**4. Membrane characteristics**

236 Distillation – Advances from Modeling to Applications

2. In air gap membrane distillation (AGMD), water vapour is condensed on a cold surface that has been separated from the membrane via an air gap. The heat losses are reduced in this configuration by addition of a stagnant air gap between membrane and

3. In sweeping gas membrane distillation (SGMD), a cold inert gas is used in permeate side for sweeping and carrying the vapour molecules to outside the membrane module where the condensation takes place. Despite the advantages of a relatively low conductive heat loss with a reduced mass transfer resistance, due to the operational

costs of the external condensation system, SGMD is the least used configuration. 4. In vacuum membrane distillation (VMD), the driving force is maintained by applying vacuum at the permeate side. The applied vacuum pressure is lower than the equilibrium vapour pressure. Therefore, condensation takes place outside of the

Each of the MD configurations has its own advantages and disadvantages for a given

Osmotic distillation (OD) is a non-thermal membrane distillation variant, in which a microporous hydrophobic membrane separates two aqueous solutions at different solute concentrations. The OD process can be operated at atmospheric pressure and ambient temperature. The driving force is the vapour pressure gradient across the membrane which is obtained by using a hypertonic salt solution on permeate side. The hydrophobic nature of the membrane prevents penetration of the pores by aqueous solutions, creating vapour/liquid interfaces at the entrance of the pores. Under these conditions, a net water flux from the high vapour pressure side to the low one occurs resulting in the concentration of feed and dilution of hypertonic salt solution. The water transport through the membrane can be summarized in three steps: (1) evaporation of water at the dilute vapour–liquid interface; (2) diffusional or convective vapour transport through the membrane pore; (3) condensation of water vapour at the membrane/brine interface (Jiao et al., 2004; Peinemann et al., 2010). In the literature the OD technique is also termed as isothermal membrane distillation, osmotic membrane distillation, osmotic evaporation and gas membrane

The basic requirements of osmotic agent are to be non-volatile, to have high osmotic activity in order to maintain a lower vapour pressure and to maximize the driving force and to be thermally stable to allow reconcentration of diluted stripping solution by evaporation. Other factors that should be taken into consideration are solubility, toxicity, corrosivity and cost. Although NaCl or CaCl2 have chosen as osmotic agent in most of the reported studies, both of these salts have the disadvantage of being corrosive to ferrous alloys (Celere & Gostoli, 2004; Shin & Johnson, 2007). MgCl2, MgSO4, K2HPO4, and KH2PO4 are some other commonly used osmotic agents in OD. Potassium salts of ortho- and pyrophosphoric acid

other three configurations.

condensation surface.

membrane module.

**3. Osmotic distillation** 

extraction (Gryta, 2005b).

application.

convenience to set up in laboratory. However, direct contact of the membrane with the cooling side and poor conductivity of the polymeric material results heat losses throughout the membrane. Therefore, in DCMD the thermal efficiency which is defined as the fraction of heat energy used only for evaporation, is relatively smaller than the

> The selection of the membrane is the most crucial factor in MD separation performance. As stated earlier, the membrane used for MD process must be hydrophobic and porous. There are various types of membranes meeting these expectations; however the efficiency of a given MD application depends largely on additional factors such as resistance to mass transfer, thermal stability, thermal conductivity, wetting phenomena and module characterization. Membrane and module related characteristics affecting selection of the appropriate membrane are summarized in this section.

#### **4.1 Membrane materials**

A large variety of membranes including both polymeric and inorganic membranes of hydrophobic nature can be used in MD process; however polymeric membranes have attracted much more attention due to their possibility to modulate the intrinsic properties. Polytetrafluoroethylene (PTFE), polypropylene (PP) and polyvinylidenefluoride (PVDF) are the most commonly used polymeric membranes due to their low surface tension values (Table 1). Hydrophobic porous membranes can be prepared by different techniques like sintering, stretching, phase inversion or thermally induced phase separation depending on the properties of the materials to be used. The useful materials should be selected according to criteria that include compatibility with the liquids involved, cost, ease of fabrication and assembly, useful operating temperatures, and thermal conductivity (Li et al., 2008; Liu et al., 2011). Among them, PTFE membranes are the most hydrophobic ones showing outstanding thermal stability and chemical resistance properties (they are low soluble in practically all common solvents). The main disadvantage of PTFE membranes is the difficulty of processing. PTFE membranes are generally prepared by sintering or stretching. PP exhibits


Table 1. Critical surface tension values of some polymers (Adapted from Oliver, 2004; Pabby et al., 2009)

Membrane Distillation: Principle, Advances,

degree of crystallinity (Alklaibi & Lior, 2005).

developed for MD applications.

**4.2 Membrane modules** 

used by MD researchers.

Phattaranawik et al., 2003a)

(Susanto, 2011).

Limitations and Future Prospects in Food Industry 239

the difficulties in measuring its real value for the membranes used in MD. In general a value of 2 is frequently assumed for tortuosity factor. (El-Bourawi et al., 2006; Khayet et al., 2004a;

*Membrane thickness:* Permeate flux is inversely proportional to the membrane thickness in MD. Therefore, membrane must be as thin as possible to achieve high permeate flux. Thickness also plays an important role in the amount of conductive heat loss though the membrane. In order to reduce heat resistances, it should be as thick as possible leading to a conflict with the requirement of higher permeate flux. Hence membrane thickness should be optimized in order to obtain optimum permeate flux and heat efficiency. The optimum thickness for MD has been estimated within the range of 30–60 µm (Lagana et al., 2000).

*Pore size distribution*: Pore size distribution affects uniformity of vapour permeation mechanism. In general, uniform pore size is preferable rather than distributed pore size

*Thermal conductivity:* Thermal conductivity of the membrane should be small in order to reduce the heat loss through the membrane from feed to the permeate side. Conductive heat loss is inversely proportional to the membrane thickness. However selection of a thicker membrane decreases both the flux and permeability. One promising approach may be selection of a membrane with higher porosity since thermal conductivity of polymer membrane is significantly higher than thermal conductivity of water vapour in the membrane pores (Khayet et al., 2006). The thermal conductivities of polymers used in MD generally varies in the range of 0.15–0.45 W m-1K-1 depending upon temperature and the

Table 2 summarizes the commercial membranes commonly used by various researchers up to date together with their principal characteristics. In fact, there is a lack of commercially available MD units and most of the MD researches use modules actually designed for other membrane operations (i.e. microfiltration) rather than MD. Design of novel membranes fabricated especially for MD purposes have been recommended by MD investigators since commercially available membranes does not meet all the requirements listed above. Novel hydrophobic membranes for MD applications can be manufactured either by hydrophobic polymers or by surface modification of hydrophilic membranes. Various surface modification applications including coating, grafting and plasma polymerization (Brodard et al., 2003; Bryjak et al., 2000; Chanachai et al., 2010; Huo et al., 2010; Kong et al., 1992; Krajewski et al., 2006; Lai et al., 2011; Li & Sirkar, 2004; Vargas-Garcia et al., 2011; Wu et al., 1992; Yang et al., 2011b) have been attempted until now. However, there is very limited number of studies on the design of MD membranes (Khayet, 2011; Khayet et al., 2010; Phattaranawik et al., 2009; Wang et al., 2009; Yang et al., 2011a). Therefore, new generation of membranes promising required features should be

Choice and arrangement of the membrane module in a MD application is based on economic considerations with the correct engineering parameters being employed. Plate and frame, spiral wound, tubular, capillary and hollow fiber membrane modules are commonly

excellent solvent resistant properties and high crystallinity. PP membranes are generally manufactured by stretching and thermal phase inversion. PVDF membranes exhibit good thermal and chemical resistance; however this polymer easily dissolves at room temperature in a variety of solvents including dimethylformamide (DMF) and triethylphosphate (TEP). PVDF membranes are generally prepared by phase inversion (Curcio and Drioli, 2005).

There are some additional criteria that should be taken into consideration for selection of the appropriate membrane for a given MD application such as pore size, tortuosity, porosity, membrane thickness and thermal conductivity. The relationship between the transmembrane flux and the different membrane characteristic related parameters is given by (Lawson & Lloyd, 1997)

$$N \; a \; \frac{ \varepsilon}{\pi \delta} \tag{1}$$

where *N* is the molar flux, <*rα*> is the mean pore size of the membrane pores where *α* equals 1 for Knudsen diffusion and equals 2 for viscous flux, *ε* is the membrane porosity, *τ* is the membrane tortousity and *δ* is the membrane thickness.

*Membrane pore size:* Membranes with pore sizes ranging from 10 nm to l µm can be used in MD (Pabby et al., 2009). The permeate flux increases with the increase in pore size as determined by Knudsen model. However, in order to avoid wettability, small pore size should be choosen (El-Bourawi et al., 2006; Khayet, 2011). Thus, an optimum value for pore size has to be determined for each MD application depending on the type of the feed solution.

*Membrane porosity:* Membrane porosity is determined as the ratio between the volume of the pores and the total volume of the membrane. Evaporation surface area increases with the increase in porosity level of the membrane, resulting in higher permeate fluxes (Huo et al., 2011; Susanto, 2011). Membrane porosity also affects the amount of heat loss by conduction (Lawson & Lloyd, 1996b):

$$Q\_m = \mathbf{h}\_m \Delta T\_m \tag{2}$$

$$
\hbar \mathbf{l}\_m = \varepsilon \hbar \mathbf{l}\_{mg} + (\mathbf{1} - \varepsilon) \hbar \mathbf{l}\_{ms} \tag{3}
$$

where *ε* is the membrane porosity, *hmg* is the conductive heat transfer coefficient of the gases entrapped in the membrane pores; *hms* is the conductive heat transfer coefficient of the hydrophobic membrane material.

Conductive heat loss can be reduced by increasing porosity of the membrane, since *hmg* is generally an order of magnitude smaller than *hms*. In general, the porosity of the membranes used in MD operations lines in the range of 65%-85%.

*Pore tortuosity:* Tortuosity is the average length of the pores compared to membrane thickness. The membrane pores do not go straight across the membrane and the diffusing molecules must move along tortuous paths, leading a decrease in MD flux. Therefore, permeate flux increases with the decrease in tortuosity. It must be pointed out that this value is frequently used as a correction factor for prediction of transmembrane flux due to

excellent solvent resistant properties and high crystallinity. PP membranes are generally manufactured by stretching and thermal phase inversion. PVDF membranes exhibit good thermal and chemical resistance; however this polymer easily dissolves at room temperature in a variety of solvents including dimethylformamide (DMF) and triethylphosphate (TEP). PVDF membranes are generally prepared by phase inversion (Curcio and Drioli, 2005).

There are some additional criteria that should be taken into consideration for selection of the appropriate membrane for a given MD application such as pore size, tortuosity, porosity, membrane thickness and thermal conductivity. The relationship between the transmembrane flux and the different membrane characteristic related parameters is given

*<sup>r</sup> <sup>N</sup>*

 

where *N* is the molar flux, <*rα*> is the mean pore size of the membrane pores where *α* equals 1 for Knudsen diffusion and equals 2 for viscous flux, *ε* is the membrane porosity, *τ* is the

*Membrane pore size:* Membranes with pore sizes ranging from 10 nm to l µm can be used in MD (Pabby et al., 2009). The permeate flux increases with the increase in pore size as determined by Knudsen model. However, in order to avoid wettability, small pore size should be choosen (El-Bourawi et al., 2006; Khayet, 2011). Thus, an optimum value for pore size has to be determined for each MD application depending on the type of the feed

*Membrane porosity:* Membrane porosity is determined as the ratio between the volume of the pores and the total volume of the membrane. Evaporation surface area increases with the increase in porosity level of the membrane, resulting in higher permeate fluxes (Huo et al., 2011; Susanto, 2011). Membrane porosity also affects the amount of heat loss by conduction

where *ε* is the membrane porosity, *hmg* is the conductive heat transfer coefficient of the gases entrapped in the membrane pores; *hms* is the conductive heat transfer coefficient of the

Conductive heat loss can be reduced by increasing porosity of the membrane, since *hmg* is generally an order of magnitude smaller than *hms*. In general, the porosity of the membranes

*Pore tortuosity:* Tortuosity is the average length of the pores compared to membrane thickness. The membrane pores do not go straight across the membrane and the diffusing molecules must move along tortuous paths, leading a decrease in MD flux. Therefore, permeate flux increases with the decrease in tortuosity. It must be pointed out that this value is frequently used as a correction factor for prediction of transmembrane flux due to

 

(1)

*Q T m mm* h (2)

(3)

by (Lawson & Lloyd, 1997)

(Lawson & Lloyd, 1996b):

hydrophobic membrane material.

solution.

membrane tortousity and *δ* is the membrane thickness.

(1 ) *m mg ms hh h*

used in MD operations lines in the range of 65%-85%.

the difficulties in measuring its real value for the membranes used in MD. In general a value of 2 is frequently assumed for tortuosity factor. (El-Bourawi et al., 2006; Khayet et al., 2004a; Phattaranawik et al., 2003a)

*Membrane thickness:* Permeate flux is inversely proportional to the membrane thickness in MD. Therefore, membrane must be as thin as possible to achieve high permeate flux. Thickness also plays an important role in the amount of conductive heat loss though the membrane. In order to reduce heat resistances, it should be as thick as possible leading to a conflict with the requirement of higher permeate flux. Hence membrane thickness should be optimized in order to obtain optimum permeate flux and heat efficiency. The optimum thickness for MD has been estimated within the range of 30–60 µm (Lagana et al., 2000).

*Pore size distribution*: Pore size distribution affects uniformity of vapour permeation mechanism. In general, uniform pore size is preferable rather than distributed pore size (Susanto, 2011).

*Thermal conductivity:* Thermal conductivity of the membrane should be small in order to reduce the heat loss through the membrane from feed to the permeate side. Conductive heat loss is inversely proportional to the membrane thickness. However selection of a thicker membrane decreases both the flux and permeability. One promising approach may be selection of a membrane with higher porosity since thermal conductivity of polymer membrane is significantly higher than thermal conductivity of water vapour in the membrane pores (Khayet et al., 2006). The thermal conductivities of polymers used in MD generally varies in the range of 0.15–0.45 W m-1K-1 depending upon temperature and the degree of crystallinity (Alklaibi & Lior, 2005).

Table 2 summarizes the commercial membranes commonly used by various researchers up to date together with their principal characteristics. In fact, there is a lack of commercially available MD units and most of the MD researches use modules actually designed for other membrane operations (i.e. microfiltration) rather than MD. Design of novel membranes fabricated especially for MD purposes have been recommended by MD investigators since commercially available membranes does not meet all the requirements listed above. Novel hydrophobic membranes for MD applications can be manufactured either by hydrophobic polymers or by surface modification of hydrophilic membranes. Various surface modification applications including coating, grafting and plasma polymerization (Brodard et al., 2003; Bryjak et al., 2000; Chanachai et al., 2010; Huo et al., 2010; Kong et al., 1992; Krajewski et al., 2006; Lai et al., 2011; Li & Sirkar, 2004; Vargas-Garcia et al., 2011; Wu et al., 1992; Yang et al., 2011b) have been attempted until now. However, there is very limited number of studies on the design of MD membranes (Khayet, 2011; Khayet et al., 2010; Phattaranawik et al., 2009; Wang et al., 2009; Yang et al., 2011a). Therefore, new generation of membranes promising required features should be developed for MD applications.

### **4.2 Membrane modules**

Choice and arrangement of the membrane module in a MD application is based on economic considerations with the correct engineering parameters being employed. Plate and frame, spiral wound, tubular, capillary and hollow fiber membrane modules are commonly used by MD researchers.

Membrane Distillation: Principle, Advances,

**Manufacturer Trade** 

Enka Accurel

Membrana Accurel

Hoechst-Celanese **name** 

1E-PP

Accurel 2E-PP

S6/2

Q3/2

Liqui-Cel® Extra-Flow 2.5×8 in

Table 2. List of commercial membranes commonly used by various MD researchers

In plate and frame modules, the membranes which are usually prepared as discs or flat sheets are placed between two plates. The feed solution flows through flat, rectangular channels. Packing densities for flat sheet membranes may be in the range of 100–400 m3/m2 (Pabby et al., 2009). Polymeric flat sheet membranes are easy to prepare, handle, and mount. The same module can be used to test many different types of MD membranes. The membrane can be supported to enhance mechanical strength. Babu et al. (2008) used a plate and frame membrane module having a membrane area of 0.01 m2 for the concentration of pineapple and sweet lime juice. The module consists of a polyester mesh (0.25 mm) and a hydrophobic microporous polypropylene membrane (pore size 0.20 μm and thickness 175 μm) supported in between a viton gasket (3.0 mm) and two stainless steel frames. In spiral wound membranes, the membrane, feed and permeate channel spacers and the porous membrane support form an envelope which is rolled around a perforated central collection tube and inserted into an outer tubular pressure shell. The feed solution passes in axial direction through the feed channel across the membrane surface. The filtrate moves along

**Membrane module** 

**Flat sheet** 

**Capillary** 

**Hollow fiber** 

Limitations and Future Prospects in Food Industry 241

**Polymer Membrane**

**thickness (µm)** 

PP 0.48 90

Gryta, 2007) Accurel

PP 400 0.20 70

PP 50 0.044 65 PP 47 0.056 42

Self-designed PP 800 0.40 73 (Gryta et al.,

Memcor PV 375 PVDF 125 0.20 75 (Bui et al., 2004) PV 660 PVDF 170 0.20 64

**Nominal pore size (µm)** 

PP 0.25 25 (Mengual et al.,

PP 450 0.20 73 (Celere &

PP 180 40 (Bailey et al., 2000) PP 53 0.074 50

**Porosity (%)** 

**References** 

1993; Narayan et al., 2002)

Gostoli, 2004;

2000b)


**Polymer Membrane**

**thickness (µm)** 

3MB PP 81 0.40 76 3MC PP 76 0.51 79 3MD PP 86 0.58 80 3ME PP 79 0.73 85 Gelman TF1000 PTFE/PP 60 0.1 80 (Khayet et al.,

> TF450 PTFE/PP 60 0.45 80 TF200 PTFE/PP 60 0.20 80 TF 200 PTFE/PP 178 0.20 80 TF 200 PTFE/PP 165 0.20 60

Milipore Durapore PVDF 110 0.45 75 (Banat &

Sartorious PTFE 70 0.20 70 (Phattaranawik

Gore PTFE 64 0.20 90 (Garcia-Payo et

Osmonics PP 150 0.22 70 (Cath et al., 2004) PTFE 175 0.22 70 PTFE 175 0.45 70 PTFE 175 1.0 70

PTFE 77 0.45 89

al., 1993) Celgard

PP 25 0.02 38

PP 28 0.05 45 (Barbe et al.,

Durapore PVDF 100 0.20 70 GVHP PVDF 125 0.20 80 GVHP PVDF 125 0.22 75 HVHP PVDF 116 0.45 66

**Nominal pore size (µm)** 

3MA PP 91 0.29 66 (Kim & Lloyd,

**Porosity (%)** 

**References** 

1991; Lawson et al., 1995; Lawson & Lloyd, 1996a)

2004b; Martinez-Diez et al., 1998; Martinez et al., 2002; Rincon et al., 1999; Rodrigues et al., 2004)

Simandl, 1999; Ding et al., 2003; Khayet et al., 2004b; Phattaranawik et al., 2003b; Phattaranawik et al., 2001)

et al., 2003b; Warczok et al., 2007)

al., 2000; Izquierdo-Gil et al., 1999; Phattaranawik et al., 2003b)

2000; Mengual et

**Membrane module** 

**Flat sheet** 

**Manufacturer Trade** 

3M Corporation

> Hoechst Celanese

Celgard 2400

2500

**name** 


Table 2. List of commercial membranes commonly used by various MD researchers

In plate and frame modules, the membranes which are usually prepared as discs or flat sheets are placed between two plates. The feed solution flows through flat, rectangular channels. Packing densities for flat sheet membranes may be in the range of 100–400 m3/m2 (Pabby et al., 2009). Polymeric flat sheet membranes are easy to prepare, handle, and mount. The same module can be used to test many different types of MD membranes. The membrane can be supported to enhance mechanical strength. Babu et al. (2008) used a plate and frame membrane module having a membrane area of 0.01 m2 for the concentration of pineapple and sweet lime juice. The module consists of a polyester mesh (0.25 mm) and a hydrophobic microporous polypropylene membrane (pore size 0.20 μm and thickness 175 μm) supported in between a viton gasket (3.0 mm) and two stainless steel frames. In spiral wound membranes, the membrane, feed and permeate channel spacers and the porous membrane support form an envelope which is rolled around a perforated central collection tube and inserted into an outer tubular pressure shell. The feed solution passes in axial direction through the feed channel across the membrane surface. The filtrate moves along

Membrane Distillation: Principle, Advances,

Matsuura, 2011; Sigurdsson & Shishoo, 1997)

Hwang et al., 2011; Tomaszewska, 2000).

Lawson & Lloyd, 1997).

**5.1 Theory of heat transfer** 

**4.4 Liquid entry pressure and wetting phenomena** 

pore coefficient (equals 1 for cylindrical pores), *<sup>L</sup>*

Heat transfer in the MD includes three main steps:

i. Heat transfer through the feed side boundary layer

the contact angle and *rm* is the maximum pore size.

Limitations and Future Prospects in Food Industry 243

Table 3. Contact angle values of water on some materials at ambient temperature (Khayet &

different materials in water at ambient temperature. For example, the parameter measured on PTFE or PVDF membrane surface was 108° or 107°, respectively (Curcio et al., 2010;

The hydrophobic nature of membranes used in membrane distillation prevents penetration of the aqueous solutions into the pores unless a critical penetration pressure is exceeded, as stated earlier. Liquid entry pressure (LEP) is the minimum transmembrane hydrostatic pressure that must be applied before liquid solutions penetrate into the membrane pores. LEP can be calculated using the Laplace-Young equation (Burgoyne & Vahdati, 2000;

*F D*

where *PF* and *PD* are the hydraulic pressure of the feed and distillate side,

**5. Transport mechanisms and polarization phenomena** 

*Cos PP P*

LEP depends on membrane characteristics and prevents wetting of the membrane pores during MD experiments. LEP increases with a decrease in maximum pore size at the surface and an increase at the hydrophobicity (i.e., large water contact angle) of the membrane material. The presence of strong surfactants or organic solvents can greatly reduce the liquid surface tension therefore causing membrane wetting. Therefore, care must be taken to prevent contamination of process solutions with detergents or other surfacting agents.

2 *<sup>L</sup>*

*m*

 

(4)

is the surface tension of the liquid,

is the geometric

is

*r* 

**Ordinary glass** 20 **Polycarbonate** 70 **Polyamide** 69 **Polyethersulphone** 54 **Polyethylene** 96 **Polypropylene** 100 **PTFE** 123 **PVDF** 111 **Teflon** 112

**Material Contact Angle, °** 

the permeate channel and is collected in a perforated central collection tube. Spiral-wound modules have a packing density of 300–1000 m2/m3 depending on the channel height, which is greater than that of the plate and frame module (Pabby et al., 2009). However, the spiral-wound module is quite sensitive to fouling. Tubular, capillary or hollow fiber membrane modules are shell and tube type modules housing pressure-tight tubes. The support is not needed in this type of modules. The membranes are usually a permanent integral part of the module and are not easily replaced. Tubular membrane modules provide much higher membrane surface area to module volume ratio than plate and frame modules (Khayet, 2011). The diameter of membranes in tubular module varies within the range of 10- 25 mm. The packing density is around 300 m2/m3 (Pabby et al., 2009). These modules offer higher cross-flow velocities and large pressure drop and generally used for MD of high viscous liquids. The diameters of membranes in capillary modules typically vary between 0.2-3 mm with packing densities of about 600-1200 m2/m3 (Li et al., 2008) . The production costs are very low and membrane fouling can effectively be controlled by the proper feed flow and back-flushing of permeate in certain time intervals. The main disadvantage of the capillary membrane module is the requirement of low operating pressure (up to 4 bars). The inner diameters of hollow fiber membranes is around 50-500 µm with very high packing densities of about 3000 m2/m3. Hollow fiber module has the highest packing density of all module types. Its production is very cost effective and hollow fiber membrane modules can be operated at pressures in excess of 100 bars (El-Bourawi et al., 2006). The main disadvantage of the hollow fiber membrane module is the difficult control of membrane fouling. Therefore, a proper pretreatment should be applied for separation of macromolecules. For example, in the case of fruit juice concentration by MD using a hollow fiber module, clarification is a crucial pretreatment step to enhance MD flux (Cassano & Drioli, 2007; Onsekizoglu et al., 2010b).

#### **4.3 Contact angle**

The contact angle is a common measurement of the hydrophobic or hydrophilic behaviour of a material. It provides information about relative wettability of membranes. The contact angle is determined as the angle between the surface of the wetted solid and a line tangent to the curved surface of the drop at the point of three-phase contact (Figure 2). The value of contact angle is greater than 90° when there is low affinity between liquid and solid; in case of water, the material is considered hydrophobic and is less than 90*°* in the case of high affinity. Wetting occurs at 0*°*, when the liquid spreads onto the surface (Curcio et al., 2010; Curcio & Drioli, 2005; Pabby et al., 2009). The wettability of a solid surface by a liquid decreases as the contact angle increases. Table 3 lists the contact angle values for few

Fig. 2. Schematic representation of contact angle

the permeate channel and is collected in a perforated central collection tube. Spiral-wound modules have a packing density of 300–1000 m2/m3 depending on the channel height, which is greater than that of the plate and frame module (Pabby et al., 2009). However, the spiral-wound module is quite sensitive to fouling. Tubular, capillary or hollow fiber membrane modules are shell and tube type modules housing pressure-tight tubes. The support is not needed in this type of modules. The membranes are usually a permanent integral part of the module and are not easily replaced. Tubular membrane modules provide much higher membrane surface area to module volume ratio than plate and frame modules (Khayet, 2011). The diameter of membranes in tubular module varies within the range of 10- 25 mm. The packing density is around 300 m2/m3 (Pabby et al., 2009). These modules offer higher cross-flow velocities and large pressure drop and generally used for MD of high viscous liquids. The diameters of membranes in capillary modules typically vary between 0.2-3 mm with packing densities of about 600-1200 m2/m3 (Li et al., 2008) . The production costs are very low and membrane fouling can effectively be controlled by the proper feed flow and back-flushing of permeate in certain time intervals. The main disadvantage of the capillary membrane module is the requirement of low operating pressure (up to 4 bars). The inner diameters of hollow fiber membranes is around 50-500 µm with very high packing densities of about 3000 m2/m3. Hollow fiber module has the highest packing density of all module types. Its production is very cost effective and hollow fiber membrane modules can be operated at pressures in excess of 100 bars (El-Bourawi et al., 2006). The main disadvantage of the hollow fiber membrane module is the difficult control of membrane fouling. Therefore, a proper pretreatment should be applied for separation of macromolecules. For example, in the case of fruit juice concentration by MD using a hollow fiber module, clarification is a crucial pretreatment step to enhance MD flux (Cassano &

The contact angle is a common measurement of the hydrophobic or hydrophilic behaviour of a material. It provides information about relative wettability of membranes. The contact angle is determined as the angle between the surface of the wetted solid and a line tangent to the curved surface of the drop at the point of three-phase contact (Figure 2). The value of contact angle is greater than 90° when there is low affinity between liquid and solid; in case of water, the material is considered hydrophobic and is less than 90*°* in the case of high affinity. Wetting occurs at 0*°*, when the liquid spreads onto the surface (Curcio et al., 2010; Curcio & Drioli, 2005; Pabby et al., 2009). The wettability of a solid surface by a liquid decreases as the contact angle increases. Table 3 lists the contact angle values for few

Drioli, 2007; Onsekizoglu et al., 2010b).

Fig. 2. Schematic representation of contact angle

**4.3 Contact angle** 


Table 3. Contact angle values of water on some materials at ambient temperature (Khayet & Matsuura, 2011; Sigurdsson & Shishoo, 1997)

different materials in water at ambient temperature. For example, the parameter measured on PTFE or PVDF membrane surface was 108° or 107°, respectively (Curcio et al., 2010; Hwang et al., 2011; Tomaszewska, 2000).

#### **4.4 Liquid entry pressure and wetting phenomena**

The hydrophobic nature of membranes used in membrane distillation prevents penetration of the aqueous solutions into the pores unless a critical penetration pressure is exceeded, as stated earlier. Liquid entry pressure (LEP) is the minimum transmembrane hydrostatic pressure that must be applied before liquid solutions penetrate into the membrane pores. LEP can be calculated using the Laplace-Young equation (Burgoyne & Vahdati, 2000; Lawson & Lloyd, 1997).

$$
\Delta P = P\_F - P\_D - \frac{2\beta \gamma\_L \text{Cost} \theta}{r\_m} \tag{4}
$$

where *PF* and *PD* are the hydraulic pressure of the feed and distillate side, is the geometric pore coefficient (equals 1 for cylindrical pores), *<sup>L</sup>* is the surface tension of the liquid, is the contact angle and *rm* is the maximum pore size.

LEP depends on membrane characteristics and prevents wetting of the membrane pores during MD experiments. LEP increases with a decrease in maximum pore size at the surface and an increase at the hydrophobicity (i.e., large water contact angle) of the membrane material. The presence of strong surfactants or organic solvents can greatly reduce the liquid surface tension therefore causing membrane wetting. Therefore, care must be taken to prevent contamination of process solutions with detergents or other surfacting agents.

#### **5. Transport mechanisms and polarization phenomena**

#### **5.1 Theory of heat transfer**

Heat transfer in the MD includes three main steps:

i. Heat transfer through the feed side boundary layer

Membrane Distillation: Principle, Advances,

can be estimated by following equations:

2001; Schofield et al., 1990a)

preferred ones are given below;

Limitations and Future Prospects in Food Industry 245

*Heat transfer through the membrane:* Heat transfer through the membrane appears as a combination of latent heat of vaporization ( *QV* ) and conductive heat transfer across both the membrane matrix and the gas filled membrane pores ( *QC* ). The corresponding values

> *<sup>m</sup> <sup>C</sup> fm pm <sup>k</sup> Q TT*

Therefore, the heat flux can be estimated by the following expression (El-Bourawi et al., 2006; Khayet & Matsuura, 2011; Lawson & Lloyd, 1997; Phattaranawik & Jiraratananon,

> *<sup>m</sup> <sup>m</sup> fm pm V <sup>k</sup> Q T T JH*

where *km* is the thermal conductivity of the membrane, *δ* is the membrane thickness, *J* is the

Various models have been proposed for estimation of *km* in Equation [10]. Two of the most

*<sup>m</sup>* 1 *<sup>g</sup> <sup>s</sup> kk k* 

*g s*

*k k* 

*Heat transfer through the permeate side boundary layer:* Heat transfer from the membrane surface to the bulk permeate side across the boundary layer is also related with the temperature polarization phenomenon. The temperature of membrane surface at the permeate side is

Both feed and permeate side boundary layers are function of fluid properties and operating conditions, as well as the hydrodynamic conditions. There are some convenient approaches in the literature to reduce the temperature polarization effects like mixing thoroughly, working at

 

 <sup>1</sup> 1

 

permeate water vapour flux and *HV* is the latent heat of vaporization.

*m*

higher than that of bulk permeate due to the temperature polarization effect.

where *hp* is the heat transfer coefficient of the permeate side boundary layer.

Heat transfer through the permeate side boundary layer is given as:

*k*

where *hf* is the heat transfer coefficient of the feed side boundary layer.

*Q hT T f ff <sup>b</sup> fm* (6)

*Q JH V V* (7)

(8)

*QQQ mVC* (9)

(11)

(12)

(10)

*Q hT T p pp <sup>m</sup> pb* (13)


*Heat transfer through the feed side boundary layer* Heat transfer from the feed solution to the membrane surface across the boundary layer in the feed side of the membrane module imposes a resistance to mass transfer since a large quantity of heat must be supplied to the surface of the membrane to vaporize the liquid. The temperature at the membrane surface is lower than the corresponding value at the bulk phase. This affects negatively the driving force for mass transfer. This phenomenon is called temperature polarization (El-Bourawi et al., 2006; Pabby et al., 2009; Qtaishat et al., 2008). Temperature polarization becomes more significant at higher feed temperatures (Burgoyne & Vahdati, 2000; Lagana et al., 2000; Phattaranawik et al., 2003b).

The temperature polarization coefficient (TPC) is determined as the ratio of the transmembrane temperature to the bulk temperature difference:

$$\text{TPPC} = \frac{T\_{fm} - T\_{pm}}{T\_{fb} - T\_{pb}} \tag{5}$$

where *T*fm, *T*pm, *T*fb and *T*pb are membrane surface temperatures and fluid bulk temperatures at the feed and permeate sides, respectively. A schematic diagram of the temperature polarization in MD is shown in Figure 3.

Fig. 3. Schematic diagram of temperature polarization in MD. *T*fm, *T*pm, *T*fb and *T*pb are membrane surface temperatures and fluid bulk temperatures at the feed and permeate sides, respectively.

Heat transfer through the feed side boundary layer can be calculated using:

*Heat transfer through the feed side boundary layer* Heat transfer from the feed solution to the membrane surface across the boundary layer in the feed side of the membrane module imposes a resistance to mass transfer since a large quantity of heat must be supplied to the surface of the membrane to vaporize the liquid. The temperature at the membrane surface is lower than the corresponding value at the bulk phase. This affects negatively the driving force for mass transfer. This phenomenon is called temperature polarization (El-Bourawi et al., 2006; Pabby et al., 2009; Qtaishat et al., 2008). Temperature polarization becomes more significant at higher feed temperatures (Burgoyne & Vahdati, 2000; Lagana et al., 2000;

The temperature polarization coefficient (TPC) is determined as the ratio of the

*TPC*

*fm pm fb pb*

(5)

*T T*

*T T*

where *T*fm, *T*pm, *T*fb and *T*pb are membrane surface temperatures and fluid bulk temperatures at the feed and permeate sides, respectively. A schematic diagram of the temperature

Fig. 3. Schematic diagram of temperature polarization in MD. *T*fm, *T*pm, *T*fb and *T*pb are membrane surface temperatures and fluid bulk temperatures at the feed and permeate

Heat transfer through the feed side boundary layer can be calculated using:

ii. Heat transfer through the membrane

Phattaranawik et al., 2003b).

sides, respectively.

polarization in MD is shown in Figure 3.

iii. Heat transfer through the permeate side boundary layer

transmembrane temperature to the bulk temperature difference:

$$Q\_f = h\_f \left( T\_{\text{fb}} - T\_{\text{fm}} \right) \tag{6}$$

where *hf* is the heat transfer coefficient of the feed side boundary layer.

*Heat transfer through the membrane:* Heat transfer through the membrane appears as a combination of latent heat of vaporization ( *QV* ) and conductive heat transfer across both the membrane matrix and the gas filled membrane pores ( *QC* ). The corresponding values can be estimated by following equations:

$$Q\_V = f \Delta H\_V \tag{7}$$

$$Q\_{\mathbb{C}} = \left(\frac{k\_m}{\mathcal{S}}\right) \left(T\_{fm} - T\_{pm}\right) \tag{8}$$

Therefore, the heat flux can be estimated by the following expression (El-Bourawi et al., 2006; Khayet & Matsuura, 2011; Lawson & Lloyd, 1997; Phattaranawik & Jiraratananon, 2001; Schofield et al., 1990a)

$$Q\_m = Q\_V + Q\_\mathbb{C} \tag{9}$$

$$Q\_m = \frac{k\_m}{\delta} \left( T\_{fin} - T\_{pm} \right) + f \Delta H\_V \tag{10}$$

where *km* is the thermal conductivity of the membrane, *δ* is the membrane thickness, *J* is the permeate water vapour flux and *HV* is the latent heat of vaporization.

Various models have been proposed for estimation of *km* in Equation [10]. Two of the most preferred ones are given below;

$$k\_m = \varepsilon k\_{\mathcal{K}} + (1 - \varepsilon)k\_s \tag{11}$$

$$k\_m = \left[\frac{\varepsilon}{k\_g} + \frac{\left(1 - \varepsilon\right)}{k\_s}\right]^{-1} \tag{12}$$

*Heat transfer through the permeate side boundary layer:* Heat transfer from the membrane surface to the bulk permeate side across the boundary layer is also related with the temperature polarization phenomenon. The temperature of membrane surface at the permeate side is higher than that of bulk permeate due to the temperature polarization effect.

Heat transfer through the permeate side boundary layer is given as:

$$Q\_p = h\_p \left( T\_{pm} - T\_{p\Phi} \right) \tag{13}$$

where *hp* is the heat transfer coefficient of the permeate side boundary layer.

Both feed and permeate side boundary layers are function of fluid properties and operating conditions, as well as the hydrodynamic conditions. There are some convenient approaches in the literature to reduce the temperature polarization effects like mixing thoroughly, working at

Membrane Distillation: Principle, Advances,

(Babu et al., 2008).

respectively.

depending on the hydrodynamics of the system:

2000; Martinez & Rodriguez-Maroto, 2007).

where *L*: characteristic length, *D*: diffusion coefficient,

Limitations and Future Prospects in Food Industry 247

where *ks* is the diffusive mass transfer coefficient through the boundary layer. Several empirical correlation of dimensionless numbers, namely, Sherwood (*Sh*), Reynolds (*Re*), Schmidt (*Sc*), Nusselt (*Nu*) and Prandtl (*Pr*) numbers can be used to estimate the value of *ks*

> Re= Sc= Nu= Pr= *<sup>P</sup> kL Lu hL C Sh D Dk k*

 

velocity, *k*: thermal conductivity, *CP*: specific heat, *h*: boundary layer heat transfer coefficient

In other membrane separation process such as microfiltration, ultrafiltration and reverse osmosis, concentration polarization is usually considered a major cause for flux decline (Agashichev, 2006; Morao et al., 2008; Song, 2010; Wang & Tarabara, 2007; Zaamouche et al., 2009). On the other hand, it is agreed upon that concentration polarization is insignificant compared to temperature polarization in DCMD (Khayet & Matsuura, 2011; Lagana et al.,

It is worth pointing out that in osmotic distillation process, concentration polarization exists at each side of the membrane. During osmotic distillation, as mass transfer proceeds, solute concentration increases at the membrane surface due to evaporation of water vapour at the feed side. On the other hand, the solute concentration decreases due to the condensation of water vapour on the permeate side, giving rise to the difference in brine concentrations (Figure 4). The existence of concentration polarization layers at each side of the membrane

Fig. 4. Schematic diagram of concentration polarization in MD. Cfm, Cpm, Cfb and Cpb are membrane surface and bulk solute concentrations at the feed and permeate sides,

 

(18)

: density, *µ*: viscosity, *u*: feed

high flow rates or using turbulence promoters (Cath et al., 2004; Chernyshov et al., 2005; El-Bourawi et al., 2006; Lawson & Lloyd, 1996a; Martinez & Rodriguez-Maroto, 2006).

#### **5.2 Theory of mass transfer**

As mentioned above, the mass transfer in MD is driven by the vapour pressure gradient imposed between two sides of the membrane. Mass transfer in membrane distillation consists of three consecutive steps:


The mass flux (*J*) can be expressed as (Close & Sorensen, 2010; Zhang et al., 2010):

$$J = K\Delta P\tag{14}$$

where *K* is the overall mass transfer coefficient which is the reciprocal of an overall mass transfer resistance. This overall resistance is the sum of three individual resistances:

$$K = \left[\frac{1}{K\_f} + \frac{1}{K\_m} + \frac{1}{K\_p}\right]^{-1} \tag{15}$$

where *Kf*, *Km* and *Kp* are the mass transfer coefficients of feed layer, membrane and permeate layer, respectively.

*Mass transfer trough feed side boundary layer:* In membrane distillation, only water vapour transport is allowed due to the hydrophobic character of the membrane. Therefore the concentration of solute(s) in feed solution becomes higher at the liquid/gas interface than that at the bulk feed as mass transfer proceeds. This phenomenon is called concentration polarization and results in reduction of the transmembrane flux by depressing the driving force for water transport. Concentration polarization coefficient (*CPC*) is determined as the ratio of the solute concentration at the membrane surface (*Cfm*) to that at the bulk feed solution (*Cfb*):

$$\text{CPC} = \frac{\text{C}\_{fm}}{\text{C}\_{\text{ft}}} \tag{16}$$

The concentration gradient between the liquid/gas interface and the bulk feed results a diffusive transfer of solutes from the surface of the membrane to the bulk solution. At steady state, the rate of convective solute transfer to the membrane surface is balanced by diffusion of solute to the bulk feed.

The molar flux is expressed as follows (El-Bourawi et al., 2006; Khayet & Matsuura, 2011):

$$J = k\_s \ln \left(\frac{\mathbf{C}\_{fm}}{\mathbf{C}\_{fb}}\right) \tag{17}$$

high flow rates or using turbulence promoters (Cath et al., 2004; Chernyshov et al., 2005; El-

As mentioned above, the mass transfer in MD is driven by the vapour pressure gradient imposed between two sides of the membrane. Mass transfer in membrane distillation

i. Evaporation of water at the liquid/gas interface on the membrane surface of the feed side

iii. Condensation of water vapour at the gas/liquid interface on the membrane surface of

where *K* is the overall mass transfer coefficient which is the reciprocal of an overall mass

111 *f mp*

*KKK* 

where *Kf*, *Km* and *Kp* are the mass transfer coefficients of feed layer, membrane and permeate

*Mass transfer trough feed side boundary layer:* In membrane distillation, only water vapour transport is allowed due to the hydrophobic character of the membrane. Therefore the concentration of solute(s) in feed solution becomes higher at the liquid/gas interface than that at the bulk feed as mass transfer proceeds. This phenomenon is called concentration polarization and results in reduction of the transmembrane flux by depressing the driving force for water transport. Concentration polarization coefficient (*CPC*) is determined as the ratio of the solute concentration at the membrane surface (*Cfm*) to that at the bulk feed

> *fm fb*

*C*

The concentration gradient between the liquid/gas interface and the bulk feed results a diffusive transfer of solutes from the surface of the membrane to the bulk solution. At steady state, the rate of convective solute transfer to the membrane surface is balanced by diffusion

The molar flux is expressed as follows (El-Bourawi et al., 2006; Khayet & Matsuura, 2011):

*J k <sup>C</sup>* 

ln *fm s*

*C*

*fb*

*CPC*

1

*J KP* (14)

*<sup>C</sup>* (16)

(17)

(15)

Bourawi et al., 2006; Lawson & Lloyd, 1996a; Martinez & Rodriguez-Maroto, 2006).

The mass flux (*J*) can be expressed as (Close & Sorensen, 2010; Zhang et al., 2010):

transfer resistance. This overall resistance is the sum of three individual resistances:

*K*

**5.2 Theory of mass transfer** 

the permeate side

layer, respectively.

solution (*Cfb*):

of solute to the bulk feed.

consists of three consecutive steps:

ii. Water vapour transfer through the membrane pores

where *ks* is the diffusive mass transfer coefficient through the boundary layer. Several empirical correlation of dimensionless numbers, namely, Sherwood (*Sh*), Reynolds (*Re*), Schmidt (*Sc*), Nusselt (*Nu*) and Prandtl (*Pr*) numbers can be used to estimate the value of *ks* depending on the hydrodynamics of the system:

$$\text{Shl} = \frac{\text{kL}}{D} \qquad \text{Re} = \frac{\text{L}\mu\rho}{\mu} \qquad \text{Sc} = \frac{\mu}{\rho D} \qquad \text{Nu} = \frac{\text{hL}}{k} \qquad \text{Pr} = \frac{\mu C\_P}{k} \tag{18}$$

where *L*: characteristic length, *D*: diffusion coefficient, : density, *µ*: viscosity, *u*: feed velocity, *k*: thermal conductivity, *CP*: specific heat, *h*: boundary layer heat transfer coefficient (Babu et al., 2008).

In other membrane separation process such as microfiltration, ultrafiltration and reverse osmosis, concentration polarization is usually considered a major cause for flux decline (Agashichev, 2006; Morao et al., 2008; Song, 2010; Wang & Tarabara, 2007; Zaamouche et al., 2009). On the other hand, it is agreed upon that concentration polarization is insignificant compared to temperature polarization in DCMD (Khayet & Matsuura, 2011; Lagana et al., 2000; Martinez & Rodriguez-Maroto, 2007).

It is worth pointing out that in osmotic distillation process, concentration polarization exists at each side of the membrane. During osmotic distillation, as mass transfer proceeds, solute concentration increases at the membrane surface due to evaporation of water vapour at the feed side. On the other hand, the solute concentration decreases due to the condensation of water vapour on the permeate side, giving rise to the difference in brine concentrations (Figure 4). The existence of concentration polarization layers at each side of the membrane

Fig. 4. Schematic diagram of concentration polarization in MD. Cfm, Cpm, Cfb and Cpb are membrane surface and bulk solute concentrations at the feed and permeate sides, respectively.

Membrane Distillation: Principle, Advances,

Drioli, 2005; Li & Sirkar, 2005; Schofield et al., 1990b).

the transmembrane vapour pressure difference.

in feed temperature (Moon et al., 2011).

(Hwang et al., 2011; Khayet et al., 2006).

*Feed flow rate & stirring* 

**6. Process parameters** 

*Feed concentration* 

*Feed temperature* 

Limitations and Future Prospects in Food Industry 249

Permeate flux decreases with an increase in feed concentration. This phenomenon can be attributed to the reduction of the driving force due to decrease of the vapour pressure of the feed solution and exponential increase of viscosity of the feed with increasing concentration. The contribution of concentration polarization effects is also known, nevertheless, this is very small in comparison with temperature polarization effects (Lagana et al., 2000; Pabby et al., 2009). As it is well known, MD can handle feed solutions at high concentrations without suffering the large drop in permeability observed in other pressure-driven membrane processes and can be preferentially employed whenever elevated permeate recovery factors or high retentate concentrations are requested (i.e. concentration of fruit juices) (Curcio &

Various investigations have been carried out on the effect of the feed temperature on permeate flux in MD. In general, it is agreed upon that there is an exponential increase of the MD flux with the increase of the feed temperature. As the driving force for membrane distillation is the difference in vapour pressure across the membrane, the increase in temperature increases the vapour pressure of the feed solution, thus results an increase in

It is worth quoting that working under high feed temperatures was offered by various MD researches, since the internal evaporation efficiency (the ratio of the heat that contributes to evaporation) and the total heat exchanged from the feed to the permeate side is high. Nevertheless, the increase in quality losses and formation of unfavorable compounds (i.e. hydroxymethyl furfural and furan) in fruit juices due to high operation temperatures restricts the temperature levels (Ciesarova & Vranova, 2009; Crews & Castle, 2007; Onsekizoglu et al., 2010b). Temperature polarization effect also increases with the increase

In MD, the increase in flow and/or stirring rate of feed increases the permeate flux. The shearing forces generated at high flow rate and/or stirring reduces the hydrodynamic boundary layer thickness and thus reduce polarization effects. Therefore, the temperature and concentration at the liquid-vapour interface becomes closer to the corresponding values at the bulk feed solution (Winter et al., 2011). Onsekizoglu et al. (2010a) studied the effects of various operating parameters on permeate flux and soluble solid content of apple juice during concentration through osmotic distillation (OD) and membrane distillation (MD) processes. They observed that the effect of feed flow rate on transmembrane flux was less

The effect of flow rate on MD flux becomes more noticeable at higher temperatures especially associated with higher temperature drop across the membrane (Walton et al., 2004). Consequently, higher productivity can be achieved by operating under a turbulent flow regime. On the other hand, the liquid entry pressure of feed solution (LEP) must be taken into account in order to avoid membrane pore wetting when optimizing feed flow rate

than half of the influence of temperature difference across the membrane.

results in the reduction of driving force for water vapour transport leading a decrease in transmembrane flux (Babu et al., 2006; Babu et al., 2008; Nagaraj et al., 2006b).

*Mass transfer through the membrane pores:* The main mass transfer mechanisms through the membrane in MD are Knudsen diffusion and molecular diffusion (Figure 5). Knudsen diffusion model is responsible for mass transfer through the membrane pore if the mean free path of the water molecules is much greater than the pore size of the membrane and hence, the molecules tend to collide more frequently with the pore wall (Li et al., 2008; Nagaraj et al., 2006b; Pabby et al., 2009; Srisurichan et al., 2006).

Fig. 5. Mass transfer mechanism involved in water vapour transport through membrane pores of MD module.

In this case, the membrane diffusion coefficient is calculated using equation:

$$K\_m = 1.064 \frac{r\varepsilon}{\pi \delta} \left(\frac{M}{RT}\right)^{0.5} \tag{19}$$

where *ε* is the fractional void volume, *δ* is the membrane thickness, *τ* is the tortuosity, *M* is the molecular weight of water, *R* is the gas constant and *T* is the absolute temperature.

On the other hand, when the pore size is relatively large, the molecule–molecule collisions are more frequent and molecular diffusion is responsible for mass transfer through the membrane pores (Khayet & Matsuura, 2011).

$$K\_m = \frac{1}{Y\_{\ln}} \frac{D\varepsilon}{\pi \mathcal{S}} \frac{M}{RT} \tag{20}$$

where Yln is the log mean of mole fraction of air and D is the diffusion coefficient.

Both models were successfully applied for predicting the mass transfer through the membrane in DCMD systems (Babu et al., 2006; Bandini & Sarti, 1999; Chen et al., 2009; Lawson & Lloyd, 1996b; Nagaraj et al., 2006b; Srisurichan et al., 2006).
