteristics of the polymers frequently used as material for microporous membranes are given in table 1.
3. Preparation methods
Different methodologies are available to prepare membranes. This paragraph will provide a brief description of sintering of powders, stretching of films, track-etching and template leaching techniques. The most common method for preparing porous membranes, the phase inversion process, is discussed with more details.
42 Chapter 2 Table 1. Frequently used materials for microporous membranes Polymer
Chemical structure
Main characteristics
Polycarbonate
o\\
)?-o-o
\-'-~/
CH 3
\-----/
Cellulose acetate CH2OAc o
High wet/dry strength; mechanical properties suitable for track-etching preparation method Very hydrophilic; sensitive to thermal and chemical degradation; low tensile strength
OAc Nylon H
I N
~
(CH2) s ~
C
Polysulfone
\ -- I
CH 3
\ -- I
\ -
I
Inherently wettable; subject to hydrolytic degradation; better chemical stability when using aliphatic polyamides pH and temperature resistant; poor hydrocarbon resistant
Membrane Materials 43
Polyethersulfone
High thermal and chemical stability
F
Polyetherketone
High thermal and chemical resistance
Polyetheretherketone
High thermal and chemical resistance; only soluble at room temperature in concentrated inorganic acids. Excellent thermal stability; good chemical resistance
to,O- ~ Polyimide 0
/c NX
0
c\ C//N
c
,,
0
0
0
Polypropylene
HI CH3 1 I C--C H
H
Polyvinylidenefluoride F
H
I
I
C--C
I
F
Chemically resistant; hydrophobic
I
H
High temperature resistant; inherently hydrophobic
44 Chapter 2 Polytetrafluoroethylene F:
F
I
I
t2--C
I
F
I
F
High temperature and chemical (acid) resistant; cannot be irradiated; inherently hydrophobic
3.1. Sintering Sintering is a simple technique: a powder of polymeric particles is pressed into a film or plate and sintered just below the melting point. The process yields to a microporous structure having porosity in the range of 10-40% and a rather irregular pore size distribution (figure 1). The typical pore size, determined by the particle size of sintered powder, ranges from 0.2 to 20 ~tm.
Figure 1. Scanning electron micrograph of a PTFE membrane prepared by sintering.
3.2. Stretching Microporous membranes can be also prepared by stretching a homogeneous polymer film made from a partially crystalline material. Films are obtained by extrusion from a polymeric powder at temperature close to the melting point coupled with a rapid draw-down. Crystallites in the polymers are aligned in the direction of drawing; after annealing and cooling, a mechanical stress is applied perpendicularly to direction of drawing. This manufacturing process gives a relatively uniform
Membrane Materials 45 porous structure with pore size distribution in the range of 0.2-20 ~tm and porosity of about 90% (figure 2).
Figure 2. Gore-Tex PTFE membrane prepared by stretching (pore size ~ 0.2 ~tm).
3.3. Track-etching Microporous membranes with uniform and perfectly round cylindrical pores can be obtained by track-etching. Homogeneous thin films, usually with thickness of 5-15 ~tm, are exposed to the irradiation of collimated charged particles, having energy of about 1 MeV. These particles damage the polymeric matrix; the film is then immersed in an acid or alkaline bath, where the polymeric material is etched away along the tracks so leaving perfect pores with a narrow size distribution Figure 3). Typical pore size ranges between 0.02 and 10 ~tm; however, the surface porosity generally is below 10%.
Figure 3. Polycarbonate membrane prepared by track-etching.
46 Chapter 2 3.4. Template leaching Porous structures can be obtained by leaching out one of the component from a film. This technique allows producing porous glass membranes suitable for emulsification process. A homogeneous melt of three components (i.e. SiO2, B203, and Na20) is cooled from 1300-1500~ down to 500-800~
As a consequence, demixing is induced in the system that splits into two
phases: one consisting mainly of Si02 which is not soluble in mineral acids, and the other phase is richer in B203, that is subsequently leached out of the structure resulting in a microporous matrix. Porous alumina membranes made by anodic oxidation contain parallel circular pores with a narrow pore size distribution. They are formed by an electrochemical process involving the oxidation of high purity aluminium foils in presence of an acid electrolyte, followed by etching in a strong acid bath. In this process, an electrical circuit is established between a carbon cathode and a thin film of aluminium which serves as the anode, resulting in the oxidation of the aluminium to form alumina according to the reaction: 2AI + 3 H 2 0 --~ Al202 + 3 H 2
(1)
In appropriate electrolyte solutions, the film that is formed has a uniform columnar array of hexagonally close packed alumina cells, each containing a circular pore (figure 4). Pores form in the oxide film because of field assisted dissolution of the alumina from the base of each pore. With appropriate process conditions, membranes can be formed with pore diameters between 0.01 and 0.3 pm, pore densities between 108 and 10 II cm "2 and thicknesses up to 200 ~tm (figure 5).
Membrane Materials 47
Figure 5. A microporous aluminum membrane prepared by anodic oxidation.
Microlithography and reactive ion etching is a further technique to produce porous membranes. A silicon nitride coating (= 1 ~m) is deposited on a silicon wafer by chemical vapor deposition. By spin-coating, on the top of the nitride layer a photosensitive lacquer is applied. The lacquer is then exposed to UV radiation and developed in a NaOH solution resulting in a print of the mask pattem in the lacquer layer; perforations are extended to silicon nitride layer by reactive ion-etching. The
48 Chapter 2
resulting membranes are characterized by a narrow pore size distribution, with pore diameters typically in the range of 0.5-10 pm. Alternatively, the exposed polymer layer can be degraded by irradiation with X-rays (figure 6).
Figure 6. A silicon microsieve prepared by X-ray lithography process.
3.5. Phase inversion technique
Membranes are prepared by phase inversion technique from polymers that are soluble at a certain temperature in an appropriate solvent or solvent mixture, and that can be precipitated as a continuous phase by changing temperature and/or composition of the system. These changes aim to create a miscibility gap in the system at a given temperature and composition; from a thermodynamic point of view, the free energy of mixing of the system becomes positive. The formation of two different phases, i.e. a solid phase forming the polymeric structure (symmetric, with porosity almost uniform across the membrane cross-section, or asymmetric, with a selective thin skin on a sub-layer) and a liquid phase generating the pores of the membrane, is determined by few and conceptually simple actions: 1. by changing the temperature of the system (cooling of a homogeneous polymer solution which separates in two phases): temperature-induced phase separation technique (TIPS); 2. by adding non-solvent or non-solvent mixture to a homogeneous solution: induced phase separation (DIPS);
diffusion-
Membrane Materials 49
3. by evaporating a volatile solvent from a homogeneous polymer solution prepared using solvents with different dissolution capacity. Although these procedures are practically dissimilar, the basic of membrane formation mechanism is governed, in all cases, by similar thermodynamic and kinetic concepts: variations in the chemical potential of the system, diffusivities of components in the mixture, Gibbs free energy of mixing and presence of miscibility gaps. TIPS and DIPS processes, often utilized also in combination to prepare membranes, are discussed in details in the following paragraphs.
3.5.1. Phase separation: a thermodynamic description
Free Gibbs energy of a system is defined as a state function of enthalpy (H) and entropy (S)" (2)
G = H - TS
where T is the temperature of the system. In general, G depends on temperature, pressure and number of moles ni of each components in the system: (3)
G = G ( r , P , nl,n 2 ..... nk)
and the change in Gibbs free energy for a multi-component systems is given by: dG = OG
dT +
dP +
P,ni
T,n,
dn~ i=1
(4)
T,P,nj
In equation (3)"
= ~t~
"~
P,n, = - S
"~
T,ni = V
~
(5)
T,P,nj
and, therefore: k dG = - S d T + VdP + ~ l.tidn i i=l
(6)
50 Chapter 2 For a two-component mixture, being T and P constant, the Gibbs free energy per mole Gm is given by the sum of the chemical potentials of both components 1 and 2: G m = Xl,s
-'b
X2,L/2
(7)
When nl moles of component 1 are mixed to n2 moles of component 2, the change in the free energy of mixing AGm per mole of mixture is: A G m = x1A].I 1 +
x2A,L/2
(8)
For an ideal solution, the chemical potential of each component is expressed by: (9)
/.ti =/.t o + R T In x~
where/.t o is the molar free energy of pure components. This circumstance is graphically illustrated in figure 7.
Gm 0
x2
~10
~2
Figure 7. Gibbs free energy of mixing for a two-components system at constant T and P.
Membrane Materials 51
From equation (8) follows that: A/~ i = RTlnx
(10)
i
and A G m = R T ( x I In x I + x 2 In
X2 )
(11)
Since lnxi is negative (being xi
= An
m -
TAS
m
(12)
For polymer/solvent systems, an expression of AHm valid for small apolar solvents is [2]:
AHm= Vm(~11 -. ~/~/2 r r )
(13)
where Vm, Vl and V2 are the molar volumes of the solution and the two components, AE the energy of vaporization, and r the volume fraction. The solubility parameter 8 is defined as the square root of the cohesive energy density AE/V. The cohesive energy per unit volume is the energy necessary to remove a molecule from its neighboring molecules. For the polymer/solvent system, if 81~82,the value of AHm tends to zero and polymer is miscible in the solvent. The solubility parameter consists of three contributions [3]: 62 = S J +82 +8~ where
(14)
6d is the solubility parameter due to dispersion forces, 8p is the solubility parameter due to
polar forces, and 8h is the solubility parameter due to hydrogen bonding.
52 Chapter 2 The entropy of mixing ASm can be described using the lattice model proposed by Flory [4]. Referring to the lattice of figure 8, it is assumed that macromolecules consist of segments identical in size, each occupying a site. If co is the number of segments of a macromolecule, n2 the number of polymeric chains, and nl the number of solvent molecules, the total number of molecules N is therefore: N = n 1 + con2
(15)
gENBEELI~~,'ltslmt'IB gEr~wEEEEI~Ir-'~-~-s |EEEEFIWEEI~IEEED lr r mmr |mmmEmEmn OEIEEIEIr -'nmmEIEEIEIlf llEIEEEEI.I~~."IEI'IIlE D~IIEIIEllEIEllE~JEI[|El )mE ,'JC:.'IWWI~II~IllE[ )NLI~D ,IW[ )GIWGImwwrJEE[ )D 'lEt )WWWWWWWEE[ )E ,IW~~~."IWEIWWWB )E gHWHIH~raEEC-~-LII~ Figure 8. A lattice representation of a polymer-solvent mixture.
Statistical considerations suggest that the total number of possible combinations f~ to arrange all the molecules in the lattice is given by:
~2 = r"2 (F - 1) n2(•-2) N_(~,_x),,,[ ',2 (N / ~r~t]" rt2 .t o-nz ~ (rtl/~)tJ
(16)
where ~ is a constant (=2 for asymmetric macromolecules), and F is the degree of coordination (= 4 for a bi-dimensional lattice). Using the Boltzmann equation (S=kln~) to calculate the entropy of the solution, and subtracting the entropy of the single components, the entropy of mixing ASm results:
ASm = -R(?'/1 #'/~1 "+"n2 h'102) where
(17)
Membrane Materials 53
~1 --
nl nl + tO"/'/2
and
~2 =
m n2 HI + ~r/'/2
(18)
Under the hypothesis that AHm =0, the Gibbs energy of mixing is:
NRTAG--"=~m- Z~mR = (~l l ln~bl+ (~2 3 ln#2
(19)
In this case, it can be shown that athermal polymer solutions never demix. Demixing occurs in presence of a positive enthalpy term AHm>0. In the most general case of AHm:/:0, the expression for Gibbs free energy of mixing is generalized by including an enthalpic contribution (nl~b2Z): AG m =
RV(n 1ln~bl+ n2 lnqk2+ HI{~2X)
(20)
Z is the Flory interaction parameter. Figure 9 shows diagrams of AGm versus ~b and its derivatives up to the second order at a given temperature and pressure for a bi-components system. The mixture is stable over a certain composition range if
0 Gm
~ < 0
(21)
The first derivative becomes zero when the binodal curve is reached, and reaches a positive maximum when the spinodal curve is attained. In this region, the mixture is metastabile: there is not a driving force to a spontaneous demixing. A thermodynamic instability is observed when
OZGm < 0
(22)
0r 2
and the system will demix spontaneously. The critical point is individuated by a zero value of the first derivative and a minimum in the second derivative.
54 C h a p t e r 2
Figure 9. Plot of the Gibbs free energy of mixing as a function of the composition, referred to a system exhibiting a miscibility gap.
Membrane Materials 55 3.5.2. Diffusion-induced phase separation In a DIPS process, the membrane is formed by polymer precipitation caused by concentration variations due to diffusive interchange between the solvent and the non-solvent. The final structure of the membrane is determined by the rate of polymer precipitation. A low precipitation rate results in the formation of a symmetric structure, whereas a high precipitation rate leads to an asymmetric membrane, with large voids, spongy sublayer and/or finger-like cavities below a microporous or dense upper layer. From a practical point of view, the diffusion-induced phase separation process is routinely used to prepare integral asymmetric membranes. This process is articulated in few simple steps: 1. dissolution of the polymer in an appropriate solvent to form a solution typically containing 1030% wt% of polymer; 2. casting the solution on a suitable support into a film of 100-500 ~m thickness; 3. quenching of the film in the non-solvent (typically water or an aqueous solution), or evaporation of the solvent to increase the polymer concentration. During the third step, the homogeneous polymeric solution demixes into two phases: a polymerrich solid phase, which forms the structure of the membrane, and a solvent-rich liquid phase, which results in the formation of liquid-filled membrane pores. For a more quantitative explication of the DIPS process, let us consider the equilibrium diagram of the ternary system polymer/solvent/non-solvent reported in figure 10. The three-component mixture exhibits a miscibility gap in a certain composition range; under these conditions, the system splits into two distinct phases. Starting from an homogeneous mixture (point A), if the solvent is completely evaporated, the final composition of the system will be represented by the point B. Here, the system consists of only two components (polymer and non-solvent) and it is distributed in two phases, whose compositions are indicated by point B' (polymer-rich phase, solid membrane structure) and point B" (non-solvent rich phase, liquid-filled pores).
56 Chapter 2
Figure 10. A three component system isothermal diagram showing the formation of a membrane by solvent evaporation.
Figure 11 illustrates the formation of the m e m b r a n e -
based on the assumption of
thermodynamic equilibrium - for a phase separation process induced by addition of a non solvent to a homogeneous polymer solution. Starting from the point A on the solvent-polymer axis, if the solvent is completely removed from the mixture at about the same rate as the non-solvent enters, the final composition of the system will be represented by point B. Liquid-liquid demixing occurs when the line A-B intersects the binodal; the polymer concentration in the polymer-rich phase is high enough to be considered as solid when the line A-B intersects the tieline in correspondence of the vitrification point.
Membrane Materials 57
Figure 11. A three component system isothermal diagram showing the formation of a membrane by addition of non-solvent.
Thermodynamic information about the system are useful but not sufficient to predict the resulting morphology of the membrane, the pore size distribution and the occurrence of symmetric or asymmetric structure. A complete understanding of membrane formation is difficult because of the high number of involved mechanisms and phenomena, i.e. thermal effects, demixing kinetics, eventual presence of additives, mutual interaction parameters between polymer/solvent/non-solvent, temary diffusivities, initial and boundary conditions etc. Referring to the thermodynamic approach proposed in the previous paragraph, the solubility parameters for some polymers and solvents are reported in tables 2 and 3, respectively.
58 Chapter 2 Table 2. Solubility parameters of some polymers (* expressed in MPa 1/2 [5]; ** expressed in cal/cm 3 [6]) Polymer ~d ~p 6h 6 Polyvinylidene
17.2
12.5
9.2
23.2
Polyethylene**
8.6
0
0
8.6
Nylon 66**
9.1
2.5
6.0
11.6
Polysulfone**
9.0
2.3
2.7
9.6
Polyacrylonitrile** 8.9
7.9
3.3
12.3
Cellulose
7.9
3.5
6.3
10.7
9.4
1.3
2.4
9.8
fluoride*
acetate* * Poly(phenylene oxide)**
Membrane Materials 59
Table 3. Solubility parameters of some solvents (expressed in MPa 1/2 [5]) Solvent
t~d
t~p
8h
t~
N,N-dimethylacetamide
16.8
11.5
10.2
22.7
17.4
13.7
11.3
24.8
18.4
16.4
10.2
26.7
18.4
8.6
11.3
23.2
18.0
12.3
7.2
22.9
Tetramethylurea (TMU)
16.8
11.5
9.2
22.3
Triethyl phosphate (TEP)
16.8
16.0
10.2
22.3
(DMA) N,N-dimethylformamide (DMF) Dimethylsulphoxide (DMSO) Hexamethylphosphoramide (HMPA) N-methyl-2-pyrrolidone (NMP)
As practical case, cloud point data at 20~ for PVDF-solvent-water systems are illustrated in the ternary phase diagram shown in figure 12 [5]; at fixed concentration of polymer, the amount of water required to precipitate PVDF increases with the following order for different solvents: HMPA> DMA> DMF> TEP> TMP.
60 Chapter 2
Figure 12. Ternary phase diagram at 20~ for PVDF-solvent-water system. After [5].
A typical asymmetric structure of PVDF membrane prepared by DIPS technique is shown in figure 13.
Figure 13. Cross section of a PVDF membrane prepared by immersion precipitation.
Membrane Materials 61
The morphology of the membrane is determined by the properties of the system used to form the membrane itself. Polymer-solvent interactions have been widely investigated by various authors [7,8,9]. In general, lower interactions correspond to a higher rate of polymer precipitation, thus resulting in the formation of finger-like structures. The compatibility of polymer and solvent can be evaluated in terms of the three component of the solubility parameter ~ (see equation 13): (23) where subscripts P and S indicate the polymer and the solvent, respectively. The tendency of a solvent to mix with the non-solvent also affects the membrane porosity and structure [9, 10, 11, 12, 13]. For asymmetric membranes prepared by immersion in water, in most cases the higher the difference of the solubility parameter of solvent and water (Sw = 47.8 MPa 1/2 [5]), and hence the lower tendency to mix, the higher the membrane water content. A low solution viscosity generally determines the occurrence of cavities in the membrane. On the contrary, an increase of the solution viscosity due to and increase of polymer concentration obstructs the penetration of the nonsolvent during the immersion step. The rate of phase separation depends on the degree of penetration in the demixing gap. An instantaneous liquid-liquid demixing results in the formation of porous membranes. When a delay in liquid-liquid demixing occurs, dense membranes are produced. Particularly during the first moments subsequent to the immersion of the casting solution in the precipitation bath, mass transfer (solvent and nonsolvent interdiffusion) could become the controlling mechanism for skin formation. It has been evaluated that, for a solvent-nonsolvent diffusivity in the order of 10-5 cm2/s and a typical skin thickness of about 0.1 ~m, the characteristic time for mutual diffusion td is 10-5 sec [ 14]. Modeling studies report that, when the value of the solvent-nonsolvent diffusivity increases, the concentration path in the temary diagram should lead to entry into the demixing gap at higher polymer concentration [ 11, 15].
62 Chapter 2
3.5.3. Thermally -induced phase separation Thermally-induced phase separation gives rise to solid-liquid phase separation by removing thermal energy from the system. TIPS process basically consists of four simple steps [ 16]" 1. formation of a homogeneous solution by melt-bending the polymer with a high-boiling, lowmolecular weight diluent; 2. casting of the solution; 3. cooling of the cast solution to induce phase separation and solidification of the polymer; 4. removal of the diluent (typically by solvent extraction) to produce the membrane structure. TIPS can be applied to a wide range of polymers, also if their low solubility prevents the use of non-solvent induced phase inversion. This preparation technique allows to obtain isotropic microporous structures. The formation of a membrane can be explained by referring to appropriate equilibrium phase diagrams, and to the theory of phase equilibria in polymer systems. For binary polymer-solvent systems in which the polymer is semi-crystalline, the melting point of the polymeric compound is related to the mixture composition as follows [4]:
1
1
R
V 2 2~' . q~
#~)
(24)
where Tm and T~ are the melting temperatures of the crystalline polymer in solution and the pure crystalline polymer, respectively; V1 is the molar volume of the solvent, V2 is the molar volume for the repeating unit, ~2 is the volume fraction of the solvent, AHf is the enthalpy of fusion and Z is the Flory-Huggins interaction parameter. Solving equation (23) for Tm:
1
rm=
~z
Vl
(25)
r~
and plotting it as function of the volume fraction of the polymer r
(1 - ~b2), it is possible to derive
a temperature-composition diagram for a semi-crystalline polymer-diluent system. A qualitative
Membrane Materials 63
version is reported in figure 14 for polypropylene and three diluents having different strengths of interaction with the polymer.
-i- m
+
0
0
r
1
Figure 14. Temperature-composition phase diagram for polypropylene- diluent system. After [ 16].
As shown in figure 14, the temperature at which phase separation occurs is increased in presence of lower strength of interactions polymer-solvent (X increases). Equation (24) also shows that, all parameters being constant, the smaller the molar volume of the solvent with respect to that of the polymer, the larger the melting point depression. Referring to figure 15, let us to consider an homogeneous polymer-solvent solution (point A) at temperature TA. If the solution is cooled at the same composition, the system loses its stability and a solid-liquid separation occurs. At the point B (final temperature TB) the system separates in a polymer-rich p h a s e - the composition of which is indicated by the point B", and in a solvent-rich phase - represented by the point B'. According to the classical lever rule, segments B'-B and B-B"
64 Chapter 2 represent the ratio of the amounts of the two phases in the mixtures, from which it is possible to estimate the porosity of the membrane.
TA HOMOGENEOUS LIQUID PHASE
(D
~._ (D c~
B
E
r
(D
i--
~
LIQUID-LIQUID DEMIXING
\ Membrane composition
SOLID-LIQUID DEMIXING
solvent
polymer
Figure 15. The phase diagram for a polymer-solvent binary system as a function of temperature.
The polymer-rich phase forms the solid membrane structure, and the solvent-rich phase the liquid filled pores. Table 4 lists some physicochemical properties of polymeric materials usually employed in TIPS process.
Membrane Materials 65
Table 4. Polymers for membrane prepared by TIPS. After [16] Polymers
Density (g/cm3)
Average
molecular Melting point (~
weight (Da) Polypropylene
(Himont 0.903
243,000
176
High density polyethylene 0.954
224,000
130
2.050
N.A.
197
0.835
N.A.
230
1.780
N.A.
169
Density (g/cm3)
Average
Pro-fax 6723)
(American
Hoechst
Hostalen GM-9255-F2) Polychlorotrifluoroethylene, (Kel-F,
3M
Company,
Grade 6300) Poly (4-methyl- 1-pentene) (Mitsui Chemicals, Grade RT-18) Poly(vinylidene fluoride) (Soltex Solvey 1011)
Diluent
Mineral ,oil (Plough Inc., 0.866
molecular Initial boiling point
weight (Da)
(~
N.A.
- 320
630
270
278.4
340
Nujol) Kel-F oligomineral oil (3M 1.930 Company, KF-3) Dibutyl phthalate (Aldrich Chemicals)
1.043
66 Chapter 2 The diagram of solid-liquid phase separation for polypropylene-mineral oil at different polymer concentrations is reported in figure 16 [16]. Crystallization curves are affected by cooling rates, since TIPS preparation method is a non-equilibrium process. Results in figure 16 show that the temperature of demixing significantly decreases if the cooling rate is increased" the solution may cool to temperatures below its corresponding equilibrium crystallization temperature prior to the actual crystallization of the polymer from solution. A scanning electron micrograph of a microporous polypropylene membrane prepared by TIPS is shown in figure 17.
120
110
100 0o v
cooling rate:
90 fl)
E !-
20 ~
80 ~
70 80 ~
I
60 0.0
0.2
0.4
0.6
0.8
1.0
Weight Fraction Polymer
Figure 16. Crystallization temperature-concentration curves for PP-mineral oil at cooling rates indicated in the diagram. After [ 16].
Membrane Materials 67
Figure 17. Microporous polypropylene membrane obtained by thermally induced phase separation technique.
4. Membrane modification
A large part of commercial microporous polymeric membranes available in capillary and flatsheet forms that are used for membrane contactors applications were originally manufactured and optimized for microfiltration purposes. The possibility to prepare new membranes for specific operations is recently increasing in interest, and some significant results reached in the preparation and modification of polymeric membranes have provided to an increase of the reliability of membrane contactors technology.
4.1. Additives in the casting solution
The use of additives to the casting solution, e.g. in the form of water-soluble polymers such as polyvinyl pirrolidone (PVP), polyethylene glycol (PEG) or inorganic salts (LiC1), represents a practical way to modulate the structure of a membrane. This aspect has been investigated in the preparation of microporous PVDF membranes for membrane distillation (MD) applications, where high porosity is requested in order to obtain a significant flux [ 17, 18, 19]. In particular, it has been observed that the addition of significant amounts of LiC1 increases the rate of PVDF precipitation during the immersion step: this causes the formation of an open structure with large macrovoids and
68 Chapter 2 cavities. The accelerated precipitation is related to the high tendency of the additive to mix with water and to the interactions of the additive with polymer and solvent [20]. The effect of LiC1 content on the porosity and mean pore size of membranes prepared from DMA/PVDF = 88/12 is illustrated in figure 18. Porosity progressively increases from 79 to 83 % in the range of 1-7 wt.% of additive, while the mean pore size achieves a maximum of 0.04 ~tm in correspondence of 3.5 wt.% LiC1 concentration. On the other hand, membranes prepared by using high amounts of LiCI exhibited low values of water entry pressure with a consequent increase of the risk of wettability. 86
4
3.6 84 3.2 82
x,~ .~
2.8
-~
2.4
~;
o
80
78
2 2
4
6
LiCl c o n c e n t r a t i o n (wt.%)
Figure 18. Effect of LiCl concentration of the porosity and mean pore size of the membrane. After [18].
4.2. Use of copolymers It is not mandatory to use a single type of monomer when preparing a membrane. Copolymers of tetrafluoroethylene (TFE) and 2,2,4-trifluoro-5-trifluoromethoxy-l,3-dioxole (TTD), commercially known as HYFLON AD, have been used to obtain asymmetric and composite membranes showing a high hydrophobic character and contact angles to water higher than 120 ~ [21 ]. Asymmetric hydrophobic microporous membranes from the copolymer of PTFE and PVDF have been prepared by phase inversion process [19]. According to the experimental analysis, these
M e m b r a n e Materials 69 membranes exhibit excellent mechanical properties (stretching strain and extension ratio at break approximately 6-8 times higher PVDF) and good hydrophobicity (contact angle to water of about 870).
4.3. Composite membranes Composite membranes generally show an asymmetric structure, generated by the deposition of a thin toplayer on a porous sublayer of a different material. Composite membranes have the advantage that the properties of each layer can be modulated and optimized independently to obtain the required selectivity, permeability, chemical and thermal stability etc. The preparation procedures for composite membranes can be grouped in four classes: 1) casting of the thin layer separately (e.g. by spreading a very dilute polymer solution on the surface of a water bath) and then laminating it on a microporous support; 2) coating of the microporous support by a polymer, a reactive monomer or a pre-polymer solution (e.g. by immersion in an appropriate solution with low solute concentration- often less than 1%) followed by drying, heat treatment or radiation (figure 19); 3) plasma polymerization (figure 20); 4) interfacial polymerization of reactive monomers on the surface of the microporous support (figure 21). Details about each of these techniques can be found elsewhere [6]. Some specific examples related to the preparation of composite membranes for membrane contactors applications are reported below.
J POLYMERIC FILM
Figure 19. Grafting by radiation.
RADIATION
IMMERSION IN MONOMER BATH
GRAFT POLYMER CHAIN
70 Chapter 2 Monomer(s)
Q ?-J M e m b r a n e 9
V A C U U M PUMP
.
.
...
v~..,.~
REACTOR
DISCHARGE COIL
Figure 20. Plasma polymerization reactor.
V//////,~ V//////A~ (/'//////,~
K//////Z f//////// V/////// [////////, Y///////, POROUS SUPPORT
~//////',,~I IMMERSION IN AQUEOUS SOLUTION OF REACTIVE MONOMER OR PREPOLYMER
IMMERSION IN A WATER IMMISCIBLE SOLVENT WHERE ANOTHER REACTIVE MONOMER IS DISSOLVED
INTERFACIAL POLYMERIZATIONAND FORMATION OF THE COMPOSITE MEMBRANE
Figure 21. Interfacial polymerization.
The work of Xu and colleagues [22] showed that hydrophobic PTFE membranes with a protective hydrophilic sodium alginate coating were resistant to wet-out at least for 300 minutes during osmotic distillation tests using feeds containing 0.2, 0.4, and 0.8 wt.% orange oil. The reduction in the overall mass transfer coefficient due to the coating was less than 5%. In order to prepare a hydrophilic/hydrophobic composite membrane, the surface of hydrophilic porous cellulose acetate was treated via radiation graft polymerization of styrene by Wu et al. [23].
Membrane Materials 71
Low pressure plasma polymerization permits to apply a thin layer upon a porous sublayer: this generally results in a change of the chemical composition and properties of a material, such as wettability, dyeability, refractive index, hardness, etc. Plasma is obtained by ionising a gas using high frequency (up to 10 MHz) electrical discharges. The pressure inside the reactor varies between 0.1 and 10 mbar. Collisions between monomers and ionised gas generate radicals: the products of the resulting reactions precipitate on the membrane when their molecular weight is high enough. A very high hydrophobicity, somewhat higher than that of PTFE, was achieved by fluorinated coatings also named "Teflon-like" [24]. Kong and co-workers [25] have modified hydrophilic microporous cellulose nitrate membranes by plasma polymerization of octafluorocyclobutane. The performance of these membranes, tested in membrane distilation applications, was found comparable with that of usual hydrophobic polymers.
4.4. Surface modifying molecules
Generally, an increase of membrane porosity and pore size improves the flux. The analogous effect can be obtained if membrane thickness and tortuosity is decreased. When considering thermal driven membrane contactors operations, such as in the case of membrane distillation, the conductive heat loss increases for thinner membranes and the efficiency of the process is therefore reduced. In order to resolve the conflict between the requirements for high mass transfer and low heat transfer through the membrane, composite microporous hydrophobic/hydrophilic membranes can be prepared: the top hydrophobic thin layer is responsible for the mass transport, while the hydrophilic sublayer increases the resistance to the conductive heat flux. Khayet et al. [26] have modified the surface of hydrophilic membranes by adding oligomeric fluoropolymers synthesized by polyurethane chemistry and tailored with fluorinated end-groups. During membrane formation, surface-modifying molecules (SMMs) migrate to the air-film surface according to the thermodynamic tendency to minimize the interfacial energy. These
72 Chapter 2 modified membranes exhibit low surface energies, good mechanical strength and high chemical resistance [27].
5. Inorganic membranes Inorganic membranes have received limited attention for applications as membrane contactors, except that in membrane emulsification. The preparation of glass membranes by leaching has been briefly considered in paragraph 4. Ceramic membranes, being aluminium oxide (T-A1203) and zirconium oxide (ZrO2) are usually obtained by sintering or by sol-gel processes. Sol-gel process is usually carried out by following two different procedures: the colloidal suspension route and the polymeric gel route (figure 22). In both cases, a precursor for hydrolysis and polymerization reactions is commonly employed: it is often an alkoxide (such as aluminium trisec butoxide) in case of colloidal dispersion. In the polymer gel route, the precursor is selected with a low hydrolysis rate. After the partial hydrolyzation of the alkoxide by addition of water, the reaction of condensation leads to the formation of a polyoxometallate. The sol is peptized by addition of an inorganic acid; the viscosity of the solution can be further increased by addition of polyvinylalcohol (PVA). A gel is formed when the concentration of particles becomes sufficiently high. After drying, the membrane is sintered at a definite temperature in order to stabilize the final morphology. More extensive and detailed information on the preparation of inorganic membranes can be found in [28].
Membrane Materials 73 colloidal particles
colloidal gel
o
o
colloidal gel
o
o ~ o o o 0 0
0
0
SOL
GEL DRYING AND SINTERING
ALKOXI DE PRECURSOR inorganic polymer
polymer gel
polymeric gel
%
Figure 22. Schematization of the sol-gel process. From [6] with kind permission of Springer Science and Buniness Media.
6. Membrane characterization
It is well known that the transport phenomena in membrane contactors are strictly related to the structure of the membrane. In next chapters, correlations between transmembrane flux, energetic efficiency, permeate or product characteristics, and structural membrane properties such as thickness, porosity, pore size distribution etc. will be described in details for each membrane operation considered in this book. The knowledge of such correlations permits to predict and to optimize the membrane performance for a given application. Membrane characterization procedures allow to determine the structural and morphological properties of a membrane. The characterization of the surface chemistry is a critical issue in membrane contactors technology, since their performance depends on hydrophobicity or hydrophilicity character, surface charge, interactions between membrane and solutes or solvents, etc. Different membranes (porous, non porous, organic, inorganic etc.) require different procedures of characterization. In this section, the most familiar methods used for microporous membranes will be described. In particular, attention will be focused on the determination of structure-related
74 Chapter 2 parameters. Usual methodologies aiming to evaluate the permeation-related parameters (pure-water flux under hydrostatic pressure gradient, solute retention, molecular-weight cut-off, bacteria challenge test etc.) are not included in this section.
6.1.Contact angle measurements The contact angle measurement is a traditional method to describe the hydrophobic or hydrophilic behaviour of a material. In principle, it provides information about the wettability of an ideal surface. In most cases, the intrinsic value of contact angle is perturbed by surface porosity and roughness, heterogeneity, etc. The value of the contact angle made by a liquid droplet deposited on a smooth surface (figure 23) is greater than 90 ~ if the affinity between liquid and solid is low; in case of water, the material is considered hydrophobic. Wetting occurs at 0 ~ when the liquid spreads onto the surface.
Figure 23. Contact angle (0) of a liquid droplet deposited on the surface of a solid. Representation of the thermodynamic equilibrium at the triple point C.
At the triple point C where solid-liquid vapour interfaces are in contact, the thermodynamic equilibrium is expressed by the Young's equation: YLv cosO = Ysv -YsL
(26)
Membrane Materials 75
where ]tLV , "[SV , and ]tSL are the surface tension for liquid-vapour, the surface energy of the polymer, and the solid-liquid surface tension, respectively. Surface tension values "}tLVfor different test liquids are reported in table 5.
Table 5. Surface tension values ~'LVfor different test liquids Test liquid
'}tLV(mJ/m)
Water
72.8
Glycerol
64
Ethylene glycol
48
Formamide
58
Dimethylsulfoxide
44
Chloroform
27.2
Diiodomethane
50.8
A-bromonaphthalene
44.4
Because surface tensions involving a solid cannot be measured directly, a second equation is required to determine the hydrophobicity of the material, as given by the surface energy 7sv. Using a thermodynamic approach, Newmann [29] established the following equation of state to relate the three interfacial tensions:
YsL = YSV + YLV - 24YsvYLV exp[- fl(YSV - YLV )2 ]
(27)
and, combining it with the Young's equation: cos (9 = - 1 + 2x/'ys v/"YLv exp[- fl(Ysv - YLV )2 ]
(28)
where 13is a parameter independent of the solid and liquid used. The Young's equation is rigorously applicable if the solid substrate is smooth, if the surface is homogeneous and rigid, chemically inert and insoluble to contacting liquids. The effect of surface
76 Chapter 2
heterogeneity on contact angle is generally established by relation (29) that predict the contact angle 0* of a rough surface from the contact angle 0 of the equivalent smooth surface [30]: cos O* = fl cos 0 - f2
(29)
where fl and f2 are the fractions of liquid-solid and liquid-air surfaces, respectively. Courel et al. [31 ] demonstrated that the application of Young's equation to a porous surface leads to an expression similar to (29), where fl = y and f2 = l-y, being y the fraction of membrane surface made of solid material. For MD membranes with surface porosity lower than 0.5, it is generally assumed 1-y = ~/z, where is the porosity and T the pore tortuosity. For PTFE membranes, a more specific model has been developed:
costg* = y2 costg_ (l_y)2-2y(1-Y)IYsv?'I~V -cost9
(3o)
Under the assumption that the contact angles on the three-phase lines both on the outer drop border and over the pores are equal (as exemplified in figure 24), Troger and colleagues [32] have obtained a general relation between the observed contact angle 0' and the ideal one 0 (to be observed on ideally smooth surface): 4e cosO'+l cos 0 = cos 0'- - 1 - e cosO'-I
(31)
where e is the porosity of the porous material; the validity of equation (31) has been tested on porous PTFE membranes with appreciable results.
Membrane Materials 77
Figure 24. The assumption that the contact angles on the three phase lines occurring in the porous structure are equal. After [32].
6.2. Good-van Oss-Chaudhury method The Good-van Oss-Chaudhury method [33] represents a more complex approach to the determination of the surface tension components by contact angle measurements. In this case, three reference liquids (typically water, di-iodomethane, and glycerol) are used to determine the apolar Lifshitz-vane der Waals component yLW, the acid- base component yAB , the acid (electron acceptor) y+, and the base (electron donor) component y- of the surface energy. For instance, di-iodometane (apolar test liquid) allows the evaluation of the Lifshitz-van der Waals component y LWof the membrane surface tension reflecting the dipole interactions"
y.,
=
(1- oso)
(32)
4
Subscripts s and 1 indicate solid (membrane) and liquid, respectively. Other components are calculated by the following equations:
Ycv(l +c~ yAB = 24 y ,-y , §
2 [I ,~vy~w
- +
(33) (34)
78 Chapter 2 Typical contact angle of water are close to 120 ~ on PP [34] and are about 108 ~ and 107 ~ on PTFE and PVDF, respectively [35]. Additional data conceming various polymeric membranes are reported in table 6.
Table 6. Contact angles in water (W), glycerol (G) and diiodomethane (D), and surface tension parameters of different polymeric membranes: yLw =Lifshitz-van der Waals component, y - = electron donor component, 7'*= electron acceptor component, y,~e= acid-base component of the liquid surface tension 7'. After [36] Membrane g~ (o) 6~W (o) gG (o) yLw 7'Y+ yA~ y (mJ/m2) (mJ/m 2) (mJ/m 2) (mJ/m2) (mJ/m 2) Nylon
24•
49•
75•
47
57
4
29
75
Polyester
41+2
75•
81+1
39
15
0.9
72
67
Polyethersulfone
30.4+ 1
54+ 1
69+2
44
41
1
14
58
Polyurethane
35+3.6
94.4+1.4
100.9+1.6 42.0
9.0
5.52
14.1
56
Polyetheretherketone-
24.6+2.4 69.9+2.2 69.5+1
46.3
15.7
0.32
4.5
51
26.2•
71.2+2.8 68.4•
45.7
13.7
0.15
2.9
49
52+1.3
82+2.2
33.4
2.3
1.1
3.2
37
WC-20 PolyetheretherketoneWC-60 Collagen
67+7.6
6.3.Contact angle and wettability Comparing experimental results and theoretical calculations, Franken and colleagues [30] concluded that contact angle measurements on homogeneous smooth materials are not suitable for an accurate description of the membrane wetting phenomenon in MD. Wettability criteria based on the concept of penetration surface tension Y~ (defined as the surface tension of the liquid on the verge of penetrating a porous medium, and measured by penetrating drop method) provided more
Membrane Materials 79
satisfying results. Table 7 collects some values of y~" as obtained by Franken for two porous hydrophobic polymers, PVDF and PP, and measured using various aqueous solutions.
6.4. The
breakthrough pressure
In membrane contactors operations, in oder to efficiently work, the interface between phases must carefully controlled. Generally, non-wetting fluid does not pass through pores as long as the pressure is kept below a critical threshold known as breakthrough pressure. The Laplace' s equation offers a relationship between the largest pore radius of the membrane rp.max and the breakthrough
pressure APentry: 2| APentry
-- -
~
rp,max
(35)
where y is the interfacial tension, | is a geometric factor related to the pore structure (equal to 1 for cylindrical pores), and 0 the liquid-solid contact angle. This angle increases with increasing polarity difference between the polymeric membrane and the liquid.
80
Chapter
2
It is reported that, for a typical water-hydrophobic membrane contact angle of 130 ~ the penetration pressure of a cylindrical pore with 1 mm diameter is only 185 kPa [37]. Breakthrough pressure data for several membranes types and fluids can be found in literature [38]" in the most part of considered cases, AP values range between 100 and 400 kPa (figure 25).
10
'
'
'
'
'
o+.~..~
v
'
'
' I
+
++~ 9
L_ C~
Jig+
9
+
C" L
++ +
PP-Accurel PVDF-Accurel PTFE-Poreflon PTFE-Gore Tex
I
I
I
9
I
I
I
-
I I I
0.1
1
Maximum pore size (tam) Figure 25. Water pressure entry for different membranes as a function of the maximum pore size. After [39].
The breakthrough pressure is drastically reduced in presence, even at trace level, of detergents and surfactants (because they reduce the surface tension), or solvents that exhibit the same behaviour. Experimental investigations [40] demonstrated that, once a membrane is wetted by the penetrating liquid, a decreasing in hydrostatic pressure is not able to restore the original un-wetted condition. For mixtures of water and ethanol, Gostoli and Sarti [41 ] observed that the liquid entry pressure decreased linearly with alcohol concentration until the membrane was completely wetted at ethanol concentration of 75 wt%.
Membrane Materials 81 If the liquid penetrates through the micropores, a reduction of the hydrostatic pressure does not restore the un-wetted condition of the membrane. This phenomenon is illustrated in figure 26: the liquid floods the largest pores if the pressure
overcomes
APentry; as the pressure is increased further,
all the pores are flooded and the transmembrane flux N obeys the Darcy's law: N = kAP
(36)
being k a constant. Experimental investigations [40] showed that, if the applied pressure is reduced, the flux decreases linearly.
F--
Hydrostatic pressure (AP) Figure 26. The characteristic trend of the liquid transmembrane flux versus pressure drop in microporous hydrophobic membranes. After [37].
6.5. Microscopic techniques Microscopy observation and image processing of micrographs directly furnish visual information about the membrane morphology. Various microscopic techniques are used to
82 Chapter 2 investigate the structure of a membrane; three of them will be considered: the Scanning Electron Microscopy (SEM), the Transmission Electron Microscopy (TEM) and the Atomic Force Microscopy (AFM). In SEM, a beam of electrons (with kinetic energy of 1-25kV) is produced at the top of the microscope by heating of a metallic filament. The electron beam passes through electromagnetic lenses which focus and direct it down towards the membrane sample. Once it hits the sample, secondary electrons are ejected from the surface of the sample. Detectors collect the secondary or backscattered electrons, and convert them to a signal that is sent to a viewing screen. SEM has a resolution up to 5 nm. Membrane can be damaged by the electron beam; to prevent it, the sample needs to be pre-treated by coating with a conducting layer. Obtaining the pore size distribution from SEM micrographs is an extremely time consuming work; a more convenient method was developed by Manabe et al. [42] for membranes with pore size larger than 10 nm. They adopted some geometrical pore models and derived theoretical equations relating the pore radius distribution function N(r) to the distribution function F(x) of the length x of test lines cut off by pores in an electron micrograph. For straight-through cylindrical pores:
F(x)=
oo N(r) )dr Xx/2I4(4r 2 _ x 2
(37)
In TEM, a tungsten filament (the cathode) is heated and a high voltage (40- 100,000 kV) in order to emit electrons. These negatively charged electrons are accelerated to the anode to form an electron beam that is focussed onto the specimen by electro-magnets and double condenser lenses. As result, some electrons are scattered whilst the remainder are focused by the objective lens either onto a phosphorescent screen or photographic film to form an image. TEM has a resolution of 0.40.5 nm; however, this technique requires a complex preparation procedure [43]. AFM (figure 27) operates by measuring attractive or repulsive forces between a tip and the sample. In the repulsive "contact" mode, the instrument lightly touches a tip at the end of a leaf spring or "cantilever" to the sample. As a raster-scan drags the tip over the sample, and detection
Membrane Materials 83
apparatus measures the vertical deflection of the cantilever, which indicates the local sample height. In non-contact mode, the AFM derives topographic images from measurements of London-van der Walls forces; the tip does not touch the sample. Membranes can be scanned in air without pretreatments. AFM has a resolution of about 1 nm and offers useful information about the mean surface roughness Ra, that represents the mean value of the surface relative to the center plane for which the volumes enclosed by the images above and below this plane are equal. This parameter is calculated by [44]: L Lv
(38) where f(x, y) is the surface profile relative to the centre plane and Lx and Ly are the dimensions of the surface in the x and y directions, respectively.
Surfaceprofile(f)
Direction(x
Cantilever f ............................~ deflecti~ ~ / , ~
"
7
/
/ Tip
Surface Figure 27. Schematic representation of the AFM technique.
84 Chapter 2 6.6. Pore size distribution of microporous membranes The knowledgement of the mean pore size and pore size distribution of a microporous membrane is necessary for an accurate prediction of the transmembrane flux.
6. 6.1. Bubble point test The bubble point test is a simple method for determining the size of the largest pore in a membrane by measuring the pressure needed to blow air through a liquid-filled membrane. It is based on the equation that gives the pressure p needed to displace one fluid by another through a pore diameter dp: 4ycosO p =~
dp
(39)
where 7 is the interfacial tension of the air-liquid interface, and 0 is the wetting angle with the solid matrix of the membrane. From a practical point of view, the membrane is in contact with the liquid (which wets the membrane) on the top, while the gas flows at the bottom at increasing pressure (figure 28). The air bubble penetrates through the membrane pores when its radius equals the pore radius; this means that the contact angle is 0 ~ If using water, a pressure of 1.4 bar has to be applied for penetrating a pore radius of 1 ~tm, and 14.5 bar for a pore radius of 0.1 p.m. In order to avoid the use of high pressures, other liquids with low surface tensions (e.g. ethanol, iso-propanol, n-propanol etc.) are preferentially used.
Membrane Materials 85
Figure 28. The principle of bubble-point method.
According to the same principle of the bubble-point method, the pore size distribution of a membrane can be obtained by gas-liquid displacement technique. For further details, readers are referred to the specific literature (e.g. [45]).
6. 6. 2. Mercury intrusion porosimetry In mercury porosimetry, an amount of mercury is forced through the pores of a membrane at increasing pressures. The required pressure corresponds to a certain pore diameter, according to equation (39), and the total amount of mercury that disappears in the membrane allows evaluating the total volume of pores. Hysteresys of the extrusion-intrusion path and no- loop closing due to some portion of mercury retained by the sample are common features present in the porograms. In polymeric membranes, only pores having a diameter greater than 2 ~tm can be reliably detected. As disadvantage, this technique does not distinct between dead-end pores and interconnective pores.
86 Chapter 2
6.6.3. Liquid permeation The liquid permeation technique - in conjunction with appropriate mathematical models represents one of the most reliable methods for effectively determining pore size distribution. This technique is based on measurements of the flux of a non-wetting fluid as a function of the applied transmembrane pressure. At the beginning, pores are completely dry and the application of a low pressure drop does not cause the flooding of the membrane pores with the non-wetting fluid. If the applied transmembrane pressure exceeds a threshold value (APmin), liquid starts to penetrate through largest pores. Further increases in pressure drop give rise to increases in flow, acoording to the typical behavior shown in figure 26. With the aim to minimize numerical problems, a smooth curve through experimental flow-pressure data is required. For this purpose, a smoothing spline S is used: b
(40) a
subjected to the constraint: .
(41)
where a and b are the boundaries abscissa of data, S" is the second derivative of S (geometrically, its curvature), n the number of data points, wi the weights given to data, and cy the smoothing parameter. In order to perform the analysis of the flow-pressure curve according to the method proposed by Grabar and Nikitine [46], let us consider a normalized function f(r) representing the pore size distribution. If NTOT is the total number of pores, rmaxis the radius of the largest pore flooded first, and r(AP) the radius of the smallest flooded pore, the number of flooded pores N is given by"
N ~ =
NTOT
rm~
If(x)dx r(ae)
(42)
Membrane Materials 87 From a mathematical point of view, the fraction of pores N/NTOT that are flooded at pressure AP is represented by the shaded area reported in figure 29.
Figure 29. A Gaussian pore size distribution. The number fraction of pores with radii between r and rmaxis represented by the shaded area.
When assuming that pores have a circular shape, the Cantor's equation allows to correlate the radius of the smallest loaded pore to the flooding pressure AP: r(AP)=
2yLc~ AP
=~f~' AP
(43)
where ~/L is the liquid surface tension, 0 is the contact angle, AP is the transmembrane pressure and ~i a constant for a given membrane-liquid pair. In case of cylindrical pores having a tortuosity x, the Hagen-Poiseuille's equation can be used to quantify the flowrate Q through the membrane pores: # r 4 AP Q=~
8~t6r
(44)
where g is the fluid viscosity, 8 is the thickness of the membrane and r the pore diameter. Under the hypothesis of uniform membrane thickness, the total transmembrane flow Q at a given AP>APmin is:
88 Chapter 2 rmax
~x4Ap
_z \ _
rmax
Q= I Xr~ -8ft6-rr:['x)dx:n2AP r(AP)
I X4 f(x)dX r(AP)
(45)
Derivative of equation (44) with respect to AP, with opportune rearrangements and substitutions give the final expression for the pore size distribution function (mathematical details in [47]):
f(r)=
d(AP)
AP
In equation (45), constants
(46)
2 ~'~1~"~2
take into account information about the structural properties of the
membrane, the testing fluid properties and the fluid membrane interactions. For a normalized distribution, the n-th moment
(r") is mathematically defined as:
rmax
(47)
rmm where rminand rmaxare the radii of the smallest and largest pores in the membrane. The first moment of the distribution corresponds to the average pore radius. As disadvantage, the characterization method shows a loss in resolution in the pore size distribution (that can be offset by opportune adjustments of the weighting factors) as the pore sizes decrease to values well below the largest pore size. Moreover, this method needs an appropriate pore model describing the membrane structure (eq. (46) is valid for non-interconnecting, cylindrical pores). Liquid-liquid displacement represents a variant of the method above described. In this case, membrane pores are filled by a liquid that is displaced by a second immiscible liquid. A typical liquid pair is water/iso-butanol. Pores with diameters in the range of 5-100 nm can be adequately detected. With respect to gas-liquid displacement, liquid pairs are characterized by lower interfacial tensions compared to gas-liquid pairs, and reduced pressures are needed to penetrate pores with the same size. Further details can be found in literature [45, 48, 49].
Membrane Materials 89 6. 6. 4. P e r p o r o m e t r y
Perporometry is based on the phenomenon of capillary condensation of liquid in micropores. The vapour pressure of a liquid depends on the radius of curvature of its surface, according to Kelvin's equation: (48)
ln P_f-l_ = 27"V cosO Po RTrk
where p and p0 are the vapour pressures in the capillary and under standard conditions, respectively, y is the surface tension between the capillary liquid and air, V is the molar volume of the liquid, 0 the contact angle, R the gas constant, T the absolute temperature and rk the Kelvin radius, little smaller than the actual pore radius due to the presence of an absorbed layer of condensable gas. By applying a partial pressure difference across the membrane, pores can be blocked with liquid by capillary condensation; this principle is coupled to the measurement of the free diffusive transport through the open pores. A scheme of the experimental set-up is reported in figure 30. A mixture of oxygen and nitrogen (e.g. air) is applied on the feed side, while nitrogen flows on the permeate side as carrier gas. This creates a concentration gradient of oxygen across the membrane. On both lines, an organic compound (e.g ethanol) is also applied as condensable gas; in order to avoid swelling phenomena, the organic vapour should exhibit a low affinity with the membrane. At both sides of the membrane, the absolute pressure is 1 atm and the relative pressure of the organic vapour is the same. Evaporator IP GC Analysis
N2, Ethanol
I N2, O2, Ethanol
Evaporator
Figure 30. A permporometry setup.
. DIFFUSION CELL
[
Membrane
90 Chapter 2 The size distribution of active pores is therefore obtained by measuring the gas flow through the membrane. For pore radii of 1-25 nm and at atmospheric pressure, the flux of the i-th component through a pore with radius ri, determined by Knudsen diffusion, can be expressed as: j, = 2 [ 8~ Ap n,r, 3 V MwRT A mr 6
(49)
where Mw is the molecular weight of the gas, R the gas constant, T the absolute temperature, Ap the partial pressure gradient across the membrane, Am the membrane surface area, x the tortuosity of pores, 8 the membrane thickness, and ni the number of pores having radius ri. Integrating over the entire distribution of pore radii, few manipulations allow obtaining the pore size distribution:
-d-~-~ rnun
L drm,n -3V 8---~ Apr3mm
(50)
Quantitative analysis are preferentially carried out during desorption process, since it is more difficult to reach equilibrium during adsorption process: the gas (oxygen, in the discussed case) flux as a function of the Kelvin radius through Nucleopore membranes (pore size given by manufacturer: 15 nm) is reported in figure 31.
Membrane Materials 91 '
I
'
I
i
I
'
I
i
i
'
i
E ~
3
6 i
0
0
4
8
I
12
16
Kelvin radius (nm)
Figure 31. Oxygen flux versus Kelvin radius for a Nucleopore membrane. After [45 ].
This technique characterizes only active pores in the range of 2-40 nm. More details are in [50, 51,
52,53].
6.6.5. Thermoporometry Thermoporometry is based on the calorimetric measurement of a solid-liquid transition in a porous material in order to determine the pore size distribution [54, 55, 56, 57]. In pores totally filled with a liquid, the curvature of the liquid-solid interface Cs is related to the change of temperature T by: r~
,
92 Chapter 2 where V is the volume of the pore, AS is the surface area of the solid-liquid interface, ), is solidliquid surface tension. The liquid-solid interface is almost spherical and its curve Cs is: 2
(52)
Cs " - ~
r-t
where t is the thickness of the layer of condensate fixed to pore wall. Equations (51) and (52) link the pore radius r to a decrease in solidification temperature T-T0. In case of water, in the range of-40
r(nm)= - ~32.33 +0.68 T-To
(53)
The energy of solidification W is related to undercooling AT by 9 W = -0.155.10 -2 AT 2 - 11.39AT- 332
(54)
The differential change of pore volume dV corresponding to d(AT) is given by:
dV =
1 dW pW
- - ~
(55)
where 9 is the water density. Differentiating eq. (53) and coupling with (55), the equation that permit to calculate the pore volume distribution function from water thermograms is derived:
dV _ (AT) 2 dW dr 64.67pWd(AT)
~
(56)
m
For practical analysis, the heat flux required for melting is measured by Differential Scanning Calorimetry (DSC), and equation (56) is more conveniently used in the form:
d___V_V=
dr
(AT)~
q
64.67pW d(AT)/dt
where q is the heat flux obtained by DSC and d(AT)dt is temperature changing rate.
(57)
M e m b r a n e Materials 93
Thermoporometry is suited to characterize pores with diameters in the range of 2-50 nm; all pores, also those not active, are included in the characterization.
7. The influence of pore size distribution on the transmembrane flux
An inadequate knowledge of the morphology of a microporous membrane can lead to inaccuracy when modelling the mass transfer [58]. A good agreement between theoretical and experimental results was obtained by Martinez-Diez and Vazquez-Gonzalez [59] using pore sizes measured by mercury porometry and liquid displacement methods. The attention to the structural properties of microporous polymeric membranes involved in MD operations is today increasing significantly. If f(r ) is the normalized distribution, and J(r ) the transmembrane flux through all pores with radii equal to r, the total flow rate JT (r') through the membrane is obtained by the following integral relation: oo
Jr = IJ(r')~r(r')2 f(r') dr'
(58)
0
Typically, a lognormal distribution in the form reported above, is sufficiently accurate to model the size distribution of membrane pores:
f(r)=
_
, exp SD,ogr.~_~
7L SD,og J
(59)
where f(r) is the number of pores with pore radius r, ~ the mean pore radius, and SDiog the standard deviation of lognormal function. Figure 32 depicts the pore size distribution of a PP Accurel| hydrophobic membrane, frequently used in membrane contactors experiments.
94 Chapter 2 800
I
'
I
'
600 0 0 d) t'~
400
m
E t-
or200
0.2
0.4
0.6
Pore diameter (gin) Figure 32. Pore size distribution of an Accurel| PP membrane. After [60].
Laganh and coworkers [61] studied the effect of the shape of pore size distribution with Gaussian (symmetric) and logarithmic (asymmetric) distribution functions; in this investigation, non-symmetrical distribution achieved better agreement with the experimental results. Several mathematical models aiming to examine the influence of both pore size distribution and air flux in DCMD were presented by Phattaranawik et al. [62]. In particular, the log-normal distribution was used to represent the shape of pore size distribution. In conclusion of their work, authors reported that the predictions of the fluxes and MD coefficients showed good agreement with the experimental results for GVHP-PVDF (Millipore, 0.22 mm) and excellent agreement for HVHPPVDF (Millipore, 0.45 mm) and PTFE (Sartorius, 0.2 mm). Additionally, the models predicted fluxes with less than 8% discrepancy. The investigation carried out by Martinez-Diez and colleagues [63] on three commercial membranes frequently used in MD applications showed that the MD water vapour transfer coefficients, calculated considering the pore size distributions, are similar to the ones obtained
M e m b r a n e Materials 95
assuming an average pore size model, and the permeabilities calculated from air-liquid displacement measurements agree well with those obtained in literature MD models.
8. Estimation of the membrane distillation coefficient
For characterization purposes, a number of works involved flat membranes assembled in the Lewis test cell having stirring capabilities (figure 33).
Figure 33. Schematic representation of a Lewis test cell.
Another common configuration considers the recirculation of both (hot) feed and (cold) permeate streams through flat membranes using channel devices (figure 34).
96 Chapter 2
Permeate inlet
Permeate outlet
Membrane
/1 Feed inlet
Feed outlet
/1
Figure 34. Schematic representation of a flat channel test cell.
In thermal membrane distillation, the mass transport process through a microporous hydrophobic membrane can be described by the following equation [64]:
cFdp ]
J = -~t"ffT-jrATm
(60
where J is the molar flux per unit area, fi is the membrane thickness, C the membrane distillation coefficient, APv the vapor pressure difference across the membrane, and ATm the temperature difference between the membrane interfaces. The value of ATm is related to bulk temperature difference ATb:
M e m b r a n e Materials 97 1 AT m = - - - - - ~ A T b
(61)
l+-h where H = CA dPv + ~km dT 6
(62)
being H the overall heat transfer coefficient, ;~ the latent heat of vaporization, km the thermal conductivity of the membrane, and 5 the membrane thickness. Combining the previous equations:
ATb
l+km eta
1
J,~
c,~ dev
h
(63)
dT
Plotting the experimental data in terms of A T b / d2 vs. 1/(dP v / d T ) , h can be evaluated from the intercept (1/h) and C is obtained from the slope (1/CL)/(1 +km/Sh). As example, figure 3 5 refers to the calculus of membrane coefficient C and heat transfer coefficient h operated by Martinez-Diez and colleagues [65] starting from permeation experiments with a PTFE membrane conducted by using a flat membrane module with channels for recirculation of hot and cold water.
98 Chapter 2
I
i
I
i
h = 1515+ 90 Wm-2K -~ C(x107)=21.0_+1.1 kg m-2s-~Pa -~
E
d X
<1
I
2
,
I
,
4
I
6
l/(dPv/dT) x 10 3, K/Pa
Figure 35. A plot of ATb/JX versus 1/(dPv/dT) corresponding to results obtained by [65] with a recirculation rate of 9.0 cm3/s.
Membrane Materials 99
References [1] J. Randon, P.P. Mardilovich, A. N.Govyadinov and R. Paterson. Computer simulation of inorganic membrane morphology. Part 3. Anodic Alumina Films and Membranes, J. Colloid and Interf. Sci., 169 (1995) 335-341 [2] J. Hildebrand and R. Scott. Solubility of Nonelectrolytes, Reinhold, New York, 1949 [3] C. Hansen. The universality of the solubility parameters. Ind. Eng. Chem. Prod. Res. Dev., 8 (1969) 2 [4] P.J. Flory. Principles of Polymer Chemistry. Cornell University Press, Itaca, 1953 [5] A. Bottino, G. Camera-Roda, G. Capannelli and S. Munari, The formation of microporous polyvinylidene difluoride membranes by phase separation. J. Membrane Sci., 57 (1991) 1-20 [6] M.H.V. Mulder. Basic principles of Membrane Technology. Kluwer Academic Publishers, Dordrecht, The Netherlands, 2000 [7] M.A. Frommer, I Feiner, O. Kedem and R. Bloch. The mechanism for the formation of 'skinned' membranes. II. Equilibrium properties and osmotic flows determining membrane structure. Desal., 7 (1970) 393 [8] M. Guillotin, C. Lemoyne, C. Noel and L. Monnerie. Physicochemical processes occurring during the formation of cellulose diacetate membranes. Research of criteria for optimizing membrane performance. IV. Cellulose diacetate-acetone-additive casting solutions. Desal., 21 (1977) 165 [9] H. Strathmann and K. Kock. The formation mechanism of phase inversion membranes. Desal., 21 (1977) 241 [10] M. A. Frommer and R.M. Messalem. Mechanism of membrane formation. VI. Convective flows and large void formation during membrane precipitation. Ind. Eng. Chem. Prod. Res. Dev., 12 8 (1973) 328 [11] C. Cohen, G.B. Tanny and S. Prager. Diffusion-controlled formation of porous structure in ternary polymer systems. J. Polym. Sci., Polym. Phys. Ed., 17 (1979) 477 [12] G. Friedrich, A. Driancourt, C. Noel and L. Monnerie. Asymmetric reverse osmosis and ultrafiltration membranes prepared from sulfonated polysulphone. Desal., 36 (1981) 39 [ 13] J.G. Wijmans, J. Kant, M.H.V. Mulder and C.A. Smolders. Phase separation phenomena in solutions of polysulfone mixtures of a solvent and a nonsolvent: relationship with membrane formation. Polymer, 26 (1985) 1539
100 C h a p t e r 2 [14] K. Hakagawa and Y. Ishida. Annealing effect in poly(vinylidene fluoride) as revealed from specific volume measurements, differential scanning calorimetry and electron microscopy. J. Polym. Sci. Polym. Phys. Ed., 11 (1973) 2153 [ 15] L. Yilmaz and A.J. McHugh. Modelling of asymmetric membrane formation. I. critique of evaporation models and development of diffusion equation formalism for the quench period. J. Membrane Sci., 28 (1986) 287 [16] D.R. Lloyd, K.E. Kinzer and H.S. Tseng. Microporous membrane formation via thermally induced phase separation. I. Solid.liquid phase separation. J. Membrane Sci., 52 (1990) 239-261 [17] M. Tomaszewska. Preparation and properties of flat-sheet membranes from poly(vinylidene fluoride) for membrane distillation. Desal., 104 (1996) 1-11 [18] C. Feng, B. Shi, G. Li and Y. Wu. Preliminary research on microporous membrane from F2.4 for membrane distillation. Sep. Purif. Technol., 39 (2004) 221-228 [19] C. Feng, B. Shi, G. Li and Y. Wu, Preparation and properties of microporous membrane from poly(vinylidene fluoride-co-tretrafluoroethylene) (F2.4) for membrane distillation. J. Membrane Sci., 237 (2004) 15-24 [20] A. Bottino, G. Capannelli, S. Munari and A. Turturro. High performance ultrafiltration membranes cast from LiC1 doped solutions. Desal., 68 (1988) 167 [21 ] V. Arcella, P. Colaianna, P. Maccone, A. Sanguinetti, A. Gordano, G. Clarizia and E. Drioli. A Study on a Perfluoropolymer Purification and its Application to Membrane Formation. J. Membrane Sci., 163 (1999) 203-209 [22] J.B. Xu, S. Lange, J.P. Bartley and R.A. Johnson. Alginate-Coated Microporous PTFE Membranes for Use in the Osmotic Distillation of Oily Feeds. J. Membrane Sci., 240 (2004) 81-89 [23] Y. Wu, Y. Kong, X. Lin, W. Liu and J. Xu. Surface-modified hydrophilic membranes in membrane distillation. J. Membrane Sci., 72 (1992) 189-196 [24] P. Favia and R. D'Agostino. Plasma Treatments and Plasma Deposition of Polymers for Biomedical Applications. Surface and Coatings Tech.nol., 98 (1998) 1102-1106
Membrane Materials 101 [25] Y.Kong, X. Lin, Y. Wu, J. Chen and J. Xu. Plasma Polymerisation of Octafluorocyclobutane and Hydrophobic Microporous Composite Membranes for Membrane Distillation. J. Appl. Polym. Sci., 46 (1992) 191-199 [26] M. Khayet, J.I. Mengual and T. Matsuura. Porous hydrophobic/hydrophilic composite membranes. Application in desalination using direct contact membrane distillation. J. Membrane Sci., 252 (2005) 101113 [27] M. Khayet, D.E. Suk, R.M. Narbaitz, J.P. Santerre, T. Matsuura, Study on surface modification by surface-modifying macromolecules and its applications in membrane-separation processes, J. Appl. Polym. Sci. 89 (2003) 2902-2916 [28]A.J. Burgraaf and K. Keizer. Synthesis of inorganic membranes, in: Inorganic Membranes, Synthesis, Characteristics and Applications, Ed. Bhave, van Nostrand, New York, 1991 [29] D. Li and A.W. Newmann. Equation of State for Interfacial Tensions of Solid-Liquid Systems. Adv. Colloid Interf. Sci., 39 (1992) 299-345 [30] A.C.M.Franken, J.A.M. Nolten, M.H.V. Mulder, D. Bargeman, D., and C.A. Smolders. Wetting Criteria for the Applicability of Membrane Distillation. J. Membrane Sci., 33 (1987) 315-328 [31] M. Courel, E. Tronel-Peyroz, G.M. Rios, M. Dornier and M. Reynes. The Problem of Membrane Characterization for the Process of Osmotic Distillation. Desal., 140 (2001) 15-25 [32] J. Troger, K. Lunkwitz and W. Burger. Determination of the Surface Tension of Microporous Membranes Using Contact Angle Measurements. J. Colloid and Interf. Sci., 194 (1997) 281-286 [33] R.J. Good and C.J. van Oss. Modem Approach to Wettability; The Modern Theory Contact Angle and the Hydrogen Bond Components of Surface Energies. M.E. Schrader and G.L. Loeb (Eds.); Plenum Press: New York (1992) [34] L. De Bartolo, S. Morelli, A. Bader and E. Drioli. Evaluation of Cell Behaviour Related to PhysicoChemical Properties of Polymeric Membranes to be Used in Bioartificial Organs. Biomaterials, 23 (2002) 2485-2497 [35] M. Tomaszewska. Membrane Distillation. Envir. Protect. Eng., 25 (1999) 37--47
102 Chapter 2 [36] L. De Bartolo, S. Morelli, M. Rende, A. Gordano and E. Drioli. New modified polyetheretherketone membrane for liver cell culture in biohybrid systems: adhesion and specific functions of isolated hepatocytes. Biomaterials. 25 (2004) 3621-3629 [37] K.W. Lawson and D.R. Lloyd. Membrane distillation. J. Membrane Sci., 124 (1997) 1-25 [38] B.W. Reed, M.J. Semmens and E.L. Cussler, E.L. Membrane contactors. In: Membrane Separation Technology. Principles and Applications. R.D. Noble and S.A. Stem, S.A.(Eds.), Elsevier: Amsterdam (1995) [39] K. Schneider, W. Holz and R. Wollbeck. Membranes and modules for transmembrane distillation. J. Membrane Sci., 39 (1988) 25-42 [40] L. Pena, J.M. Ortiz de Zarat and J.I. Mengual. Steady States in Membrane Distillation: Influence of Membrane Wetting. J. Chem. Soc. Faraday Trans., 89 (1993) 4333--4338 [41] C. Gostoli and G.C. Sarti. Separation of Liquid Mixtures by Membrane Distillation. J. Membrane Sci., 41(1989) 211-224 [42] S. Manabe, Y. Shigemotoand
K. Kamide. Determination of pore radius distribution of porous
polymeric membranes by electron microscopic method. Polym. J., 17 (1985) 775-785 [43] A.G. Fane, C.J.D. Fell and A.G. Waters. The relationship between membrane surface pore characteristics and flux for ultrafiltration membranes. J. Membrane Sci., 9 (1981) 2455-262 [44] M. Khayet, C.Y. Feng, K.C. Khulbe and T. Matsuura. Preparation and characterization of polyvinylidene fluoride hollow fiber membranes for ultrafiltration,. Polymer, 43 (2002) 3879-3890 [45] S. Nakao. Determination of pore size and pore size distribution. 3. Filtration membranes. J. Membrane Sci., 96(1994) 131-165 [46] P.Grabar and S. Nikitine. Le diameter des pores des membranes en collodion utilisees en ultrafiltration. J. Chem. Phys. (Paris), 33 (1936) 721 [47] K. S. McGuire, K.W. Lawson and D.R. Lloyd. Pore size distribution determination from liquid permeation through microporous membranes. J. Membrane Sci., 99 (1995) 127-137 [48] S. Munari, A. Bottino, G. Capannelli and P. Moretti. Membrane morphology and transport properties. Desal., 53 (1985) 11-23
Membrane Materials 103 [49] G. Capannelli, F. Vigo and S. Munari. Ultrafiltration membranes - characterization methods. J. Membrane Sci., 15 (1983) 289-313 [50] F.P. Cuperus, D. Bargeman and C.A. Smolders. Permporometry. The determination of the size distribution of active pores in UF membranes. J. Membrane Sci., 71 (1992) 57-67 [51 ] M.G. Katz and G. Baruch. New insights into the structure of microporous membranes obtained using a new pore size evaluation method. Desal., 58 (1986) 199-211 [52] A. Mey-Marom and M.G. Katz. Measurement of active pore size distribution of microporous membranes- a new approach. J. Membrane Sci., 27 (1986) 119-130 [53] G.Z. Cao, J. Meijerink, H.W. Brinkman and A.J. Burgraaf. Permporometry study on the size distribution of active pores in porous ceramic membranes. J. Membrane Sci., 83 (1993) 221-235 [54] M. Brun, A. Lallemand, J.F. Quinson and C. Eyraud. A new method for the simultaneous determination of the size and the shape of pores : the thermoporometry. Thermochim. Acta, 21 (1977) 59-88 [55] J.F. Quinson, N. Mameri, L. Guihard and B. Barious. The study of the swelling of an ultrafiltration membrane under the influence of solvents by thermoporometry and measurement of permeability. J. Membrane Sci., 58 (1991) 191-200 [56] L. Zeman, G. Tkacik and P. Le Parlouer. Characterization of porous sublayers in UF membranes by thermoporometry. J. Membrane Sci., 32 (1987) 329-337 [57] F.P. Cuperus, D. Bargeman and C.A. Smolders. Critical points in the analysis of membrane pore structures by thermoporometry. J. Membrane Sci., 66 (1992) 45-53 [58] R.W. Schofield, A.G. Fane and C.J.D. Fell. Gas and Vapour Transport through Microporous Membranes. I: Knudsen-Poiseuille Transition. J. Membrane Sci., 53 (1990) 159-171 [59] L. Martinez-Diez and M.I. Vazquez-Gonzalez. Temperature Polarization in Mass Transport through Microporous Membranes. AIChE J., 42 (1996) 1844-1852 [60] K. Schneider, W. Holz and R. Wollbeck. Membranes and modules for transmembrane distillation, J. Membrane Sci., 39 (1988) 25-42 [61] F. Lagan~, G. Barbieri and E. Drioli. Direct Contact Membrane Distillation: Modelling and Concentration Experiments. J. Membrane Sci., 166 (2000) 1-11
104 Chapter 2 [62] J. Phattaranawik, R. Jiraratananon and A.G. Fane. Effect of Pore Size Distribution and Air Flux on Mass Transport in Direct Contact Membrane Distillation. J. Membrane Sci., 215 (2003) 75-85 [63] L. Martinez-Diez, F.J. Florido-Diaz, A. Hernandez and P. Pradanos. Estimation of Vapour Transfer Coefficient of Hydrophobic Porous Membranes for Applications in Membrane Distillation. Sep. Purif. Technol. 33: (2003) 45-55 [64] J.M. Ortiz de Zarate, L. Pena and J.I. Mengual. Characterization of membrane distillation membranes prepared by phase inversion. Desal., 100 (1995) 139-148 [65] L. Martinez-Diez, M.I. Vazquez-Gonzalez and F.J. Florido-Diaz. Study of membrane distillation using channel spacers. J. Membrane Sci., 144 (1998)45-56
Chapter 3. Module configurations and design 1. Introduction
In this Chapter module configurations and design for membrane contactors operations are discussed. Different kinds of modules are presented and compared in terms of mass transfer efficiency. Problems such as flow maldistribution and pressure drops and attempts to solve them are reported. Being the hollow fiber module the configuration of major interest for industrial applications, particular attention is devoted to it and to the research efforts made worldwide for improving its performance. These include both theoretical studies, that consist in the development of mathematical models able to describe the hollow fibers behaviour, and new module configurations that have been conceived as alternative to the first tube-in-shell modules with parallel flow of the streams. Modules layout is also discussed: the choice of a particular module assembly depends on both the economic and the specific operating conditions. The Chapter ends with information on commercial modules.
2. Modules used for membrane contactors applications
The performance of membrane contactors has been studied in different module configurations. Modules with flat membranes are mainly used for laboratory tests, because of
106 Chapter 3
their easier building at lab scale and the simple and fast sheet replacement. Usually, a single flat sheet is located between two plates that are equipped with the inlets/outlets of the involved phases (Figure 1). Inlets/Outlets
Membrane~~,N-~xx,
x~
Figure 1. Flat membrane contactor module. For pilot plants as well as industrial scale applications, higher membrane area per volume ratios are needed and the flat membranes are used in plate and frame or in spiral wound configurations. For desalination tests a plate and frame system has been used by Andersson et al. [1 ], whereas Gore [2] and Koschikowshi et al. [3] proposed a spiral wound configuration. Tubular membranes have also been tested in membrane distillation with high viscous fluids. However, at industrial scale, hollow fibers are mainly preferred, due to their high packing density. An important parameter to consider in the module design is the length needed to achieved the desired results. As for conventional devices, also for membrane contactors, this length can be expressed by:
Module Configurations and Design 107 L = HTU x N T U
(1)
where: HTU, height o f the transfer unit; NTU, number o f transfer units. With respect to packed towers, membrane contactors can operate also in orizontal position and LTU (length of the transfer unit) can be used instead of HTU. LTU can be expressed as
[4]: LTU = v/Ka
(2)
where: v, fluid velocity; K, module-averaged overall mass transfer coefficient; a, interfacial area. Equation (2) states that, due to the very high interfacial area, membrane contactors lead to lower values of LTU than conventional units. Whatever kind of module is adopted, the principal targets to pursue for an efficient performance are: -
to maximize the mass transfer;
-
to reduce and control the fouling;
-
to work with low pressure drops,
-
to guarantee a constant performance of the module for all its length.
In order to limit the polarization phenomenon, several improvements of the first-type developed modules have been proposed during years. First of all, turbulence promoters have
!
! i i84
i 84184 U been added to reduce the boundary layer resistances [5]. Ultrasonic treatments have been proposed as a mean to increase the trans-membrane flux (up to 200%) in membrane iiii!
i i84
distillation [6]. The formation of Dean vortices, which help in mixing the streams, has been achieved by using several designs of curved membrane modules [7]. Being the hollow fiber of big interest for industrial applications, significant efforts have been addressed to the improvement of the hollow fiber module design. When the module design has to be carried out, several parameters have to be taken into account, such as membrane properties (porosity, thickness, tortuosity), packing density, fibers length and diameters, operative flowrates, pressures and concentrations, fluid physical properties, pressure drops, breakthrough pressure. The type of flow inside the module has also to be carefully chosen. For example, crossflow design leads to higher mass transfer coefficients than co/counter-current one, but the pressure drops increase too. As it will be mentioned in Chapter 4, one major limitation in the membrane contactor scale-up is the unavailability of a general correlation for the prediction of the mass transfer coefficient at the shell side. This lack is related to the non-uniform flow that can occur because of channeling, bypassing, mixing, entry region phenomena, often caused by fiber deformation, non uniform fiber distribution, polydispersity of fiber diameters, presence of stagnant zones. The first typology of hollow fiber used was the tube-in-shell configuration with parallel flow of the involved streams (Figure 2).
i 84 i i~ ii~III~iili~I;Iii~iiiiliI i~i
i~ii~iiiiiil ~ iiiiiiiiiIiii!i~ !~iI
Module Configurations and Design 109
I
Stream 2 OUT
53
~
Stream 2 IN
_
Stream 1 IN
~ ,-
Stream 1 OUT
-=
)
Figure 2. Tube-in-shell configuration with parallel flow of the involved streams.
This configuration suffers of a maldistribution of the liquid fed at the shell side that can occur for the reasons above exposed and that leads to a reduction of the mass transfer efficiency. The maldistribution of flow can be limited if the feed stream is sent in the lumen of the fibers, but this implies higher pressure drops. Based on these results, researchers dedicated a lot of efforts to improve the module design in terms of higher mass transfer coefficients at the shell side. For example, several authors introduced baffles that deviated the fluid leading to a transverse flow and to a local turbulence, with consequent increase of the mass transfer [8-9]. Figure 3 shows a baffles-containing module. Both theoretical evaluations and experimental tests carried out worldwide are presented and discussed in sections 3 and 4, respectively.
I
Stream 2 OUT
Stream 1 ~N
Figure 3. Example of a baffled module.
I
~
I
Stream 2 IN
Stream 1 OUT
110 Chapter 3
3. Theoretical studies on hollow fiber modules Several are the mathematical models developed to analyze the performance of hollow fiber modules when a not uniform flow is established at the shell side. Works on the calculation of the fiber and flow distribution in randomly packed fibers have been carried out by different authors [ 10-11 ]. More recently, Wu and Chen [ 12] evaluated for the water deoxygenation the effect of flow maldistribution on mass transfer in a randomly packed hollow fiber module, considering a parallel flow. They used a random fiber distribution model and the L6v~que's equation and, by comparing the theoretical predictions with experimental data, they concluded that the model was able to well estimate the performance of the system only when an axial laminar flow is established. The shell side mass transfer coefficient for axial flow in hollow fiber modules has been calculated by Lipniski and Field [13] who have taken into account the influence of the entrance effects and packing fraction. The influence of maldistribution has been analysed by using local mass transfer coefficients. The trends they obtained are similar to those predicted by empirical correlations over a wide range of operating flowrates and packing densities. Lemanski and Lipscomb [ 14] analysed the effect of the shell-side flow on the performance of hollow fiber gas separation modules. In their system the fluid is sent across the fibers through the inlet and outlet ports, while flowing along the fibers between the ports. In order to include both the parallel and the cross flow that are established in the module, the analysis was two-dimensional. They also considered the effect of fiber packing, pressure and velocity fields by using the Darcy's law. The model predictions
Module Configurations and Design 111 were in closer agreement with experimental data (obtained for a shell-fed air separator) than one dimensional plug flow model. The variation in fiber size is another important cause of the not uniform flow. For example, Lemanski et al. [ 15] studied the effect of the fiber properties (size, permeance and selectivity) on the performance of a cross-flow hollow fiber gas separation module, finding that the size distribution had the greatest influence on the process. Wickramasinghe et al. [ 16] proposed a correlation to take into account the difference in fiber size, assuming a Gaussian distribution of fiber radii:
k~v = k [1-9kV/ Q r~v + 7) coe + .....]
(3)
where: k, tube side mass transfer coefficient for a uniform distribution of fiber radii," kay, average tube side mass transfer coefficient; V, average volume occupied by one fiber; ray, average fiber radius; eo, standard deviation of fiber radii divided by the mean.
The costs of a membrane module are necessarily related to the membrane area packed inside. In hollow fiber modules, the membrane area depends on the fiber size and number. Wickramasinghe et al. [ 17] performed an analysis to determine the optimum fiber diameter and pointed out that lower membrane costs are possible by using small fiber diameters. However, small fibers lead to high pumping costs, thus an optimum exists and authors found it around a few hundred microns. Table 1 summarizes the theoretical studies above reported.
112 Chapter 3 Table 1. Theoretical studies on hollow fiber modules Studies
References
Random fiber distribution model
[12]
Analysis of the influence of the entrance effects and packing fraction on the shell side mass transfer coefficient for axial flow
[13]
Two-dimensional analysis of the effect of packing fraction, pressure and velocity fields on the performance of hollow fiber gas separation modules
[14]
Studies of the effect of the fiber properties on the performance of a cross-flow hollow fiber gas separation module
[15]
Analysis to determine the optimum fiber diameter
[17]
4. New module configurations In order to improve the efficiency of membrane contactors, different types of modules have been developed. As previously stated, one of the first attempts made was to insert baffles inside tube-in-shell modules, with the aim to promote transversal flow and to enhance the mass transfer coefficient. Wang and Cussler [8] used baffles in both rectangular and cylindrical modules containing a woven fabric of fibers. The rectangular module led to a lower mass transfer because of stagnant zones that were created between adjacent fibers. Woven fibers have been also tested by Wickramasinghe et al. [18], who noticed the better
Module Configurations and Design 113
performance of the system with respect to the modules built with individual fibers, because of the more uniform fiber distribution achievable. The transverse flow has been obtained by Bhaumik et al. [19] by means of a device containing the fibers in a mat wrapped around a central tube (distributor of the liquid). The authors demonstrated the efficiency of the system for the CO2 absorption in water. Figure 4 shows how this module looks like.
T Stream 2 our
T Stream 2 OUT
Hole ':~' 9 o q
T StreamliN
0 " ....o.. .....
~
Stream 2
StreamlouT
Figure 4. Transverse flow in a device containing the fibers in a mat wrapped around a central tube (From [ 19], Copyright (1998), with permission from Elsevier).
The water deoxygenation has been carried out by Wickramasinghe et al. [16] in four alternative configurations, all operating in cross-flow, and the results have been compared with those achievable in a parallel flow cylindrical module. All modules led to higher removal than the parallel flow cylindrical module (7%). The highest removal was achieved with the
114 Chapter 3
rectangular bundle module (98%), followed by the helical bundle (86%) and the cylindrical bundle (82%). The crimped flat plate led to 72% of removal. A three-phase hollow fiber membrane contactor with frame elements has been recently proposed by Vladisavljevic and Mitrovic [20]. The module is made of stacks of polygonal plates containing internal frames packed with hollow fibers. For each stack, the inlets and outlets of the fluids flowing inside the fibers are provided on an external frame. Plates can be monoaxial, if a two-phase contact is needed, or biaxial for allowing a three-phase contact. In figure 5 the monoaxial and biaxial internal frames are depicted.
Iltl!l l l (a)
(b)
Figure 5. Monoaxial (a) and biaxial (b) internal frames (From [20], Copyright (2001), with permission from Elsevier).
Authors claimed several advantages of their system over conventional parallel flow modules, such as the possibility to adjust the fiber length independently on the module length, the regular positioning of the fibers within the sheets that prevents the flow maldistribution outside the fibers, the possibility to replace the only submodules that contain the failed fibers.
Module Configurations and Design 115 Concerning the pressure drops, their value at the shell side was lower than the tube side. In latter case, authors found that the pressure drops were mainly related to the local obstacles in the module rather than the resistance in the fibers. Patents on new module designs have been recently presented. TNO (The Netherlands) patented [21] a rectangular module containing fibers located at well defined positions that ensure a good flow distribution. The system performs with high mass transfer coefficients, low pressure drops, it is easy in the scale-up and has been successfully tested in pilot plants for different applications [22]. The same Company patented a new type of module, able to operate with gaseous streams at high pressures [23]. The module houses hollow fiber membranes inside a pressure vessel and can be adopted for absorbing species such as CO2 and H2S usually present in natural gas or petrochemical streams. A spiral-wound design for membrane contactor applications has been patented by Nitto Denko of Japan [24]. The device consists of a central feed pipe around which membranes are wound. Tests on water ozonation demonstrated the possibility to obtain a water with an ozone content 10% higher than that achievable with hollow fibers. The different kind of modules developed are reported in Table 2.
116 Chapter 3
Table 2. Module design developments ml
|,
Module design
References
Baffles in both rectangular and cylindrical modules containing a woven fabric of fibers
[8]
Transverse flow by means of a device containing the fibers in a mat wrapped around a central tube
[19]
Crimped flat plate, rectangular, helical and cylindrical bundle modules operating in crossflow
[16]
Three-phase hollow fiber membrane contactor with frame elements
[201
Rectangular module containing well-located fibers
[21]
Module for high pressure gas treatments
[23]
Spiral-wound module with a central feed pipe
[24]
5. Modules layout
In practical applications, it is usually impossible to achieve the desired target by using a single module and a combination of different modules is often required. As other membrane units, membrane contactors can be assembled in series or in parallel. The former type of layout leads to higher efficiency (e.g., higher purities of the treated stream) whereas the latter increases the system capacity. In order to reach the desired performance, both types of assembling are sometimes necessary and a combination of parallel and series layouts is made (Figure 6).
Module Configurations and Design 117
v I
~
j
""t
~
J
v t
~
j
"L
1-4
l---I
,"-
--
1
Figure 6. Membrane contactors assembled in a series/parallel fashion. In defining the combinations of modules, it is important to take under control the pressure drops inside the system. This constraint is more pronounced in membrane contactors with respect to other membrane devices because of the breakthrough pressure limitations. Generally, the number of parallel modules decreases as the the number of modules in series increases, but more modules in series mean also an increase of pressure drops. Therefore, an optimum will exist. Moreover, higher the number of modules in series lower the velocity of the fluid through them and, thus, lower the mass transfer efficiency. The "tree-assembly" is sometimes adopted for working with the same fluid velocity in all modules. In this layout the fluid encounters, during its flow, a reduced cross sectional area (Figure 7). The total membrane area is now reduced, but the pressure drops are increased.
118 Chapter 3
The module layout has, then, to be optimized depending on the specific application and on the capital and operating costs. In Table 3 the main factors that influence the choice of a particular module layout are summarized.
-L
y
,-[
j___.
Figure 7. "Tree assembly" of membrane contactors.
Table 3. Main factors that influence the choice of the module layout Process capacity Pressure drops Membrane area Pumping costs Level of purity required Mass transfer efficiency
Module Configurations and Design 119 6. Commercial modules Several are the commercial modules that can be used for membrane contactors applications. Some of them are produced by Companies that are not enterely devoted to membrane contactors design. This is the case, for example, of the Microdyn Technologies, Enka (Germany) that mainly produces modules for filtration purposes but offers also modules equipped with polypropylene capillaries that have been used in membrane distillation experiments [25-26]. In order to reduce problems of fouling and headlosses that can occur in membrane contactors, Membrane Corporation (Minneapolis, MN) developed a bubble-free gas-liquid mass transfer module containing various fiber bundles that are fluidized by the liquid flowing outside. A 100% of gas transfer is obtained by sealing the fiber at one end. By working with low packing densities high turbulence, no plug for suspended solids and low pressure drops are achieved. Modules for bubble-free ozonation of water (DISSO3LVE TM) are commercialized by W.L. Gore&Associates (Elkton, MD). A helix arrangement of the ozone-resistant fibers in PTFE characterizes this device. Both hydrophobic and hydrophilic hollow fibers microporous membranes in shell-and-tube modules are commercialized by Sepracor Inc. GVS Spa (Italy) manufactures flat membrane contactors specifically developed for controlling the air humidity. The membranes used are microporous super-hydrophobic (PTFE,
120 Chapter 3 PVDF or PP treated to increase the hydrophobicity) and the module has a plate and flame configuration. Several are the advantages with respect to traditional dehumidifiers, such as up to 50% lower capital and operating costs, low pressure drops and low noise emissions [27]. Membrana-Charlotte, a division of Celgard LLC, commercializes different hollow fiber modules, among which the hollow fiber Liqui-Cel Extra Flow module. The module has been developed mainly for gas-liquid applications, such as ultrapure water production in electronical industry, water carbonation, water deareation, etc., but it has been also successfully used for liquid-liquid extractions and osmotic distillation tests [28-29]. The system has been designed for avoiding large pressure drops and for enhancing the mass transport coefficient. It is characterized by a central baffle that forces the liquid stream (sent to the shell side) to flow perpendicularly to the fibers. This implies a reduction of both mass transport resistances and shell side bypassing with respect to parallel flow devices. Furthermore, a more uniform fiber spacing is achieved thanks to a woven fabric of the fibers. Figure 8 refers to the water deoxygenation process and shows the flow pattern of the liquid stream at the shell side. The picture contains also information on the fiber size and structure and the hollow fiber array.
Module Configurations and Design 121
Figure 8. Liqui-Cel| Extra-Flow Membrane Contactor. Liqui-Cel is a registered trademark of Membrana-Charlotte, a division of Celgard LLC (with permission).
Fotos of some commercial modules of different size are reported in Figure 9. Sengupta et al. [30] developed for the LiquiCel devices a procedure for predicting the separation performance of similar contactors (similar packing densities, same hollow fiber properties) of different size.
122 Chapter 3
Figure 9. Membrane Contactors commercialized by Membrana. Liqui-Cel is a registered trademark of Membrana-Charlotte, a division of Celgard LLC (with permission).
Compact Membrane Systems, Inc. (USA) has developed composite membranes by coating different microporous supports (polypropylene, PVDF, polysulfone, etc.), both as hollow fibers and flat sheets, with a thin nonporous perfluoro layer. The composite membranes are characterized by higher contact angles than the supports and possess an excellent thermal stability. Furthermore, the presence of the dense layer allows to operate at high pressures without the problem of the breakthrough. The membranes have high gas fluxes and find application in different areas such as bubbleless introduction/removal of gases, portable oxygen for respiratory care, emissions reduction et. CMS commercializes both hollow fiber modules and flat sheet configurations. Gas permeable membranes made using patented hollow fiber technology are also employed in a contactor designed for water ozonation by Mykrolis Corp. (USA).
Module Configurations and Design 123
The main characteristics of some of the commercial modules described are summarized in Table 4.
Table 4. Main characteristics of some commercial modules Company Microdyn Technologies (Enka) GVS SpA
Membrane area (m 2)
Pore size (l.tm)
Capillary polypropylene
0.1 - 10
0.2
Flat superhydrophobic
0.5 - 5
0.02-5
Membrane type
Membrana-Charlotte Hollow fiber polypropylene
0.18 - 220
0.03
124 Chapter 3 References [1] S.I. Andersonn, N. Kjellander and B. Rodesjo. Design and field tests of a new membrane distillation desalination process. Desal., 56 (1985) 345-354 [2] D.W. Gore. Gore-Tex Membrane distillation. Proc. of the 10th Ann. Con. Water (1982) 25-29 [3] J. Koschilowski, M. Wieghaus and M. Rommel. Solar thermal-driven desalination plants based on membrane distillation. Desal., 156 (2003) 295-304 [4] A. Gabelman and S.T. Hwang. Hollow fiber membrane contactors. J. Membrane Sci., 159 (1999) 61-106 [5] L. Martinez-Diez, M.I. Vasquez-Gonzalez, F.J. Florido-Diaz. Study of membrane distillation using channel spacers. J. Membrane Sci., 144 (1998) 45-56 [6] C. Zhu and G. Liu. Modelling of ultrasonic enhancement on membrane distillation. J. Membrane Sci., 176 (2000) 31-41 [7] J.N. Ghogomu, C. Guigui, J.C. Rouch, M.J. Clifton and P. Aptel. Hollow fibre membrane module design: comparison of different curved geometries with Dean vortices. J. Membrane Sci. 81 (2001) 71-80 [8] K.L. Wang and E.L. Cussler. Baffled membrane modules made with hollow fiber fabric. J. Membrane Sci., 85 (1993) 265-278 [9] A.F. Seibert and J.R. Fair. Scale-up of hollow fiber extractors. Sep. Sci. Technol., 32 N.1-4 (1997) 573-583 [ 10]
V. Chen and M. Hlavacek. Applications of Voronoi tessellation for modeling randomly
packed hollow fiber bundles. AIChE J., 41 (1995) 2322
Module Configurations and Design 125 [ 11]
J.D. Rogers and R. Long. Modeling hollow fiber membrane contactors using film theory,
Voronoi tessellations and facilitation factors for systems with interface reactions. J. Membrane Sci., 134 (1997) 1 [12]
J. Wu and V. Chen. Shell-side mass transfer performance of randomly packed hollow fiber
modules. J. Membrane Sci. 172 (2000) 59-74 [13]
F. Lipniski and R.W. Field. Mass transfer performance for hollow fibre modules with shell-
side axial feed flow: using an engineering approach to develop a framework. J. Membrane Sci., 193 (2001) 195-208 [14]
J. Lemanski and G.G. Lipscomb. Effect of shell-side flows on the performance ofhollow-fiber
gas separation modules. J. Membrane Sci., 195 (2001) 215-228 [ 15]
J. Lemanski, B. Liu and G. G. Lipscomb. Effect of fiber variation on the performance of
cross-flow hollow fiber gas separation modules. J. Membrane Sci. 153 (1999) 33-43 [ 16]
S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Mass transfer in various hollow fiber
geometries. J. Membrane Sci., 69 (1992) 235-250 [17]
S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Better hollow fiber contactors. J.
Membrane Sci., 62 ( 1991) 371-388 [ 18]
S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Hollow fiber modules made with
hollow fiber fabric. J. Membrane Sci., 84 (1993) 1-14 [ 19]
D. Bhaumik, S. Majumdar and K.K. Sirkar. Absorption of C O 2 in a transverse flow hollow
fiber membrane module having a few wraps of the fiber mat. J. Membrane Sci., 138 (1998) 77-82 [20]
G.T. Vladisavljevic and M.V. Mitrovic. Pressure drops and hydraulic resistances in a three-
phase hollow fiber membrane contactor with flame elements. Chem. Eng. Proc., 40 (2001) 3-11
126 Chapter 3 [21]
B. Meulen Ph. Ter. Transfer device for the transfer of metter and/or heat from one medium
flow to another medium flow. EU Patent 0509031 US Patent 5230796 [22]
P.H.M. Feron and A.E. Jansen. CO2 separation with polyolefin membrane contatcors and
dedicated absorption liquids: performances and prospects. Sep. Puris Technol., 27 (2002) 231-242 [23]
TNO (The Netherlands), US 6,355,092
[24]
Nitto Denko (Japan), US 6,168,648
[25]
E. Drioli, A. Criscuoli and E. Curcio. Integrated membrane operations for seawater
desalination. Desal., 147 (2002) 77-81 [26]
A. Capuano, B. Memoli, V.E. Andreucci, A. Criscuoli and E. Drioli. Membrane distillation of
human plasma ultrafiltrate and its theoretical applications to haemodialysis techniques. Int. J. Artis Org., 23 N.7 (2000) 415-422 [27]
S.N. Gaeta. Membrane contactors: application to industrial processes. Proc. of 1st Workshop
Italy-Russia, Cetraro (CS), Italy, September 17-20 (2003) 51-55 [28]
B.W. Reed, R. Klassen, A.E. Jansen, J.J. Akkerhuis, B.A. Bult and F.I.H.M. Oesterholt.
Removal of hydrocarbons from wastewater by membrane extraction. Proc. AIChE Spring National Meeting, Atlanta, April 17-21 (1994) [29]
P.A. Hogan, R.P. Canning, P.A. Peterson, R.A. Johnson and A.S. Michaels. A new option:
osmotic distillation. Chem. Eng. Prog., (1998) 49-61 [30]
A. Sengupta, P.A. Peterson, B.D. Miller, J. Schneider and C.W. Fulk Jr. Large-scale
application of membrane contactors for gas transfer from or to ultrapure water. Sep. Purif. Technol., 14 (1998) 189-200
Chapter 4. Gas- liquid systems 1. Introduction In this Chapter the transport of species from a gas phase to a liquid phase, and viceversa, is discussed in detail. The gas-liquid equilibria and the resistances to the mass transport involved are illustrated and the mass transfer coefficients calculated. Mass balances across the membrane are reported. As example of calculation, the sparkling water production by membrane contactors is described. This specific application of membrane contactors includes a simultaneous transfer of species from/to the gas phase to/from the liquid phase. The aim of the Chapter is to furnish the knowledge about the phenomena that regulate, in membrane contactors, the mass exchange between a gas and a liquid, and the equations required for describing these systems. The analysis hereinafter presented is general and can be extended to all the applications of membrane contactors where a gas and a liquid are involved, such as gas streams purification, water ozonation, water deoxygenation, and so on.
128 Chapter 4
2. Mass transfer equations As already stated, both hydrophobic and hydrophilic microporous membranes can be used to put in contact a gas with a liquid. Composite (microporous - dense or hydrophobic hydrophilic) membranes are also useful for this purpose.
2.1. Hydrophobic membranes Figure 1 shows the concentration profiles that are formed, because of the resistances to the mass transfer, when a species i is transferred from the gas phase to the liquid and a hydrophobic flat membrane is used. If i is transferred from the liquid to the gas, the concentration profile is that depicted in Figure 2. The resistances to the mass transfer encountered in both cases are those offered by the boundary layers and the membrane and can be drawn, as in Figure 3, by considering an electrical analogy.
Gas- Liquid Systems 129
Figure 3. Mass transfer resistances involved in the transport.
Referring to Figure 1, at steady-state the flux of the species i through the gas film equals its flux through the membrane, as well as its flux through the liquid film and the following equation can be written for the flat membrane:
J, = k~g (C,g - G m d = k~.. (C~mg- G e d = k, (C,e- GO
where:
(1)
130 Chapter 4
k~g, mass transfer coefficient in the gas phase for the species i," kim, mass transfer coefficient in the hydrophobic membrane for the species i; kit, mass transfer coefficient in the liquid phase for the species i; C~g, concentration of the species i in the gas phase; C~mg, concentration of the species i at the g a s - membrane interface; C~eg, concentration of the species i at the g a s - liquid interface; C~e, concentration of the species i at the liquid-gas interface; Ca, concentration of the species i in the liquid phase. In the case of transfer of the species i from the liquid to the gas phase, the same equation can be used, by only changing the sign in each flux (the transfer occurs in the opposite direction).
The pressure and the concentration of the species i at the gas - liquid interface can be related each other by equilibrium expressions. When a gas is in contact with a liquid it dissolves until equilibrium is established. The concentration of a gas into a liquid, under equilibrium conditions, is related to its partial pressure by the Henry' s constant [ 1]:
Pt =/-/~ C~.
where:
P, gas partial pressure; H, Henry's constant; C, gas content into the liquid.
(2)
Gas- Liquid Systems 131
Figure 4. Gas-liquid equilibrium. Equation (2) is useful to calculate the gas content into a liquid, once its partial pressure is known, or to calculate the partial pressure of the gas, once its concentration into the liquid is available. In the development of the equations that describe the mass transport in membrane contactors, the Henry's law can be used to link the pressure and the concentration of the species i at the g a s - liquid interface as follows:
Pie : Hi Cie
(3)
The flux of the species i can be also expressed in terms of the overall mass transfer coefficient:
Ji = Kg (Cig- Cigiaeat) = Kt (C/aeat- Ca)
(4)
where:
Kg, overall mass transfer coefficient based on gas phase," Kt, overall mass transfer coefficient based on liquid phase; C,gideal., concentration of the species i in the gas phase ideally in equilibrium with its concentration in the liquid phase;
132 Chapter 4
Ci ideat, concentration o f the species i in the liquid phase ideally in equilibrium with its concentration in the gas phase.
By considering equation (2): p/dear= Hi Gt
(5)
Ci ideat =P i~/Hi
(6)
The overall mass transfer coefficients can be expressed in terms of single mass transfer coefficients, by combining equations (1) and (4) and taking into account the equilibrium expressions previously reported:
1/Kt = 1/ka + 1/(kim nadi)
-k-
1/(kig nad~
1/Kg = Haa~/kit + 1/k~m+ 1/kig
(7)
(8)
where: Haas, adimensional Henry's constant.
Equations (7) and (8) state that the overall resistance offered to the mass transport is due to the liquid film resistance, the membrane resistance and the gas film resistance, as depicted in Figure 3, and are valid also for the transport of the species i from the liquid to the gas phase.
Gas- Liquid Systems 133 When an hollow fiber configuration is considered with the liquid phase in the shell side and the gas phase at the lumen side the interface is located at the outer diameter of the tubes and equations (7) and (8) change into:
1/(Kt do) = 1/(kits do) + 1/(kim nadi dtm)+ 1/(kigt Haaiat.)
1/(Kgdo) = Haai /(kasdo) + 1/(kimdtm)+ 1/(kigt dO where:
kits, mass transfer coefficient for the species i in the liquid at the shell side; kigt, mass transfer coefficient for the species i in the gas at the tube side; di, inner diameter of the tube; do, outer diameter of the tube; dtm, logarithmic mean of the hydrophobic membrane diameters.
(9)
(10)
134 Chapter 4 2.2. Hydrophilic membranes Figures 5 and 6 show the concentration profiles for the transfer of the species i from the gas and from the liquid, respectively, in a hydrophilic flat membrane.
Figure 5. Concentration profile for the species i when it moves from a gas phase towards a liquid phase through a microporous hydrophilic flat membrane.
Figure 6. Concentration profile for the species i when it moves from a liquid phase towards a gas phase through a microporous hydrophilic flat membrane.
Referring to Figure 5, as for the hydrophobic membrane, at steady state the flux of the species i through the gas film equals its flux through the membrane and its flux through the liquid film:
J, = k,g (c,g- G ~ = k',m (C,~- C,.~ = k, (C,m- G~ where:
Cim, concentration of the species i at the liquid- membrane interface," k'im, mass transfer coefficient in the hydrophilic membrane for the species i.
(11)
Gas- Liquid Systems 135 By considering the equilibrium expression for the gas - liquid interface, the overall mass transfer coefficients can be related to the single mass transfer coefficients by:
1/KI : 1/ka + 1/k'im + 1/(kig Had)
(12)
1/Kg = Hadi/ka + Hadi/k'im + 1/k~g
(13)
In an hollow fiber configuration with the liquid phase in the shell side and the gas phase at the lumen side the interface is located at the inner diameter of the tubes and equations (12) and (13) become:
1/(Kt d) : 1/(kas do) + 1/(k'im d'tm) + 1/(k, gt Had, d)
(14)
1/(Kg di) = Hadi /(kits do) + Hadi /(k'im d'ln) + 1/(kigt dt)
(15)
where:
d'tm, logarithmic mean o f the hydrophilic membrane diameters.
In the case of asymmetric membranes, if the membranes are hydrophobic or hydrophilic along all the thickness, the equations that describe the mass transport and the mass transfer coefficients are those previously reported.
136 Chapter 4 2.3. Partially wetted and hydrophobic-hydrophilic composite membranes When the membrane micropores are partially hydrophobic, for example, because of the different breakthrough pressure values along the thickness of asymmetric membranes or, more generally, because of a partial lose of the hydrophobic character, the equations describing the mass transport slightly vary with respect to those illustrated. Let consider the transport of the species i from the gas to the liquid phase through a partially wetted asymmetric flat membrane. As shown in Figure 7, the species encounters four resistances during its transfer: the gas film resistance, the hydrophobic membrane resistance, the hydrophilic membrane resistance and the liquid film resistance. The g a s -
liquid
equilibrium now occurs within the membrane pores.
Figure 7. Concentration profile for the transport of the species i from the gas towards the liquid phase through a partially wetted asymmetric fiat membrane.
At steady state, the following equation is valid for the flux:
J~ = k~g (C~g - C,mg) = k~m (C~mg -- C,~g) = k',m (C~e - Ctm) = kit (C~m - C,t)
(I 6)
Gas- Liquid Systems 137 The overall mass transfer coefficients are linked to the single mass transfer coefficients by:
I/Kt = I/ka + I/('ki m Hadi) + 1/k'im + 1/(kig Haai) (17)
1/Kg = Haai /kit + 1/kim + Hadi /k'im + I/kig
(18)
The same equations are valid for describing what occurs in hydrophobic-hydrophilic composite flat membranes, as well as in symmetric partially wetted flat membranes. However, it has to be pointed out that usually for partially wetted membranes (both asymmetric and symmetric), it is quite difficult to determine the location of the interface within the membrane pores and, then, the calculation of the membrane mass transfer coefficient becomes hard. On the contrary, in hydrophobic - hydrophilic composite membranes the interface coincides with the hydrophobic-hydrophilic interface and the membranes mass transfer coefficients can be easily determined.
For the hollow fiber configuration with the liquid phase at the shell side and the gas in the lumen side, the above equations change into:
1/(Kt di,,t) = if(kits do) + 1/(kim H.ai dtm) + 1/(k'imd'tm) + 1/(kigt Hadi dO
(19)
1/(Kg d,,t) = H,,di/(kits do) + 1/(k~m dt~ + H,,d~/(k'im d'lm) + 1/(k~gt dO
(20)
138 Chapter 4 where"
di,,t, interfacial diameter. Figure 8 shows the diameters involved in equations (19) and (20) when a hydrophobichydrophilic composite hollow fiber membrane section is considered.
Hydrophilic meml~rane section "222
Liquid phase
Liquid phase ~ / ~ G a s phase ....
,.x~yr I I I [~, I~
r--------_~ ~=Ts2 I~.. --.~
di dint do
~1 v
I I
~l
Hydrophobic membrane section
[ ,~1
Figure 8. Section of the hydrophobic-hydrophilic composite hollow fiber membrane.
Gas- Liquid Systems 139 2.4. Microporous- dense composite membranes Figure 9 shows the profiles that are established in a composite m i c r o p o r o u s - dense flat membrane during the transport of the species i from the gas to the liquid phase.
Figure 9. Concentration profile for the transport of the species i from the gas towards the liquid phase through a composite microporous-dense fiat membrane.
The species i has now to diffuse also through the dense layer and the gas comes in contact with the liquid on the wetted dense surface. The equality of the flux, at steady state, leads to:
Ji = kig (Gig- Cimg) = kim (Cimg- Cim int) = kdim (Cim int- Cieg) = kil ( f i e - Cil)
(21)
where: Cim int, concentration of the species i at the microporous- dense interface; kaim, mass transfer coefficient in the dense membrane for the species i.
The equations that correlate the overall mass transfer coefficients to the single mass transfer coefficients are:
140Chapter4 1 / g l -- 1/kil nt- 1/(kim nadi) nt- 1/(kdim
1/Kg = Had~/kit + 1/kim
+
Hadi) + 1/(k,g Had#
(22)
1/kaim + 1/k~g (23)
With this type of membranes the liquid is always in contact with the dense layer. In a hollow fiber configuration this means that the location of the dense layer (inside or outside the microporous tube) determines the side where the liquid phase has to be sent. Figures 10 and 11 show the case of a coating of the dense skin on the external and on the inner surface of the microporous tube, respectively.
Liq
~ Gas~- Liq
Gas
Liq - ~
Gas
B I
I
L. did I~ ,-.q
do
~1 ,v.
Figure 10. Dense skin coated on the external surface of the microporous tube.
L d~ d do
I .J ,v
Figure 11. Dense skin coated on the inner surface of the microporous tube.
Gas - Liquid Systems 141
Referring to the above Figures, the equations to be used to link the overall mass transfer coefficient to the single mass transfer coefficients are reported below.
External coating I/(Kt do) = 1/(ka~ do) + 1/(k~m Haai dtm) + 1/(kdim nadi ddt,n) + 1/(kigt Hadi dO
(24)
I/(Kgdo) = Haai/(kitsdo) + 1/(k, mdtm)+ 1/(kdmddlnt) + 1/ (k~g~lO
(25)
where:
datm, logarithmic mean of the dense membrane diameters.
Internal coating 1/(Kt dd = 1/(kitt dd + 1/(kim nadi dtm) + 1/(kdim nadi ddlm) + 1/(kigs Hadi do)
(26)
1/(Kg di) = Haai /(kitt dO + 1/(kim dtm)+ 1/(kdm dlm) + 1/(kigs do)
(27)
where:
kin, mass transfer coefficient for the species i in the liquid at the tube side; kigs, mass transfer coefficient for the species i in the gas at the shell side.
Table 1 and 2 summarize the expressions of the overall mass transfer coefficients for the different types of membranes described.
142 C h a p t e r 4 Table 1. Expressions of the overall mass transfer coefficients for the different flat membranes I
Membrane
/
I|
1/Kl
l/Kg
Hydrophobic
1/kil + 1/(kim Hadi) + 1/(kig Hadi)
Hadi/kil + 1/kim+ 1/kig
Hydrophilic
1/kil + l/k'im + 1/(kig Hadi)
Hadi/kil + Hadi/k'im + 1/kig
Partially wetted (symmetric and asymmetric) and hydrophilichydrophobic composite
1/kil + 1/(kim Hadi) + 1/k'im + 1/(kig Hadi)
Hadi/kil + 1/kim + Hadi/k'im + 1/ kig
Microporous-dense composite
1/kil + 1/(kim Hadi) + l/(kdim Hadi) + 1/(kig Hadi)
Hadi/kii + 1/kim + 1 &dim+ l/ kig .
.
.
.
.
.
Table 2. Expressions of the overall mass transfer coefficients for the different hollow fiber membranes. Operating conditions: liquid phase at the shell side and gas phase in the lumen ,,
,,
Membrane Hydrophobic
l/(Ki do) = 1/(kils do) + 1/(kim Hadi dim) l/(Kgdo) = Hadi/(kilsdo) + + 1/(kigt Hadidi) l/(kim dim)+ 1/(kigt di)
Hydrophilic
1/(Kl di) = 1/(kils do) + 1/(k'im d'lm) + 1/(kigt Hadi di)
Partially wetted (symmetric and asymmetric) and hydrophilichydrophobic composite
1/(Kl dint) = 1/(k,s do) + 1/(kim Hadi dim) 1/(Kg dint) = Hadi/(kils do) + + 1/(k'im d'lm) + 1/(kigt Hadi di) 1/(kim dim)+ Hadi/(k'im d'lm) + 1/( kigt di)
Microporous-dense composite
1/(Ki do) = 1/(kils do) + l/(kim Hadi dim) 1/(Kg do) = Hadi/( kilsdo) + + l/(kdim Hadi ddim) + 1/(kigt Hadi di) 1/(kim dim)+ 1/(kdim ddjm) + 1/ (kigt di)
1/(Kgdi) = Hadi/(k.s do) + Hadi/(k'im d'lm) + 1/(kigt di)
* 1/(Kl di) = 1/(kilt di) + 1/(kim Hadi dim) * 1/(Kg di) = Hadi/( kilt di) + + 1/(kdim Hadi ddlm) + 1/(kigs Hadi do) 1/(kim dim)+ 1/(kdim ddlm) + 1/ (kigs do) ,
* Liquid in the lumen and gas at the shell side
The equations that express the overall mass transfer coefficient in terms of single mass transfer coefficients can be simplified for specific situations. For example, the resistance
.,
Gas- Liquid Systems 143 offered by a gas - filled membrane for the transfer of a gaseous species i can usually be neglected with respect to the boundary layers resistances. However, when the species i is highly soluble or rapidly reacts in the liquid phase, the membrane resistance becomes important. In gas-liquid operations, reactive liquids are often used as absorbers in order to increase the removal efficiency. In these cases, the enhancement factor (E) is introduced to express the effect of the chemical reaction on the absorption and the overall mass transfer coefficient based on the gas phase for a flat hydrophobic membrane is calculated by [2]:
1/Kg = Ho~ /(ka E) + 1/kim+ 1/kig
(28)
In particular, for an instantaneous reaction the enhancement factor assumes the following form [3]:
E = 1 + DRR/(vDcC~,~)
(29)
where: Cint, interfacial concentration o f the species that is absorbed; R, reactant concentration; v, stoichiometric coefficient.
Figure 12 compares the fluxes of carbon dioxide achievable through a hollow fiber when the absorbing liquid is water (physical absorption) or an aqueous NaOH solution [4]. The
144 Chapter 4 effect of chemical reaction on carbon dioxide absorption is also reported in Figure 13 in terms of overall mass transfer coefficient [5]. 10-2
o
9
NaOH
10 -3
9
H20
10 -4 0.0
I
I
0.2
0.4
I 0.6
0.8
V l ( m s -1)
Figure 12. Effect of chemical reaction on carbon dioxide absorption (From [4], Copyright (2002), with permission from Elsevier). 0.04
DEA 1 M '~
0.02 _ DEA 0.5 M
H20
9
0.00 0.0
90
180
Q1 (c m3 s -l)
Figure 13. The overall mass transfer coefficient for the carbon dioxide absorption with different absorption liquids (From [5], Copyright (2003), with permission from Elsevier).
G a s - L i q u i d Systems 145 Tables 3 and 4 summarize the different simplifications that can be made for the several types of membrane contactors described in this Chapter.
Table 3. Simplifications of the expressions for the calculation of the overall mass transfer coefficients for the different fiat membranes Membrane Hydrophobic Hydrophilic Partially wetted (symmetric and asymmetric) and hydrophilichydrophobic composite Microporous-dense composite (with high permeable dense skin)
Low soluble species in the liquid
High soluble species or rapid chemical reaction in the liquid
1/Kl = 1/kil
1/KI = 1/(kim Hadi) + 1/(kig Hadi)
1/Kg = Hadi/kil
1/Kg = 1/kim+ 1/kig
I/KI = l/kil + 1/k'im
1/Ki = 1/(kig Hadi)
1/Kg = Hadi flKil + Hadi/k'im
1/Kg = l/kig
l/K1 = 1/kil +l/k'im
1/KI = l/(kim nadi) + 1/(kig Hadi)
1/Kg = Hadi/kil + Hadi/k'im
1/Kg = 1/kim + 1/ kig
1/Kl = 1/kil
1/Kl =1/(kim Hadi) + 1/(kdim Hadi) + 1/(kig Hadi)
1/Kg = Hadi/kii
1/Kg = 1/kim + 1/kdim + 1/kig
146 C h a p t e r 4 Table 4. Simplifications of the expressions for the calculation of the overall mass transfer coefficients for the different hollow fiber membranes. Operating conditions: liquid phase at the shell side and gas phase at the lumen side Membrane
Hydrophobic
Hydrophilic
Partially wetted (symmetric and asymmetric) and hydrophilichydrophobic composite
Microporous--dense composite (with high permeable dense skin)
Low soluble species in the liquid
High soluble species or rapid chemical reaction in the liquid
1/(Kl do) = 1/(kils do)
1/(KI do) = 1/(kim Hadi dim) + 1/(kigt Hadi di)
1/(Kg do) = Hadi/(kils do)
1/(Kg do) = 1/(kim dim)+ 1/(kigt d~)
1/(Ki di) = 1/(k, ls do) + 1/(k'im d'lm)
1/(Ki di) = 1/(kigt Hadi di)
1/(Kg di) d'lm)
1/(Kg di) = 1/(kigt di)
= Hadi/(kils
do)
+ Hadi/(k'im
1/(Kl dint) = 1/(kils do) + 1/(k'im d'lm)
1/(Kl dint)= 1/(kim Hadi dim) + 1/(kigt Hadi di)
1/(Kg dint) = Hadi/(kils do) + Hadi/(k'im d'lm)
1/(Kg dint) = 1/(kim dim) + 1/( kigt di)
1/(K1 do) - 1/(ki~sdo)
1/(Kl do) = 1/(kim Hadi dim) + 1/(kdim Hadi ddlm) + 1/(kigt Hadi d~)
1/(Kg do) = Hadi/( kils do) * 1/(Kl di) = l/(kilt di)
* 1/(Ks di) = Hadi/( kilt di)
* Liquid at the lumen side and gas at the shell side
1/(Kg do)= 1/(kim dim)+ 1 /(kdim ddlm) + 1/(kigt di) * 1/(Kl di)= 1/(kim Hadi dim) + 1/(kdim Hadi ddlm) + 1/(kigs Hadi do) * 1/(Kg di) = 1/(kim dim)+ 1 /(kdim ddlm) + 1/(kigs do)
G a s - Liquid Systems 147 2.5. Individual mass transfer coefficients
The individual mass transfer coefficients can be calculated by means of appropriate correlations. In the following part, the equations to determine the membrane mass transfer coefficients for the different types of membrane contactors are reported. The equations to calculate the mass transfer coefficients into the phases are also furnished for a hollow fiber configuration of the membrane module (tube side mass transfer coefficient and shell side mass transfer coefficient) that is the configuration more used for industrial applications. These equations are based on the dependence of the Sherwood number (Sh = k dD 1) on Reynolds
(Re = p v dp -1) and Schmidt (Sc - / 3 p - l D "1) numbers, as well as the operating mode (e.g., parallel or crossflow).
2.5.1. Membrane mass transfer coefficient The membrane mass transfer coefficient strongly depends on the phase present into the pores. For microporous membranes, three different cases can occur: 1. gas-filled micropores, hydrophobic membranes; 2. liquid-filled micropores, hydrophilic membranes; 3. gas and liquid-filled micropores, partially wetted membranes.
When a microporous-dense composite membrane is considered, the resistance offered by the dense layer has also to be included in the evaluation of the overall membrane mass transfer coefficient.
148 Chapter 4
Hydrophobic mem.branes In hydrophobic membranes the pores are gas-filled and, depending on the ratio between the membrane pore radius and the mean free path of the species i (rp/~,i), the species can be transferred mainly by Knudsen or/and viscous flow [6, 7]. When the ratio is much smaller than 1, then Knudsen flow occurs, whereas, if the ratio is much bigger than 1, viscous flow dominates; both flows can co-exist for intermediate values of the ratio. Very often, in membranes used for gas-liquid transfer, Knudsen flow prevails and the expression for the membrane mass transfer coefficient is"
kiz = Dkig e/r8
(3 0)
with [6, 7]." Dkig = 2 rJ3 (8 RTl(zcMi))~
(31)
where:
Dkig, Knudsen diffusion coefficient for the species i," e, membrane porosity; r,membrane tortuosity; 6, membrane thickness; Mi, molecular weight of the species i; R, gas constant," T, temperature.
Gas- Liquid Systems 149 Hydrophilic membranes Hydrophilic membranes have micropores filled by the liquid phase and the transport of the species i generally depends on its diffusion coefficient into the liquid. The membrane mass transfer coefficient can be calculated by [6]'
k'im = Dit e/rfi
(32)
where:
Dit, diffusion coefficient of the species i in the liquid phase.
Partially wetted membranes In partially wetted membranes the micropores are both gas and liquid-filled and the membrane mass transfer coefficient is a combination of the hydrophobic and hydrophilic membrane mass transfer coefficients:
II kpW~m= 1/kim + 11 k'im
(33)
with: kim = lJ'ig e/r6dry
(34)
k 'im
(35)
=
Da e/rlff~vetted
where:
kpWim,partially wetted mass transfer coefficient,"
150 Chapter 4
6dry, dry thickness of the membrane; t~wetted, wetted thickness of the membrane.
Microporous-dense composite membranes For a microporous-dense composite membrane the membrane mass transfer coefficient is expressed as function of the individual microporous and dense mass transfer coefficients:
1/kmdim = 1/kim + 1/kd~m
(36)
where:
kmaim, microporous-dense composite mass transfer coefficient," kaim, dense skin mass transfer coefficient. The mass transfer coefficient for the dense skin depends on its permeability to the species i:
kdim
=f (Pei)
(37)
2.5.2. Tube side mass transfer coefficient The tube side mass transfer coefficient can be usually well described by the L6v6que's equation [8]:
Sh = 1.62 (d2 v/(LD)) ~ where: v, fluid velocity; d, fiber diameter; L, length of the fiber;
(38)
Gas- Liquid Systems 151 D, diffusion coefficient of the species i into the fluid
Some discrepancies between theoretical predictions and experimental results have been reported for Graetz number (Gz = mp -1 D 1 L -1) less than 10, due to the un-uniform flow that sometimes occurs because of the polydispersity in hollow fiber diameter [9].
2.5.3. Shell side mass transfer coefficient For the shell side mass transfer coefficient, no general expression is available to describe the mass transport, probably due to maldistribution of flow, presence of stagnant zones, splitting and remixing of streams, bypassing or channeling phenomena that can occur more frequently than in the tube side and that can be caused by the un-uniform fiber distribution and/or fiber deformation. For example, although the Reynolds values are usually in the laminar range, localized turbulence or channeling around the fibers can lead to a higher dependence of the mass transfer coefficients on the Reynolds number than that typical of a laminar regime. The packing density also affects the performance of the system. Costello et al. [10] have found that the mass transfer coefficient increases with packing density until a density of 65% is reached and then decreases for further increases in density. Furthermore, the coefficient varies also with the type of flow (parallel or crossflow) that is established inside the module and the expressions available in literature depend on the particular system analyzed [8-10, 18-28]. In order to obtain a more uniform spacing of membrane fibers, fibers woven into a fabric have also been studied [11,12]. The role of
152 Chapter 4 packing density, entrance effects, fibers maldistribution and properties (e.g., size, permeance, selectivity, etc.) and pressure fields has been studied in details by several authors [ 13-16], as already reported in Chapter 3. In Table 5 some of the correlations developed to calculate the shell side mass transfer coefficients are summarized. Information on the systems for which some of them were derived has been reported in more detail by Gabelman and Hwang [ 17].
Gas- Liquid Systems 153 Table 5. Correlations for the calculation of the shell side mass transfer coefficient Equation
Applicability range
Reference
Parallel flow Sh* = 13de/L (1-~) Re ~176S c 033
0
Sh = 1.25 (Re dffL)093 S c 0"33
0.5
[8]
Sh =0.019 Gz
Gz<60; closely packed fibers
[9]
21
[10]
Re = 100; loosely-closely packed fibers
[ 19]
Sh = 0.15 Re ~176S c 033
Re>2.5
[9]
Sh = 0.12 Re
Re<2.5
[9]
Sh = (0.53-0.58~) Re ~ Sh = (0.31 ~2-0.34~ +0.10)Re~176176
Sc 0'33
[ 18]
Crossflow
S c 0'33
Sh = 1.38 Re TM Sc ~
l
[8]
Sh = 0.90 Re ~176Sc ~
l
[8]
Sh = ~ Re !2 Sc ~
0.03
[20]
Re based on the hydraulic diameter of the bed; ~, 0.50
[21 ]
0.01
[22]
Sh = 1.76 Re ~
Sc ~
Sh** = 0.57 Re TM
S c 033
* fl is 5.8for hydrophobic and 6.1for hydrophilic membranes ** hollow fiber fabric
The difference in results achievable by using the various correlations proposed in literature is further evidenced in Figure 14. The Figure reports the variation of Sherwood number with Reynolds for three of the expressions developed for crossflow at low Reynolds values: Sh = 0.12 Re SC TM
(39)
154 Chapter 4 S h = ~ R e 12 S c 0"33
(40)
S h = 0 . 5 7 R e 031 Sr TM
(41)
The fiber packing fraction in equation (40) is that of the 2.4" x 8" LiquiCel module for which the correlation has been developed [20]. The system considered here is the removal of oxygen from water and the physical properties of the water, as well as the diffusion coefficient of the oxygen in water, refer to 25~
3
(41)
2,52t,-,
J
1,51-
(40) o
0,5-
_~
0 0
(39)
I
I
I
0,1
0,2
0,3
0,4
Re Figure 14. Sherwood number vs Reynolds for three of the expressions developed for crossflow at the shell side.
Gas- Liquid Systems 155 A simultaneous gas transfer from /to a liquid phase characterizes the sparkling water production by membrane contactors. In sparkling water production carried out in a hydrophobic membrane contactor dissolved oxygen is transferred from the liquid water to the C02 stream while C02 dissolves into the water. Figure 15 shows the streams involved in the operation.
Ql
Cil in
.11
"l .,i Ci gout
"
MembraneContactor
I .~
I
I
Ci lout
Qz
"" Ci gin
TM
Figure 15. Streams involved in the sparkling water production by membrane contactors.
with:
Qt, water flow rate," Qg, carbon dioxide flow rate; Ci t;,, inlet concentration of the species i in the liquid," Ci tout, outlet concentration of the species i in the liquid," Ci gi,, inlet concentration of the species i in the gas; Ci go,t, outlet concentration of the species i in the gas.
Water, 02
Water, C02
co=
0
~176 z
()
r C02, 02
Figure 16. Transfer of oxygen and carbon dioxide through a membrane fiber.
156 Chapter 4
Referring to a hollow fiber module, with the water flowing at the shell side and the CO2 at the lumen side (as in Figure 16), a differential mass balance along the module for the generic species i can be written as follows:
Liquid phase dCa/dz = J~ 2 ~rfNf/Qt
(42)
boundary conditions." z = O, C~t=Ca~ z =L, Ca=Ci tout
..Gas phase
dCig/dz = Jj 2 rOTNI/Qg
(43)
boundary conditions." z = O, C~g=Cig~ z =L, C~g=C~gout where: rf, outer fiber radius; Nfi number of fibers; L, length o f the fibers; z, module local length. By neglecting any interaction among species, the fluxes of the different species involved (02,
C02) can be calculated in terms of overall mass transfer coefficient based on the liquid phase and difference in concentration between the phases:
Gas- Liquid Systems 157 Jo2=Kto2 (Co2t- Po2/Ho2)
(44)
Jco2=Klco2 (Pco2/Hco2- Cco2l)
(45)
For a counter-current flow, considering the equations (44) and (45), the differential mass balances become:
Liquid phase
dCo2t/dz = ,]02 2 rof Nf /Ql
(46)
dCco2l/dz = - dco2 2 rof Nf /Ql
(47)
Gas phase dCo2~dz = ,lo2 2 roTN/Qg
(48)
dCco2g/dz = -Jco2 2 roy N/Qg
(49)
The process is controlled by the liquid resistance and the overall mass transfer coefficient coincides, for each species, with the liquid mass transfer coefficient:
1/Kli = 1/kli
(50)
By using an appropriate correlation for kli, it is possible to solve the differential mass balances and, thus, to determine theoretically how a given membrane contactor behaves in
158 Chapter 4 terms of oxygen removal and water carbonation, and how the operating conditions affect the performance of the system. Figures 17 and 18 show, respectively, an example of the influence of Reynolds and gas flow rate on the oxygen removal [20].
~,
90
=
80-
~-
70-
0
6050
0
I
I
I
0,1
0,2
0,3
Re
Figure 17. Inflence of Reynolds on the oxygen removal. Afier [20].
0,4
G a s - Liquid Systems 159
80
m
m > 70 0
E !_.
e,,
~
60
X
o
50
I
0
50
I
i
100
150
I
I
I
200
250
300
Gas flowrate (ml/min)
Figure 18. Influence of the gas flow rate on oxygen removal. After [20].
350
160 Chapter 4 References [ 1] Perry's Chemical Engineers' Handbook. R.H Perry, D. W. Green and J.O. Maloney (Eds), 6th edition, McGraw-Hill Book Co., New York (1984) [2] H. Kreulen, C.A. Smolders, G.F. Versteeg and W.P.M. van Swaaij. Microporous hollow fibre membrane modules as gas-liquid contactors. Part 2. Mass transfer with chemical reaction. J. Membrane Sci., 78 (1993) 217-238 [3] Z. Qi and E.L. Cussler. Microporous hollow fibers for gas absorption. II. Mass transfer across the membrane. J. Membrane Sci., 23 (1985) 333-345 [4] P.S. Kumar, J.A. Hogendoorn, P.H.M. Feron and G.F. Versteeg. New absorption liquids for the removal of CO 2 from dilute gas streams using membrane contactors. Chem. Eng. Sci., 57 (2002) 1639-1651 [5] M. Mavroudi, S.P. Kaldis and G.P. SakeIlaropoulos. Reduction of CO2 emissions by a membrane contacting process. Fuel, 82 (2003) 2153-2159 [6] K.K. Sirkar. Other new membrane processes, in: W.S.W. Ho and K.K. Sirkar (Eds.), Membrane Handbook, Chapman and Hall, New York (1992) 885-912 [7] M.H.V. Mulder. Basic Principle of Membrane Technology., second edition, Kluwer Academic Publishers, The Netherlands (1996) 225-228 [8] M.-C. Yang and E.L. Cussler. Designing hollow-fiber contactors. AIChE J., 32 (1986) 1910-1915 [9] S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Mass transfer in various hollow fiber geometries. J. Membrane Sci., 69 (1992) 235-250 [10]
M.J. Costello, A.G. Fane, P.A. Hogan and R.W. Schofield. The effect of shell side
hydrodynamics on the performance of axial flow hollow fibre modules. J. Membrane Sci., 80 (1993) 1-11
Gas - Liquid Systems 161
[ 11 ]
K.L. Wang and E.L. Cussler. Baffled membrane modules made with hollow fiber fabric. J.
Membrane Sci., 85 (1993) 265-278 [ 12]
S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Hollow fiber modules made with
hollow fiber fabric. J. Membrane Sci., 84 (1993) 1-14 [ 13]
S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Better hollow fiber contactors. J.
Membrane Sci., 62 (1991) 371-388 [ 14]
J. Lemanski, B. Liu and G. G. Lipscomb. Effect of fiber variation on the performance of
cross-flow hollow fiber gas separation modules. J. Membrane Sci. 153 (1999) 33-43 [ 15]
J. Lemanski and G.G. Lipscomb. Effect of shell-side flows on the performance of hollow-fiber
gas separation modules. J. Membrane Sci., 195 (2000) 215-228 [16]
F. Lipniski and R.W. Field. Mass transfer performance for hollow fibre modules with shell-
side axial feed flow: using an engineering approach to develop a framework. J. Membrane Sci., 193 (2001) 195-208 [ 17]
A. Gabelman and S.T. Hwang. Hollow fiber membrane contactors. J. Membrane Sci., 159
(1999) 61-106 [18]
R. Prasad and K.K. Sirkar. Dispersion-free solvent extraction with microporous hollow-fiber
modules. AIChE J., 34 N. 2 (1988) 177-188 [19]
J. Wu and V. Chen. Shell-side mass transfer performance of randomly packed hollow fibre
modules. J. Membrane Sci., 172 (2000) 59-74 [20]
A. Criscuoli, E. Drioli and U. Moretti. Membrane contactors in the beverage industry for
controlling the water gas composition. Annals of New York Acad. Sci., 984 (2003) 1-16 [21 ]
P. Schoner, P. Plucinski, W. Nitsch and U. Daiminger. Mass transfer in the shell side of cross
flow hollo fiber modules. Chem. Eng. Sci., 53 N. 13 (1998) 2319-2326
162 Chapter 4 [22]
D. Bhaumik, S. Majumdar and K.K. Sirkar. Absorption of C O 2 in a transverse flow hollow
fiber membrane module having a few wraps of the fiber mat. J. Membrane Sci., 138 (1998) 77-82 [23]
L. Dahuron and E.L. Cussler. Protein extraction with hollow fibers. AIChE J., 34 (1988) 130
[24]
T. Ahmed and M.J. Semmens. Use of sealed end hollow fibers for bubbleless membrane
aeration: experimental studies. J. Membrane Sci., 69 (1992) 1-10 [25]
T. Ahmed and M.J. Semmens. The use of independenlty selaed microporous hollow fiber
membranes for oxygenation of water: model development. J. Membranse Sci., 69 (1992) 11-20 [26]
R.M.C. Viegas, M. Rodriguez, S. Luque, J.R. Alvarez, I.M. Coelhoso and J.P.S.G. Crespo.
Mass transfer correlations in membrane extraction: Analysis of Wilson-plot methodology. J. Membrane Sci., 145 (1998) 129-142 [27]
T. Leiknes and M.J. Semmens. Vacuum degassing using microporous hollow fiber
membranes. Sep Purif. Technol., 22-23 (2000) 287-294 [28]
R. Gawronski and B. Wrzesinska. Kinetics of solvent extraction in hollow-fiber contactors. J.
Membrane Sci., 168 (2000) 213-222
Chapter 5. Liquid - liquid extractions
1. Introduction The liquid - liquid extractions are hereinafter presented in a similar way of g a s - liquid systems (Chapter 4). As for the gas-liquid systems, in fact, the mass transfer of solutes between two liquid phases in membrane contactors is regulated by the phase equilibria and the mass transfer resistances involved. The non-polar phase (usually, an organic phase) replaces now the gas phase and the liquid -liquid equilibria are considered instead of the gas liquid equilibria. The Chapter provides an analysis of the mass transport in terms of mass balances and calculation of the mass transfer coefficients. This analysis is general and valid for all the applications of membrane contactors where two liquid phases are involved.
2. Mass transfer equations In l i q u i d - liquid extractions, both hydrophobic and hydrophilic microporous, as well as hydrophobic - hydrophilic composite membranes can be used to put in contact the two phases, as discussed in detail below.
164 Chapter 5
2.1. Hydrophobic membranes When a species i is transferred from the non polar phase to the polar phase and a hydrophobic flat membrane is used, the resistances to the mass transfer lead to a concentration profile, as shown in Figure 1. If i is transferred from the polar to the non polar phase, the concentration profile is that of Figure 2. In both cases, the mass transfer resistances are those offered by the boundary layers and the membrane and can be drawn, as in Figure 3, by considering the electrical analogy.
Figure 3. Mass transfer resistances involved in the transport.
Liquid- Liquid Extractions 165 Referring to Figure 1, at steady-state the flux of the species i through the non polar film equals its flux through the membrane, as well as its flux through the polar film and the following equation can be written for the flat membrane:
Ji = kinp (Ci np - f im np) = kim np (Cim np - f ie np) -- kip (Cie p - f i p)
(1)
where:
kinp, mass transfer coefficient in the non polar phase for the species i; kimnp, mass transfer coefficient in the hydrophobic membrane for the species i; kip, mass transfer coefficient in the polar phase for the species i; Cinp, concentration of the species i in the non polarphase; Cimnp, concentration of the species i at the non p o l a r - membrane interface; Cie~p, concentration of the species i at the non p o la r - p o la r interface, non polar side; Ciep, concentration of the species i at the polar - non polar interface, polar side; Cip, concentration of the species i in the polar phase. For the transfer of the species i from the polar to the non polar phase, the fluxes are calculated by the same equation by only changing the sign in each flux (the transfer occurs in the opposite direction). When two liquids are in contact, a generic solute contained into one liquid diffuses through the second until equilibrium is established. The concentration of the solute in the two liquids at the interface, under equilibrium conditions, is related to its distribution coefficient
[1]: C l = m Cz
(2)
where: CI, solute concentration at the interface in the liquid 1; C2, solute concentration at the interface in the liquid 2;
166 Chapter 5 m, solute distribution coefficient.
Figure 4. Liquid-liquid equilibrium. The concentrations of a species i at the liquid - liquid interface (Cie np and Cie p) can be, then, calculated by equation (2):
Cie ,,p = mr Cie p
(3)
The flux of the species i can be also expressed in terms of the overall mass transfer coefficient: J~ ._ K,p (C , ,p - Cinpidea 9 __ Kp (C , pideal - C , p) (4) where: Knp, overall mass transfer coefficient based on non polar phase; Kp, overall mass transfer coefficient based on polar phase; Ci npiaeat, concentration of the species i in the non polar phase ideally in equilibrium with its concentration in the polar phase; Ci piaeat, concentration of the species i in the polar phase ideally in equilibrium with its concentration in the non polar phase.
Liquid- Liquid Extractions 167 Considering equation (2):
Cinp
ideal
= mi Cip
Ci pideal _ Ci n / mi
(5) (6)
The overall mass transfer coefficients can be expressed in terms of single mass transfer coefficients, by combining equations (1) and (4) and taking into account the equilibrium expressions previously reported: 1/Kp = l/kip + 1/(kim np mO + 1/(ki,,p mJ
(7)
1/Knp = m/kip + 1/kimnP+ 1/kinp
(8)
Equations (7) and (8) state that the overall resistance offered to the mass transport is due to the liquid film resistances and the membrane resistance as shown in Figure 3, and are valid also for the transport of the species i from the polar to the non polar phase.
When an hollow fiber configuration is considered with the polar phase in the shell side and the non polar phase at the lumen side the interface is located at the outer diameter of the tubes and equations (7) and (8) change into" 1/(Kp do) = 1~(kips do) + 1/(kim np mi dl,n) + 1/(kinpt mi dO
(9)
168 Chapter 5 (10)
1/(Knp do) : m/(kips do) + 1/(kimnp dim)+ 1/(kinpt dO where:
kips, mass transfer coefficient for the species i in the polar phase at the shell side; kmpt, mass transfer coefficient for the species i in the non polar phase at the tube side; di, inner diameter of the tube; do, outer diameter of the tube; dim, logarithmic mean of the hydrophobic membrane diameters.
2.2. Hydrophilic m e m b r a n e s Figures 5 and 6 show the concentration profiles that occur when the species i is transferred from the non polar and from the polar, respectively, in a hydrophilic flat membrane.
Figure 5. Concentration profile for the species i when it moves from a non polar phase toward a polar phase through a microporous hydrophilic flat membrane.
Figure 6. Concentration profile for the species i when it moves from a polar phase towards a non polar phase through a microporous hydrophilic flat membrane.
Liquid - Liquid Extractions 169
Referring to Figure 5, as for the hydrophobic membrane, at steady state the flux of the species i through the non polar film equals its flux through the membrane and its flux through the polar film:
J, = k,.~ ( c , .p- G . d = k J
(C,~ p - C,m p) = k,p (C, m p - C, p)
(11)
where: Cim p, concentration o f the species i at the p o l a r - membrane interface; k J , mass transfer coefficient in the hydrophilic membrane for the species i.
By considering the equilibrium expression for the liquid - liquid interface, the overall mass transfer coefficients can be related to the single mass transfer coefficients by:
+ 1/(ki.p mO
(12)
1/K.p = mi /kip + mi /ki p + 1/kinp
(13)
1/Kp = 1~kip + 1/ki p
In an hollow fiber configuration with the polar phase in the shell side and the non polar phase at the lumen side the interface is located at the inner diameter of the tubes and equations (12) and (13) become:
1/(Kp dO = if(kips do) + 1/(kimp d'lm) + 1/(kinp, mi dO
(14)
I/(Knp di) = mi/(kips do) + m./(ki p d'tm) + 1/(kinpt dO
(15)
170 Chapter 5 where:
d'tm, logarithmic mean of the hydrophilic membrane diameters. The equations reported until now for symmetric hydrophobic and hydrophilic membranes are valid also in the case of asymmetric membranes, when the membranes are hydrophobic or hydrophilic along all the thickness.
2.3. Partially wetted and hydrophobic-hydrophilic composite membranes Sometimes the membranes can be partially wetted by the liquid phase that should be blocked at the membrane interface and both liquids can coexist into the micropores of symmetric or asymmetric membranes, where the liquid-liquid equilibrium is established. Figure 7 shows the concentration profile for the transport of the species i from the non polar to the polar phase through a partially wetted asymmetric flat membrane. The resistances during the transport are those offered by the non polar film, the hydrophobic membrane, the hydrophilic membrane and the polar film.
Figure 7. Concentration profile for the transport of the species i from the non polar towards the polar phase through a partially wetted asymmetric flat membrane.
Liquid- Liquid Extractions 171 At steady state, the following equation is valid for the flux:
Ji : kinp (Ci rip- Cim np) = kim np (Cim np-Cie np)= ki p (Cie p -- Cim p) : kip (Cim p -- Ci p)
(16)
The overall mass transfer coefficients are calculated by:
1/Kp : I/kip + 1 / ( k J p mO + 1 / k J
+ I/(ki,,pmO
(17)
1/Knp = m~ /kip + 1/kimnp+ m~ / k J +
I/kinp
(18)
For hydrophobic - hydrophilic composite flat membranes the same equations are valid. As already stated in Chapter 4, while for partially wetted membranes (both asymmetric and symmetric), it is quite difficult to determine the location of the interface within the membrane pores, in hydrophobic - hydrophilic composite membranes the interface coincides with the hydrophobic-hydrophilic interface and it is easier to determine the membranes mass transfer coefficients.
For the hollow fiber configuration with the polar phase at the shell side and the non polar phase in the lumen side, the overall mass transfer coefficients are calculated by:
1/(Kp di,,t) = if(kips do) + l/(kim np mr dim) + 1 / ( k J d'lm) + 1/(ki~pt mr di)
(19)
1/(Knp dint) = mr/(kipsdo) + 1/(kim np dhn)+ mr/(ki,,, p d'tm) + 1/( ki,,ptdd
(20)
172 Chapter
5
where:
dmt, interfacial diameter. The hydrophobic - hydrophilic composite hollow fiber membrane section indicating the diameters present in equations (19) and (20) is reported below (Figure 8).
Hydrophilic membrane section .. _. _. _. . . . .
Polar phase
!!!i ::::::i Non polar phase
. . . .
~ x . , .
. . . .
, %,~,.
,\\,q Polar phase
. . . .
. . . . . . . .
. . . .
_... --~
d i
dint
.. w.
I I
I
Hydrophobic membrane section
I
d o
Figure 8. Section of the hydrophobic-hydrophiliccomposite hollow fiber membrane. The overall mass transfer coefficients for the different types of membranes described are summarized in Table 1 and 2.
Liquid- Liquid Extractions 173 Table 1. Expressions of the overall mass transfer coefficients for the different flat membranes Membrane Hydrophobic
1/Kp
1/Knp
1/kip + 1/(kim np mi) + 1/(kinp mi)
mi/kip + 1/kimnP+ 1/kinp
Hydrophilic
1/kip + 1/kimp + 1/(kinp mi)
mi/kip + mi/kim p+ 1/kinp
Partially wetted (symmetric and asymmetric) and hydrophilichydrophobic composite
1/kip+ 1/(kimnp mi) +l/kim p + 1/(kinp mi) mi/kip + 1/kimnp + mi/kim p+ 1/ kinp
Table 2. Expressions of the overall mass transfer coefficients for the different hollow fiber membranes. Operating conditions: polar phase at the shell side and non polar phase in the lumen Membrane Hydrophobic
1/(Kp do) = 1/(kips do) + 1/(kimno mi dim) 1/(Knpdo) = mi/(kips do) + + 1/(kinpt mi di) 1/(kimnp dim)+ 1/(kinpt di)
Hydrophilic
1/(Kp di) = 1/(kips do) + 1/(kimp d'lm) + 1/(kinpt mi di)
Partially wetted (symmetric and asymmetric) and hydrophilichydrophobic composite
1/(Kpdint)= 1/(kips do) + 1/(kimnp mi dim) 1/(Knpdint) = mi/(kipsdo) + + 1/(kimp d'lm) + 1/(kinpt mi di) 1/(kimnp dim)+ mi/(kim p d'lm) + 1/( kinptdi)
1/(Knpdi) - mi/(kips do) + mi/(kimp d'lm) + 1/(kinpt di)
,,,
As in gas-liquid operations, reactive liquids can be used as extractants also in liquid-liquid systems in order to increase the removal efficiency. The enhancement factor (E) is then introduced to takes into account the effect of the chemical reaction on the extraction. If the transfer of the species i occurs from a non polar phase towards a polar phase in which the species reacts, the overall mass transfer coefficient based on the non polar phase for a flat hydrophobic membrane can be calculated by:
i
••••••••••••••••••••••••••••!••••••il•••••••••••••
174 Chapter 5 1/Knp = m/(kipE)
!~i!i i i~ii i iiliii!~I~ii~!ii~!iIi!i iiii/iiii!i/!ili/iii!ii!iil/!iiill~I~I~IIIIIII/
+ l/kimnP+ l/kinp
(21)
Generally, depending on the value of the solute distribution coefficient, the solute might prefer one of the two phases. When the solute strongly prefer a phase (mi is >> 1 or <<1) or rapidly reacts in it, the expressions describing the overall mass transfer coefficients can be simplified, as reported in Tables 3 and 4. In the case of a solute that strongly prefers a phase (phase l) but rapidly reacts in the other phase (phase 2), usually the resistance offered by the phase 1 controls the process. A comparison among mass transfer coefficients is required for not well defined cases.
Table 3. Simplifications of the expressions for the calculation of the overall mass transfer coefficients for the different flat membranes Membrane Hydrophobic
Hydrophilic
mi >> 1
mi << 1 or rapid chemical reaction in the polar phase
1/Kp = 1/kip
1/Kp = 1/(kim np mi) + 1/(kinp mi)
1/Knp = mi/kip
1/Knp = 1/kimrip+ 1/ kinp
1/Kp = 1/kip + 1/kimp
1/Kp = 1/(kinp mi)
1/Knp = mi/kip + mi/kim p
1~ . p = 1/kinp
Partially wetted (symmetric and
1/Kp- 1/kip + 1/kimp
1/Kp = 1/(kimnp mi) + 1/(kinpmi)
asymmetric) and hydrophilichydrophobic composite
1/Knp= mi/kip + mi/kim p
1/Knp = 1/kimnp+ 1/ ki.p
i84 / / il i~i iiiiiii!iI~iii/ii/iiiiii/i
Liquid- Liquid Extractions 175 Table 4. Simplifications of the expressions for the calculation of the overall mass transfer coefficients for the different hollow fiber membranes. Operating conditions: polar phase at the shell side and non polar phase at the lumen side Membrane Hydrophobic
Hydrophilic
mi >> 1
mi << 1 or rapid chemical reaction in the polar phase
1/(Kp do) = 1/(kips do)
1/(Kp do) = 1/(kimnp mi dim) +
l/(Knp do) = mi/(kips do)
1/(kinpt mi di) 1/(Knp do) = 1/(kimnp dim)+ 1/(kinpt di)
1/(Kp di) = 1/(kips do) + 1/(kimp d'lm)
1/(Kp di) = 1/(kinpt mi di)
1/(Knp di) = mi/(kipsdo) + mi/(kimp d'lm) 1/(Knp di) = 1/(kinpt di) Partially wetted (symmetric and
1/(Kpdint)= 1/(kips do) + 1/(kimp d'lm)
asymmetric) and hydrophilichydrophobic composite
1/(Knpdint)= mi/(kipsdo) + mi/(kim p d'lm) 1/(kinpt mi di) 1/(Knpdint) = 1/(kimnp dim)+
l/(Kpdint) = 1/(kimnp mi dim) +
1/( kinptdi)
In Figure 9 it is reported the overall mass transfer coefficient, based on the aqueous phase, versus Reynolds of the organic phase for the extraction of phenol from an aqueous solution with 1-decanol [2 ]. For this system, authors report a distribution coefficient of 25.4.
176 Chapter 5
r~
4
--
0.0
0.4
0.t
Reo
Figure 9. Effect of Reynolds on the overall mass transfer coefficient based on aqueous phase for the extraction of phenol with 1-decanol (From [2], Copyright (2003), with permission from Elsevier).
2.4. Variable distribution coefficient
The studies on liquid-liquid extractions are mainly based on the assumption that the distribution coefficient is constant during the process. This hypothesis is valid when the changes in the concentration is small, as for the case of continuous operations, but can lead to significant deviations from the reality for batch operations, usually characterized by a large variation of solute concentrations. Being the mass transport through the membrane dependent on the distribution coefficient value, the evaluation of its change during the process is, in fact, important for the correct description of the performance of the system. Coelhoso et al. [3] studied this aspect for the extraction and stripping of lactate by ionexchange with a quaternary ammonium salt. They obtained the equilibrium curve as function of the solute concentration in the feed (see Figure 10).
L i q u i d - Liquid Extractions 177
0.5
0.0 0.0
I
0.5
1
x
Figure 10. Equilibrium curve for lactate extraction (From [3], Copyright (1997), with permission from Elsevier).
The region of linearity of the graph gives information on the operating range in which the distribution coefficient can be considered constant, whereas, the non-linear region indicates the operating range for which the variation of the distribution coefficient has to be taken into account. Authors point out that this analysis can be extended to all systems where an ionexchange reaction is involved. Coelhoso et al. (2000) [4] studied the effect of the variable distribution coefficient for extraction of fermentation and pharmaceutical products.
2.5. Individual mass transfer coefficients
In liquid - liquid extractions, the mass transfer coefficients for the two phases involved (in a hollow fiber configuration, tube side and shell side mass transfer coefficients) can be
178 Chapter 5
calculated by using the correlations already illustrated for the gas -liquid systems (Chapter 4) with the specific properties of the phases [5-19]. For what concerns the membrane mass transfer coefficient, the equations to be used for the different types of membranes, are reported below.
2.5.1. Membrane mass transfer coefficient
As for gas - liquid systems, the membrane mass transfer coefficient strongly depends on the phase present into the pores. For microporous membranes, three different cases can occur: 1. non polar phase-filled micropores, hydrophobic membranes; 2. polar phase-filled micropores, hydrophilic membranes; 3. non polar and polar phase-filled micropores, partially wetted membranes.
Hydrophobic membranes In hydrophobic membranes the non polar phase wets the pores and the transport of the species i depends on its diffusion through the non polar phase. The membrane mass transfer coefficient is calculated by [ 16]:
k,m"p = Dinp g/r6
(22)
where:
Dinp,diffusion
coefficient of the species i in the non polar phase.
Liquid- Liquid Extractions 179 Hydrophilic membranes Hydrophilic membranes have micropores filled by the polar phase and the transport of the species i generally depends on its diffusion into this phase. The membrane mass transfer coefficient can be now calculated by [16]:
kip = Dip e/r6
(23)
where: Dip, diffusion coefficient o f the species i in the polar phase.
Partially wetted membranes Both non polar and polar phases are present into the pores of a partially wetted membrane and the membrane mass transfer coefficient is a combination of the hydrophobic and hydrophilic membrane mass transfer coefficients:
1/kpwllim
-'-
1/kimnp + 1/k~mp
(24)
with:
k i m np --
D i.p c/rdary
kip = Dip e/rC~etted
( 2 5)
(26)
180 Chapter 5 where: kpwllim, partially wetted mass transfer coefficient.
2.6. The Wilson-plot method The mass transfer coefficients can be also evaluated by experimental procedures based on the so-called Wilson plot [20]. In literature this approach has been used mainly for liquidliquid operations, but it can be extended also to gas-liquid transfers. The method consists in the following steps: 1. the overall mass transfer coefficient is experimentally determined; 2. the overall mass transfer resistance is plotted versus the inverse of velocity (Wilson plot); 3. the membrane mass transfer coefficient is calculated as the intercept of the Wilson plot; 4. the liquid mass transfer coefficients are obtained by substracting the membrane mass transfer coefficient from the overall mass transfer coefficient.
The overall mass transfer coefficient can be calculated by considering the overall mass balance on the membrane contactor.
For water-organic streams, the mass balance on the
water phase is:
Qw (c,w~,,- Gwo,,t) = Kw A AC,t,,, where:
(27)
Liquid- Liquid Extractions 181 Qw, water flow rate," ciwin, concentration of i in the water at the inlet of the module; Ciwout, concentration of i in the water at the outlet of the module; Kw, overall mass transfer coefficient based on the water phase; A, membrane area; ACam, logarithmic mean driving force.
The equation (27) allows to calculate the value of Kw, being all the other terms known (or easily measurable). The plot of 1/Kw versus 1/Vw~ represents the Wilson plot. The value of the exponent ct is chosen to obtain a straight line through the experimental points.
Y 1/Kw
1/Vw t~
Figure 11. Example of Wilson plot.
If we consider, as a particular case, a hydrophobic membrane and a high organic flow rate (or a high distribution coefficient), the equation (9) can be simplified as:
182 Chapter 5 1/(Kw do) = 1/(kiws do)
+
1/(kim ~ m, dim)
(28)
and the intercept (1/Vwa= O) of the straight line represents the membrane mass transfer coefficient.
Once the membrane mass transfer coefficient is known, by properly acting on the flow rates of the streams involved, it is possible to calculate the individual mass transfer coefficients. For example, by working at high water flow rate, the organic mass transfer coefficient will be obtained by substracting the membrane mass transfer coefficient from the overall one (calculated by (27)), whereas the water mass transfer coefficient will be achieved, with the same procedure, by working at high organic flow rates.
One limitaton of the Wilson-plot method is that it takes into account only the effect of the fluid velocity on the mass transfer, the other parameters being constant. Furthermore, Viegas et al. [ 18] pointed out that another weak point of the method is that it is based on two-steps calculations: 1. overall mass transfer coefficient calculation from experimental data (first fitting); 2. plot of the overall mass transfer resistance versus the inverse of the velocity (second fitting) that involve a double fitting of data and that could lead to a possible accumulation of errors.
Liquid- Liquid Extractions 183 Authors proposed an altemative method for performing the calculations in one step" the overall mass transfer coefficient is expressed as a function of time and concentration and only one equation is subjected to fitting. Readers can find further details of the work in the above reference.
184 Chapter 5 References [ 1] Perry's Chemical Engineers' Handbook. R.H Perry, D. W. Green and J.O. Maloney (Eds), 6th edition, McGraw-Hill Book Co., New York (1984) [2] M.J. Gonzales-Munoz, S. Luque, J.R. Alvarez and J. Coca. Recovery of phenol from aqueous solutions using hollow fibre contactors. J. Membrane Sci., 213 (2003) 181-193 [3] I.M. Coelhoso, J.P.S.G. Crespo and M.J.T. Carrondo. Kinetics of liquid membrane extraction in systems with variable distribution coefficient. J. Membrane Sci., 127 (1997) 141-152 [4] I.M. Coelhoso, M.M. Cardoso, R.M.C. Viegas and J.P.S.G. Crespo. Transport mechanisms and modelling in liquid membrane contactors. Sep. Purif. Technol., 19 (2000) 183-197 [5] M.-C. Yang and E.L. Cussler. Designing hollow-fiber contactors. AIChE J., 32 (1986) 1910-1915 [6] S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Mass transfer in various hollow fiber geometries. J. Membrane Sci., 69 (1992) 235-250 [7] M.J. Costello, A.G. Fane, P.A. Hogan and R.W. Schofield. The effect of shell side hydrodynamics on the performance of axial flow hollow fibre modules. J. Membrane Sci., 80 (1993) 1-11 [8] R. Prasad and K.K. Sirkar. Dispersion-free solvent extraction with microporous hollow-fiber modules. AIChE J., 34 N. 2 (1988) 177-188 [9] J. Wu and V. Chen. Shell-side mass transfer performance of randomly packed hollow fibre modules. J. Membrane Sci., 172 (2000) 59-74 [ 10]
A. Criscuoli, E. Drioli and U. Moretti. Membrane contactors in the beverage industry for
controlling the water gas composition. Annals of New York Acad. Sci., 984 (2003) 1-16 [11]
P. Schoner, P. Plucinski, W. Nitsch and U. Daiminger. Mass transfer in the shell side of cross
flow hollo fiber modules. Chem. Eng. Sci., 53 N. 13 (1998) 2319-2326
Liquid- Liquid Extractions 185 [12]
D. Bhaumik, S. Majumdar and K.K. Sirkar. Absorption of CO2 in a transverse flow hollow
fiber membrane module having a few wraps of the fiber mat. J. Membrane Sci., 138 (1998) 77-82 [ 13]
T. Ahmed and M.J. Semmens. Use of sealed end hollow fibers for bubbleless membrane
aeration: experimental studies. J. Membrane Sci., 69 (1992) 1-10 [ 14]
T. Ahmed and M.J. Semmens. The use of independenlty selaed microporous hollow fiber
membranes for oxygenation of water: model development. J. Membranse Sci., 69 (1992) 11-20 [15]
T. Leiknes and M.J. Semmens. Vacuum degassing using microporous hollow fiber
membranes. Sep Purif. Technol., 22-23 (2000) 287-294 [16]
R. Prasad and K.K. Sirkar. Membrane-based solvent extraction, in: W.S.W. Ho and K.K.
Sirkar (Eds.), Membrane Handbook, Chapman and Hall, New York (1992) 727-763 [ 17]
L. Dahuron and E.L. Cussler. Protein extraction with hollow fibers. AIChE J., 34 (1988) 130
[ 18]
R.M.C. Viegas, M. Rodriguez, S. Luque, J.R. Alvarez, I.M. Coelhoso and J.P.S.G. Crespo.
Mass transfer correlations in membrane extraction: analysis of Wilson-plot methodology. J. Membrane Sci. 145 (1998) 129-142 [ 19]
R. Gawronski and B. Wrzesinska. Kinetics of solvent extraction in hollow-fiber contactors. J.
Membrane Sci., 168 (2000) 213-222 [20]
A. Gabelman and S.T. Hwang. Hollow fiber membrane contactors. J. Membrane Sci., 159
(1999) 61-106
Chapter 6. Membrane Distillation and Osmotic Distillation
1. Membrane Distillation: operational principles In Membrane Distillation (MD), a microporous hydrophobic membrane is in contact with an aqueous heated solution on one side ("feed" or "retentate"). The hydrophobic nature of the membrane prevents the mass transfer in liquid phase and creates a vapour-liquid interface at the entrance of each pore. Here, volatile compounds (typically water) evaporate, diffuse and/or convect across the membrane, and are condensed and/or removed on the opposite side (permeate or distillate) of the system (figure 1).
FEED
DISTILLATE
or
or
RETENTATE
PERMEATE
I...........VAPOU
Aqueous solution
Figure 1. A general scheme of the MD process.
Air gap Aqueous solution Vacuum Sweep gas
M e m b r a n e Distillation and Osmotic Distillation 187
The basic principles of MD can be already individuated in the US patent application granted to Bodell in 1966 [ 1], who offered some guidelines for the realization of an "...apparatus and methods for converting impotable aqueous fluids to a potable water". Although Bodell did not present quantitative results nor discuss about the microporous structure of the membrane used (tubular silicone membranes), he clearly stated that the polymeric barrier was permeable to vapour but impermeable to liquid. The preliminary results of a more systematic experimental activity carried out by using different hydrophobic materials (including paper, gum wood, glass fibers, cellophane and nylon treated with silicone, Teflon and other water repellent materials), as well as the elaboration of a quite simple and basic theory, were published by Findley a year later [2]. The specific method used to activate the vapour pressure gradient across the membrane (the driving force for this membrane operation) characterizes four main different MD configurations. In the most common arrangement- known as Direct Contact Membrane Distillation (DCMD) - the permeate side of the membrane consists of a condensing fluid (often pure water) that is directly in contact with the membrane. Alternatively, the vaporized solvent can be recovered on a condensing surface separated from the membrane by an air gap (AGMD), vacuum (VMD), or removed by a sweep gas (SGMD). All these configurations are schematised in figure 2 and discussed in details in the paragraphs 5-8.
188 Chapter 6
Figure 2. Schematic representation of the various MD configurations. After [6].
The selection of a specific configuration depends upon feed and permeate compositions, and requested transmembrane fluxes. In general, DCMD (the simplest to operate) is the best choice for applications in aqueous environments, SGMD and VMD are used to remove
Membrane Distillation and Osmotic Distillation 189
volatile organic components from aqueous solutions, AGMD (the most versatile) is employed to concentrate various non-volatile solutes whenever high fluxes are not required. The nature of the driving force, coupled with the hydro-repellent character of the membrane, a l l o w s - at least theoretically - the complete rejection of non-volatile solutes such as macromolecules, colloidal species, ions etc. Lower temperatures and pressures with respect to those usually used in conventional distillation columns are generally sufficient to establish a quite interesting transmembrane flux (1-20 kg/m2h), with consequent reduction of energy costs and mechanical requirements of the materials. Typical feed temperatures vary in the range of 30-50~
thus permitting the efficient recycle of low-grade or waste heat streams, as
well as the use of alternative energy sources (solar, wind or geothermal). In addition, the possibility to use plastic equipments also reduces or avoids erosion problems. With respect to Reverse Osmosis (RO) process, MD does not suffer limitations of concentration polarization (see paragraph 3.1) and can be therefore employed when high permeate recovery factors or retentate concentrations are requested [3]. Figure 3 compares transmembrane fluxes between RO and MD observed during concentration experiments of orange juices: while RO flux drastically drops off because osmotic pressure increases with concentration, MD flux slightly decreases as consequence of both reduction of the activity coefficient and increase of the viscosity in the feed solution.
190 Chapter 6 25
20 t---
E
Reverse Osmosis
r) x" i
:3
15
rI,--
..Q
E E
10
t-L_
Membrane Distillation
,
10
I
15
,
I
,
20
I
25
30
solute concentration, ~
Figure 3. Comparison of RO and MD fluxes for concentrating orange juice, Treed= 40~ Tpe,n=20~ After [3].
In addition, the high flexibility and compatibility of MD operation offer the attractive opportunity to integrate it in various important industrial production cycles, with consequent benefits due to synergic effects that can be reached. On the other hands, Membrane Distillation suffers from some drawbacks. With respect to RO, MD fluxes of permeate are usually lower, and a higher energy consumption is necessary to drive this thermal membrane operation. Moreover, only a restricted class of polymeric materials present a sufficient chemical resistance and operational stability and, despite the decreasing trend of membrane costs, commercial modules are still quite expensive.
Membrane Distillation and Osmotic Distillation 191 2. Driving force to mass transfer
In general, heat and mass transport through membranes occur only if the overall system is not under thermodynamic equilibrium. In membrane processes, two homogeneous subsystems (with defined values of chemical potential Pl and PI') are separated by a membrane. For small changes of the number of moles in the two phases (caused by the mass transfer across the membrane), the variation of the Gibbs free energy (G) is:
da = ~[ (l.t; - la;' )dn I
(1.a)
t
Equation (1 .a) suggests that the driving force for the mass transport of a component from one phase to the other is given by the difference in the chemical potential of the two phases caused by changes in temperature, pressure and activity; n I is the mole of i-th component transferred and is related to transmembrane flux Ji by:
--dn----z-'= A J, dt
(1 .b)
where t indicates the time and A the membrane area. The hydrostatic pressure gradient across the membrane usually is negligible in MD, and the driving force of the process is the partial pressure difference across the membrane, established by a temperature difference between the two contacting solution, or by vacuum, air gap or sweep gas in the permeate side. In case of non-ideal mixtures, the vapour-liquid equilibrium is mathematically described in terms of partial pressure (Pi), vapour pressure of pure i (pO), and activity coefficient ~i, according to the thermodynamic relationships: 0
P, = P Yi = P, a~ = pO ~',x~
(2)
where P is the total pressure, ai the activity of i-th component, Xi and Yi are the liquid and vapour mole fraction, respectively.
192
Chapter
6
The vapour pressure of a pure substance varies with temperature according to the ClausiusClapeyron' s equation: dp ~ dT
p~ = ~
(3)
RT 2
where k is the latent heat of vaporization (k =9.7 cal/mole for water at 100~ [4]), R the gas constant, and T the absolute temperature (figure 4). Vapour pressure data (if expressed in Pa) are often available in the form of Antoine's equation:
~
C+T
]
(4)
where T is expressed in Kelvin, and A, B, C are constant deduced by regression of experimental measurements (e.g.: for water A=23.1964, B=3816.44, C=-46.13) [4]. 80O
~. c~
'
I
'
I
'
I
'
I
'
l
600
"1-
400
200
O0
-50
0
temperature
50
1O0
150
(~
Figure 4. Dependence of vapour pressure on temperature for different compounds.
Membrane Distillation and Osmotic Distillation 193
At the pore entrance, the curvature of the vapour-liquid interface is generally assumed to have a negligible effect on the vapour pressure; however, possible influences can be evaluated by the Kelvin's equation:
pcOonvexsurface =
p~ LrpR 2ycT1
(5)
where r is the curvature radius, YL the liquid surface tension, and p the liquid molar density. For instance, in the case of water-ethanol mixtures, the liquid surface tension 7L (N/m) can be estimated by the following empirical relation (at 298 K): YL =
3.35 + 20.6x '10-3 0.0465 + x
(6)
where x is the mole fraction of ethanol [6]. Activity coefficients can be deduced by a large variety of equations aiming to evaluate the excess Gibbs function of mixtures G z [5], being:
L
an,
(7)
~,~,.,
The most popular of them are listed in table 1.
194 Chapter 6
Table 1. Empirical expressions for the activity coefficients (~:~ is the activity coefficient at infinite dilution) [5] G ~ / RT = (A2,x , + A,2x2)x,x 2
Margules
In 4 --- x2 [A12 dr- 2(A21 - A,2)x,]
In ~ = A,2
ln~2 = x?[A2, + 2(A,2- A21)x2]
ln~2 = A21
Equation
G e / R T = A12A21XlX2 AI2X 1 + A'2tx 2
Van Laar Equation
l n ~ = A12
AI2A;IXlX2 ln~l = (Ai2xl + A2tx2 )2
ln~:2 = A;1
A;,AI/ (Ai2x 1 + A;lX2) 2
GE/RT=-xlln(x,+A12x2)-x21n(xz+A21x,)
Wilson Equation ln~l = - l n ( x 1 + A12x2)+ x2
A12 x I + A12x2
ln~2 = - l n ( x 2 + A21xI)+
Xl
121 /
ln~
= -/nA12 + l - A 2 1
X2 + A21x1
A21
A12
X2 + A21X1
Xl + At2x 2
'/
ln~2 = -lnA21 + 1 - A12
Margules equation is the simplest one, and has been found to give similar results to Van Laar equation for several organic solutions. For asymmetric systems showing large positive deviation from ideality, Van Laar equation is preferentially used. Wilson equation is a powerful tool for systems that do not exhibit liquid-liquid phase splitting.
Membrane Distillation and Osmotic Distillation 195
NRTL model [7] is also frequently used in the description of vapour-liquid equilibrium of binary mixtures; for ethanol-water mixtures [8], Sarti et al. (1993) reported the following expressions:
ln~=xk
2
Gk, rk, ~ x, + xkGk,
+
r,kG,k )2 (Xk + x,G,k
r,k = a,k + b,k /T ; G,k = exp(-otr,k )
(8.a)
(8.b)
where xi and Xk are the mole fractions of ethanol and water, respectively; constant values are aik = 0.49854, bik=-456.00, C~= 0.24. A most flexible and complex approach for determining the activity coefficients is given by UNIQUAC equation; it is based on a combinatorial term that contains pure-components parameters, and a residual contribute depending on adjustable parameters that are characteristic for each binary system. For diluted aqueous ionic solutions, the value of activity coefficients can be deduced by the Debye-Hfickel' s theory: log~
=
-Iz+z lA47
(9.a)
where ~ is the activity coefficient of the electrolyte, A is a constant which depends on the temperature and solution permittivity, z is the ion valence, and I the ionic strength of the solution, given by:
I =lZzzc,
(9.b)
2i
Ci is the concentration of the i-th ion; in an aqueous solution at 25~ the constant A is 0.509 (mol kgl) u2 [9]. For NaC1 aqueous solutions, an empirical correlation between the activity coefficient and molar fraction of solute has been derived by Schofield (1989) as reported in [6]:
196 C h a p t e r 6 2
(10)
Ywater = 1 --0.5XNacl -- 10XN,,C t
For aqueous solution of CaC12, in the range of mass fraction (w) 32.2-46.2 %, Courel (1999) [ 10] proposed the following empirical correlation: a w = 1.6941 - 0.0410wcact 2 + 2.4 910 -4 Wc,,ct ~2
(11)
where aw is the activity of the solution. Activity coefficients for water in a sucrose solution (from 0 to 6.18 m) at 25~ are reported in figure 5.
0.98 r d~ o U:: r 0 o
0.96
D
> t9 t~
0.94
m
I
0.92 0
I 2
i
I 4
molal concentration of
i
I
I
6
8
C12Hz2Oll
Figure 5. Activity coefficient of water in sucrose solutions at 25~
After [ 11].
3. M a s s t r a n s f e r
According to an electrical analogy, mass transport for MD process can be conveniently described in terms of serial resistances upon the transfer between the bulks of two phases contacting the membrane (figure 6). The mass transfer coefficient K is defined as the inverse of the total resistance to the mass transport, expressed as combination of the mass transfer coefficients in the feed side (kf), in the membrane (kM) and in the distillate side (kd):
Membrane Distillation and Osmotic Distillation 197
K =
1
(12)
1/kz +l/k M +l/k d
Mass transfer boundary layers adjoining the membrane generally offer a negligible contribution to the overall mass transfer resistance, whereas diffusion across the polymeric membrane often represents the controlling step. The resistance to mass transfer on the distillate side can be omitted whenever MD operates with pure water as condensing fluid in direct contact with the membrane, or in VMD. The resistances within the membrane are associated to Knudsen, molecular and surface diffusion mechanisms, and viscous transport. A more detailed description of these mechanisms is proposed in the next paragraphs.
~'~
v,scous .~s,s.~.c~
f
BOUNDARY LAYERI RESISTANCE
BOUNDARY LAYER RESISTANCE
-'X
~,,,v
#
Figure 6. Serial and parallel arrangement of resistances to mass transport in MD.
3.1. Mass transfer: boundary layer resistances When solvent molecules are transferred through the membrane, the retained solute tends to accumulate at the membrane surface where its concentration gradually increases. Such a concentration gradient generates a diffusive counterflow that, under steady-state conditions, balances the net convective solute flow into the system: a concentration profile localized in the boundary layer adjacent to the membrane is therefore established (concentration
198 Chapter 6
polarization). In a thermally-driven membrane distillation process, the concentration polarization has generally a limited effect on the process performance [ 12]. Jc
._
Cb
D(dc/dx)
Cp=O
I_.., Figure 7. The development of a concentration profile into the boundary layer under steady-state conditions is known as concentration polarization phenomenon.
Referring to figure 7, and assuming that the solute is completely retained by the membrane, a mass balance across the feed side boundary layer yields to a relation between the molar flux J, the mass transfer coefficient kx (given by D/f, being D the diffusion coefficient and 8 the boundary layer thickness), and solute concentrations Cm and Cb at the membrane interface and in the bulk, respectively:
J=kxlnCm p
(13)
Cb
Concentration polarization phenomenon is usually quantified by a CPC coefficient, defined as: C P C = cm
Cb
(14)
Membrane Distillation and Osmotic Distillation 199
Literature provides several correlations (see review [13]), often derived by analogy with those evaluated for the heat transport, that are practical for determining the mass transfer coefficient. These empirical relationships are usually expressed in the form: S h = ct
Re p S c r
(15)
where: Sh, Sherwood number
Sh =
Re, Reynolds number Re
9
Sc, Schmidt number
kxdh
D
= p v dh
(dh: hydraulic diameter, D: diffusion coefficient)
(p: fluid density, v: fluid velocity, g: fluid viscosity)
Sc = pD
A brief list of mass transfer correlations for Newtonian fluids is given in table 2. The equations for the calculation of the shell side mass transfer coefficient have been already reported in Chapter 4.
Table 2. Predictive equations for mass transfer coefficients for shell and tube configuration Equation (tube side flow)
Comments
Reference
Sh=1.62 (d2v/(LD)) ~
L6v6que equation
[ 14]
Sh = 0.023 Re~
Chilton-Colburn, Re> 105, Sc>0.5
[ 14]
Chilton-Colbum, 1040.5
[ 14]
Sh = 0.34 Re~176
0"33
200 Chapter 6
Prandtl-Taylor
[ 14]
Von-Karman
[ 14]
Sh = 0.023 Re~176
300<Sc<700
[ 15]
Sh = 0.0149Re~176
Sc> 100
[ 16]
Sh = 0.023 Re~176
0.6<Sc<2.5
[ 17]
Equation (shell side flow)
Comments
Reference
0.5
[ 18]
Sh = 0.019 Gz
Gz<60 (*), closelypacked fibers
[19]
Sh = 1.3 8 Re~
1
[ 18]
Sh = 0.90 Re~176
1
[ 18]
Sh = 0.57 Re~176
0.01
[20,21]
Sh
( f l2)ReSc 1 + 5 f ~ - 7 2 ( 8 6 -1) ( f / 2)Re Sc
Sh
l + 5 f~-~[Sc- l + lnO - 5Sc)/ 6 ]
Sh = 1.25 (Re
de/L)~
(*) Gz, Graetz number
Gz =
pDL
(rh : mass flow rate, L: tube length).
Starting from traditional correlations, some researchers involved in MD studies have estimated coefficients related to the specific module geometry and hydrodynamics of the investigated system; the table 3 collects some of these correlations found in MD literature.
Membrane Distillation and Osmotic Distillation 201
Table 3. Examples of specific predictive equations for mass transfer coefficients in MD Correlation
c~
[3
~/
kx=13Q~
-
4.02-10 .5
0.38
2.0
0.48
0.33
p Sc r Sh=aRe a Set
1.86 0.960.45cp
0.38 0.55
0.38 0.33
17.5
p Sc r
0.023
0.33
0.33
6.6-7.4
Sh = a
Re p S c r
Sh = aRe
Sh = aRe
kx(10 -s m/s) 3.5 - 7.6
Comment
Reference
Q" volumetric feed flowrate (L/min) VMD Stirred cell Feed side Stirred cell Stirring rate: 200800 rpm With aqueous LiBr solution (055 % w / w ) Tangential flux Helicoidal hollow fibers q~: angle of inclination With oxygen 50
[22]
[23]
[24] [25]
[26]
Figure 8 illustrates the mass transfer coefficient values - given by the Dittus-Boelter's equation (15) - as a linear function of the feed temperature over the range of interest in MD. Data are from an analysis based on pure water VMD experiments and extended by computer simulation to NaC1 aqueous mixtures [26].
202 Chapter 6
20
l
'
I
'
I
'
I
,
i
I
i
I
i
I
i
1
16
oo E
C~
12
v
4
30
40
50
60
70
Feed temperature (~ Figure 8. Mass transfer coefficient for use in MD as calculated in [26]. Feed flowrate: 1 gpm; permeate pressure" 3kPa.
3.2. Mass transfer: transport across the membrane
In a porous medium, assuming surface diffusion negligible [27], the mass transfer is affected by viscous resistance (resulting from the momentum transferred to the membrane), Knudsen diffusion resistance (due to collisions between molecules and membrane walls) and/or ordinary diffusion (due to collisions between diffusing molecules) [28]. Predominance, coexistence or transition between all of these different mechanisms are estimated by comparing the mean free path t of diffusing molecules to the mean pore size of the membrane (Knudsen number). Kinetic theory of ideal gases calculates t as"
kBT p~/-2zcr 2
(16)
where kB is the Boltzmann's constant (1.380.10 .23 JKl), and cr is the collision diameter of the molecule (cr=2.7A for water).
Membrane Distillation and Osmotic Distillation 203
For the binary mixture of water vapottr in air, the free mean path t~/w can be evaluated at the u
average membrane temperature T , by [29]'
kBT -
((Ow +
oo)/2)
1
P 41 + (M /Mo)
where aa (= 3.7 A) and Crw the collision diameters, and
(17 Ma
and Mw the molecular weight for
air and water, respectively. For an average temperature T =60~
Phattaranawik and
colleagues [30] reported a mean free path of O. 11 gm. In the continuum region, the free mean path of the gas is small if compared with the average membrane pore diameter, molecule-molecule collisions predominate over molecule-wall collisions and Knudsen number (Kn) - defined as the ratio of the free path of the gas to the pore diameter- is < 1. In the Knudsen region this situation is reversed: the mean free path of the gas is large with respect to the average membrane pore diameter (Kn>l) and moleculewall collisions predominate over molecule-molecule collisions. In many practical cases, t is comparable to the typical pore size of MD membranes. Dusty Gas Model (DGM) is frequently used for describing gaseous molar fluxes through porous media; the most general form (again neglecting surface diffusion) is expressed as [31 ]"
j D
~_,pjjD_p,jD.
O ke t-Z.,a J--;~;
-1_)-0 "-'~e
1 = -~R V pTi
(18.a)
2
v o~r Pi J, = VP
(18.b)
D ke = 2er 1_[8RT
(18.c)
8RTr ct
3r ~ zM i o
Doe
E
o
rPD~
(18.d)
204 Chapter 6 where jD is the diffusive flux, jv the viscous flux, D k the Knudsen diffusion coefficient, D Othe ordinary diffusion coefficient, p the partial pressure, R the gas constant (8.314 J mol 1 K "1 ), T the temperature, P the total pressure, ~ the gas viscosity, r the membrane radius, e the membrane porosity and x the membrane tortuosity. Underscript e indicates the "effective" diffusion coefficient, calculated by taking into account the structural parameters of the membrane as shown in equations (18.c) and (18.d). Although DGM was derived for an isothermal systems, it is successfully applied in MD working under relatively small thermal gradients by assuming an average value of temperature across the membrane. Simplifications related to particular MD configurations are summarised in table 4.
Table 4. Simplified equations for the mass transfer in MD Configuration
Assumption
Vacuum membrane
Mean pore size << t
Transport equation
Dike__
20" ~ T
J, = --~--f Vp, = 3-~-T ~ ~ V p ,
distillation (limited by Knudsen
(IV. 1)
[22]
diffusion) Air gap membrane distillation
1
Air is considered as stagnant film
(limited by diffusion
J~ = D?_air, e
RTlPa,r[,,Vp, (IV.2)
[32]
through stagnant air) Air filled pores Direct Contact Membrane Distillation (Knudsen-molecular diffusion transition)
]_,
Mean pore size ~ t
1 J, = - - - ~
[6]
1 Pa.. "if-; + o D,e Dqe
Vp, (IV.3)
Membrane Distillation and Osmotic Distillation 205
De-aerated Direct Contact
Knudsen diffusion
Membrane Distillation
resistence dominant.
(Knudsen-Viscous
Existence of viscous
transition)
flux
J,
F 2 c r _]8R7
-Vp, + ~
P___LVP]
8r r
(IV.4) [61
Despite the modest complexity of the proposed models, the adoption of even simpler empirical correlations is in some cases preferred. The transmembrane flux is thus expressed as a linear function of the vapour pressure difference across the membrane [33]:
J = CAp
(19)
where C is the membrane distillation coefficient, and Ap the vapour pressure gradient evaluated at the membrane surfaces. In equation (19), the membrane distillation coefficient is a function of the membrane properties (pore size, thickness, porosity, and tortuosity), properties of the vapour transported across the membrane (molecular weight and diffusivity), and operative temperatures [34 - 36].
4. Heat transfer 4.1. Heat transfer: model of resistances
The complex relations between simultaneous heat and mass transfer are generally described in terms of a set of serial and parallel resistances through the boundary layers of the membrane and through the membrane itself. Figure 9 illustrates the heat transfer process in the case of DCMD; simplifications deriving from the possibility to omit one or more resistances can be made for specific MD configurations. Boundary-layers reduce the transport efficiency, and several efforts (use of spacers, turbulence promoters or turbulent flow) aim to minimize such extemal resistances. From a
206 Chapter 6 practical point of view, heat transport across the membrane occurs on the basis of two different mechanisms: conduction across the polymeric material, and latent heat flow related to vaporized components. The total heat flux Q transferred across the membrane is expressed as:
Q
E 1 ~+
1
hf
h m + Ji/~ / A r m
+
+1
AT
(20)
where AT is the (bulk) temperature difference among feed and permeate sides, ATm is the transmembrane temperature, Ji is the transmembrane molar flux, ~, is the molar heat of vaporization, and hf, hm and hp are the heat transfer coefficients on the feed side, membrane, and permeate side, respectively.
CON~D~/~,~' UCTO I N//////////////i~ ( ~/..////////////////'
_) LATENT~IZATION ( Figure 9. Serial and parallel arrangement of resistances to heat transfer in MD.
4.2. Heat transfer: boundary layer resistances The most unfavourable effect due to boundary layer resistances is the creation of a temperature difference between the bulk and the membrane surface where vapour-liquid transition occurs. A temperature polarization coefficient (TPC) is commonly used to quantify
Membrane Distillation and Osmotic Distillation 207
the extent of the boundary layer resistances over the total heat transfer resistance. It is defined as:
T P C = 7'7 - T p
rr
(21)
where superscript m indicates the temperature at the membrane surface. TPC is also employed to evaluate indirectly the efficiency for MD process: it falls between 0.4 and 0.7 for well-designed systems, and approaches to unity for mass transfer limited operations. TPC was used by Schofield et al. (1987) [33] as a tool in designing membrane distillation systems: table 5 reports calculated TPC for different module configurations and fluid-dynamic conditions.
Table 5. Temperature Polarization Coefficients (TPC) for various module configurations. After [33] Module and Fluid-dynamic Nusselt Hydraulic h (W/m 2 K) TPC membrane characteristics number diameter geometry (mm) Stirred cell Re=8,000 54 50 710 0.2 Re=32,000 120 50 1,600 0.4 Parallel plates 2mm thick Laminar 5.4 8 450 0.15 0.5 mm thick Laminar 5.4 2 1,800 0.4 0.1 mm thick Laminar 5.4 0.4 8,900 0.7 tube Re=5,000 29 1 19,000 0.9 Re=3,000 20 1 13,000 0.85 Re=l,000 4.4 1 2,900 0.54 Re=300 4.4 0.3 9,700 0.8 Channel 0.5 v=2 m/s 970 500 1,300 0.51 m m long
A large variety of empirical correlations that allow evaluating boundary layer heat transfer coefficients can be found in literature. In addition to the previously defined Reynold's number, most of these correlations involve some additional dimensionless numbers:
208 Chapter 6
9
hD
Nu, Nusselt number N u =
k
Gr, Grashof number Gr =
(h: heat transfer coefficient, k: thermal conductivity)
0 3t9 2 g f l A T
/12
(g: gravity acceleration, [3: thermal expansion
coefficient) 9
Pr, Prandtl Pr =
Cp l.t
k
(%: specific heat)
Gz, Graetz number Gz =
rhCp kL
A brief summary of useful empirical relationships is proposed in table 6.
Table 6. Predictive correlations for heat transfer coefficients in MD (tubular conducts) Equation (tube side)
Comments
Reference
Nu = 0.13 Re~
Laminar flow
[37]
Laminar flow
[3 7]
Laminarflow,
[38]
Nu = 0.097 Re~
Pr 038 (VI. 1) Pr~
(VI.2)
Nu = 1.62 Re Pr
(VI.3)
/0.14 Nu = 0.023 Re~ prO33
N u = 3.66 +
0.067Gz 1 + 0.04Gz~
~
Turbulent liquid
[39]
Laminar flow, VMD
[8]
Turbulent flow
[40]
(VI.4)
(VI.5)
id 0.055 (VI.6)
Nu = 0.036 Re 0s Pr 0.33
tangential flux
Membrane Distillation and Osmotic Distillation 209
Nu = 0.027(1 + - ~ / R e ~ Pr ~
At
Nu = 1.75{Gz + O.04[(d/L)Gr Pr] ~
10.1(VI.7) 4
Turbulent flow
}0.3(WI.8) 3
[40]
Important infuence of [41] free convection Transition region
[42]
Nu = 0.298 Re 0646 Pr 0"316 (VI.10)
Tangential flux
[24]
Nu = 0.74 Re ~176Pr ~176Gr ~176(VI.11)
Tangential flux
[37]
Nu = 0.023 Re o.8oPr 0.33 (VI. 12)
VMD
[26]
Nu = 2.0 Re ~
Stirred cell
[23]
Nu= O.116(Re~176176
TM
(VI.9)
Pr 0"33 (VI.13)
High feed flowrates improve the heat transport efficiency of MD operations. Also spacers have benefits in MD, since they destabilize the flow and create eddy currents in the laminar regime enhancing momentum, heat, and mass transfer. An heat transfer correlation for DCMD with spacers was proposed by Phattaranawik [43]:
Nu = kacO.664 Re~S Pr~
"s
(22.a)
with 00 9
kac =1.654
~o~ ~sin~-~)j
(22.b)
Here, kae is the correction factor for spacer geometry, dh,s the hydraulic diameter for the spacer-filled channel, df filament size, lm the mesh size, H the spacer thickness, q) the spacer voidage, and 0 the hydrodynamic angle. Figure 10 better clarifies the meaning of these parameters.
210 Chapter 6
Figure 10. Flow direction in spacer-filled channel. After [43].
4.3. Heat transfer: transport across the membrane
The total heat flux Q is transferred across the membrane by two mechanisms: conduction across the membrane material, and as latent heat associated to the vaporized solvent. A differential balance of energy gives: Q = J H v(T)-
Td k m .-....s._-
dx
(23)
where Hv is the (vapour) enthalpy at temperature T, km is the thermal conductivity of the membrane, and x the coordinate. Assuming To as a reference temperature, and considering that temperature inside membrane changes within few degrees (so that the specific heat can be supposed constant), the vapour enthalpy at a generic temperature T is given by: Hv (T) = 2,(To ) + Cpv (T - To )
(24)
where Cpv is the specific heat of vapour (for water: Cpv=8.22 + 0.00015 T + 0.00000134 T 2 (cal/Kmol), [5]) and ~ the heat of vaporization. Whereas the latent heat of vaporization is effectively used to promote the permeate flux, the conduction of heat across the membrane is an energetic loss that need to be minimized.
212 Chapter 6
The thermal conductivity of polymers strongly depends upon their degree of crystallinity. Data at 296K for some common hydrophobic polymers span in a relatively narrow range: polypropylene (PP): 0.11-0.16 W m l K 1, polyvinylidenedifluoride (PVDF): 0.17-0.19 W m l K 1, polytetrafluoroethylene (PTFE): 0.25-0.27 W m l K 1 [50]. Lost by conduction is reduced by increasing membrane porosity, since the water vapour thermal conductivity is one order of magnitude lower than that of polymeric materials in the range of typical MD working temperatures. Heat transferred by convection is generally considered negligible with exception of AGMD process [34].
5. Direct contact membrane distillation
DCMD represents the oldest and simplest configuration of membrane distillation. The liquid feed and the liquid distillate (or permeate) are kept in contact with the membrane and maintained at different temperatures. This temperature difference between the two liquid phases generates a transmembrane vapor pressure gradient that drives the flux of volatiles. The possibility to carry out DCMD in any desired membrane configuration (flat sheets, spiral wound, tubular, capillaries, hollow fibers) is a significant advantage of this configuration. A schematic representation of DCMD is shown in figure 11.
Membrane Distillation and Osmotic Distillation 213 feed
distillate
retentate
Figure 11. Schematic representation of DCMD.
In order to predict the mass transfer of water (in vapour phase) across the air-filled microporous hydrophobic membrane, the Dusty Gas Model reduces to the Knudsenmolecular diffusion transition form (equation IV.3) under the assumptions that i) the net flux of air across the membrane is very small with respect to the flux of water; ii) the viscous flux is negligible. Integration of eq. (IV.3) gives:
O6RT
r , ~KPermeat e _ eDK + Oair-water'e 0 1 Jwater -_ Dair-water'e lnlF'p ~,.~feed k Pair LJwater,e-- Oair-water,e --
0
(28)
where J is the molar flux, D Othe molecular diffusion coefficient, D K the Knudsen diffusion m
coefficient, T the average membrane temperature, 8 the membrane thickness, p the partial pressure.
214 Chapter 6
FEED (RETENTATE)
DISTILLATE (PERMEATE)
T1 T2
or>
,,=,
Figure 12. Typical temperature profile in DCMD.
Heat transport in DCMD is not easy to describe because of its interrelations with mass transfer. Temperature polarization coefficient TPC describes the dissipation of the driving force through the boundary layers adjoining the membrane on both side. With respect to fig. 12: TPC
= T( - T 2
(29)
T~ - V~_
This coefficient assumes different values for different plant geometries and operative conditions" about 0.3 for well stirred cells, from 0.4 to 0.7 for flat membrane cross-flow systems, from 0.6 to 0.9 for tubular or hollow-fiber configurations [33]. One of the major drawbacks in DCMD is related to the high amount of heat lost by conduction through the membrane and the membrane module, function of module geometry, thermal conductivities of the module and membrane materials.
Membrane Distillation and Osmotic Distillation 215
Working with pure water, and considering negligible the amount of heat lost to the environment (a valid assumption for well insulated modules), Lawson et a1.(1996) [51] have clarified the relationship existing between transmembrane flux, temperature drop along the module, and percent of heat lost (figure 13).
100
80
I
'
I
'
I
'
I
'
I
'
m
60 0 ...J
,,i,-,
i
40
~
ule'-O'3~~--
20
~mmm~0"1~ i
0
I
0.5
\
J
I
1
I
~ I
1.5
i
I
2
i
_
2.5
Transmembrane flux (mol/m 2 s) Figure 13. Percentage of heat lost by conduction in DCMD experiments with pure water (membrane module: 3M Corporation, feed flowrate: 1 gpm). After [51].
The magnitude of the mass fluxes achieved in literature varied between-~ 1 L/m 2 h [52] to 40 L/m 2 h [52]. The main operating parameters that have a significant influence on DCMD performance are feed and permeate temperatures and flowrates. Figure 14 shows the effect of the bulk temperature difference between retentate and permeate AT on the transmembrane flux at two different concentrations; data refer to DCMD experiments carried out on apple juice [12]. The shape of the curves depends on the way to set AT; in this case, the retentate temperature was keep constant and the distillate temperature was progressively decreased.
216 Chapter 6 The net flux inverts its direction whenever AT is not sufficient to compensate the reduction of activity on the retentate side. '
'
I
'
I
i
I
'
I
i
tO4
1 -
E v
x m
t-
0
.t3
E I
E co t"-
-1
-
II i
-2 -10
I
0
i
10
20
30
AT (~ Figure 14. Effect of the feed apple juice concentration on the transmembrane flux versus AT (feed temperature: 32~ After [12].
The trans-membrane flux increases with increasing the axial flow rate both in the feed and distillate sides. Figure 15 illustrates this behaviour as a result of the improvement of heat and mass transfer coefficients. Consequently, both concentration and temperature polarisation phenomena are reduced. Mass flux is more sensitive to the flow-rate variation induced in the stream having the highest concentration and/or viscosity.
Membrane Distillation and Osmotic Distillation 217
30
'
I
'
I
'
I
i
I
i
Retentate
E
20
e x
lO
Distillate
0
100
I
150
~
I
200
~
[
250
i
I
300
i
350
Flowrate (L/h) Figure 15. Variation of the transmembrane flux as function of feed and distillate flowrates (feed temperature: 29~ AT =20~ during apple juice concentration experiments. After [ 12].
The permeate flux moderately decreases with increasing feed concentration, mainly due to the concomitant reduction of the solution activity (figure 16).
218
Chapter
6 10
~"
i
i
i
i
i
i
i
i
i
t
lllll
i i i
'
'
'
'
' '"1
t
i
i
i
itlll
'
8-
E v
x m
6
-
c" i,,,..
E
4-
E t.I,.,..
I-
2
-
0 0.1
1
10
NaCI concentration (%wt) Figure 16. Flux drop with raising NaCI concentration (feed temperature: 100-->58~ distillate temperature: 42----~86~ After [53].
A short list of typical operative conditions in DCMD for selected aqueous systems is presented in table 7.
Table 7. Usual experimental conditions in DCMD Module Plate frame
Membrane
Liquid system
Operative Flux conditions and Polypropylene Raw sugarcane Feed flowrate: 15-26kg/m2h PP syrup (20~ 250-1250 (Akzo, ml/min Wuppertal) Membrane Distillate area: 75 cm2 flowrate: 500 ml/min Feed temperature" 75~
Reference [54]
Membrane Distillation and Osmotic Distillation 219
Tangential flow
Tangential flow
Plate frame
PTFE Water-glycole Membranes (37% w/w) (Gelman). TF200: pore radius 0.2 ~tm, thickness 178 gm, porosity 80%; TF450: pore radius 0.45 ~tm, thickness 178 ~m, porosity 80%; TF- 1000: pore radius 1 ~tm, thickness 178 gm, porosity 80% Effective membrane area: 30 cm 2 Millipore Protium FGLP: pore /deuterium (size 0.2 gm, 180ppm) thickness 0.175 mm, Oxygenporosity 70%, 16/oxygen-18 tortuosity 2. (-~ 1800 ppm)
PTFETarflen: pore size 0.84 gm, thickness 0.06 mm, porosity 20%, tortuosity 2. Effective membrane area: 15.2.10 "4 m2 and PVDF Pure water Durapore HVHP45: pore size 0.45
Distillate temperature: 25~ Hot side Maximum flux:~ [55] temperature: 20 1/m2h 65~ Cold side temperature: 25-45~ Tangential velocity: 15 cm/s
Distillate and Millipore feed FGLP:0.0004flowrates: 0.003 kg/m2s 4.10 -3 m3/h PTFEFeed Tarflen:0.0004temperature: 0.0024 kg/m2s 323-353 K
[56]
Distillate temperature: 287,293
Feed temperature: 20-50~
Flux: kg/m2s
0-0.008
[57]
220 Chapter 6
Capillary membrane module
Stirred cell
gm, thickness 125 gm, porosity 70%)
Distillate temperature (inlet): - 14~
Durapore GVHP22:pore size 0.22 gm, thickness 125 gm, porosity 75%) Polypropylene HC1 solution membrane (Accurel) Pore size: 0.2 lam, porosity 73%
Flowrates: 820 cm3/s
Effective area: 120 cm 2 PTFE membrane NTF-1122
solution Stirring rate: (0-55wt%) 200-800 rpm
(Nitto
Effective surface area: 2.38.10 "3 m 2 Polyethylene: N a C 1 aqueous inner diameter solution (35 g/l) 267.5 l.tm, thickness 50 ~tm, porosity 66.3 %, average pore diameter 0.087 ~tm Polypropylene inner diameter 342.5, 275.0, 357.5 lam,
600
[58]
Distillate inlet temperature: 20~
LiBr
Denko): pore diameter 0.2 gm, thickness 80 gm, porosity 0.75)
Hollow fiber
Feed inlet Up to mol/m2d temperature: 40~
From 5' 10-3 2.10 "2 kg/m2s
-[23]
Hot solution temperature: 308-347 K Cold solution temperature: 288K
Feed flowrate: 75 1/1 Distillate flowrate: 200 1/h Feed temperature: 40-60~
0-1 1/m2h
[59]
M e m b r a n e Distillation and O s m o t i c Distillation 221
thickness 50,65,42 pm, porosity 53.5, 50.0, 47.3 %, average pore diameter 0.074, 0.044, 0.056 gm; Capillary membrane module
Capillary membrane module
Polypropylene membrane (Accurel): pore size 0.2 gm, porosity 73% Effective area: -~ 220 cm 2 Polypropylene membrane (Enka MD020-2N-CP):
Water-Oil 3-10 (Horiba) ppm
Feed inlet temperature: 40-80~
Up to 250 kg/m2d
[60]
Up to 2 1/m2h
[12]
Distillate inlet temperature: 20~
Apple juice
Inner/Outer diameter 1,5/2,8 mm, thickness 120 gm, nominal pore diameter 0.45 pm, 40 fibers) Total membrane area: 0.1 m 2
Retentate flowrate: 100300 1/h Distillate flowrate: 100300 l/h Feed temperature: 32~ Transmembra ne temperature difference: 8+ 28~
6. Vacuum membrane distillation
In Vacuum Membrane Distillation, a low pressure is applied at the downstream side of the apparatus, as schematized in figure 17. Condensation of the permeate takes place outside the module. In VMD, transmembrane fuxes are the highest for the different MD setups, because the driving force that can be reached is close to the vapor pressure on the retentate side.
222 Chapter 6
feed
retentate
l
membrane
oQ
permeate
VACUUM PUMP Figure 17. Schematic representation of VMD process.
In VMD, assuming that the mass transfer through the membrane is dominated by Knudsen diffusion mechanism, the molar flux Ji of a permeating i-th specie - as described by equation (IV. 1) - is linearly related to the partial pressure gradient Ap across the membrane. The total molar flux Jt, sum over all the components, can be expressed as:
Km J, =~Ap
(30)
where Km is a permeability coefficient that depends on the membrane properties and temperature, and M is an average molecular weight in the permeating stream, evaluated as
[8]. J' 4M,
(31)
where Jt is the total flux across the membrane. Mass transfer through the boundary layers is adequately described by the film theory model discussed in paragraph 3.1. The mole fractions in the liquid bulk x~ and at the interface x I are related to the molar fluxes by the following relationship"
Membrane Distillation and Osmotic Distillation 223 r
Jt = In xi - Ji / Jt k Cr x~-J,/J t
(32)
where k is the mass transfer coefficient and Cv the total molar concentration in the liquid phase. The temperature at the vapour-liquid interface can be estimated by a simple energetic balance, assuming that the heat required for the interfacial evaporation is supplied by the heat flux through the liquid stream:
~-'Ji2, =h(T: -T':)
(33)
i
where ~i is the latent heat of vaporization of i-th component, h the heat transfer coefficient (estimated, i.e., by equation VI.5), Tf the feed temperature; apex refers to the membrane interface temperature. Sarti et al (1993) suggested to evaluate mass transfer coefficient kx by analogy with equation V.5, by substituing Nusselt number (Nu) with Sherwood number (Sh), and using Graetz number (Gz) for diffusion. The effect of permeate pressure on the transmembrane flux of benzene (a volatile compound) in aqueous mixture is illustrated in figure 17. When the permeate pressure exceeds the vapour pressure of water, the separation efficiency is enhanced, resulting in a highly concentrated benzene permeate. The effects of feed flowrate, concentration and temperature are similar to those discussed in DCMD.
224
Chapter
6
0.01
O0
E
0.001
'
_--
I
'
I
'
-
,~,
i X _
_
_
c
t~
0.0001
-
E
\
Jbenzene
E t"
9I.-
_
k le-005
\ \ \ -
Jwater
_
i
le-006
0
I
i
1 O0
\
I
, 300
200
Downstream pressure Figure 1 7 . Transmembrane fluxes of water and benzene ( 1 0 0 0 After [8].
.
.
(mbar)
ppm) as function of permeate pressure.
A short list of typical operative conditions in VMD for selected systems is presented in table
Table 8. Usual experimental conditions in VMD Module Tangential flow
Stirred cell
Membrane
Liquid system
Operative conditions Flat sheet Pure water Feed 3M temperature: Corporation: 2-4%wt NaC1- 30-75~ maximum water pore size: Permeate 0.29- 0.73~tm; 5% wt ethanol- pressure: free volume: water 3kPa 0.66-0.85; thickness: 79Flowrate: O.191 lam) 1 gpm
Flux
Reference
0.2-10 mol/m2s (depending on feed temperature)
[26]
Flat sheet Pure water PVDF: Millipore GVHP" pore
O.1 mol/m 2 s
Feed temperature" 25~
Membrane Distillation and Osmotic Distillation 225
Tubular
Tangential flow
size 0.28 ~tm, tortuosity 2.1, porosity 70%, thickness 118 lam
Permeate pressure: 1.66 kPa
Polypropylene Benzene-water (Enka Microdyn MD020TP2n) : internal tube diameter 5.5 mm, thickness 1.5 mm, pore size 0.2 ~m, number of tubes 3
Temperature: 27.5 ~
Effective membrane area: 0.036 m 2 PTFE Black membrane juice Surface area: 37 cm 2
Tangential flow
Stirring rate: 53.3 rps
Vacuum pressure: mmHg
-5" 10-4 kg/m2s
[62]
4.1-48 1/m2h
[63]
Recovered aroma (%) at 10~ and 2% volume reduction: Methyl butanoate 36.5• Ethyl butanoate 40.2• 1.5; 1,8 Cineole: 34.5+1.9; Ethyl hexanoate: 50.2+ 1.8 0-3.5 kg/m2s
[64]
5
Flowrate: 4-5 L/min
currant Temperature: 10-45~ Feed flowrates: 100-500 l/h
Flat sheet Water/Trichloro Temperature: PVDF ethylene(in 35-50~ (Millipore traces) GVHP): pore Permeate size 0.22 ~tm, pressure: 1-7 porosity 75%, kPa) thickness 125 ~tm
226 Chapter 6
7. Sweeping gas membrane distillation
-!
feed
retentate membrane
sweeping
r
gas
~.._/~ permeat~
Figure 18. Schematic representation of Sweeping Gas Membrane Distillation (SGMD).
The configuration of Sweeping Gas Membrane Distillation combines a relatively low conductive heat loss with a reduced mass transfer resistance. SGMD involves four serial stages: i) evaporation of the volatile compound (usually water) at the hot feed side; ii) transport of vapour through the membrane pores; iii) collection of the permeate by an inert cold sweeping gas - generally humid air; iv) condensation of the permeate out of the membrane module (in a external condenser).
The efficiency of this technical solution is
generally low, since a little amount of permeate is condensed from a large volume of sweep gas. In order to predict the mass flux in SGMD, it is convenient to assume that the combined Knudsen and molecular diffusive flux of molecules trough the membrane is expressed by (IV.3). Both mass and heat transport are involved simultaneously in the SGMD process; under steady-state conditions, the total heat flux transferred through the liquid boundary layer, the membrane, and the air boundary layer is given by: -
+
(34)
Membrane Distillation and Osmotic Distillation 227
where h is the heat transfer coefficient, J the molar transmembrane flux, ~. the heat of vaporization, 6 the membrane thickness, km the thermal conductivity of the membrane, T the temperature; apex refers to membrane surface, and subscripts f and a to feed and air sides, respectively. For the hot feed recirculating under laminar flow in the lumen side of an hollowfiber module, correlations in table 6 can be used to calculate the heat transfer coefficient hf. For the cold air recirculating in the shell side of the membrane module, Khayet et al. (2003) used the following equation [65]:
Sh = 0.206(Re c o s 6t) 063 Pr 0"36
(35)
where tx is the yaw angle (c~=0~ pure cross-flow; ~=90~ pure parallel flow). According to equation (19), the flux J is proportional to the transmembrane partial pressure difference; equation (2) is applied to describe the vapour-liquid equilibrium on the feed side. On the air side, the water vapour pressure
Pw,a is usually written as function of the total
pressure P and the umidity ratio w [66]:
Pw,a(Ta) =
wP
w + O.622
(36)
The humidity ratio along the membrane module length increases its inlet value Win as a result of the transmembrane flux of vapour: JA wi. = w + ~
(37)
,ha
where rha is the massive flux of air and A the membrane area. Khayet et al. (2003) combined equations (19), (36) and (37) to get the theoretical transmembrane mass flux:
J
2 + (wm +0.622)
+C(P-pw.f
rha[pw,, -Pw.a (w,. + 0.622)] = 0 J+C---~
(38)
Experimental data and calculations show that temperature polarization in the feed side has a little effect with respect to the influence of the air flow rate. The increase of the SGMD flux
228 Chapter 6 as function of the sweeping air velocity is illustrated in figure 19: this is due to the enhancement of the heat transfer coefficient, and to the consequent reduction of temperature
polarization effect at the shell side of the membrane module. Anyway, because the pressure of the sweep gas increases with air velocity, the resistance in the boundary layer increases as
well, and the flux versus air velocity curve is expected to attain a maximum value [66]. The effects of feed flowrate, concentration and temperature are similar to those discussed in DCMD.
80
'
I
'
I
'
I
'
I
Ckl E
60
m
mair,in=20oc mwater,in=60~
X
~
_
J
40
E .Q
E E m E
20
mair,in=20~ mwater,in--50~ C
0
i 0.4
I
i
I
i
0.8
1.2
sweeping
air v e l o c i t y
I 1.6
i
I 2
(m/s)
Figure 19. SGMD flux versus sweeping air velocity. After [67].
A short list of typical operative conditions in SGMD for selected systems is presented in table
Membrane Distillation and Osmotic Distillation 229 Table 9. Usual experimental conditions in SGMD Module Plate frame
Hollow fiber
Membrane
Liquid system
and PTFE Water (Gelman Science). TF200: pore size 0.2 ~tm, porosity 80%, thickness 178 ~tm; TF-450: pore size 0.45 ~m, porosity 80%, thickness 178 ~tm
Reference Flux Operative conditions Sweeping air -~10-60 9 10-4 [68] velocity: 0.5- kg/m2s 2.1 m/s Feed circulation velocity: 0.070.21 m/s Feed inlet temperature: 40-70~ Air inlet temperature: 10-30~
WaterSweep N2 0.2-2 kg/mEh PTFE (Poreflon)" Isopropanol (3- velocity: 8.7 pore size 0.8 10 wt%) m/s ~m, porosity Sweep N2 62% temperature: 25~ Effective membrane Feed area 9602 cm 2 temperature: 20-50~
[69]
Feed flowrate: 10-150 kg/h Tubular
TF200 Water-formic membranes. acid (0.5-1 PTFE wt%) (Gelman Sciences): pore size 0.2 ~m, porosity 79%) Effective membrane surface of the module: 0.019m 2
Feed flow Pure water [70] rate: 200 kg/h permeate flux (AT=I 0-40~ Air flow rate: 1-19kg/m2h 16.5-64.3 mol/h Pure formic acid permeate Sweeping air flux (AT=10velocity: 0.7- 40~ 2-40 2.6 rn/s kg/m2h Cooling water flow rate:
230 Chapter 6 500kg/h External condenser temperature: 2~
8. Air gap membrane distillation In AGDM, an air gap is placed between the condensing surface and the membrane surface in order to increase the conductive heat transfer resistance, and therefore to decrease the conductive heat loss through the membrane (see figure 20). AGMD is generally more versatile than DCMD: the vapour is condensed against a cooling surface rather than directly in the cooled permeate. This permit, for instance, to efficiently remove trace of volatile compounds from aqueous solutions, since an eventual partial wetting of the membrane has not a drastic effect on the permeate quality.
Membrane Distillation and Osmotic Distillation 231
feed y
cooling fluid retentate
permeate y
Figure 20. Schematic representation of Air Gap Membrane Distillation (AGMD).
As disadvantage, one should consider that the additional resistance represented by the air gap decreases also the mass transfer rate: AGMD transmembrane fluxes are consequently lower than in other MD configurations. The rather complicated design and construction of the module (due to the presence of the cooling surface) is an additional drawback of AGMD, and applications are practically limited to the use of plate-and-frame or spiral-wound membrane modules. Mass transfer in AGMD configuration is modelled on the basis of the general formulation of multicomponent diffusion in ideal gas mixture by Stefan-Maxwell equations. If water is the sole specie diffusing through a stagnant air gap, the molar flux can be expressed by equation
232 Chapter 6
(IV.2), modified in order to take into account of some structural parameters such as porosity (e), thickness (5), tortuosity (z) and air gap thickness (13):
Owar[ 1 ]
(39)
J w - RT[Plm,ln 8"r+ fl Ap
The diffusion coefficient depends on temperature; for water vapour/air mixtures [71]: D . . . . =2.16.10_5( T ~lS
[m 2 / s]
(40)
With respect to DCMD, a further unstirred layer adjoining the cooler surface exists (see figure 21). When modelling AGMD, a more complex system has to be considered, and stages of heat transfer in air gap membrane distillation include the transport: (i) from the feed bulk to the membrane, (ii) from the membrane to the condensing permeate; (iii) from the condensing permeate to the cooling surface, (iv) from the cooling surface to the coolant bulk. Contributions (i) and (iv) can be determined by following the guidelines in paragraph 4.
Membrane Distillation and Osmotic Distillation 233
Figure 21. Temperature profile in Air Gap Membrane Distillation (AGMD).
The heat flux (Q) through the air gap is made of two parts" the sensible heat flux Qs, and the vaporization heat (JwX):
Q = Qs +Jw A = - k d T + Jwcp(T-Tp)+JwA dz
(41)
where Cp is the specific heat of the gas, k is the gas-phase thermal conductivity, z the coordinate and X the latent heat of vaporization. For convenience, it is useful to focus the attention on the calculus of the sensible heat flux. In this respect, the energy balance in a differential element in the air gap gives:
dOs
~=0 dz
(42)
According to the solution proposed by Banat et al. (1998) [32] the following dimensionless variables are introduced:
234 Chapter 6 Z
q = --
(43)
P
where z is the distance from the membrane interface in contact with the feed solution and [3 the air gap thickness; k
hy = --
(44)
and 0=
JwC p
(45)
hy
where 0 is in the form of Peclet number which is the ratio of heat transfer by convection to conduction. Combining equations (41-45):
-
hy d2T ~+hyO dq 2
dT =0 dq
(46)
to be integrated for the following boundary conditions: q=O, rI = 1,
T =Tz T = Tp
(47)
where subscripts m and p refer to membrane and permeate, respectively. The resulting temperature profile is: T-T m
1-e ~
Tp - T m
1-e ~
(48)
The sensible heat flux Qs is evaluated by integrating equation (46) for ri=0 and T=Tm: Q = hy
0
[
\
[Tm - Tp)
1 - e -~
and the total heat flux Q becomes"
(49)
Membrane Distillation and Osmotic Distillation 235
( o
-rp
Q=hy l_e_ o
The heat flux Q from the condensed permeate layer to the cooling surface is: (50.b) where kc and b are the thermal conductivity and the thickness of the cooling plate, respectively, and hd is the condensation heat transfer coefficient. For a vertical surface [72]:
hu = 0.943 r
(51)
AT
where p and k are the fluid density and thermal conductivity at the film temperature, respectively, g the gravitational acceleration, [3 the
thermal expansion coefficient, ~t the
viscosity, L the latent heat of vaporization. The effects of retentate temperature, flowrate, and solute concentration are similar to those described in paragraph 5 for DCMD. The effect of air gap width on transmembrane flux rate is shown in figure 21, adapted from [32]: a reduction of the air gap increases the temperature gradient within the vapour compartment and, therefore, improves the flux. Theoretical and experimental studies [38, 73, 74, 75, 76] also evidence a linear correlation between the flux and the reciprocal of the air gap width.
236 Chapter 6
'
I
'
I
'
(.-
E v X :3 m ,.i-. (-
..Q
E E (D (--
!-
,
I
,
0.4
1
,
0.8
1.2
Air gap width (cm)
Figure 21. AGMD flux versus air gap thickness (Tb= 60~
Tc=20~ ). After [32].
A short list of typical operative conditions in SGMD for selected systems is presented in table 10.
Table 10. Usual experimental conditions in SGMD Module
Membrane
Tangential flow
PVDF Water-sucrose (Millipore): (90,150,300 pore size 0.22- g/L) 0.45ktm) PTFE (Gore Tex): pore size 0.20-0.45~tm. Two membranes in parallel.
Air gap thickness: -0.9mm
Membrane dimensions: 0.119m
Liquid system
Operative conditions Flowrate: 58 L/h Feed temperature: 35~ AT= 25~
8-
Flux
Reference
70-150 ml/h
[75]
Membrane Distillation and Osmotic Distillation 237
Tangential flow Air gap thickness:4mm
Tangential flow
0.0595m PTFE Water-NaCl (Millipore (0.5,1,5 %wt) FALP): pore size 1 ~m, porosity 85%, thickness 150 ~tm Membrane area: 160 cm 2 Microporous hydrophobic membrane.
Water-NaC1 concentration (3-35 g/L)
Air gap thickness: 2 Effective area: mm 64+0.04 cm 2 Tangential PVDF Water-NaCl flow (Durapore): (1-10 wt%) pore size 0.45 Air gap ~tm, thickness thickness: 110 lam, 8mm porosity 75%.
Plate-andframe Gas gap: 4 mm
Composite PE/PTFE: pore size 0.5 ~m, porosity 75%, overall thickness: 175 ~tm PTFE membrane: pore size 0.2 ~tm Membrane area 0.04 m 2
Water-formic acid
Feed temperature: 35-75~
3-16 kg/m2h
[77]
0-8 kg/m2h
[78]
5-75.10 -4 kg/m2s
[73]
Depending on the gap gas. SF6: 1.5-2 kg/m2h
[79]
Cooling water temperature: 25~
Feed flowrate: 0.25-2 L/min AT: 12-43~
Feed temperature: 30-80~ Cooling water temperature: 7oc
Feed temperature: 60~ Cooling water temperature: 10~
Air: kg/m2h
3-4.5
He: kg/m2h
10-14
238 Chapter 6 9. Osmotic distillation Osmotic distillation (OD) is a concentration technique for aqueous mixtures in which a volatile component (commonly water) is removed from the feed by a hypertonic solution flowing downstream a microporous hydrophobic membrane. Vapour diffuses through the membrane pores under a driving force given by the vapour pressure difference between both membrane sides, related to the activities in the two streams (figure 22).
FEED SOLUTION
I :ili plwl>Plw2 . . . .
STRIPPING SOLUTION
i i :
"iiiii:: i-i ......-........ 9 .....-........
:-i
i!!i!i!i!iii ii! -.-.-.-.-.-.... 9 .-.-.-.-.-.... -..-.-....-.-. -..-.-.-..-... -..-.-.-..-.-.
Cw2
. : . : . :. .: .: . : . : .
i : : : : : : : .: i : i : : :
Clwl . . . . . . . . . . . . . . . . -.-.-.-....-.
i!iiiiiiiiii: iii . . . . . . . . . . . . . . .
Cwl . . . . . . . . . . . . . . . ..-.-.-..-.-.-. .
.
9
.
.
.
.
..
,i:i: w ::?: .:.:. m - , : . : 9:.:. U J . : . : , iii.: ~ 12:1', . . . . . . . . . . . . . . . . .. . . . . . . -.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........-.-.-
Figure 22. Concentration profile in Osmotic Distillation.
OD is not a purely mass transfer operation: transport involves an evaporation at the feed side and a condensation at the stripping side. A temperature difference at the membrane interfaces is thus created, even if the bulk temperatures of the two liquids are equal. Mengual et a1.(1993) [80] estimated the temperature difference in their aqueous system to be lower than 1~ which would lead to a negligible decrease of the vapour flux.
Membrane Distillation and Osmotic Distillation 239
Assuming a simplified approach, the transmembrane flux can be expressed in terms of membrane permeability Km, partial pressures P'w at feed (1) and stripping (2) fluid-membrane interfaces, and log mean pressure of air entrapped into pores Poir "
j ._ Km
P'wl_-P'w2
(52)
Pair
Water vapour pressure is related to activity by eq. (2). Referring to a series expansion cut at the first order term, and assuming that the temperature difference through the membrane is small, the following relationship can be derived [81 ]"
J = -Pa~,. gm{P~ (T)" [a'~ - o'~2 ]- a~ (T)" IdP~ 1 [ . - ~ . J f (r~-~()}
(53)
where p0 is the vapour pressure of pure liquid, aw is the water activity, and T is the average temperature between the two membrane interfaces. Film-theory model can be used to describe the mass transport through boundary layers: t
t
j = kxPln x ~ kx p a w - a._.......~ x
(54)
~w~
assuming that water activity coefficients ~w are constant in the layers' x is the solute mole fraction, ~ its logarithmic mean value, kx the mass transfer coefficient, p the solution density. Apex refers to value at the membrane interface. Some useful empirical correlations for the calculus of kx are reported in table 11.
240 Chapter 6 Table 11. Correlations for mass transfer coefficient in OD Correlation
ot
[3
k x = co r
-
-
7 1.09+0.17
kx(10 5 m/s) 54+78.10 -3
S h = ct R e # S c r
1.86
0.33
0.33
kf=0.17;0.9;0.37
S h = ct R e ~ S c r
1.62 (shell)
0.33
0.33
1.86 (tube)
Comment (o=0+350rpm aqueous NaC1 solution 0.5+5M For NH3, SO2, HzS: 1+20.10 .3 M Shell: watersucrose 0+70%wt
Reference [80]
[821
[83]
Tube: waterCaC12: 26+40 %wt
Temperature and activity gradients can act in a synergistic way, or can operate in an antagonistic way to each other. Heat transfer phenomena are described as in MD operations. The salts chosen as osmotic pressure agents are in general NaC1 (because of its low cost), MgC12, CaC12 and MgSO4 [84]. In osmotic evaporation carried out at room temperature, transmembrane fluxes generally range between 0.2 and 1 L/mZh [85]. Table 12 offers a short list of typical operative conditions in OD for selected systems.
o . ....,
o o ....,
o m
.....,
d [-
s
o
o
0~
o tr
d~ ~176
o
r~
o~ s
o
e
9.
.~
N.~
8"~" ~o
,~,.~
o
~--.=
e'q
e~
0
t~~ ",~
~r3 e~
O
~
O r
oo
O
~
c;
o
=
Membrane Distillation and Osmotic Distillation 241
~
~--
r o,..~
0
0
r-i
a.
o
!
242 C h a p t e r 6
0 oo
r o
n~ o9
~
~
~
0
0
r-i
t~ 0
t~~ o~.,~
o
0
,~k c~
uaZ
~
0
~c~l
~~
.~.~
oo
~
0 0
o o
~ ~.~.~
~Z
0
r ,.Q
0
o
~
~
~
--
r
0'~
~
~ OD
9
o O
C'4
~
tt% ("4
~
t"--
Membrane Distillation and Osmotic Distillation 243
r....ii @',1 i..iii
t,,i
("4
"u o
C',l
~
::L "~
~ ~Z ~
=~ o
"7 ,5
0
9
~
o ~
~D
L~
t",,I
ri-i
~,5~ru
iii!
~s ~
Oeq
0", i....,.ii
e~
& 0
..
r
~
~. ~ -~ ~
0
9,.,
~
~D
244 Chapter 6 I0. Coupling thermal and osmotic distillation
In order to describe the action of a thermal gradient and a osmotic gradient simultaneously imposed across a microporous hydrophobic membrane, let express the transmembrane flux as in equation (19), and the dependence of the water vapour pressure on temperature and solute concentration by equation (3). In addition, the Clausius-Clapeyron's equation established an exponential dependence between the vapour pressure of a pure component and the absolute temperature of the form:
p~
exp(- R-~-f)
(55)
Let T1 and T2 be the temperatures at the corresponding liquid-vapour interfaces at sides 1 and 2, AT the transmembrane temperature difference, and T the mean temperature. Therefore: T1 = T +AT/2 and T2 = T-AT/2. Analogously, let Aa and 6 the transmembrane activity difference and the mean activity, respectively. Under the usual operative conditions of MD and OD, it can be assumed that:
AT - <<1 2T 2AT Aa <<1 2 R T z 4-d
(56)
Developing the series expansion of the vapour pressure gradient considering approximations (56), the transmembrane vapour pressure difference Ap splits into two terms representing, respectively, the "osmotic" and the "thermal" contributions to the driving force:
AP ~ p ~ (T )Aa + p ~ (T )?i R T 2 AT
(57)
Equation (57) cannot be used in the present form, due to the existence of boundary layers adjoining the membrane at both sides and giving rise to both temperature and concentration polarization phenomena. In order to take into account of these effects, the term
Membrane Distillation and Osmotic Distillation 245
p~
=-Ap~
can be expressed - in a first order approximation- as function of the
corresponding value in the bulk phase ( Apb ):
(58)
Ap~ = Ap~ + dAp~ 1(Cm-Cb)+...
de
where C is the solute concentration, and subscripts m and b refer to membrane and bulk value, respectively. In a similar way, the mean value of the vapour pressure p~
in equation (57) can be
written as:
P~
= P~
AP~ = P~
--2-- Idap~ lCm 2
(59)
According to equation (13), the concentration of solute at the membrane interface Cm is related to the bulk concentration Cb as follows:
Cm = Cbexp(- J / kx)
(60)
where kx is the mass transfer coefficient. In the limit of validity of the approximations above discussed, the combination of eqs. (55-60) allows deriving a final relationship for the transmembrane flux when temperature and concentration gradients act in proactive or counteractive ways [87] :
j
E+
+C bdAp ~ =Ap o_ 2 , A T ApOATb + ~ kx dC 2RATb~2 RATbf2
(~)ATb
(61)
The synergistic effect of both concentration and temperature gradients on the transmembrane flux is reported in figure 23 for aqueous solutions of sodium chloride.
246 Chapter 6
0.4
"E r
s
'
I
'
I
'
0.2
x
o
=~
I--
-0.2
~ /
" Ac = 1 mol/L
Ac = 0 mol/L -0.4 I i I -20 -10
1
i
i 0
I 10
L 20
AT (~ Figure 23. Volume flux versus bulk temperature difference at different concentration gradients of NaCI. Experiments performed at 100 rpm in [86].
Membrane Distillation and Osmotic Distillation 247 References [1] B.R. Bodell. Silicone rubber vapour diffusion in saline water distillation. US Patent. 285,032 (1963) [2] M.E. Findley. Vaporization through microporous membranes. Ind. Eng. Chem. Proc. Res. Dev., 6 (1967) 226-230 [3] E. Drioli, B.L. Jiao and V. Calabrb. The preliminary study on the concentration of orange juice by membrane distillation. Proc. Int. Soc. Citriculture, 3 (1992) 1140-1144 [4] The properties of gases and liquids. R.C. Reid, J.M. Prausnitz and T.K. Sherwood, 3rd Ed. McGraw-Hill, New York (1977) [5] Perry's Chemical Engineers' Handbook. R.H Perry, D. W. Green and J.O. Maloney (Eds), 6th edition, McGraw-Hill Book Co., New York (1984) [6] K.W. Lawson and D.R. Lloyd. Membrane distillation. J. Membrane Sci., 124 (1997) 1-25 [7] Molecular thermodynamics of fluid-phase equilibria. J.M. Prausnitz. Prentice-Hall, Englewood Cliffs- NJ, (1969) [8] G.C. Sarti, C. Gostoli and S. Bandini. Extraction of organic components from aqueous streams by vacuum membrane distillation. J. Membrane Sci., 80 (1993) 21-33 [9] Physical chemistry. P.W. Atkins. 4th Ed. Oxford University Press, Oxford (1990) [10] M. Courel. Etudes des transferts de mati6re en 6vaporation osmotique: application a la concentration des jus de fruits. PhD Thesis: University of Mompelier, 1999 [11] O.A.Hougen, K.M.Watson and R.R. Ragatz. Chemical Process Principles,Wiley & Son, Inc.: New York, 1966 [12] F. Laganb, G. Barbieri and E. Drioli. Direct contact membrane distillation: modelling and concentration experiments. J. Membrane Sci., 166 (2000) 1-11 [ 13] V. Gekas and B. Hallstrom. Mass transfer in the membrane concentration polarization layer under turbulent cross flow. I. Critical literature review and adaptation of existing Sherwood correlations to membrane operations. J. Membrane Sci., 30 (1987) 153-170 [14] Momentum heat and mass transfer. C. Bennet and J. Myers. McGraw-Hill, New York (1982)
248 Chapter 6 [15] R. Deissler. Analysis of turbulent heat transfer, mass transfer and friction in smooth tubes and high Prandtl and Schmidth numbers. In Advances in heat and mass transfer; Hartnett J.P. Ed. McGraw-Hill: New York, 1961 [16] R.H. Notter and C.A. Sleicher. The eddy diffusivity in the turbulent boundary layer near a wall. Chem. Eng. Sci., 26 (1971) 161-171 [17] E. Gilliland and T. Sherwood. Diffusion of vapors into a air stream. Ind. Eng. Chem., 26(1934) 516-523 [ 18] M.C. Yang and E.L.Cussler. Designing hollow-fiber contactors. AIChE J., 32 (1986) 1910-1915 [19] S.R. Wickramasinghe, M.J. Semmens and E.L. Cussler. Mass transfer in various hollow fiber geometries. J. Membrane Sci., 69 (1992) 235-250 [20] K.L.Wang and E.L. Cussler. Baffled membrane modules made with hollow fiber fabric. J. Membrane Sci., 85 (1993) 265-278 [21 ] D. Bhaumik, S. Majumdar and K.K. Sirkar. Absorption of CO2 in a transverse flow hollow fiber membrane module having a few wraps of the fiber mat. J. Membrane Sci., 138 (1998) 77-82 [22] S. Bandini, C. Gostoli and G.C. Sarti, Separation efficiency in vacuum membrane distillation process. J. Membrane Sci., 73 (1992) 217-229 [23] M. Sudoh, K. Takuwa, H. Iizuka and K. Nagamatsuya. Effects of thermal and concentration boundary layers on vapor permeation in membrane distillation of aqueous lithium bromide solution. J. Membrane Sci., 131 (1997) 1-7 [24] M. Tomaszewska, M. Gryta and A.W. Morawski, Study on the concentration of acids by membrane distillation. J. Membrane Sci., 102 (1995) 113-122 [25] M.J. Costello, P.A. Hogan and A.G. Fane, Proc. Euromembrane '97, 23-27 Jun 1997, The Netherlands [26] K.W. Lawson
and D.R. Lloyd. Membrane distillation. I. Module design and performance
evaluation using vacuum membrane distillation. J. Membrane Sci., 120 (1996) 111-121
Membrane Distillation and Osmotic Distillation 249 [27] Y. Fujii, S. Kigoshi, H. Iwatani and M. Aoyama. Selectivity and characteristics of direct contact membrane distillation type experiment. I. Permeability and selectivity through dried hydrophobic fine porous membranes J. Membrane Sci., 72 (1992) 53-72 [28] W. Kast and C.-R. Hohenthanner. Mass transfer within the gas-phase of porous media. Int. J. Heat and Mass Trans., 43 (2000) 807-823 [29] Principles of Physical Chemistry. H. Kuhn and H.-D. Fostering. Wiley, New York (2000) [30] J. Phattaranawik, R. Jiraratananon and A.G. Fane. Effect of pore size distribution and air flux on mass transport in direct contact membrane distillation. J. Membrane Sci., 215 (2003) 75-85 [31] Gas transport in porous media: the Dusty-Gas Model. E. A. Mason and A.P. Malinauskas. Elsevier, New York (1983) [32] F.A. Banat and J. Simandl. Desalination by Membrane Distillation: a Parametric Study. Sep. Sci. Technol., 33/2 (1998) 201-226 [33] R.W. Schofield, A.G. Fane and C.J.D. Fell. Heat and mass transfer in membrane distillation. J. Membrane Sci., 33 (1987) 299-313 [34] L. Pena, M.P. Godino and J.I. Mengual. A Method to Evaluate the Net Membrane Distillation Coefficient, J. Membrane Sci., 143 (1998) 219-233 [35] L. Martinez-Diez and M.I. Vazquez-Gonzalez. A method to evaluate coefficients affecting flux in membrane distillation. J. Membrane Sci., 173 (2000) 225-235 [36] V. V. Ugrozov and I. B. Elkina. Mathematical modeling of influence of porous structure a membrane on its vapour-conductivity in the process of membrane distillation. Desal., 147 (2002) 167171 [37] M. Gryta, M. Tomaszewska and A.W. Morawski. Membrane distillation with laminar flow. Sep. Purif. Technol., 11 (1997) 93-101 [38] S. Kimura, S. Nakao and S. Shimatani. Transport phenomena in membrane distillation. J. Membrane Sci., 33 (1987) 285-298 [39] Unit operations of chemical engineering. W.L. McCabe, J.C. Smith and P. Harriot. 4th Ed. McGraw-Hill, New York (1985)
250 Chapter 6 [40] Heat Transfer. M.N. Ozisik. McGraw-Hill, New York (1985) [41 ] M. Gryta, M. Tomaszewska. Heat transport in the membrane distillation process. J. Membrane Sci., 144 (1998) 211-222 [42] J.I. Mengual, M. Khayet and M.P. Godino. Heat and mass transfer in vacuum membrane distillation. Int. J.Heat and Mass Trans., 47 (2004) 865-875 [43] J. Phattaranawik, R. Jiraratananon and A.G. Fane. Heat transport and membrane distillation coefficients in direct contact membrane distillation. J. Membrane Sci., 212 (2003) 177-193 [44] A.G. Fane, R.W. Schofield and C.J.D. Fell. The efficient use of energy in membrane distillation. Desal., 64 (1987) 231-243 [45] L. Martinez-Diez, F.J. Florido-Diaz and M.I. Vasquez-Gonzalez. Study of evaporation efficiency in membrane distillation. Desal., 126 (1999) 193-198 [46] V. Calabr6, E. Drioli and F. Matera. Membrane Distillation in the Textile Wastewater Treatment. Desal., 83 (1991) 209-224 [47] M.C. de Andr6s, J. Dria, M. Khayet, L. Pena and J.I. Mengual. Coupling of a membrane distillation module to a multieffect distiller for pure water production. Desal., 115 (1998) 71-81 [48] Fiber Science. S.B. Wamer. Prentice-Hall, Englewood Cliffs-NJ (1995) [49] R.W. Schofield, A.G. Fane, C.J.D. Fell. Gas and Vapour Transport through Microporous Membranes. II: Membrane Distillation. J. Membrane Sci., 53 (1990) 173-185 [50] Polymer Handbook. J. Brandrup and E.H. Immergut. 3rd Ed. Wiley, New York (1989) [51 ] K.W. Lawson and D.R. Lloyd. Membrane distillation. II. Direct contact MD. J. Membrane Sci., 120 (1996) 123-133 [52] E. Drioli and Y. Wu. Membrane distillation: an experimental study. Desal., 53 (1984) 339-346 [53] K. Schneider, W. Holz and R. Wollbeck. Membranes and modules for transmembrane distillation. J. Membrane Sci., 39 (1988) 25-42 [54] S. Nene, S. Kaur, K. Sumod, B. Joshi and K.S.M.S. Raghavarao. Membrane distillation for the concentration of raw cane-sugar syrup and membrane clarified sugarcane juice. Desal., 147 (2002) 157-160
Membrane Distillation and Osmotic Distillation 251 [55] C. Rincon, J.M. Ortiz de Zarate and J.I. Mengual. Separation of water and glycols by direct contact membrane distillation. J. Membrane Sci., 158 (1999) 155-165 [56] G. Zakrewska-Trznadel, AG. Chmielewski and N.R. Miljevic, Separation of protium/deuterium and oxygen-16/oxygen-18 by membrane distillation. J. Membrane Sci., 113 (1996) 337-342 [57] L. Martinez-Diez and F.J. Florido-Diaz. Theoretical and experimental studies on desalination using membrane distillation. Desal., 139 (2001) 373-379 [58] M. Tomaszewska, M. Gryta and A.W. Morawski. Recovery of hydrocloric acid from metal pickling solutions by membrane distillation. Sep. Purif. Technol., 22/23 (2001) 591-600 [59] J.-M. Li, Z.-K. Xu, Z.-M. Liu, W.-F. Yuan, H. Xiang, S.-Y. Wang and Y.-Y. Xu. Microporous polypropylene and polyethylene hollow fiber membranes. Part 3. Experimental studies on membrane distillation for desalination. Desal., 155 (2003) 153-156 [60] M. Gryta, K. Karakulski and A.W. Morawski. Purification of oily wastewater by hybrid UF/MD. Water Research, 35/15 (2001) 3665-3669 [61] M. Khayet, K.C. Khulbe and T. Matsuura. Characterization of membranes for membrane distillation by atomic force microscopy and estimation of their water vapour transfer coefficients in vacuum membrane distillation process. J. Membrane Sci., 238 (2004) 199-211 [62] F.A. Banat and J. Simandl. Removal of benzene traces from contaminated water by vacuum membrane distillation. Chem Eng. Sci., 51/8 (1996) 1257-1265 [63] R. Bagger-Jorgensen, A.S. Meyer, C. Varming and G. Jonsson. Recovery of volatile aroma compounds from black currant juice by vacuum membrane distillation. J. Food Eng., 64 (2004) 23-31 [64] N. Couffin, C. Cabassud and V. Lahoussine-Turcaud. A new process to remove halogenated VOCs for drinking water production: vacuum membrane distillation. Desal., 117 (1998) 233-245 [65] M. Khayet, M.P. Godino and J.I. Mengual. Theoretical and experimental studies on desalination using the sweeping gas membrane distillation module. Desal., 157 (2003) 297-305 [66] I. Basini, G. D'Angelo, M. Gobbi, G.C. Sarti and C. Gostoli. A desalination process through sweeping gas membrane distillation. Desal., 64 (1987) 245-257
252 Chapter 6 [67] M. Khayet, P. Godino and J.I. Mengual. Nature of flow on sweeping gas membrane distillation, J. Membrane Sci., 170 (2000) 243-255 [68] M. Khayet, P. Godino and J.I. Mengual. Theory and experiments on sweeping gas membrane distillation. J. Membrane Sci., 165 (2000) 261-272 [69] C.H. Lee and W.H. Hong. Effect of operating variables on the flux and selectivity in sweep gas membrane distillation for dilute aqueous isopropanol. J. Membrane Sci., 188 (2001) 79-86 [70] M.C. Garcia Payo, C.A. Rivier, I.W. Marison and U. von Stockar. Separation of binary mixtures by thermostatic sweeping gas membrane distillation. II. Experimental results with aqueous formic acid solutions. J. Membrane Sci., 198 (2002) 197-210 [71] C.M. Guijt, I.G. Racz, J.W. van Heuven, T. Reith and A.B. de Haan. Modelling of a transmembrane evaporation module for desalination of seawater, Desal., 126 (1999) 119-125 [72] Heat Transfer. M. Jacob. Wiley, New York (1958) [73] F.A. Banat and J. Simandl. Theoretical an experimental study in membrane distillation. Desal., 95 (1994) 39-52 [74] H. Kurokawa, K. Ebara, O. Kuroda and S. Takahashi. Vapor permeate characteristics of membrane distillation. Sep. Sci. Technol., 25 (1990) 1349-1359 [75] M.A. Izquierdo-Gil, M.C. Garcia-Payo and C. Fernandez-Pineda. Air gap membrane distillation of sucrose aqueous solutions. J. Membrane Sci., 155 (1999) 291-307 [76] S. Bouguecha, R. Chouikh and M. Dhahbi. Numerical study of the coupled heat and mass transfer in membrane distillation. Desal., 152 (2002) 245-252 [77] C. Zhu, G. Liu, C.S. Cheung, C.W. Leung and Z.C. Zhu. Ultrasonic stimulation on enhancement of air gap membrane distillation. J. Membrane Sci., 161 (1999) 85-93 [78] S. Bouguecha and M. Dhabhi. Fluidised bed crystallizer and air gap membrane distillation as a solution to geothermal water desalination. Desal., 152 (2002) 237-244
Membrane Distillation and Osmotic Distillation 253
[79] F.A. Banat, F.A. AI-Rub, R. Jumah and M. Shannag. On the effect of inert gases in breaking the formic acid-water azeotrope by gas-gap membrane distillation. Chem, Eng. J., 73 (1999) 37-42 [80] J.I. Mengual, J. Ortiz de Zarate, L. Pena and A. Velazquez. Osmotic distillation through porous hydrophobic membranes. J. Membrane Sci., 82 (1993) 129-140 [81] M. Celere and C. Gostoli, C. The heat and mass transfer phenomena in osmotic membrane distillation. Desal., 147 (2002) 133-138 [82] Q. Zhang and E.L.Cussler. Hollow fiber gas membranes. AIChE J. 31/9 (1985) 1548-1553 [83] P.A. Hogan. Thermal and isothermal membrane distillation. PhD Thesis, University of New South Wales (AU) 1996 [84] J. Sheng. Osmotic distillation technology and its applications. Austral. Chem. Eng. Conf., 3 (1993) 429-432 [85] W. Kunz, A. Benhabiles, R. Ben-Aim. Osmotic evaporation through macroporous hydrophobic membranes: a survey of current research and applications. J. Membrane Sci., 121 (1996) 25-36 [86] M.P. Godino, L. Pena, J.M. Ortiz de Zarate and J.I. Mengual. Coupled phenomena membrane distillation and osmotic distillation through a porous hydrophobic membrane. Sep. Sci. Technol., 30 (1995) 993-1011 [87] P. Godino, L. Pena and J.I. Mengual. Membrane distillation: theory and experiments. J. Membrane Sci., 121 (1996) 83-93 [88] C. Gostoli and S. Bandini. Gas membrane extraction of ethanol by glycols: experiments and modelling. J. Membrane Sci., 98 (1995) 1-12 [89] R.J. Durham and M.H. Nguyen. Hydrophobic membrane evaluation and cleaning for osmotic distillation of tomato puree. J. Membrane Sci., 87 (1994) 181-189 [90] M.M. Vahdati and G.H. Priestman. Reducing boundary layer affects in membrane osmotic distillation. ICHEME Research Event London (1994) 177-179 [91] J. Sheng, R.A. Johnson and M.S. Lefebvre. Mass and heat transfer mechanisms in the osmotic distillation process. Desal., 80 (1991) 113-121 [92] C. Gostoli. Thermal effects in osmotic distillation. J. Membrane Sci., 163 (1999) 75-91
Chapter 7. Membrane Crystallization
1. Introduction to crystallization from solution
Crystallization is an excellent technique for purification of chemical species by solidification from liquid mixtures: many materials are marketed in crystalline form, and a large amount of product may be obtained from impure solutions even in a single step. The attraction for this method is heightened when considering that crystallization may operate at lower temperatures and requires lower energy than other conventional separation processes [1]. In 1999, the world production of crystals limited to use in electronic industry has been estimated at more than 20,000 tons per year, of which the largest fraction are semiconductors silicon, GaAs, InP, GaP, CdTe and its alloys, scintillation and optical crystals [2]. Crystallization technology also provides powerful aids in the manufacture of pharmaceutical compounds, with particular interest in chiral discrimination and polymorphism [3]. In biochemistry, the detailed description of protein structure at the atomic level is achieved by X-ray diffraction analysis as applied to single crystals [4]. In general, the particle size distribution and morphologies produced are a result of the relative rates of nucleation and growth of crystals from a supersaturated solution. Supersaturation may be created in a number of ways: by cooling or heating (if solubility decreases with temperature) a saturated solution, by evaporating the solvent, by addition of another component to the reaction environment thus reducing the solubility of the solute, by salting out, by reaction to form the solute in situ etc.
Membrane Crystallization 255 When a substance is transformed from phase 1 to phase 2, the change in molar Gibbs free energy ( A(~ ) of the transformation, at constant pressure and temperature, is given by: A ~ = (~,~ - ~ , )
(1)
where ~ti and ~2 are the chemical potentials of the phases considered. The molar Gibbs free energy can also be expressed in terms of activity as:
where R is the universal gas constant, T the absolute temperature, a is the solute activity, ao is the activity of solute in equilibrium with a macroscopic crystal, and S the supersaturation. Assuming that activity coefficients are unity, the supersaturation is usually expressed in term of concentration driving force: C S = -C*
(3)
C being the solute concentration and C* the equilibrium solubility at the temperature and pressure of the system [5]. Any design procedure starts from the knowledge of the thermodynamic properties of the system: in the case of crystallization from solution, solubility data and their dependence on temperature are mandatory. Complications arise if a specie can crystallize in different forms; depending on temperature, for instance, magnesium sulphate crystallizes with different numbers of water molecules per anhydrous molecule (figure 1).
256 Chapter 7 0.7
~
i
~
i
'
[
~
i
O 0.6 ~D3 ¢¢-
d co Cb v
0.5
0.4
O
09 (I) t~
/•gS04* 7H20
0.3
0.2
,
0
1
20
,
1
MgS04*H20
,
40 Temperature
I 60
~
I 80
, 100
(°C)
Figure 1. Solubility diagram of saturated aqueous magnesium sulphate solutions as a function of temperature. Redraw from [6] with permission.
2. Membrane Crystallizers Although crystallizers are in operation since many decades in the chemical industry, their design and operation still pose many problems. Large-scale plants often produce crystals which do not satisfy the specific quality criteria, that are becoming more and more stringent due to the shift of the market trend from base chemicals towards life-science products. A critical issue in the design of an industrial crystallization process is the strong interaction existing between kinetic aspects and hydrodynamics. At large-scale, process parameters such as temperature, supersaturation and turbulence are typically not uniformly distributed in the crystallizer body, and the slurry is not well-mixed and suspended: this type of unhomogeneity has a major impact on the final product quality [7]. For evaporative crystallization from solution, two types of industrial crystallizers are frequently in operation: the Forced Circulation (FC) and the Draft-Tube-Baffled (DTB). In a
Membrane Crystallization 257 FC crystallizer, the externally heated suspension reaches the boiling zone through either a tangential or an axial inlet. Due to the high recirculation rates imposed, the main design problem is that not the full cross sectional area of the crystallizer is used for evaporation (vortexing); this leads to thermal shot-circuiting that results in higher levels of supersaturation throughout the body, variable nucleation rates and consequent cycling of the crystal size distribution (CSD) between fine and coarse particles as the crystal surface varies [8,9]. In a DTB crystallizer, the suspension of crystals is maintained by a large, slow-moving propeller surrounded by a draft tube within the body. The propeller directs the slurry to the liquid surface so as to prevent solids from short-circuiting the zone of the most intense supersaturation. In this apparatus, the final shape of the crystals is often unsatisfactory, as consequence of the fact that larger crystals are rounded off by abrasion and attrition with the mechanical parts of the apparatus. Membrane Crystallization has been recently proposed as an interesting and promising extension of the Membrane Distillation concept [10]. This innovative technology uses the evaporative mass transfer of volatile solvents through microporous hydrophobic membranes in order to concentrate solutions above the saturation limit, thus attaining a supersaturated environment where crystals may nucleate and grow. Slurry leaving the crystallizer body is mixed to the feed and pumped through an heat exchanger where its temperature raises 35-45°C. The heated solution passes through the membrane module where the solvent (typically water) is partially evaporated according to the principles of MD (figure 2). Recirculation rates, size of the body and temperatures are critical design items in order to prevent the membrane blocking due to crystals deposition.
258 Chapter 7 I I
i I
feed inlet
i
CIRCULATION PUMP
CRYSTALLIZER BODY
steam inlet --t~1~ ---THEAT EXCHANGER steam outlet ~1----I~
I llll]lll
water inlet ~ 1 ~
_5
[lllllll
7
MEMBRANE MODULE
water outlet ~ 1 - ~ I
product discharge
Figure 2. Membrane crystallizer.
With respect to conventional devices, a membrane crystallizer is characterized by a laminar flow of the mother liquor through the capillary fibers of the membrane module(s); a low shear stress and an improved homogeneity of the solution are expected to promote a well ordered organization of the molecules, resulting in the formation of a crystalline lattice with good structural properties. A comparison between usual shapes of NaC1 crystals produced by the mentioned crystallizers is shown in figure 3.
Membrane Crystallization 259
Figure 3. Crystalline habit of NaCl from: a) Forced Circulation Crystallizer; b) Draft Tube Baffle Crystallizer; c) Membrane Crystallizer.
3. Product characterization: Crystal Size Distribution Crystal size distribution (CSD) is a critical parameter for precipitation and crystallization processes. Relatively large crystals limit the adhesion of mother liquor after filtration and, therefore, show a low tendency towards caking. On the contrary, small crystals are preferentially requested whenever reduced dissolution times are needed. In all cases, narrow distributions are necessary. The mass distribution W(Li) is generally expressed in term of the mass fraction o f crystals belonging to a certain size class divided by the width of the size class (ALi) [6]:
AL,
(4)
where c~ is the volume shape factor, 9c is the crystal density, AN the number of crystals counted in a unit volume of slurry - having length between L and L+AL. Apart from the average sizes evaluated by distribution diagrams, the width of a distribution around the mean size is often characterized by the coefficient of variation (CV), equal to the standard deviation divided by the mean size:
260 Chapter 7
~o
~o
(L
fw(L)dL (5)
o
L50% is the crystal length corresponding to a value of W(L) equal to 50%. A narrow CSD is characterized by a low CV. Experimental CV values measured in membrane crystallization tests carried out on NaC1 range within 15-30% [11].
25
I
[ ~'
;\/
~
' ~
t:90 rain
[]
,=,2om,n
20
g
_~
15
N ~
lO
0
5
0
T
0
I
0.05
T
I
,
I
,
1
0.1 0.15 0.2 crystal length (mm)
,
I
0.25
T
0.3
Figure 4. CSD, evaluated at times indicated in the legend, of NaCl crystals produced in a membrane crystallizer working at supersaturation ratio S=0.065.
As term o f comparison, a CV of about 50% is obtained when using a conventional mixedsuspension crystallizers [ 12].
Membrane Crystallization 261 4. Role of the membrane: heterogeneous nucleation
The role of a membrane is not limited to furnish a support for solvent evaporation. A crystallizing solution can be imagined as a certain number of solute molecules moving among the molecules of solvent and colliding with each other, so that some of them converge forming clusters, or growth units. In general, these clusters have a larger probability of being dissolved than of continuing to grow but, under specific conditions, they achieve a critical size having the same probability of growth as of dissolution. It can be shown that a maximum in Gibbs free energy (sum o f the contributions from the bulk and the surface molecular free energies) occurs at a critical cluster size, corresponding to the size at which further growth of the cluster leads to a decrease in free energy. Clusters of critical size are called critical nuclei; all those clusters with a size larger than the size of the critical nucleus (r*) will be likely to grow spontaneously. There is, therefore, an energetic barrier AG* (the nucleation barrier, see figure 5) that must be crossed in order to induce the formation of stable nuclei [13]: AG* =
161rv2y3
3[kBT In S] 2
(6)
where v is the molar volume occupied by a growth unit, ~, the surface energy, ks the Boltzmann constant, T the absolute temperature and S the supersaturation. The values of r* and AG* vary inversely with supersaturation: r* tends to infinity as S tends to 1. As the supersaturation increases, the height of the energetic barrier decreases, eventually reducing enough to induces spontaneous nucleation (figure 5).
262 Chapter 7
AG* 3
>,
c" III
It. ¢9 .0 .13 .
\ \,
t
(.9
Cluster Radius Figure 5. Dependence of the Gibbs free energy on the cluster size. The energetic barrier to the formation of a critical nucleus reduces with supersaturation.
The existence of this energy barrier explains why a solution that should experience homogeneous precipitation under thermodynamic conditions does so only if a certain value of supersaturation is exceeded. The presence of a foreign interface in the crystallizing system - specifically a polymeric membrane - decreases the work required to create critical nuclei and increases locally the probability of nucleation with respect to other locations in the system: this phenomenon is known as heterogeneous nucleation. Considering the interaction between solute and substrate in terms of contact angle 0 that the crystallizing solution forms with the substrate, the reduction of the activation energy is given by equation [ 13]: AGhe' = AGh°mI 1 - 3 ct ° cs " °9 s +34 40 1
(7)
If the solution wets the substrate completely (0 = 180°), AGhet = AGhom; when the contact angle is 90 ° (limit between hydrophobic and hydrophilic behavior), AGhet= 1/2AGhom.
Membrane Crystallization 263 This concept, exemplified in figure 6 for a solution of lysozyme in contact with several polymeric membranes, might represent the starting point for a better engineering of the crystallization procedures.
0,8
P~n .~ /7 ~
0,6
~O
~
¢~ 0,4
T
"<
~ / 0,2
/
PDMS PVDF PVDF(KF2800)-LiCI/2.5%
PVDF(KF2800)-PVP/2.5% PVDF(K460) PVDF(K460)-PVP/2.5%
0 0
50
1 O0
150
200
contact angle (*)
Figure 6. Heterogeneousnucleation:reductionof the energetic barrier to nucleationas functionof the contact angle of solution with the polymeric surface. Protein solution : Lysozyme40 mg/ml in NaAc buffer 0.05 M pH 4.5, NaCI 2% wt; Membranes: PP: polypropylene,PDMS: polydimethylsiloxane, PVDF: polyvinylidenefluoride;K: Kynar;KF: Kynar flex.
Measurements of induction times confirm experimentally the ability of the membrane to promote heterogeneous nucleation. This is particularly evident in protein crystallization: in homogeneous systems, macromolecules display a rather asymmetric and weak bonding configuration at their surfaces, and tend to aggregate in n-mers that diversify shape and size of lattice units, thus making ordered attachment for growing units less probable.
264 Chapter 7 The induction period of nucleation is the time elapsed from the attainment of a given supersaturation up to the formation of critical nuclei; since it is quite complex to evaluate experimentally the formation of critical nuclei, it is necessary to wait until they grow to visible size (total induction period or waiting time) [6]. For instance, literature related to lysozyme crystallization clearly shows that the induction time is a function of the supersaturation ratio S. In general, high value of S are necessary to obtain low induction times. Drenth and colleagues (2003) [14] observed by NMR technique that the induction period (33.8 mg/mL of HEWL in 0.1 M AcNa, pH 4.5, 5% NaC1) was reduced from 74h to about 0.2 hrs as supersaturation increased from 4.9 to 13.7. According to data of Paxton et al. (2001) [15], who carried out crystallization experiments by using un-silanized polystyrene wells, induction time values decreased from 55.8 h at S=6.54, to 2.77 h at S=16.13 (HEWL in 0.1 M AcNa, pH 4.6, 3% w/v NaC1). Kulkarni and Zukoski (2002) [16] measured, by dynamic light scattering analysis, induction periods decreasing from 6.7 h to 0.4 h when supersaturation ratio increased from 6.2 to 11.7 (HEWL in 50 mM AcNa, pH 4.6). Under comparable and even low initial supersaturation ratios, induction times for HEWL crystals grown on polypropylene membranes (see figure 7) are lower than those reported in literature for different crystallization techniques [ 17,18].
Membrane Crystallization 265
Figure 7. Scanning Electron Microscopy image of a tetragonal lysozyme grown on a polypropylene membrane surface.
Referring to figure 8, for S =6.5, 8.9 and 14.2, induction times were 2 (24% MgC12), 2.3 and 1.2 (30% MgCI2) hours; for S = 0.7, 1.3, 2.3 and 3.9, the corresponding induction times were of t .8 (30% MgCI2), 2, 1.4 and 1.6 (24% MgC12) hours.
3,0 -
2,5.
r
I'--~-- MgCl2 30% w t ~ , + MgCl2 24% wt/v---~-- MgCI2 22% wt/v I--v--MgCI2 1 6 % w t / v + M g C I 2 10%wt,'v
/
j,r/" 2,0-
1,0.
k
0,5. 0,0
° _
o
i
g
6
NaCl% / w t v -1 Figure 8. Induction times for membrane crystallization of lysozyme vs NaCI concentration; MgC12% wt/v
is indicated in the legend.
266 Chapter 7
5. Crystallization kinetics of a membrane crystallizer Experimental results show that the nucleation rate B, also in a membrane crystallizer, depends on the concentration of crystals in the magma (roT), hydrodynamics (e.g. stirrer speed, pump impeller, rotation rate: ~o), and difference AC between the mother liquor concentration and the value at equilibrium (solubility): (8)
B = kbmiTo)JAC k
Typical values of the nucleation rate constant kb lie between 1 and 2.5; i is close to 1 whenever collisions between crystals and vessel walls dominate on crystal-crystal interactions. The overall growth rate G can be related to the overall concentration driving force Ac by an empirical relation: (9)
G = kgAC g
where g is the overall order of the growth process, and kg is the growth rate constant that is function of temperature, controlling kinetic mechanism (diffusion or integration), presence of impurity in the system (figure 9). -19,5
O
oJ
-I9,75
E
-20
-20,25 y = 1,421x - 24,488 R2 = 0,9033
-20,5 2,9
3
3,1
3,2
3,3
3,4
3,5
In (C-C*), g/L
Figure 9. Correlation between growth rate and supersaturation (c: actual concentration, c*: saturation) for NaCI in a membrane crystallizer.
Membrane Crystallization 267 The process of crystal growth occurs according to sequential stages: transport of molecule from the bulk of the mother liquor to the crystal surface, adsorption and diffusion over the surface, attachment and diffusion along a step, integration into the crystal at a kink site (figure 10).
(~
Transport of solutefrom the bulk solutionto the crystal surface
Figure 10. Sequential stages of crystal growth.
When diffusive phenomena in solution are the rate-determining step for crystal growth, a concentration gradient between bulk and crystal surface is established; in this case, first and second Fick's laws can be used to describe the diffusion of a molecule, through the boundary layer, to a stationary particle. In a membrane crystallizer under forced flow, diffusion is enhanced by convective mass transport, and the particle flux J toward a crystal growing in supersaturated environment is generally expressed as:
J = kx(c~ -Ce)= ~D- ( % - ce)
(10)
268 Chapter 7
where kx is the mass transfer coefficient, D the diffusion coefficient, 8 the boundary layer thickness, coo and cc are the bulk and surface (equilibrium) concentrations, respectively. This yield to a convective growth rate G given by:
= ~-i k x C e ( S _ l )
(11)
where ~ is the number of molecules precipitated per unit volume of solution and S the supersaturation ratio. The boundary layer thickness 8 is evaluated by using Carlson analysis of laminar flow with a slip velocity u [5] :
t A-7-
tT)
(12)
L can be approximated by the crystal linear dimension, v is the cinematic viscosity. The most interesting feature of this theory is that 8 is predicted to be inversely proportional to the squared root of the velocity u and that G should increase proportionally to u 112 Peclet number is the control parameter for the dominant mechanism, defined as the ratio of rate constants for non-linear integration kinetics and bulk transport in the form of:
Pe k = flf-'~6
(13)
where 13f is the face kinetic coefficient. Decreasing values of Pek correspond to kinetic controlled regime with significantly faster transport. Under kinetic control, the incorporation of impurities is a potential source of defects and aggregates resulting in a growth rate reduction and quality degradation. According to these considerations, a deceleration of crystal growth due to enhanced supply of impurities to the interface is expected. Membrane crystallization experiments carried out on lysozyme with flow velocity varied in the range of 210-2100 btm/s show that crystal growth rate is improved by increasing the axial flow
Membrane Crystallization 269 velocity up to 1100 ~tm/s; conversely, a substantial deceleration of the HEWL growth process is observed if this threshold value was exceeded (figure 11) [19]. The linear shape of growth rates versus the square root of the average fluid velocities lower than 1100 pJn/s suggests that HEWL crystallization occurs under diffusive control with growth rates. Assuming v=1.4.10 -2 cm2/s and D = 4.4.10 -7 cm2/s, the boundary layer thickness 5 for a HEWL crystal with size of 100 pan is reduced from 110 pm to 50 p.m when the flow velocity increases from 210 p.m/s to 1100 p.m/s; according to literature data, ~ - 300 pm in absence of forced flow resulting in Pek - 0.05 [20]. With the lysozyme parameter 13~= 10.6 cm/s [21], Peclet number is shifted from 0.011 to 0.0080 by varying the flow velocity from 1100 pm/s to 2100 pro/s, thus increasing the relative weight of interfacial kinetics in the overall process control. Growth rate reduction due to impurity integration in the crystalline lattice is instead observed in correspondence of bulk flow rates greater than 1100 p.m/s. By conventional batch crystallization, experiments carded out on contaminated protein solutions (amount of impurities> 1.0% w/v) have shown that any value of flow rate causes a decreasing of the crystal growth rate and, often, growth cessation [22]. An interesting aspect of a membranebased crystallization device working under convection is that a maximum in growth rate is observed even using a non-purified HEWL solution (total impurity content > 5% w/v).
270 Chapter 7
<
o
I
500
,
I
,
I
1000 1500 Flow velocity, I~m/s
,
I
2000
J
2500
Figure 11. Crystal growth rates as a function of the axial flow velocity. Crystallization conditions: MgCI2 10% w/v, NaC1 3% w/v, 5°C. After [19].
M e m b r a n e Crystallization 271 References
[ 1] K. Toyokura. New aspects of industrial crystallization. J. Chem. Eng. Jap., 28/4 ( 1995) 361-371 [2] Crystal Growth Technology. H.J. Scheel and T. Fukuda. John Wiley & Sons Ltd, England (2003) [3] Chirality in industry II. A.N. Collins, G.N. Sheldrake and J. Crosby. John Wiley & Sons, England (1997) [4] Introduction to macromolecular crystallography. A. McPherson. John Wiley & Sons, New Jersey (2003) [5] J. Dirksen and T.A. Ring. Fundamentals of crystallization kinetic effects on particle size distribution and morphology. Chem. Eng. Sci., 46 (1991) 2389-2425 [6] Measurement of crystal growth and nucleation rates. J. Garside, A. Mersmann and J. Nyvlt. IChemE: United Kingdom (2002) [7] E.D. Hollander, J.J. Derksen, O.S.L. Bruinsma, H.E.A. van der Akker and G.M. van Rosmaten. A numerical study on the coupling of hydrodynamics and orthokinetic agglomeration. Chem. Eng. Sci., 56 (2001) 2531-2541 [8] H.J.M. Kramer and P.J. Jansen. Tools for design & control of industrial crystallizers. Proc. of the 15th International Symposium on Industrial Crystallization, 2002, Sorrento (Italy) [9] W.I. Genck. Guidelines for Crystallizer Selection and Operation. Chem. Eng. Prog., 100/10 (2004) 26-32 [10] E. Curcio, A. Criscuoli and E. Drioli. Membrane Crystallizers. Ind. Eng. Chem. Res., 40 /12 (2001) 2679-2684 [11] E. Drioli, E. Curcio, A. Criscuoli and G. Di Profio. Integrated system for recovery of CaCO3, NaC1 and MgSO4'7H20 from nanofiltration retentate. J. Membrane Sci., 239 (2004) 27-38 [12] Theory of particulate processes, analysis and techniques of continuous crystallization. A.D. Randolph and M.A. Larson. Academic Press Inc., San Diego - CA (1988) [13] M. Volmer and A.Weber. Keimbildung in Ubersattigten Gebilden. Z. Phys. Chem., 119 (1926) 277-301
272 Chapter 7 [14] J. Drenth, K. Dijkstra, C. Haas, J. Leppert and O. Ohlenschl~iger. Effect of molecular anisotropy on the nucleation of lysozyme. J. Phys. Chem., B 107 (2003) 4203-4207 [15]
T.E. Paxton, A.
Sambanis and R.W. Rousseau. Influence of Vessel
Surfaces on
the Nucleation of Protein Crystals. Langmuir, 17 (2001) 3076-3079 [16] A.M. Kulkarni and C.F. Zukoski. Nanoparticle Crystal Nucleation: Influence of Solution Conditions. Langmuir ,18 (2002) 3090-3099 [17] E. Curcio, G. Di Profio and E. Drioli. A new membrane-based crystallization technique: tests on lysozyme. J. Cryst. Grow., 247/1 (2003) 166-176 [18] G. Di Profio, E. Curcio, A. Cassetta, D. Lamba and E. Drioli. Membrane crystallization of lysozyme: kinetic aspects. J. Cryst. Grow., 257 (2003) 359-369 [19] E. Curcio, S. Simone, G. Di Profio, E. Drioli, A. Cassetta and D. Lamba. Membrane Crystallization of Lysozyme under Forced Solution Flow, in press on J. Membrane Sci. [20] P.G. Vekilov and F. Rosenberger. Dependence of lysozyme growth kinetics on step sources and impurities. J. Cryst. Grow., 158 (1995) 540-551 [2t] H. Lin, F. Rosenberger, J.I.D. Alexander and A. Nadarajah. Convective-diffusive transport in protein crystal growth. J. Cryst. Grow., 151 (I995) 153-162 [22] P.G. Vekilov, L.A. Monaco and F. Rosenberger. High resolution interferometric technique for in-situ studies of crystal growth morphologies and kinetics. J. Cryst. Grow., 146 (1995) 289-296
Chapter 8. Membrane Emulsification
1. Introduction to membrane emulsification Emulsions are disperse multi-phase systems of two or more immiscible liquids; they consist of at least one dispersed phase, present in the form of droplets in the continuous phase. From a thermodynamic point of view, unprotected small droplets are generally unstable. Referring to the classical case of oil in water, coalescence reduces the net surface area of the oil-water interface and leads to a gain in Gibbs energy ),AA, where 3, is the interfacial tension and A the droplet surface. On the other hand, the Gibbs energy required to disperse an oil with volume V in drops of radius R is 37V/R and increases with the size of droplets. Moreover, the higher the interfacial tension, the more unstable are droplets. Emulsion can be stabilized by protective agents, such as surfactants with charged head groups (electrostatic stabilization), polymer surfactants with hydrophilic heads (steric stabilization), proteins or other biological surfactants etc. The use of surfactants decreases the interfacial tension and promotes head-head repulsions that give kinetic stability. Under the condition that the surface tension is adequately low, the gain in configurational entropy is large enough to produce a stable dispersion (figure 1).
274 Chapter 8
Figure 1. Formation of an emulsion of liquid b) in liquid a).
Referring to systems I and II depicted in figure 1, the variation in Gibbs energy AG due to droplets formation is: A G = G u - G I = y A A _ T S II
(1)
where superscripts identify the system, T the temperature and S is the configurational entropy evaluated as" SH ~
kB
= - n d l n ~ b - nr l n O -~b )
(2)
In equation (2) kB is the Boltzmann constant, ~ the volume fraction of the dispersed phase, lad and nr the number of molecules of the dispersed and continuous phase, respectively. When droplets are formed, the total surface area increases of an amount AA approximately given by: A A = A II - A I ~ A I I = n4rcR 2
(3)
being n the number of droplets supposed to have an uniform radius R. Under these assumption, the critical value of surface tension in correspondence of AG=O is:
Membrane Emulsification 275
Ycritical
--
4a.R 2
(4)
that should be very small. Emulsions are conventionally prepared in large-scale using colloid mills, rotor-stator systems and high-pressure homogenizers [1]. Minimum sizes of droplet sizes broken up by shear stress are of about 1 ~tm, and a narrow droplet size distribution can be obtained if the energy density in the space between rotor and stator is well controlled [2]. Membrane emulsification is a relatively new technology [3;4] having potential to produce emulsion droplets with a narrow size distribution, relatively low shear stress and energy input, simple and efficient design. In cross-flow membrane emulsification, a liquid phase is pressed against a porous membrane to form droplets at the permeate side; here, the droplets are carried away from the membrane surface by continuous phase flowing (figure 2). The membrane should not be wetted by the phase to be dispersed.
276 Chapter 8 DISPERSED PHASE PRESED THROUGH THE MEMBRANE
DISPERSED PHASE PRESED THROUGH THE MEMBRANE
Figure 2. Principle of membrane emulsification.
Conventional emulsifiers typically work under severe operative conditions: pressures up to 100 MPa are requested to produce 50 m 3 of emulsions per hour with droplet sizes down to 0.1 ~tm, and only a small fraction of energy input is used to produce droplets (- 0.2% in a highpressure homogenizer), the rest being converted in heat [5]. As from figure 3, membrane emulsification is a more efficient process and requires much lower energy [6], especially for producing small droplets (
Membrane Emulsification 277
1
C•LOID MILL
E
=:L
v
\
0.5
CROSS FLOW MEMBRANE k EMULSIFICATION
E
\
\ HIGH" PRESSURE
\
0
"ID .
~=03
\
z
1
I,,-
\~
\ ~b=05\
= 05
MONO \ ",~LUIDIZEF
09
o
i
1 3
I 5
7
9
log energy density (J/rn3) Figure 3. Droplet diameter as a function of the energy density supplied by different types of equipment: cross-flow membrane emulsification (~ denotes the disperse phase fraction), high-pressure homogenizer (dotted lines), colloid mill, monofluidizer. After [6].
In a membrane emulsifier, few hundreds of kPa are generally sufficient to force the dispersed phase to permeate through a membrane into the continuous dispersion medium. In some cases, a coarse pre-mixed emulsion can be pressed through the membrane in order to reduce the droplet size of the dispersed phase; this also provides a reasonable emulsification flux (about 10 m 3 m "2 h "1 at 15"10 5 Pa) [7]. Because of the low shear stresses, new shear-sensitive droplets with shear sensitive components can be produced; in recent years, research activity is mainly addressed to produce monodisperse emulsions for food, pharmaceutical and cosmetic applications [8, 9]. In food industry, membrane technology can be applied to produce O/W emulsions such as dressing, artificial milk, cream liquors etc., as well as W/O emulsions such as margarine or spreads. At present, the main obstacle to commercialization of membrane emulsifiers is the low flux of the dispersed phase that largely depends on the properties of the membrane. For instance, although an high porosity is expected to be favourable, a recent study demonstrated that a
278 Chapter 8
membrane porosity higher than 1.5% increases the probability of coalescence during droplet formation [ 10]. Fluxes for producing O/W emulsions range typically from 2 to 40 1/m2h for microporous membranes having a nominal pore size comprised between 0.2 and 0.8 gm, respectively [9]. On the other hand, fluxes of 2300 1/m2h through hydrophilic membranes with pore size of 1 btm to produce W/O emulsions were achieved by Kato and colleagues [ 11]. The overall productivity is also depressed by the low percentage of the active pores (3-40% of the pores, depending on the pore size and transmembrane pressure (reported in [ 12]).
2. Theoretical aspects
Theoretical descriptions of a membrane emulsification process focus attention on two fundamental aspects: permeation of the dispersed phase through the pores of a membrane, and dynamic mechanism of droplet detachment which arise from operating parameters (transmembrane pressure, cross-flow velocity), physical properties of the membrane (pore size and porosity, number of active pores, thickness, hydrophobic/hydrophilic character etc.), and physico-chemical properties of emulsified system (density, viscosity, surfacial tension etc.). In most cases, the aim of theoretical contributions is to theoretically support the linear relationship, experimentally observed, that relates the droplet size of the emulsion dd to the pore size of the membrane dp for a given set of operating conditions:
dd =Zdp
(5)
where ~ typically ranges between 2 and 10 [13,14], although a value near to 50 has been observed for particular operating conditions and emulsifier concentrations [15].
Membrane Emulsification 279
Modelling studies have been carried out with different approaches, most frequently including force balances, torque balances, and CFD simulation [ 16]. The flux of the dispersed phase Jd through the pores of the membrane is generally assumed to be described by the Darcy's law:
and relation (6) takes the form of the Hagen-Poiseuille equation. In eq. (7), N o is the number of active pores, Am the membrane area and x the mean tortuosity. The transmembrane pressure APtm is given by the difference between the applied pressure to the dispersed phase Pd and the mean pressure of the continuous phase Pc = (Pc.,, + P .... , )/2"
Aptm =p~ --pc
(8)
where pc,in and pc,out are the pressure of the flowing continuous phase at both inlet and outlet of the membrane module. However, due to the curvature of the interface and the resulting capillary pressure, a minimum pressure difference APy is required to blow up the drop through the pore. Capillary pressure leads to a drug reduction of the disperse flow; according to Laplace's equation, the intensity of this contribution depends on both dynamic interfacial tension 7(t) and droplet size:
APy = 4 y(t_____)) de
(9)
The driving force to the flux of dispersed phase is, therefore, conveniently described by an effective transmembrane pressure APe, given by:
280Chapter8 (10)
APe "- Aetm - z~Jgy
While a droplet is being formed at the single pore of a membrane, several forces determining its growth and detachment have to be considered [18,19]. Whereas the final droplet size is determined by the droplet volume at the end of the detachment stage [20], the final droplet shape mainly depends on the cross-flow shear force. Droplets are spherical at the pore mouth if this force is small with respect to the interfacial tension; the droplet deforms from spherical symmetry as the cross-flow velocity increases [21]. Forces acting on a droplet forming at the pore mouth are schematically depicted in figure 14. iliI!iiII ii~iiii i~ii!Iii~~~!i~
FBG ik
FD
, FR , . ~ ~ [
Fa ,,
~stat ::~:::':'
Fy-F
Figure 4. Forces during droplet formation. After [8].
The viscous drag force (FR) is generated by the continuous phase flowing parallel to the membrane surface and, for a Reynolds number Re <1, is:
Membrane Emulsification 281
FR =3nkx/.tcddV ~
(11)
where kx is the wall correction factor (kx=l.7 for a single sphere touching an impermeable wall in simple shear force ([22] and ref. therein)), vc and pc the undisturbed tangential velocity and the viscosity of the continuous phase, respectively. The dynamic lift force (FD), resulting from an asymmetric velocity profile of the continuous phase near the droplet, is given by [ 18]: 13
0.5
F D = 0.761 crL5 adPc
(12)
Pc where aw is the shear stress at the membrane surface, pc and pc are the density and the viscosity of the continuous phase, respectively. The interfacial tension force (F0, which represents the effects of dispersed phase adhesion around the edge of the pore, is:
Fr=ZdNr(t )
(13)
where dy is the droplet neck diameter. The static pressure difference force, Fstat, is due to the pressure difference between disperse and continuous phases required to overcome the capillarity:
F,,o, = z ~ = F
de
r da
(14)
The buoyancy force FBG results from the density difference between dispersed and continuous phases:
FBG :(Pc--Pd)g V
(15)
where V is the droplet volume. The inertial force F1 is the linear momentum force associated with a mass of the fluid flowing out from the opening of the pore:
282 Chapter 8
(16) AN
where Vp is the velocity of the dispersed phase inside the pore, e is the membrane porosity and AN the cross sectional area of the droplet neck.
.... -10 L _
"o .... o
~
FR
-
I i
F~
-
|
Z
-13
!
o ,9o r 0
-16 _
begining contraction
_
i
-19 ~
m I -22 1
2
3
4
droplet diameter (lam) Figure 5. Amount of forces depending on the droplet diameter. After [8].
Not all forces previously mentioned are significant for a microscopic balance: for instance, when considering the relative preponderance, it is found that inertia and buoyancy forces are approximately 9 and 6 order of magnitude smaller than drag and interfacial tension forces for micron-scale droplets. For a typical membrane-emulsification setup (pore diameter: 1 ktm, interfacial tension: 5 mN/m, dispersed phase velocity: 1 mm/s, oil density" 800 kg/m 3, viscosity: 50 mPa.s), Sugiura et al. (2001) [23] calculated: 9
Re
.
inertial force pvdp . . . viscous force l~
2.1
0- 5
(17.a)
M e m b r a n e E m u l s i f i c a t i o n 283
9
9
9
Bo
We
.
gravitational force . . . interfacial tension force
Ap g d 2 y
4.10 -7
(17.b)
.
inertial force . . . int erfacial tension force
P v2dp 7"
2 . 1 0 -7
(17.c)
Ca =
.
viscous force = p v = 1.10 -2 int erfacial tension force y
(17.d)
being Re the Reynolds number, Bo the Bond number, We the Weber number, and Ca the Capillary number. In general, the interfacial tension force dominates a membrane emulsification process. The point at which a droplet detaches after reaching a certain volume occurs when the sum of these forces acting on it equals zero. The fundamentals of droplets formation in membranes can be suitably investigated by direct visualization of the formation of individual emulsion droplets using a microscope video systems.
3. Effect o f m e m b r a n e p a r a m e t e r s
Until now, membranes studied for emulsification process include: 9
microporous glass (MPG) [24, 25, 26]
9
hydrophilic Shirasu Porous Glass (SPG), symmetric structure, pore size: 0.4-6.6 ktm, porosity: 0.5-0.6 [13, 25, 26, 27, 28]
9
ceramic ~-A1203 eventually coated with zirconium oxide, hydrophilic, asymmetric structure, pore size: 0.1-0.8 ~tm [9, 19, 21 ]
9
PTFE [7]
9
Polypropylene, hollow fibres, pore size: 0.4 ktm [29]
284 Chapter 8
9
Polycarbonate, hydrophilic, parallel pores, pore size: 0.05-12 ktm, porosity: 0.05-0.2
[30]; 9
Microsieve, hydrophilic, parallel pores, pore size: 0.2-20 ~tm [12, 31,32]
9
Cellulose acetate, hydrophilic, asymmetric structure, pore size: 0.2-0.45 ~tm [33]
9
Polyamide (10 kDa) [15]
These membranes, commonly used in separation processes, are not specifically developed and optimized for emulsion process and suffer from important drawbacks: their high porosity is the main cause for coalescence, and a non uniform pore size distribution is a serious obstacle to the achievement of a narrow droplet size distribution. In this respect, pore size-tailored membranes manufactured in the form of laser-drilled, stainless-steel sheets [34], or inorganic membranes made by laser-interference lithography and silicon micro-machining technology [35] have been also considered. The final droplet size distribution primarily depends on the choice of the membrane, although an appropriate setting of the operative parameters (such as cross-flow velocity, emulsifier concentration, dispersed phase transmembrane flux) allows to produce emulsions with relatively narrow droplet size distribution. Coalescence or steric hindrance induce polydispersity in emulsions; in order to prevent these phenomena, the membrane porosity should be low. Computational fluid dynamics calculations show that porosity should not exceed 1% for a ratio of droplet diameter over pore diameter of 7 [10]. However, hindrance is often limited also in high-porosity membranes because of the low fraction of active pores [36]. Assuming a square array of pores (all active), the maximum porosity while preventing steric hindrance and coalescence can be calculated assuming that the distance (x) between pores should be at least be equal to the droplet diameter [6]:
Membrane Emulsification 285
c = 0.25~r(132
(18)
Droplet deformation in the direction of the cross-flow is not considered by this equation. Droplets generally form at only 3-40% of the pores depending on pore size and transmembrane pressure. The low percentage of active pores is a serious disadvantage for the overall productivity. For pores having the same diameter, it could be expected that a critical transmembrane pressure (defined as the pressure at which a droplet starts to form at the pore exists). If this threshold - theoretically given by the Laplace's equation (9) - is surmounted, all pores become active. However, experimental investigations suggest that the number of active pores increase linearly with increasing transmembrane pressure, and droplets form only at few pores (figure 6). With a transmembrane pressure of 14 kPa, approximately 3 times the critical value, only the 16% of the monitored pores were active [12]. An analogous results was obtained by Sugiura et al.(2000) [37]: the percentage of active pores did not exceed 20% below a transmembrane pressure of 2.5 times the critical one. These results also showed that the velocity did not affect significantly the number of active pores.
286 Chapter 8
'
I
'
A
I
I
'
I
I
I
'
I
I
16
"i"
Np,active-(1.4
v
12
o >
. _ ,.i-,
o
8
o ..Q
E z
4
I
0
4
8
12
16
Transmembrane pressure (kPa)
Figure 6. Number of active pores as function of transmembrane pressure (evaluated on 100 pores with diameter of 7 ~tm). Cross-flow velocities are comprised between 0.011-0.039 m/s. After [ 12].
The minimum transmembrane pressure required for droplets formation differ at pores due to their size and tortuosity that cause a pressure drop distribution in the membrane pore broad. The relationship between the pressure drop APd in the membrane pore and the flux of the dispersed-phase is described by the Fanning's equation:
4f(lp l(u2) td. jr-7- )
(19)
being f the friction factor, lp the pore length and u the dispersed-phase velocity though the membrane pore. The membrane thickness indirectly influences the final droplet size: at a given transmembrane pressure, larger thickness decreases the disperse phase flux and lowers the droplet expansion rate.
Membrane Emulsification 287
The hydrophobic/hydrophilic character of a membrane plays a significant role in determining the final droplet size and monodispersity. In order to obtain droplets in the disperse phase, the membrane should be wetted with the continuous phase (i.e: wall contact angle measured in the continuous phase 0scd <90~ Young's equation provides a correlation between (lsc d and the properties of the two immiscible liquids and the solids: COS Oscd = Ysd -- Ysc
(20)
Ycd
in which ')tsd and ~/sc are the interfacial tensions of the boundary solid/disperse phase and solid/continuous phase, respectively; ~cd is the interfacial tension between the two liquid phases. The solid/liquid interfacial tensions vary for different emulsifiers; the liquid/liquid interfacial tensions mainly depend on the expansion rate of the droplet and on the type and concentration of the emulsifier used. The effect of the wall contact angle on droplet formation has been simulated with CFX 4.3, and results show that droplet spreads over the membrane surface already at a wall contact angle of 60 ~ as consequence of the droplet deformation induced by the cross-flow of the continuous phase (figure 7) [ 10].
288 Chapter 8
Figure 7. Droplet deformation for different wall contact angles as calculated with CFD. After [ 10].
The wetting properties of the membrane can be manipulated by pre-soaking the membrane in the continuous phase [38], or by chemical modification of the membrane surface [39;40].
4. Membrane resistance
For an hydrodynamic characterization, the membrane is generally considered as: 9 a layer with parallel pores supported by a certain structure below it (e.g: microsieves, silicon microchannels); 9 a structure with interconnected pores (as in the case of symmetric SPG membranes or asymmetric ceramic membranes).
Membrane Emulsification 289
In both cases, the overall resistance against the disperse phase flow is assumed to consist of two terms: the resistance in the (upper) pore layer (Rup), and the resistance in the residue structure (Rs) [41 ]. A dense support structure, resulting in a high R~, is requested for microsieves with small pore diameters. In the case of asymmetric or symmetric membranes with interconnected pores, Rs can be estimated by [6]" R~ = ~
6
(21)
AmKmb
where 8 is the membrane thickness, Am the membrane area, Km the membrane permeability and b a constant (= 1 if 8/do>5). The value of Run for layers with parallel pores is given by [41 ]" R~P ~to~( 128/p
24/r ~:d4p +-~p)(-Ea' k'+~ll2,:l
=
/
(22)
where lp is the pore length, k is the product of active pore fraction and porosity, Ntot the total number of pores, and a1=0.344, a2=0.111, a3=0.066. For interconnected pores is [6]"
R,,p = Rsa @-
(23)
where a is a constant, falling in the range of 0.5 (thick membrane) < a < 1 (very thin membrane). For membranes with interconnected pores, the total resistance offered by the membrane Rm can be estimated by Karman-Kozeny's equation [42]" Rm = C (1 - e)2 Sm26 C
3
with C=2 for uniform cylindrical pores and
(24)
290 Chapter 8 4s am=~ dp(1-c)
(25)
An analytical solution for the required membrane area can be obtained for membranes with equally sized pores. The fraction n of active pores is a function of the ratio of the resistances Rs and Rup, and of the ratio of transmembrane pressure over critical pressure:
n
= RuP (APt" -1~ e~ (Ap~ )
(26)
Substituting n=l in the previous equation gives the minimum transmembrane pressure at which all pore are active: tin
APt
Rs
=~
R,p
+1
(27)
To calculate the membrane area required to produce a certain amount of emulsion per unit of time, Gijsbertsen-Abrahamse et al. (2004) [6] have derived the following equations: -1 A m . ltlJdR; ( APtm 1 . . . . . , for
i Ap,
/( /'
Am _ ,UJd (R; + R~p Apt,. --
@y
~ Apy )
Ap,m
Ap,
<
R;
+1
,for APtm> R; Apy - R;p d- 1
(28.a)
(28.b)
where R's=Rs'Am, R'up=Rup'Am, ~t is the viscosity, and Jd the flux of the dispersed phase. From a technological point of view, the optimization of a membrane emulsifier requires to minimize the membrane area and, therefore, to reduce the membrane resistance. For a given pore size, this can be done by increasing the membrane porosity or by decreasing the membrane thickness. In general, a convenient balance between the strength of the top-layer, strength of the substructure, and the sum of both resistances is requested to obtain a maximum disperse phase flux.
Membrane Emulsification 291 5. Measurement of droplets diameter Droplet size distribution is generally measured by optical microscopy or dynamic light scattering system, allowing the detection of droplets with a minimum diameter of about 0.1 lam. The mean droplet size is typically expressed as the mean Sauter diameter d3.2 which is the diameter of a spherical droplet having the same area per unit volume, Sv, as that of the total collection of droplets in the emulsion:
(29)
d3,2 =~'-v =
where Vi is the volume fraction of droplets in the i-th range of sizes having a mean diameter di, and ks is the number of size ranges. The shape of a droplet size distribution is quantified by a Coefficient of Variation (CV), defined as:
cv(%)=
aav
x 100
(30)
where SD is the standard deviation, and dav the mean droplet diameter. In a practical way, the width of a droplet size distribution is expressed as a span ot of distribution: a -- (d90% -- a10% ) d50%
(31)
were dx0 is the diameter corresponding to x0 vol.% on a cumulative droplet size distribution curve. The more uniform the droplet diameters are, the closer to zero span values become.
292 Chapter 8
Figure 8. Oil-in-Water emulsion (SDS 2%, PVA 0.8%) and related droplet size distribution.
For statistical analysis of the droplet size distribution, data points can be often fitted with a 4-parameter Gaussian curve:
N(d)= N O+ aexp
I(d-dm)2 2~ ~
(32)
where N denotes the number of drops, d the droplet diameter, dm the most probable value (corresponding to the maximum of the Gaussian curve, b is the half-width of the peak, No characterizes the background at which the peak appears, and a is a scaling parameter. As previously reported, a linear correlation between droplet sizes and membrane pore diameters is experimentally observed (eq. 5). In order to exemplify this concept, the results of Katoh et al (1996) [13] and Asano et al. (1999) [25] for O/W monodispersed emulsions are reported in figure 9.
Membrane Emulsification 293 30
'
I
O
[] /k
v
E ~-
I
'
I
Deionizedwater/PGEat 10 wt% corn oil") Deionizedwater/SE at 0.3 wt% corn oil / /~ Deionizedwater/SDSat 2 wt% k e r o s e n e ~ -/
20
E "o
s
lO
D
0 0
2 4 Membrane pore diameter (gin)
6
Figure 9. Relationship between membrane pore size and droplets diameter. O/W emulsions produced by MPG tubes under N2 gas pressure. After [ 13, 25].
6. The role of surfactants
In an emulsification process, as a rule, a surfactant (emulsifier) is dissolved in the continuous phase to stabilize the produced emulsion against drop coalescence. The adsorption kinetics of the emulsifier determine the time needed to stabilize the droplets against coalescence during the emulsification process. Both the rate at which deformation and detachment forces act, and absorption rate of the surfactants to the growing interfacial area, are relevant over the time-scale of the process. At high relative expansion rates, diffusion rate is not sufficiently fast to supply the expanding interface with surfactants (case a in figure 10). On the contrary, surfactant molecules cover the droplet interface if the absorption kinetics is very fast (case b in figure 10); in this case, the relative expansion rate is slower than the diffusion rate of surfactant.
294 Chapter 8
"~o/
~.
o_~
T /o
o _
io
o--
/o
j/jj j/j/jj o,,,
Figure 10. Relative expansion rate of the droplet a) higher than surfactant adsorption rate; b) lower than surfactant adsorption rate. After [32].
During droplet formation, surfactants absorb onto the interface of the growing droplet, thus reducing the interfacial tension at detachment and decreasing the volume of the droplet. As exemplified in figure 11, experimental data confirm that droplet diameter decreases at higher surfactant concentration [32]. Also the droplet formation time decreases as a function of the surfactant concentration.
Membrane
120 L
'
I
'
I
'
I
'
I
Emulsification
295
'
100 E =L v
80
I
E . m
"o (1) t,-3 O
60
~II--~
m
Transmembrane pressure: 0.43 bar
tm 40 f
IL Transmembrane pressure: 0.94 bar I) i
20
0
II
I
i
I
J
I
i
I
2 4 6 8 Concentration of Tween 20 (% w/w)
i
10
Figure 11. Droplet diameter as function of surfactant concentration at different transmembrane pressures. After [32].
In membrane emulsification, the relative expansion rate and the dynamic interracial tension change in time. For a growing droplet under a constant dispersed-phase flow-rate, the relative expansion rate 0 just before detachment is calculated as [32]" ldA 2 0. . . . .
A dt
3tf
(33)
where A is the interracial area of the droplet and tf the time of droplet formation. A the initial stages the relative expansion rate is large and decreases in time. For diffusion of surfactants to a spherical interface the loading F is given by [32]:
(34)
with F the loading of the interface with surfactants, D the effective diffusion coefficient of the surfactant (e.g." for Tween 20, D=8.3.10 "ll m2/s), and c the surfactant concentration. The
296 Chapter 8 effect of a curved interface on the loading depends on the relative magnitude between the diffusion concentration depth ~ - t If ~
and the radius of droplet rd.
t <<2rd, the absorption step- st short time scale - is diffusion controlled and:
~Dt
F = 2 ~c
(35)
//-
The bursting membrane method [43] allows measurements of dynamic interfacial tension (LL) in the subsecond range. An empirical expression of the dynamic interracial tension
7F(t) for Tween 80 emulsifier has
been given by De Luca et al. ([22] and references therein): 25.1 + 17.9t YF(t)=i~b-+ 2.20t 183 TM
(36)
7F(t) is in N/m, t in seconds. The dynamic behaviour of the interfacial tension of Lacprodan 60, Tween 20 and LEO 10 are reported in figure 12; dependence of interfacial tension on SDS concentration is shown in figure 13.
Membrane Emulsification 297 30
'
. . . . . . .
I
. . . . . . . .
I
. . . . .
E Z
E
20
C O (/) C
O
C
I VEGETABLE OIL/WATER ?0=27mN/m 0.01
0.1
LEO 10, 0.7%
1 Time (s)
10
100
Figure 12. Dynamic interfacial tension of different emulsifiers measured by the bursting membrane method. After [8].
30
i
i
i
i
i
iii
I
'
'
'
' ' ' " 1
'
'
'
''
_
25
E Z
E
20
c.mO (/)
c
15
{
lO
C
SDS
0
0.001
I
I
I
I I I ill
I
I
'
' ''~I[
I
0.01 0.1 Surfactant concentration (%)
I
I
I I Ill
1
Figure 13. Dependence of interfacial tension on SDS surfactant concentration. After [44].
298 Chapter 8 The different typology of emulsifier used has a remarkable influence on the droplet size: a fast adsorption of the emulsifier at newly interfaces gives rise to smaller droplets produced and to a lower tendency to coalescence, especially during droplet formation. On the contrary, a low coverage rate of the drop surface with absorbed surfactant can lead both to a larger interfacial tension, and to a significant coalescence in the produced emulsion.
7. Typical membrane emulsification setup A schematic diagram of a membrane emulsification system (figure 14) typically consists of an emulsion reservoir, a membrane module, a gear pump and an heat exchanger for recirculating the continuous phase. The phase to be dispersed can be forced through the membrane by using gas pressure (typically nitrogen- shown in the figure) of by an additional gear pump operating on the feed side.
~> Membranemodule
Continuous
Compressedgas (nitrogen,air....)
phase
Dispersed phase
Pump
Heatexchanger
Figure 14. Membrane emulsification setup.
8. Transmembrane pressure The critical pressure is defined as the minimum pressure difference over a pore necessary to start producing droplets. According to Laplace equation (eq. 9), the critical pressure is
Membrane Emulsification 299
inversely proportional to the pore diameter and, in membranes with uniform pores with sharp edges, all the pores should be active at the same critical pressure. The wall contact angle slightly affect the critical pressure if the pore edges are rounded, and if the size at the pore opening differs from the smallest pore size [10]. A schematic representation of droplet formation at several stages of time at a pore with sharp edges is illustrated in figure 15.
,,~176176 ........ "~% ,"'4
,
""\\
1";3" '" ....."lr..... P "'\ ,.'/
DISPERSED PHASE FLUX
Figure 15. Evolution in time of droplet formation at sharp edges of a membrane pore. Critical pressure is defined for hemispherical droplet at stage 3. After [6].
For cylindrical pores of uniform diameter dp, droplet formation time tf can be expressed as a function of the disperse phase flux Jd and the mean droplet diameter do: 2 e k d4,3 tf =
3 dp2 Ja
(37)
where k is the fraction of active membrane pores. Calculations of the droplet formation times made by Schroeder et al (1999) [ 19] on the basis of mean droplet sizes experimentally observed under the assumption of no coalescence are
300 Chapter 8 reported in Table 1. The formation times of droplets stabilized with SDS (2% w/w) are shorter than those calculated and observed for Tween 20 (0.1%w/w).
Table 1. Calculation of droplet formation times for different transmembrane pressures and emulsifiers. After [ 19] Emulsifier Transmeml~rane Dispersed phase Droplet formation Average drop pressure (Pa) flux (l/m2 h) time (s) volume (~l,m3) SDS (2% w/w) 0.5 9 105 8 0.09 10.6 SDS (2 % w/w) Tween
1.3 9 105
40
0.04
22.4
20
(0.1
0.7.105
7
1.5
151
20
(0.1
1.5.105
47
1.05
693
%w/w) Tween %w/w)
The effect of increasing transmembrane pressure varies with the pore size of the membrane and the emulsifier used. If the decreasing time of interfacial tension is significantly shorter with respect to droplet formation times, the effect of increasing transmembrane pressure on the droplet size is negligible. If droplet formation and detachment occur within the period of decreasing interfacial tension, larger droplets are formed due to stronger retaining forces (figure 16).
M e m b r a n e Emulsification 301 L_
(D
E
12
'
I
'
I
'
I
'
I
'
"o i...
m
12
E
O fD_ ~D c'-
v I_
_0
E E
E
(D
tt~ 03
9 %w/w)
El. O L_
E
-
.i...a
c-
E
......
. _
SDS (2% w/w)
.l...a
O
0
i
0
I
0.4
i
I
0.8
,
I
1.2
i
I
J
1.6
0
2
-o tt~
Mean transmembrane pressure (bar) Figure 16. Influence of transmembrane pressure and dynamic interfacial tension of different emulsifiers on droplet size of the disperse phase, dp=0.8 gm. After [8].
A qualitative
agreement between this explanation
and experimental
results has been
furnished by Schroder et al (1998) [8]. In order to produce a net flow through membrane pores, the pressure applied to the disperse phase has to overcome both capillary pressure and the continuous phase pressure. When tubular or capillary membranes are used, droplets forming near the inlet of the module experience a different pressure with respect to droplets forming near the outlet; the relative pressure drop (at), compared to the effective transmembrane pressure (APe), is calculated as: (38)
where
I~ m
is the wall shear stress. These effects can be relevant with increasing the length
(Lm)
or with decreasing the diameter (din) of membrane fibres and have to be considered during scale-up.
!!
i
iii!
~ !iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii
302 Chapter 8 9. Flux of the dispersed phase and continuous phase crossflow According slope
to Darcy's
increase
increases
with
increases exceed
with
law, flux data of the dispersed transmembrane
transmembrane
pressure,
as well. This behaviour the capillary
pressure
25
iii~i~(?i~ii~/i~I~iii~!ii//~i~i~i~ i~!~/~ ii~i(ii(i~!~i~ ~ ii~iI!ii~ii!!~i~i!ii~!~!~i i~i~i~l ii~i~i~i~i~i~i!~!i
pressure.
}
probably
is observed
of the smallest
,
For
phase
generally
shows
dp<0.2
~tm,
this
slope
because
the
number
as long as the transmembrane pores (see figure
I
'
I
of
a constantprogressively active
pressure
pores
does not
17).
'
I
,,l
/
%)
2"
2o
x
15
[-
/ Lacprodan60 (0.1 Yo) S-[
Q.
10
5
10 (0.7o~/o)
0 1
F__JJ ,
I
2
3
,
I
I
,
4
l
5
1 I 6
Mean transmembrane pressure (bar) Figure 17. Influence
of transmembrane
on the flux of the disperse
phase for different
emulsifiers.
After [8].
A number in a greater 36]
of experiments mean
droplet
show that typically size and in a higher
an increase
of the transmembrane
polydispersity
of the formed
drops
flow results [8; 13; 24; .........
Membrane Emulsification 303
The presence of crossflow in the continuous phase activates an hydrodynamic drag force that helps the emulsion droplets to detach from the membrane; the final size of the produced drops is therefore decreased [9; 19; 20; 24]. In general, experiments show that shear stress at the membrane surface caused by the flow of the continuous phase has a little influence on the droplet size. According to the investigation of Schroder et al (1999) [ 19] for 0.1 lam membranes, droplet size is independent of shear stress for aw> 2 Pa; for membranes with mean pore size of 0.8 gm, there is not influence for aw> 30 Pa.
304 Chapter 8 References [1 ] H. Schubert and H. Armbruster. Principles of formation and stability of emulsions. Int. Chem. Eng., 32 (1992) 14-28 [2] W.S. Arbuckle. Emulsification, in: C.W. Hall, A.W. Farral, A.L. Rippen (Eds.), Encyclopedia of Food Engineering. Avi Publishing Company Inc.: Westport (Connecticut), 1978 [3] T. Nakashima and M. Shimizu. Advanced inorganic separative membranes and their developments. Chem. Eng. Symp. Ser., 21 (1988) 93-99 [4] Membrane Emulsification Operational Manual. T. Nakashima, M. Shimizu and M. Kukizaki. First Ed., Industrial Research Institute of Miyazaki Prefecture, Miyazaki (1991) [5] P. Walstra and P.E.A. Smulders, Emulsion formation, in: B.P. Binks (Ed.), Modern Aspects of Emulsion Science, The Royal Society of Chemistry: Cambridge (UK), 1998 [6] A.J. Abrahamse, A. van der Padt and R.M. Boom. Status of cross-flow membrane emulsification and outlook for industrial applications. J. Membrane Sci., 230 (2004) 149-159 [7] K. Suzuki, I. Fujiki and Y. Hagura, Preparation of corn oil/water and water/corn oil emulsions using PTFE membranes. Food Sci. Technol. Int., Tokyo 4 (1998) 164-167 [8] V. Schroder, O. Behrend and H. Schubert. Effect of dynamic interfacial tension on the emulsification process using microporous, ceramic membranes. J. Colloid and Interf. Sci., 202 (1998) 334-340 [9] S.M. Joscelyne and G. Tragardh. Food emulsions using membrane emulsification: conditions for producing small droplets. J. Food Eng., 39 (1999) 59-64 [10] A.J. Abrahamse, A. van der Padt, R.M. Boom and W.B.C. de Heij. Process fundamentals of membrane emulsification: simulation with CFD. AIChE J., 47 (2001) 1285-1291 [ 11] R. Katoh, Y. Asano, A. Furuya, K. Sotoyama, and M. Tomita. Preparation of Food Emulsions using Membrane Emulsification System, in: Proceedings of the 7th International Symposium on Synthetic Membranes in Science (1994) Tubingen, Germany
Membrane Emulsification 305
[12] A.J. Abrahamse, R. van Lierop, R.G.M. van der Sman, A. van der Padt and R.M. Boom, Analysis of droplet formation and interactions during cross-flow membrane emulsification. J. Membrane Sci., 204 (2002) 125-137 [13] R. Katoh, Y. Asano, A. Furuya, K. Sotoyama and M. Tomita. Preparation of food emulsions using a membrane emulsification system. J. Membrane Sci., 113 (1996) 131-135 [14] Y. Mine, M. Shimizu and T. Nakashima. Preparation and stabilization of simple and multiple emulsions using a microporous glass membrane. Colloids and Surfaces B, 6 (1996) 261-268 [15] L. Giomo, N. Li and E. Drioli. Preparation of oil-in-water emulsions using polyamide 10 kDa hollow fiber membrane. J. Membrane Sci., 217 (2003) 173-180 [16] M. Rayner and G. Tragardh. Membrane emulsification modelling: how can we get from characterization to design? Desal., 145 (2002) 165-172 [17] Basic Principles of Membrane Technology. M.H.V. Mulder. Kluwer Academic Publishers, Dordrecht (1991) [ 18] S.J. Peng and R.A. Williams. Controlled Production of Emulsions Using a Crossflow Membrane Part I: Droplet Formation from a Single Pore. Chem. Eng. Res. Des., 76 (1998) 894-901 [ 19] V. Schroeder and H. Schubert. Production of emulsions using microporous, ceramic membranes. Colloids and Surfaces A, 152 (1999) 103-109 [20] S.J. Peng and R.A. Williams. Controlled production of emulsions using a crossflow membrane. Part I. Droplet formation from a single pore. Trans. IchemE, 76 (1998) 894-901 [21] R.A. Williams, S.J. Peng, D.A. Wheeler, N.C. Morley, D. Taylor, M. Whalley and D.W. Houldsworth. Controlled production of emulsions using a cross flow membrane. Part II. Industrial scale manufacture. Trans. IchemE, 76 (1998) 894-901 [22] G. De Luca, A. Sindona, L. Giorno and E. Drioli. Quantitative analysis of coupling effects in cross-flow membrane emulsification. J. Membrane Sci., 229 (2004) 199-209 [23] S. Sugiura, M. Nakajima, S. Iwamoto and M. Seki. Interfacial Tension Driven Monodispersed Droplet Formation from Microfabricated Channel Array. Langmuir, 17 (2001) 5562-5566
306 Chapter 8 [24] I. Scherze, K. Marzilger and G. Muschiolik. Emulsification using micro porous glass (MPG): surface behaviour of milk proteins. Colloids and Surfaces B, 12 (1999) 213-221 [25] Y. Asano and K. Sotoyama. Viscosity change in oil/water food emulsions prepared using a membrane emulsification system. Food Chem., 66 (1999) 327-331 [26] G. H. Ma, M. Nagai and S. Omi. Preparation of uniform poly(lactide) microspheres by employing the Shirasu Porous Glass (SPG)emulsification technique. Colloids and Surfaces A, 153 (1999) 383394 [27] G. T. Vladisavljevic and H. Schubert. Preparation and analysis of oil-in-water emulsions with a narrow droplet size distribution using Shirasu-porous-glass (SPG) membranes. Desal., 144 (2002) 167-172 [28] L-Y. Chu, R. Xie, W.-M. Chen, T. Yamaguchi and S. Nakao. Study of SPG membrane emulsification processes for the preparation of monodisperse core-shell microcapsules. J. Colloids and Interf. Sci., 265 (2003) 187-196 [29] G.T. Vladisavljevic, S. Tesch and H. Schubert. Preparation of water-in-oil emulsions using microporous polypropylene hollow-fibres: influence of some operating parameters on droplet size distribution. Chem. Eng. Proc., 41 (2002) 231-238 [30] I. Kobayashi, M. Yasuno, S. Iwamoto, A. Shono, K. Satoh and M. Nakajima. Microscopic observation of emulsion droplet formation from a polycarbonate membrane. Colloids and Surfaces A, 207 (2002) 185-196 [31] C.J.M. van Rijn and M.C. Elwenspoek. Microfiltration membrane sieve with silicon micro machining for industrial and biomedical applications, proc. IEEE 83 (1995) 83 [32] S. van der Graaf, C.G.P.H. Schroen, R.G.M. van der Sman and R.M. Boom. Influence of dynamic interfacial tension on droplet formation during membrane emulsification. J. Colloids and Interf. Sci., 277 (2004) 456-463 [33] M. Shima, Y. Kobayashi, T. Fujii, M. Tanaka, Y. Kimura, S. Adachi and R. Matsuno. Preparation of fine W/O/W emulsion through membrane filtration of coarse W/O/W emulsion and disappearance of the inclusion of outer phase solution. Food Hydrocolloids, 18 (2004) 61-70
Membrane Emulsification 307
[34] P.J. Dowding, J.W. Goodwin and B. Vincent. Production of Porous Suspension Polymer Beads with a Narrow Size Distribution Using a Cross-flow Membrane and a Continuous Tubular Reactor. Colloids and Surfaces A, 180 (2001) 301-309 [35] S. Kuiper, C.J.M. van Rijin, W. Nijdan and M.C. Elwenspoek. Development and applications of very high flux microfiltration membranes. J. Membrane Sci., 150 (1998) 1-8 [36] H. Yuyama, T. Watanabe, G.H. Ma, M. Nagai and S. Omi. Preparation and analysis of uniform emulsion droplets using SPG membrane emulsification technique. Colloids and Surfaces A, 168 (2000) 159-174 [37] S. Sugiura, M. Nakajima, J. Tong, H. Nabetani and M. Seki. Preparation of monodispersed solid lipid microspheres using a microchannel emulsification technique. J. Colloid and Interf. Sci., 227 (2000) 95-103 [38] T. Fuchigami, M. Toki and K. Nakanishi. Membrane emulsification using sol-gel derived macroporous silica glass. J. Sol-Gel Sci. Technol., 19 (2000) 337-341 [39] T. Kawakatsu, G. Tragardh, Ch. Tragardh, N. Nakajima, N. Oda and T. Yonemoto. The effect of the hydrophobicity of microchannels and components in water and oil phases on droplet formation in microchannel water-in-oil emulsification. Colloids and Surfaces A, 179 (2001) 29-37 [40] T. Kawakatsu, Y. Kikuchi and N. Nakajima. Regular-sized cell creation in microchannel emulsification by visual microprocessing method. J. Am. Oil Chem. Soc. 74 (1997) 317-321 [41] A.J. Abrahamse, A. van der Padt and R.M. Boom. Influence of membrane morphology on pore activation in membrane emulsification. J. Membrane Sci., 217 (2003) 141-150 [42] Membrane handbook. W.S.W. Ho and K.K. Sirkar. Van Nostrand Reinhold (Ed), New York (1992) [43] M. Stang, H. Karbstein and H. Schubert. Adsorption Kinetics of Emulsifiers at Oil-Water Interfaces and their Effect on the Mechanical Emulsification. Chem. Eng. Proc., 33 (1994) 307-311 [44] S. Sugiura, M. Nakajima, T. Oda, M. Satake and M. Seki. Effect of interfacial tension on the dynamic behavior of droplet formation during microchannel emulsification. J. Colloid and Interf. Sci., 269 (2004) 178-185
Chapter 9. Supported liquid membranes 1. Introduction
In this Chapter the potentialities, as well as the drawbacks, of supported liquid membranes are discussed. The different types of facilitated transport that can be established within these systems are described. Concerning the analysis of the mass transport, the mass transfer resistances, the concentration profiles and the mass transport equations for both carrier-free and carried-charged membranes are reported. The above equations are derived for the supported liquid membrane most investigated in literature: an hydrophobic membrane with an organic phase immobilized into its micropores for the treatment of aqueous solutions. However, the analysis made is general and can be extended to the other supported liquid membrane configurations. A section devoted to the research efforts made worldwide for improving the supported liquid membrane stability is provided at the end of the Chapter.
2. Facilitated transport
The transport of a species in a supported liquid membrane can occur by simple permeation through the liquid immobilized into the micropores or by a facilitated transport.
Supported Liquid Membranes 309 In the former case the species is transferred by a solution-diffusion mechanism and the affinity between the liquid phase and the species determines the selectivity of the process (see Figure 1).
Figure 1. Transport of i through the immobilized liquid by a solution-diffusion mechanism.
The facilitated transport occurs when the species that diffuses through the liquid is subjected to a reversible chemical reaction (the so-called complexation). Usually, this type of transport is obtained thanks to a carrier dissolved in the liquid that reversibly reacts with the solubilized species and that facilitate its transfers through the membrane as a complex. The permeation as a complex takes place in parallel with the solution-diffusion of the species into the liquid phase and is usually higher than the simple permeation (see Figure 2) [ 1,2]. In order to compare the transport of the flee species with that of the complex, a facilitation factor (F) is
310 Chapter 9 often used which is defined as the ratio of the flux achievable in presence of the carrier to the flux obtained in a carrier-free membrane [3].
F = Flux in presence o f carrier~ Flux without carrier
(1)
Figure 2. Transport of i through the immobilized liquid as a complex and as a free species.
A generic form of the reversible chemical reaction between a species i and a carrier C is:
kc i+C ~
ka
C-i
(2)
Supported Liquid Membranes 311 where: kc, complexation rate coefficient; kd, decomplexation rate coefficient.
The species i complexes with the carrier C at the feed side: kc i+C ~ C-i
(3)
and decomplexes at the strip side: ka C-i ~ i + C
(4)
Depending on the specific application, the transport of the complex can be controlled by acting on different parameters, such as temperature and pH of the feed and strip streams [4,5]. The facilitated transport can be simple or coupled [1,2]. In the former case, the carrier reacts with the species at the feed-membrane interface to form the complex, the complex diffuses through the membrane, the species is released (decomplexation) at the membranestrip interface and the carrier re-diffuses to the feed-membrane interface (Figure 3).
Figure 3. Simple facilitated transport.
312 Chapter 9 When the feed stream contains charged species, a coupled facilitated transport can take place. In particular, if the species have an opposite charge we can have a co-trasport: the species form a complex with the carrier at the feed-membrane interface, the complex diffuses through the membrane, the species are released at the membrane-strip interface and the carrier re-diffuse towards the feed-membrane interface (see Figure 4).
Figure 4. Coupled facilitated transport: co-transport of i and j.
If the species have the same charge, we can have a counter-transport: the carrier complexes with one of the species at the feed-membrane interface, the complex moves towards the membrane-strip interface where the carrier releases the species. Once the species is released, the carrier complexes at the membrane-strip interface with the other species and the complex re-diffuses towards the membrane-feed interface, where the other species is released. In the
Supported Liquid Membranes 313 counter-transport it is possible to transfer a species against its concentration gradient (from a low to a high concentration side), driving the process the concentration gradient of the other species. Figure 5 shows this type of transport.
Figure 5. Coupled facilitated transport: counter-transport of i andj.
Among several carriers available, tertiary amines and quatemary ammonium salts can be used to form complexes with anions [1] and crown ethers and oximes (LIX-series) to complex cations [ 1, 6-10]. In Table 1 are reported some of the carriers used. Although the huge number of carriers already known, the synthesis of new carriers with better properties is in progress worldwide, as will be more discussed in Chapter 11.
314 Chapter 9
Table 1. Some of the carriers used Carrier
Species transferred
Reference
18-crown-6 ether
Potassium, copper(II), silver (I), gold (III), silver(I), zinc(II) ions
[1], [6], [11-13]
Tetraoctyl ammonium bromide,Trioctylmethyl ammonium chloride
Nitrate ions
[5]
LIX84, LIX864, LIX64N, LIX860, LIX984N
Copper(II), cadmiun (II), zinc(II) ions
[7-10], [14], [15]
Bis(2-ethylhexyl) hydrogen phosphate (D2EHPA)
Copper (II), zinc(II), cobalt(II), nickel(II)
[41, [161, [171
Silver(I) ions
Benzene, Ethylene
[18], [19]
Cobalt (II) salt, haemoglobin, porphyrins
Oxygen
[20-22]
3. Mass transfer equations
When a facilitated transport takes place, the total transport of a species through the membrane consists of two contributes: the facilitate and the simple transport. The former refers to the transport performed by the carrier, whereas the simple transport is related to the permeation of the species through the liquid and is regulated by a solution-diffusion mechanism. This dual mechanism leads to a total flux that is not necessarily proportional to the driving force. Therefore, appreciable fluxes can be obtained even at very low concentrations of the species to be transferred. The facilitated transport is characterized by five steps that occur in series (solubilization of the species, complexation, diffusion of the complex, decomplexation and desorption of the
Supported Liquid Membranes 315 species) and, depending on the specific case, each of them can represent the limiting step and control the rate of transfer. Generally, the solubilization and desorption steps can be supposed instantaneous and the controlling step can be established by comparing the relative rates of complexation, decomplexation and the diffusive transport of the complex. Figure 6 reports the mass transfer resistances involved in presence of a facilitated transport.
Figure 6. Resistances involved in facilitated transport.
The most widely studied type of supported liquid membrane consists of a hydrophobic support, with micropores filled by an organic phase, that is located between two aqueous phases (feed and strip) [11, 23]. The organic phase is immiscible with the aqueous streams and is retained in the pores thanks to the surfacial tensions and the capillarity.
316 Chapter 9 For the development of the mass transfer equations we will refer to this specific system, although the same iter can be followed for the case of hydrophilic membranes with w a t e r filled pores and organic feed and strip phases as well as for gaseous feed and strip streams.
A species i contained in the feed stream during its movement towards the strip phase finds three mass transfer resistances offered by the feed, the membrane and the strip (Figure 7) and its total flux can be expressed as:
J, - K ( c f
- c,9
(5)
Figure 7. Resistances encountered by the species i during its permeation from the feed to the strip phase.
The concentration profiles that are established because of the mass transfer resistances are depicted in Figure 8 for a flat membrane.
Supported Liquid Membranes 317
Figure 8. Concentration profiles of the species i during its permeation from the feed to the strip phase. Flat membrane.
The overall mass transfer coefficient can be related to the individual mass transfer coefficients by considering, as already made in Chapters 4 and 5, that at steady state the flux through the feed side equals the flux through the membrane as well as the flux through the strip side'
Ji -- k i J (Ci f- Cfm) -- kimslm ( C f m e - CSm) -- kiws (CSme- Cff)
where:
k f mass transfer coefficient in the aqueous feed for the species i; kin~tm, mass transfer coefficient in the supported liquid membrane for the species i; k J , mass transfer coefficient in the aqueous strip for the species i; Cif , concentration of the species i at the f e e d - membrane interface; Cfme, concentration of the species i at the feed-organic interface, organic side," CiSm, concentration of the species i at the membrane - strip interface," C~me, concentration of the species i at the organic - strip interface, strip side.
(6)
318 Chapter 9 The equilibrium conditions at the aqueous-organic interfaces are described by the distribution coefficients:
C f me= mi Ci f
(7)
CiSm: mi CiSme
(8)
and the overall mass transfer coefficient can be expressed in terms of single mass transfer coefficients by: 1/K = I/kwf + 1/(k,m stm mO + 1/k~w~
(9)
Concerning the distribution coefficients determination, Coelhoso et al. [24] pointed out that, when the extraction and stripping are carried out simultaneously, their values can be different with respect to those obtained for the single experiment of extraction or stripping. In particular, the osmotic pressure difference between the two aqueous streams that might occur during the process under certain conditions has to be taken into account for a correct evaluation.
In order to calculate the flux, the individual mass transfer coefficients have to be determined. Referring to the resistances offered by the phases, the same correlations reported in Chapter 4 can be used. For what concerns the membrane mass transfer resistance, we can distinguish two different situations: carrier-free membranes and carrier-charged membranes.
Supported Liquid Membranes 319 3.1.Carrier-free membranes In carrier-free membranes, the mass transfer through the micropores is function of the diffusion coefficient of the species in the organic phase and can be derived by [25]:
kimslm = Dio c/r6
(1 O)
where:
Dio, diffusion coefficient of the species i in the organic phase.
3.2.Carrier-charged membranes When the organic phase contains a carrier, the flux of the species i is given by the sum of two contributes:
J~ = D,o c/r6 (cUe - C[m) + D c-,o e/r6 (c~_fme - C~-7~
(ll)
where:
D c-io, diffusion coefficient of the complex in the organic phase; Cc-{me, concentration of the complex at the feed-organic interface," Cc_iSm, concentration of the complex at the membrane-strip interface.
The first term of the sum represents the diffusion of the free species through the organic phase, whereas the second refers to the diffusion of the species as a complex and depends on the diffusion coefficient of the complex into the organic liquid as well as the concentration difference of the complex across the immobilized liquid. By introducing the binding constant of the complexation reaction between the species and the carrier and assuming that the concentrations at the membrane-strip interface of both the
320 Chapter 9
free species and the complex are negligible, it is possible to re-write the second term as function of the concentration of the species i at the feed - organic interface, Cife [ 1]. The equation (1 l) becomes:
J~ = D,o c C L e / r 6 + kmr
C[me
(12)
with the membrane mass transfer coefficient for the complex (kmcomplex) function of different parameters such as its diffusion coefficient, the carrier concentration, the binding constant and the concentration of the species at the feed-organic interface [ 1].
The membrane mass transfer coefficient of the carrier-charged membrane can be, then, written as" kimstr" = D~o e/r6 + kmcomptex