1

An Introduction to Membrane

Distillation

Christos Charisiadis 2014

Membrane distillation: A comprehensive review

Desalination 287 (2012) 2–18

Advances in Membrane Distillation for Water Desalination and Purification Applications

Water 2013, 5, 94-196; doi:10.3390/w5010094

2

Contents

1. Introduction 4

1.1 Using MD has many attractive features 4

1.2 MD is also attended by some drawbacks 5

1.3 Modelling Aspects of Membrane Distillation 5

2. Membrane configuration 6

2.1. Direct Contact Membrane Distillation (DCMD) 6

2.2. Air Gap Membrane Distillation (AGMD) 7

2.3. Sweeping Gas Membrane Distillation (SGMD) 7

2.4. Vacuum Membrane Distillation (VMD) 8

3. Configurations of MD Modules 9

4. Membranes for MD Applications 10

4.1 Hollow fiber 10

4.2 Plate and frame 11

4.3. Tubular membrane 11

4.4. Spiral wound membrane 11

5. Membrane Material 12

6. Membrane characteristics 12

6.1 Membrane thickness 13

6.2 Liquid entry pressure (wetting pressure) 13

6.3 Membrane porosity and tortuosity 15

6.4 Mean pore size and pore size distribution 15

6.5 Thermal conductivity 16

6.6 Surface Roughness 17

7. Advances on MD Processes and Modules for Water Purification 18

7.1. MD Stand-Alone Systems 18

3

7.2. State of the Art MD Research and Systems 19

8. Mechanisms 20

8.1 Mass transfer 21

8.1.1 Direct Contact Membrane Distillation (DCMD) 21

8.1.2. Air Gap Membrane Distillation (AGMD) 24

8.1.3. Vacuum Membrane Distillation (VMD) 26

8.1.4. Sweeping Gas Membrane Distillation (SGMD) 27

8.2. Heat transfer 28

9. Thermal efficiency and energy consumption 33

10. Temperature polarization and concentration polarization 34

11. Fouling 35

12. Operating parameters 37

12.1 Parameters to Reducing Temperature Polarization 37

12.2 Feed temperature 38

12.3 The concentration and solution feature 39

12.4 Recirculation rate 40

12.5 The air gap 40

12.6 Membrane type 41

13. Future Developments and Conclusions 41

4

1. Introduction

Membrane Distillation (MD) is a thermally-driven separation process, in which only

vapour molecules transfer through a microporous hydrophobic membrane. The

driving force in the MD process is the vapour pressure difference induced by the

temperature difference across the hydrophobic membrane. This process has various

applications, such as desalination, wastewater treatment and in the food industry.

1.1 Using MD has many attractive features

1. Low operating temperatures in comparison to those encountered in conventional

process; the solution (mainly water) is not necessarily heated up to the boiling point.

2. The hydrostatic pressure encountered in MD is lower than that used in pressure-

driven membrane processes like reverse osmosis (RO). Therefore, MD is expected to

be a cost effective process, which requires less demanding of membrane

characteristics too. In this respect, less expensive material can be involved in it such

as plastic, for example, thus alleviating corrosion problems.

3. According to the principle of vapour–liquid equilibrium, the MD process has a high

rejection factor. As a matter of fact, theoretically, complete separation takes place.

In addition, the membrane pore size required for MD is relatively larger than those

for other membrane separation processes, such as RO. The MD process, therefore,

suffers less from fouling.

4. The MD system has the feasibility to be combined with other separation processes

to create an integrated separation system, such as ultrafiltration (UF) or with a RO

unit.

5. MD has the ability to utilize alternative energy sources, such as solar energy.

The MD process is competitive for desalination of brackish water and sea water. It is

also an effective process for removing organic and heavy metals from aqueous

solution, from waste water. MD has also been used to treat radioactive waste,

where the product could be safely discharged to the environment.

5

1.2 MD is also attended by some drawbacks

1. Low permeate flux (compared to other separation processes, like RO)

2. High susceptibility permeate flux to the concentration and temperature of the

feed conditions due to the concentration and temperature polarization

phenomenon.

3. The trapped air within the membrane introduces a further mass transfer

resistance, which also limits the MD permeate flux.

4. The heat lost by conduction is quite large.

1.3 Modelling Aspects of Membrane Distillation

Mass transfer in MD is accompanied by heat transfer, so MD Modelling is more

complex than that for heat exchangers. The parameters or data that should be

considered during Modelling include:

• Membrane characteristics, such as membrane thickness, pore size, tortuosity and

porosity, which can be aquired by gas permeation experiment, scanning electron

microscopy and image analysis;

• Thermal conductivity of the membrane is measurable in some cases or can be

calculated;

• Convective heat transfer coefficient of the feed and/or permeate streams, which

can be calculated by semi-emperical equations based on Nusselt numbers and by

including factors such as the structure of the spacer or module, flow velocities,

properties of feed and permeate, the operation temperature, etc. and

• An important assumption adopted in Modelling MD is that the kinetic effects at the

vapor-liquid interface are negligible. According to this assumption, vapor-liquid

equilibrium equations can be applied to determine the partial vapor pressures of

each component at each side of the membrane.

Mass transfer in MD is controlled by three basic mechanisms,

1. Knudsen diffusion

2. Poiseuille flow (viscous flow)

3. Molecular diffusion.

This gives rise to several types of resistance to mass transfer resulting from transfer

of momentum to the supported membrane (viscous), collision of molecules with

other molecules (molecular resistance) or with the membrane itself-(Knudsen

6

resistance (see Fig. 1). In this context, the dusty gas model is used to describe the

mass transfer resistances in the MD system. It is worth mentioning that the mass

transfer boundary layer resistance is generally negligible.

Similarly, the surface resistance is insignificant, because the surface area of the MD

is small compared to the pore area. On the other hand, the thermal boundary layer

is considered to be the factor limiting mass transfer.

2. Membrane configuration

In this section, different MD configurations that have been utilized to separate

aqueous feed solution using a microporous hydrophobic membrane will be

addressed.

2.1. Direct Contact Membrane Distillation (DCMD)

In this configuration (Fig. 2), the hot solution (feed) is in direct contact with the hot

membrane side surface. Therefore, evaporation takes place at the feed-membrane

surface. The vapour is moved by the pressure difference across the membrane to the

permeate side and condenses inside the membrane module. Because of the

hydrophobic characteristic, the feed cannot penetrate the membrane (only the gas

phase exists inside the membrane pores). DCMD is the simplest MD configuration,

and is widely employed in desalination processes and concentration of aqueous

solutions in food industries, or acids manufacturing. The main disadvantage for

DCMD in commercial applications is its low energy efficiency. Although polymeric

7

membranes generally have low thermal conductivity, the driving force (temperature

difference between the feed and permeate sides) for mass transfer will also lead to

significant conductive heat transfer through the membrane due to the small

membrane thickness, so only part of the supplied heat energy is used for production.

Of the four configurations, DCMD has the highest heat conduction loss because of

the higher heat transfer coefficient on the permeate side for this configuration,

which results in relatively low thermal efficiency.

2.2. Air Gap Membrane Distillation (AGMD)

The schematic of the Air Gap Membrane Distillation (AGMD) is shown in Fig. 3. The

feed solution is in direct contact with the hot side of the membrane surface only.

Stagnant air is introduced between the membrane and the condensation surface.

The vapour crosses the air gap to condense over the cold surface inside the

membrane cell. The benefit of this design is the reduced heat lost by conduction.

However, additional resistance to mass transfer is created, which is considered a

disadvantage. This configuration is suitable for desalination and removing volatile

compounds from aqueous solutions.

In AGMD, the air gap is usually the controlling factor for the mass and heat transfers

because of its greater thermal and mass transfer resistances. In comparison with the

thickness (40–250 μm) and conductivity of the membrane, the air gap is much

thicker (general 2000–10,000 μm) and has lower thermal conductivity. Therefore,

more heat energy in AGMD will be used for water evaporation than in DCMD.

Additionally, if a low temperature feed is used as the cooling stream in this

configuration, the latent heat can be recovered through the condensation of the

vapor on the cooling plate. However, AGMD typically has a low flux, due to the low

temperature difference across the membrane and therefore larger surface areas are

required.

2.3. Sweeping Gas Membrane Distillation (SGMD)

In Sweeping Gas Membrane Distillation (SGMD), as the schematic diagram in Fig. 4

shows, inert gas is used to sweep the vapour at the permeate membrane side to

condense outside the membrane module. There is a gas barrier, like in AGMD, to

8

reduce the heat loss, but this is not stationary, which enhances the mass transfer

coefficient. This configuration is useful for removing volatile compounds from

aqueous solution. The main disadvantage of this configuration is that a small volume

of permeate diffuses in a large sweep gas volume, requiring a large condenser.

In SGMD, the vapor is stripped from the hot feed by a gas stream, and then

condensed in an external condenser. It has higher mass transfer rates than AGMD,

due to the greater driving force originating from the reduced vapor pressure on the

permeate side of the membrane, and has less heat loss through the membrane than

DCMD. However, an external condenser and an air blower or compressed air are

needed to maintain operation of this configuration, which will cause an increase in

investment, energy use and running costs.

It is worthwhile stating that AGMD and SGMD can be combined in a process called

thermostatic sweeping gas membrane distillation (TSGMD). The inert gas in this case

is passed through the gap between the membrane and the condensation surface.

Part

of

vapour is condensed over the condensation surface (AGMD) and the remainder is

condensed outside the membrane cell by external condenser (SGMD).

2.4. Vacuum Membrane Distillation (VMD)

The schematic diagram of this module is shown in Fig. 5. In VMD configuration, a

pump is used to create a vacuum in the permeate membrane side. Condensation

takes place outside the membrane module. The heat lost by conduction is negligible,

which is considered a great advantage. This type of MD is used to separate aqueous

volatile solutions.

9

In VMD, the vapor permeate is removed continuously from the vacuum chamber to

form a vapor pressure difference across the membrane. Theoretically, this

configuration can provide the greatest driving force at the same feed temperature,

because the vapor pressure on the cold side can be reduced to almost zero. An

external condenser is required as for AGMD, if the liquid permeate is the product.

Of the four configurations, DCMD is the most popular for MD laboratory research,

with more than half of the published references for membrane distillation based on

DCMD. However, AGMD is more popular in commercial applications, because of its

high energy efficiency and capability for latent heat recovery

3. Configurations of MD Modules

There are two major MD module configurations, which are the tubular module and

the plate and frame module. Both of these modules have been used in pilot plant

trials.

Figure (a) shows a schematic diagram of a hollow fiber tubular module, in which

hollow fiber membranes are glued into a housing. This configuration can have a very

high packing density (3000 m2/m3). The feed is introduced into the shell side or into

lumen side of the hollow fibers, and cooling fluid, sweeping gas, or negative pressure

can be applied on the other side to form VMD, SGMD, or DCMD. Because of its large

active area combined with a small footprint, hollow fiber modules have great

potential in commercial applications. Although broken hollow fibers cannot be

replaced, they can be detected by the liquid decay test (LDT) and pinned to remove

broken fibers from service. Good flow distribution on the shell side can be difficult to

achieve, with subsequent high degrees of temperature polarization. Cross-flow

modules have been developed to reduce this effect for hollow fiber modules

Figure (b) shows the structure of the plate and frame module. This module is suitable

for flat sheet membranes and can be used for DCMD, AGMD, VMD, and SGMD. In

this configuration, the packing density is about 100–400 m2/m3. Although this

10

configuration has a relatively smaller effective area for the same volume when

compared to the tubular modules, it is easy to construct and multiple layers of flat

sheet MD membranes can be used to increase the effective area. As shown in Figure

(b), it is easy to change damaged membranes from this configuration. Thus, this

module is widely employed in laboratory experiments for testing the influence of

membrane properties and process parameters on the flux or energy efficiency of

membrane distillation. Also the flow dynamics can be improved by the use of spacers

that increase turbulence and reduce temperature polarization.

To meet the requirement of commercial applications, other configurations with large

specific areas were also developed, i.e., spiral-wound modules mainly employed for

air/permeate gap membrane distillation have a much more compact structure than

the conventional plate and frame AGMD.

4. Membranes for MD Applications

There are two common types of membrane configurations shown in Figure 3:

• Hollow fiber membrane mainly prepared from PP, PVDF, and PVDF-PTFE composite

material; and

• Flat sheet membrane mainly prepared from PP, PTFE, and PVDF.

4.1 Hollow fiber

The hollow fiber module, which has been used in MD, has thousands of hollow fibers

bundled and sealed inside a shell tube. The feed solution flows through the hollow

fiber and the permeate is collected on the outside of the membrane fiber (inside-

outside), or the feed solution flows from outside the hollow fibers and the permeate

is collected inside the hollow fiber (outside-inside). The main advantages of the

hollow fiber module are very high packing density and low energy consumption. On

the other hand, it has high tendency to fouling and is difficult to clean and maintain.

It is worth mentioning that, if feed solution penetrates the membrane pores in shell

and tube modules, the whole module should be changed.

11

Compared with flat sheet membranes, hollow fiber membranes have relatively large

specific surface areas, but the main impediment of the hollow fiber module is its

typically low flux (generally 1–4 L m−2 h−1 at 40–60 °C). The low flux is related to its

poor flow dynamics and the resultant high degree of temperature polarization.

However, high-flux hollow fiber membranes with different features suitable for

membrane distillation have been developed recently, such as dual-layer hydrophilic-

hydrophobic fibers with a very thin effective hydrophobic PVDF layer (50 μm), and

hollow fiber membranes with a sponge-like structure and thin walls, which have flux

of about 50–70 kg m−2 h−1 at about 80–90 °C. This flux is as high as that from flat

sheet membrane.

4.2 Plate and frame

The membrane and the spacers are layered together between two plates (e.g. flat

sheet). The flat sheet membrane configuration is widely used on laboratory scale,

because it is easy to clean and replace. However, the packing density, which is the

ratio of membrane area to the packing volume, is low and a membrane support is

required.The flat sheet membrane is used widely in MD applications, such as

desalination and water treatment.

The reported flux from flat sheet membranes is typically 20–30 L m−2 h−1 at inlet

temperatures of hot 60 °C and cold 20 °C. In general, the polymeric membrane

shown in Figure 3b is composed of a thin active layer and a porous support layer.

This structure is able to provide sufficient mechanical strength for the membrane to

enable the active layer to be manufactured as thin as possible, which reduces the

mass transfer resistance.

4.3. Tubular membrane

In this sort of modules, the membrane is tube-shaped and inserted between two

cylindrical chambers (hot and cold fluid chambers). In the commercial field, the

tubular module is more attractive, because it has low tendency to fouling, easy to

clean and has a high effective area. However, the packing density of this module is

low and it has a high operating cost. Tubular membranes are also utilized in MD.

Tubular ceramic membranes were employed in three MD configurations DCMD,

AGMD and VMD to treat NaCl aqueous solution, where salt rejection was more than

99%.

4.4. Spiral wound membrane

In this type, flat sheet membrane and spacers are enveloped and rolled around a

perforated central collection tube. The feed moves across the membrane surface in

an axial direction, while the permeate flows radially to the centre and exits through

the collection tube. The spiral wound membrane has good packing density, average

12

tendency to fouling and acceptable energy consumption. It is worth stating that

there are two possibilities for flow in a microfiltration system; cross flow and dead-

end flow. For cross flow, which is used in MD, the feed solution is pumped

tangentially to the membrane. The permeate passes through the membrane, while

the feed is re-circulated. However, all the feed passes through the membrane in the

dead-end type.

5. Membrane Material

The most common materials used for MD membranes are polytetrafluoroethylene

(PTFE), polypropylene (PP) and polyvinylidenefluoride (PVDF). The porosity of the

membranes used is in the range of 0.60 to 0.95, the pore size is in the range of 0.2 to

1.0 μm, and the thickness is in the range of 0.04 to 0.25 mm. The surface energies

and thermal conductivities of these materials are listed in Table 1.

Of these materials, PTFE has the highest hydrophobicity (largest contact angle with

water), good chemical and thermal stability and oxidation resistance, but it has the

highest conductivity which will cause greater heat transfer through PTFE

membranes. PVDF has good hydrophobicity, thermal resistance and mechanical

strength and can be easily prepared into membranes with versatile pore structures

by different methods. PP also exhibits good thermal and chemical resistance.

Recently, new membrane materials, such as carbon nanotubes, fluorinated

copolymer materials and surface modified PES, have been developed to make MD

membranes with good mechanical strength and high hydrophobicity and porosity.

6. Membrane characteristics

In membrane distillation, membranes on the basis of their selective properties are

not involved in the mass transport phenomena, but are involved in heat transport

from the hot side to the cold side. Therefore, compounds transferred across the

membrane in gas phase are driven by vapour pressure differences based on vapour-

liquid equilibrium, and the macro-porous polymeric or inorganic membrane

employed between the permeate and feed sides acts as a physical barrier providing

the interfaces where heat and mass are simultaneously exchanged.

A suitable membrane needs to exhibit certain characteristics in order to be viable in

MD. Although, the membrane should be porous, it should not be wetted by the

13

process liquids under the pressure applied within the membrane module.

Furthermore, no capillary condensation should occur within the pores, while the

membrane itself should not affect the vapour/liquid equilibrium of the system being

desalinated. Finally, at least one side of the membrane should be in direct contact

with the process liquids while only vapour should be transported across the pores of

the membrane.

The properties and structure of the surface of the membrane are highly important,

and the requirements may vary depending on application and the type of MD

configuration in which the membrane is being used: VMD, AGMD, SGMD or DCMD.

While SGMD and VMD are the most energy intensive MD techniques, they are

generally preferred to separate two mixed liquids having different boiling points to

avoid further treatments linked to the contamination of another carrier liquid on the

permeate side. On the other hand for water purification, desalination and

dewatering AGMD and DCMD are generally used since only water is evaporated from

the bulk feed.

Thus, the properties of membranes suitable for membrane distillation should

include:

6.1 Membrane thickness

An adequate thickness, based on a compromise between increased membrane

permeability (tend to increase flux) and decreased thermal resistance (tend to

reduce heat efficiency or interface temperature difference) as the membrane

becomes thinner;

The membrane thickness is a significant characteristic in the MD system. There is an

inversely proportional relationship between the membrane thickness and the

permeate flux. The permeate flux is reduced as the membrane becomes thicker,

because the mass transfer resistance increases, while heat loss is also reduced as the

membrane thickness increases. The optimum membrane thickness lies between 30–

60 μm. It is worth noting that the effect of membrane thickness in AGMD can be

neglected, because the stagnant air gap represents the predominant resistance to

mass transfer.

6.2 Liquid entry pressure (wetting pressure)

Reasonably large pore size and narrow pore size distribution, limited by the

minimum Liquid Entry Pressure (LEP) of the membrane. In MD, the hydrostatic

pressure must be lower than LEP to avoid membrane wetting. This can be quantified

by the Laplace (Cantor) Equation as following Equation (1):

(1)

14

where Pf and Pp are the hydraulic pressure on the feed and permeate side, B is a

geometric pore coefficient (equal to 1 for cylindrical pores), γl is liquid surface

tension, θ contact angle and rmax is the maximum pore size. The contact angle of a

water droplet on a Teflon surface varies from 108° to 115°; 107° for PVDF and 120°

for PP. It is worthwhile indicating that a flat ceramic membrane had a contact angle

varying from 177° to 179°. Moreover, ceramic zirconia and titania hydrophobic

membranes were prepared with a 160° contact angle. All these ceramic membranes

are used for desalination purposes (see Table 1). With regard to surface tension, the

impact of salt concentration (NaCl) on the water surface tension is:

(2)

γl stands for pure water surface tension at 25 °C, which is 72 mN/m

As a result, membranes that have a high contact angle (high hydrophobicity), small

pore size, low surface energy and high surface tension for the feed solution possess a

high LEP value. Typical values for surface energy for some polymeric materials are

reported in Table 2 below. The maximum pore size to prevent wetting should be

between 0,1–0,6 μm. Moreover, the possibility of liquid penetration in VMD is higher

than other MD configurations, so a small pore size is recommended.

15

6.3 Membrane porosity and tortuosity

High porosity. High porosity increases both the thermal resistance and the

permeability of MD membranes, so both the heat efficiency and flux are increased.

However, high porosity membranes have low mechanical strength and tend to crack

or compress under mild pressure, which results in the loss of membrane

performance.

Membrane porosity refers to the void volume fraction of the membrane (defined as

the volume of the pores divided by the total volume of the membrane). Higher

porosity membranes have a larger evaporation surface area. Two types of liquid are

used to estimate membrane porosity. The first penetrates the membrane pores (e.g.

isopropyl alcohol, IPA), while the other, like water, does not. In general, a membrane

with high porosity has higher permeate flux and lower conductive heat loss. The

porosity (ε) can be determined by the Smolder–Franken equation.

(3)

where ρm and ρpol are the densities of membrane and polymer material, respectively.

Membrane porosity in the MD system varies from 30 to 85%.

Tortuosity (τ) is the deviation of the pore structure from the cylindrical shape. As a

result, the higher the tortuosity value, the lower the permeate flux. The most

successful correlation is:

(4)

6.4 Mean pore size and pore size distribution

Low surface energy, equivalent to high hydrophobicity, based on Equation (1).

Material with higher hydrophobicity can be made into membranes with larger pore

sizes, or membranes made from more hydrophobic material will be applicable under

higher pressures for a given pore size;

Membranes with pore size between 100 nm to 1 μm are usually used in MD systems.

The permeate flux increases with increasing membrane pore size. The mechanism of

mass transfer can be determined, and the permeate flux calculated, based on the

membrane pore size and the mean free path through the membrane pores taken by

transferred molecules (water vapour). Generally, the mean pore size is used to

determine the vapour flux. A large pore size is required for high permeate flux, while

the pore size should be small to avoid liquid penetration. As a result, the optimum

pore size should be determined for each feed solution and operating condition.

16

In fact, the membrane does not have a uniform pore size so more than mass transfer

mechanisms occur simultaneously (depending to the pore size). There are several

investigations examine the importance of pore size distribution in MD flux. Better

understanding of membrane morphology such as pore size, pore size distribution,

porosity, and thickness directs to have an accurate mass and heat transfer modeling.

Regarding to the MD membrane, two types of characteristics can be analyzed, the

structural characteristic and the actual separation parameters (permeation).

6.5 Thermal conductivity

Low thermal conductivity. High thermal conductivities increases sensible heat

transfer and reduce vapor flux due to reduced interface temperature difference;

The thermal conductivity of the membrane is calculated based on the thermal

conductivity of both polymer ks and gas kg (usually air). The thermal conductivity of

the polymer depends on temperature, the degree of crystallinity, and the shape of

the crystal. The thermal conductivities of most hydrophobic polymers are close to

each other. For example, the thermal conductivity of PVDF, PTFE and PP at 23 °C are

0.17–0.19, 0.25–0.27 and 0.11–0.16 Wm−1K−1 respectively. The thermal conductivity

of PTFE can be estimated by

Ks = 4,86 x 10-4 x T + 0,253 (5)

The thermal conductivity of the MD membrane is usually taken a volume-average of

both conductivities ks and kg as follows:

km = (1-ε) x ks + ε x kg (6a)

However, it is suggested that thermal conductivity of an MD membrane is better

based on the volume average of both resistances (1/kg and 1/ks), i.e.,

km = [ e / kg + (1-ε) / ks]-1 (6b)

for, the thermal conductivity values for air and water vapour at 25 °C are of the same

order of magnitude. For instance, the thermal conductivity of air at 25 °C is

0.026Wm−1K−1 and for water vapour, it is 0.020Wm−1K−1. As a result, the assumption

of one component gas present inside the pores is justified. The thermal conductivity

of water vapour and air at around 40 °C can be computed by:

kg = 1,5 x 10-3 x (7)

Some ways to reduce the heat loss by conduction through the membrane; using

membrane materials with low thermal conductivities, using a high porosity

membrane, using thicker membrane, and minimizing heat losses. It is also suggested

that the permeability can be enhanced by using a composite porous

17

hydrophobic/hydrophilic membrane. In this case, the top layer is very thin

hydrophobic layer to stop liquid penetration, followed by thick hydrophilic layer.

Both layers reduce the heat losses through the membrane.

6.6 Surface Roughness

Surface roughness is also critical, because it will affect a number of properties

including surface fouling and the contact angle of water on the membrane surface. A

change of wetting behaviour will likely affect heat conduction across the top

membrane layer, therefore, clearly affecting performance of MD membranes.

Although wetting was shown to be facilitated by rough hydrophilic surfaces as more

points for spreading are offered to the liquid, this is not always true for

homogeneous hydrophobic materials and was shown to highly depend on the

composition of the surface and the shape of the roughness extrusions. As the

average roughness increases, the advancing angle of liquids on hydrophobic surfaces

tends to be increased due to the larger number of interactions between the nodules

and obstacles composing the surface of the membrane and the liquid. This tendency,

known in surface science as the lotus effect, is particularly enhanced for materials

exhibiting contact angles >150° with the wetting liquid. A convenient way to

measure roughness is typically given by the roughness factor κ, defined as:

κ = Αm / An (8)

where An and Am are respectively the area calculated as the projection of the object

on a plan normal to the main direction of the surface, and the surface area

measured by any experimental adsorption technique.

The measured surface roughness and area can be obtained by a number of

techniques, such as atomic force microscopy (AFM), gas adsorption (BET), diffuse X-

ray spectroscopy or laser light scattering depending on the size of the pores and the

accuracy sought. In the case of membrane surfaces, the difficulty resides in the

definition of what the true roughness is or, in other words, how deep one wants to

consider fluctuations from the surface as the surface or the inside of the pores. The

characterization of surface roughness is often ignored in MD as the process is

considered to be mostly unaffected by fouling. However, surface roughness as

shown does have other implications on the performance of membranes for MD and

should, therefore, be more thoroughly studied.

7. Advances on MD Processes and Modules for Water Purification

Even though membrane distillation was patented in the 1960s, it has not been

commercialised because of the success of competing technologies. However in just

the last few years, MD has emerged with numerous commercially oriented devices

18

and novel process integrations. This section focuses on the current process

arrangements and commercially available MD systems.

7.1. MD Stand-Alone Systems

A module to house a membrane and perform MD is not complicated but requires

more complexity in its connections as compared to pressurised membrane systems

(micro, ultra and nanofiltration as well as reverse osmosis). As shown in Figure 6, we

see the simplest form of DCMD configuration which will desalinate a saline water

feed to a very high quality permeate.

However, the simplest form suffers drawbacks which must be overcome to make MD

practically useful. The three key drawbacks under standard process configuration

are:

• Water recovery limit: The flux of the membrane draws a significant amount of

energy purely through the evaporation of the feed, which is deposited into the

permeate. The limiting amount of water permeated as a fraction of water fed, F,

(i.e., single pass recovery) is presented according to as Equation (9):

F = (1-t) x CP x (TF - TE)/ΔΗvap (9)

where TF and TE are the feed and exit temperatures, respectively (K or °C), CP is the

specific heat of water (4.18 kJ/kg/K), t is the proportion of conductive heat (balance

due to evaporative heat) loss through the membrane, and ΔHvap is the latent heat of

vaporisation (kJ/kg). For example, if the feed water is supplied at 80 °C, no more

than 7.7 wt % of this desalinated water will evaporate to the permeate (i.e., F) by the

time this temperature is reduced to 20 °C (assuming t = 0.3). This is typically

managed by reheating the cool brine reject and sending it back to the feed. In

DCMD, this recirculation is likewise done on the permeate side. Both pumps will now

be larger, by at least an order of magnitude, in order to achieve useful recoveries

exceeding 50%.

19

• Electrical energy constraints: The thermodynamics of the simple MD setup in turn

constrains the electrical consumption. Each pump in Figure 6 will consume electrical

energy per unit water permeated, Eelec,std (kWh/m3), according to:

Eelec,std = PF/ (η x F) x 1/3600 (10)

where PF is the MD module feed pressure (kPa), and η is pump efficiency. If we

assume PF = 20 kPa, and pump efficiency of 0.6, each pump consumes 0.12 kWh/m3

of electricity. Both pumps consume 0.23 kWh/m3. Clearly achieving low pressure

drops along the module will have an impact on the electrical energy requirement of

MD systems. This minimum is related to the point above, where F equates to around

7.7 wt %;

• Thermal energy constraints: Water evaporation energy per unit mass, ΔHvap, is

2260 kJ/kg, or 628 kWh/m3. This energy is in the form of thermal energy, which is

the standard thermal energy required to operate the MD system in Figure 6. This

value equates to a performance ratio (PR), or gain output ratio (GOR) of 1, being the

mass ratio of water produced to the amount of steam energy (i.e., latent heat) fed to

the process.

With state-of-the-art reverse osmosis requiring as little as 2 kWh/m3 of electric

energy and no thermal energy, we see that standard MD by thermodynamics uses an

order of magnitude less electricity, and nearly 300 fold the thermal energy to

desalinate the same amount of water. State-of-the-art MD systems feature

refinement of the system proposed in Figure 6, or its variants VMD, SGMD and

AGMD, primarily to reduce the thermal energy required, and more recently, the

electrical energy.

7.2. State of the Art MD Research and Systems

The principal research activities on MD can be divided broadly into two categories:

fouling/performance testing, and energy efficient process design. With

fouling/performance design, fundamental understandings of the diffusion

mechanisms coupled with heat and mass transfer has unlocked the critical science

needed to select optimal operating conditions, membrane materials and module

designs that ultimately give better flux performance for the same operational

conditions. Fouling of membranes has explored scaling issues for the classic

applications in brine concentration, and the more novel application in dairy

processing. While this research progresses to uncover further fundamental

improvements, the focus here is on the novel process configurations that address

the performance limitations defined in Section 2.1. The most notable organisations

specialising in MD modules or high efficiency systems are:

20

• Fraunhofer ISE (AGMD);

• Memstill and Aquastill (AGMD);

• Scarab (AGMD);

• Memsys (vacuum enhanced multi effect AGMD).

8. Mechanisms

21

8.1 Mass transfer

8.1.1 Direct Contact Membrane Distillation (DCMD)

Mass transfer in the DCMD process includes three steps: firstly the hot feed

vaporizes from the liquid/gas interface, secondly the vapour is driven by the vapour

pressure difference and crosses from the hot interface to the cold interface through

the pores, and thirdly the vapour condenses into the cold side stream. Therefore,

there are two major factors controlling the mass transfer: one is the vapour pressure

difference, and the other is the permeability of the membrane.

If the fluid dynamics conditions on both sides of the membrane could be considered

good, mass transfer through the membrane may be the limiting step for mass

transfer in MD. The influence of the physical properties on membrane permeability

includes:

(1) The effective area for mass transfer is less than the total membrane area because

the membrane is not 100% porous;

(2) For most practical membranes, the membrane pores do not go straight through

the membrane and the path for vapour transport is greater than the thickness of the

membrane; and

(3) The inside walls of the pores increase the resistance to diffusion by decreasing

the momentum of the vapour molecules.

The mass transport mechanism in the membrane pores is governed by three basic

mechanisms known as Knudsen-diffusion (K), Poiseuille-flow (P) and Molecular-

diffusion (M) or a combination between these known as the transition mechanism.

Knudsen diffusion takes place when the pore size is too small, so the

collision between the molecules and the inside walls of the membrane

suitably expresses the mass transport and the collision between molecules

can be ignored.

Molecular diffusion occurs when the molecules move corresponding to each

other under the influence of concentration gradients.

In Poiseuille flow (viscous flow), the gas molecules act as a continuous fluid

driven by a pressure gradient.

The Knudsen number (Kn) is used to indicate the dominant mass transfer

mechanism in the pores:

22

The Knudsen number (Kn), defined as the ratio of the mean free path (λ) of

transported molecules to the membrane pore size, provides a guideline of which

mechanism is active inside the membrane pore. According to kinetic theory of gases,

the molecules are assumed to be hard spheres with diameter de and are involved in

binary collisions only. It is worth noting that the collision diameter for water vapour

and air are about 2.64×10−10, and 3.66×10−10, respectively. The average distance

travelled by molecules to make collisions (λ) is defined as.

λ = (kΒ x T) / ( x π x P x de2) (11)

kB, T and P are Boltzman constant, absolute temperature, and average pressure

within the membrane pores respectively. The mean free path value of water vapour

at 60 °C was estimated to be 0.11 μm.

For Kn > 1 or dp < λ (Knudsen region), the mean free path of water vapour

molecules is large compared to the membrane pore size, which means the

molecule-pore wall collisions are dominant over molecule-molecule collision.

The mass transfer is:

CKn = 2π/3 x 1/ (R x T) x [(8 x R x T)/ (π x Mw)]1/2 x r3/ (τ x δ) (12)

where ε, τ, r, δ and Mw are porosity, pore tortuosity, pore radius, membrane

thickness and molecular weight of water vapour, respectively.

If the kn < 0.01 or dp > 100λ (continuum region), ordinary molecular diffusion

model represents the diffusion of the vapour flux through stationary air film

(the air which exist inside the membrane pores), ordinary molecular diffusion

is used to describe the mass transport

CD = π/ (R x T) x [(P x D)/ Pair] x r2/ (τ x δ) (13)

where Pair is the air pressure within the membrane pore, D is diffusion coefficient,

and P is the total pressure inside the pore which is equal to the partial pressure of air

and water vapour.

23

In addition, the flux of water vapour molecules, which diffuse through the

membrane pores (stagnant air), is:

J = 1/ Pair x ε/ (τ x δ) x [(D x P x Mw)/ (R x T)] x ΔP (14)

where Pair and P are the average air pressure and average gas pressure within the

membrane respectively.

Removing the stagnant air existing inside the pores by degassing the feed and

permeate will reduce the molecular diffusion resistance, so the membrane

permeability will increase.

However, If 0.01 < kn < 1 or λ < dp <100λ (transition region), the water vapour

molecules collide with each other, and also diffuse through the air film.

Consequently, the mass transfer takes place by both the Knudsen/ordinary

diffusion mechanism, where:

Cc = π/ (R x T) x 1/ (τ x δ) x [(2/3 x ((8 x R x T)/ (π x Mw)1/2 x r3)-1 + ((PD)/

Pα x r2)-1]-1 (15)

The diffusivity of water vapour through the stagnant air inside the pores is given by

PD = 1.895 x 10-5 x T2,072 (16)

In addition, the Fuller equation, which is a common equation to predict binary gas

diffusion, can be used

D = 10-7 x T1,75 x (1/ Mwα + 1/ Μwb)1/2/ [Px ((Σvα)1/3 + (Σvb)1/3)2] (17)

where Σv represents the diffusion volume, T is temperature in Kelvin and P is

pressure in atmospheres. The diffusion volume of air and water are 20.1 and 12.7

respectively.

It is stated that the molecule-pore wall collisions (Knudsen diffusion) and molecule-

molecule collisions (molecular diffusion) takes place simultaneously for pore size less

than 0.5 μm. Moreover, the flux can be expressed by molecular diffusion only for

large pores. Furthermore, the vapour flux across the membrane can be expressed by

Knudsen diffusion and Poiseuille (viscous) flow model for de-aerated DCMD. On the

other hand, the Poiseuille flow should be considered as one of the mechanisms of

mass transfer model in large pore size membrane.

It is helpful to analyze the transport in terms of resistances, to identify the

controlling role of each transport step, and as a result the flux permeate can be

improved. Table 4 shows DCMD membrane coefficients as reported by some

researchers.

24

It is believed that when the pore size is near the mean free path value (critical pore

size), the permeate flux under the Knudsen mechanism is higher than that obtained

from the combination of Knudsen and molecular diffusion mechanisms. Therefore,

choosing membranes that have small pore size may be better than membranes

having large pore size. It is worth mentioning that the effect of pore size distribution

can be neglected for large pore size.

8.1.2. Air Gap Membrane Distillation (AGMD)

The molecular diffusion theory is used to describe the transfer of vapour molecules

through the membrane and the air gap. A stagnant gas film (air) is assumed to lie

inside the membrane at the air gap side.

Kurokawa computed the flux by considering the diffusion in one direction through

both membrane and air gap, where the air gap is below 5 mm:

J = (P x Mw)/ (RT x P*) x (D/ (δ/ε3,6 + 1) x ΔP (18)

where ΔP is the water vapour pressure difference between the feed on the

membrane and the condensation surface, and P* is the partial pressure of water.

Liu estimated the permeate flux for aqueous solution when the average operating

temperature, Ta, was between 30 °C and 80 °C, thus:

J = (Tf - Tp)/ (αΤα-2,1 + β) (19)

where α and β are parameters that can be determined experimentally.

It is worthwhile stating that the air gap is about 10 to 100 times the membrane

thickness, so the effect of air inside the membrane can be neglected.

25

Stefan diffusion was used to describe the diffusion through a stagnant gas film. It can

be represented mathematically as

Ν = - (c x D)/ (1 - y) x dy/ dz (20)

where D, y, c and z are diffusion coefficient, mole fraction of the vapour phase,

molar concentration and diffusion length, respectively.

The Stefan equation was solved by Kimura and Nakad

N = (c x D)/ z x ln((1-yf)/ ((1 - ym)) (21)

where ym and yf represent the mole fraction of vapour at the membrane and the

condensation film, respectively.

However, Jonson solved the same equation by neglecting the effect of temperature

and concentration polarization. They suggested that, the value of c D for water

vapour and air at around 40 °C to be calculated using this equation:

c x D = 6,3 x 10-5 x (22)

In addition, the molar concentration can be calculated from ideal gas law:

c = P/ (R x T) (23)

According to the standard condition, the diffusion coefficient can be corrected to the

desired temperature by:

D/ D0 = (T/ T0)3/2 (24)

Bouguecha used Stefan diffusion to express the vapour flux when it is governed by

diffusion through the membrane pores and by natural convection through the air

gap:

N = KT/ R x (Pf,m - Pfilm) (25)

where KT is the overall mass transfer coefficient. Stefan diffusion was also utilized to

evaluate the molar flux of seawater as:

N = DP/ (RT x l x Plm) x (P2- P4) (26)

where, P2, P4, Dw and Plm are the vapour pressures at Tf, m, the vapour pressures at

Tfilm, diffusion coefficient and log mean partial pressure respectively. The log mean

partial pressure difference at the air gap is defined as:

Plm = (P4 - P2)/ ln(P4/ P2) (27)

26

For a multi-component mixture, the Stefan-Maxwell equation was applied by Gostoli

and Sarti to express the ethanol and water vapour diffusion in stagnant gas (air). This

was given by:

dyi/ dz = / cDij x (yiNj - yjNi) (28)

The vapour composition at evaporation and condensation interface scan be

calculated by assuming liquid-vapour equilibrium, such that:

yi = (xi x ai x P0)/ P (29)

Vapour pressure P0 can be computed by the Antoine equation at the temperature of

interest. The activity coefficient ai can be calculated by the Van Laar equation at the

temperature and composition of interest. The condensate composition xi is

determined by the components flux

xi = Ni/ ΣN (30)

On the other hand, Banat and Simandl employed Stefan diffusion (Eq. (24)) to

represent the molar diffusion flux of an ethanol-water solution. The molar diffusion

flux of ethanol and water through stagnant gas (air) in terms of pressure is given by:

Ni = (ε x Di x P)/ (RT x l x Plm) x (Pi2-Pi4) (31)

For the non-equilibrium thermodynamics case, the ordinary diffusion, which is

related to the concentration gradient, and thermal diffusion which is related to the

temperature gradient were considered to calculate the total mass flux. A linear

relation between flux and vapour pressure can be assumed, and the thermal

diffusion can be neglected.

The Stefan–Maxwell model is reported to be more accurate than the molecular

diffusion model (Fick's law) for separation of azeotropic mixtures

8.1.3. Vacuum Membrane Distillation (VMD)

In order to remove air trapped in the membrane pores, the deaeration of the feed

solution or a continuous vacuum in the permeate side should be applied.

Consequently, the ordinary molecular diffusion resistance is neglected. The Knudsen

mechanism is used to express the mass transfer, Poisseille flow or both together.

For example, when the ratio of the pore radius to the mean free path r/λ is

<0.05, the molecule-pore wall collisions control the gas transport mechanism

(Knudsen flow model) and the molar flow rate is:

Ni = 2π/3 x 1/RT x ((8RT/ (π x Mwi)1/2 x r3/ (δ x τ) x ΔPi (32)

27

If r is between 0.05λ and 50λ, both molecular-molecular and molecular-wall

collisions should be considered. The total mass transfer is described by the

Knudsen-viscous model and can be represented by the following equation:

Ni = π/ (RT x δ x τ) x (2/3 x (8RT/ (π x Mwi)1/2 x r3 + r4/ 8μi x Pavg) x ΔPi

(33)

where μi is the viscosity of species i, and Pavg is the average pressure in the pore.

When r/λ is >50, molecular- molecular collision dominates and the mass

transfer can be expressed by Poisseuille flow (viscous), such that:

Ni = (π x r4)/ 8μi x Pavg/ RT x 1/ (τ x δ) x ΔPi (34)

8.1.4. Sweeping Gas Membrane Distillation (SGMD)

The equations, which illustrate the mass transfer of DCMD can be used in SGMD.

Knudsen/molecular diffusion can be used to describe the mass transfer through the

membrane pores. Moreover, the circulation velocity and feed temperature are

significant parameters.

Sherwood correlation can be used to estimate the mass transfer coefficient, k, across

the boundary layers, then the concentration at the boundary layer can be evaluated.

The empirical form of the Sherwood correlation is

Sh = (k x d)/ D = (Constant) x Rea x Scb (35)

where Re, Sc, and D are Reynolds number, Schmidt number and diffusion coefficient

respectively (Table 5).

Schmidt numbers can be calculated by:

28

Sc= μ/ (ρ x d) (36)

where μ is the viscosity. For a non-circular channel, these correlations

can be utilized if the equivalent (hydraulic) diameter deq is employed.

deq = 4rH = 4S/ LP (37)

where rH, S and LP are the hydraulic radius, cross sectional area of the flow channel,

and length of wetted perimeter of the flow channel, respectively.

8.2. Heat transfer

In MD processes, heat and mass transfers are coupled together in the same direction

from the hot side to the cold side. Figure 4 illustrates these processes in DCMD,

which is typical for MD configurations. The feed temperature, Tf, drops across the

feed side boundary layer to T1 at the membrane surface. Some water evaporates and

is transported through the membrane. Simultaneously, heat is conducted through

the membrane to the cold (permeate) side. The cold flow temperature Tp increases

across the permeate boundary layer to T2 at the membrane surface on the cold side

as water vapour condenses into the fresh water stream and gains heat from the feed

side. The driving force is, therefore, the vapour pressure difference between T1 and

T2, which is less than the vapour pressure difference between Tf and Tp. This

phenomenon is called temperature polarization.

29

So two main heat transfer mechanisms occur in the MD system: latent heat and

conduction heat transfer and the heat transfer, which occurs in DCMD, can be

divided into three regions (Fig. 7)

Heat transfer by convection in the feed boundary layer:

Qf = hf x (Tf - Tf,m) (38)

Heat transfer through the membrane by conduction, and by movement of vapour

across the membrane (latent heat of vaporization). The influence of mass transfer on

the heat transfer can be ignored

Qm = Km/δ x (Tf,m - Tp,m) + J x ΔHv (39)

Qm = hm x (Tf,m - Tp,m) + J x ΔHv (43)

where hm represents the heat transfer coefficient of the membrane.

It is worth mentioning that hm can be rewritten for pure water or very diluted

solution, and where temperature difference across the membrane surfaces is less

than or equal to 10 °C by substituting Eq. J = Cm x dP/ dT x (Tf,m - Tp,m) into (39),

Qm = Km/δ x (Tf,m - Tp,m) + [Cm x dP/dT x (Tf,m - Tp,m)] x ΔHv (44)

Qm = [Km/δ + (Cm x dP/dT) x ΔHv] x (Tf,m - Tp,m)] (45)

Qm = hm x (Tf,m - Tp,m) (46)

For a non-linear temperature distribution assumption, Qm (for the x-dimension) is

also expressed as

Qm = -km x dT/dx + J x ΔHv (47)

30

For the permeate side, the convection heat transfer takes place in the permeate

boundary layer

Qp = hp x (Tp,m - Tp) (48)

At steady state, the overall heat transfer flux through the membrane is given by:

Q = Qf = Qm = Qp (49)

hf x (Tf - Tf,m) = km/δ x (Tf,m - Tp,m) + J x ΔHv = hp x (Tp,m - Tp) (50)

Q = U x (Tf - Tp) (51)

where U represents the overall heat transfer coefficient.

It is worth pointing out that the heat conduction can be neglected for non-sported

thin membrane and for high operating temperature as well. Moreover, the heat

transfer by convection is ignored in the MD process, except in AGMD.

The surface temperature of both sides of membrane cannot be measured

experimentally, or calculated directly. Therefore, a mathematical iterative model has

been designed to estimate these temperatures:

Tf,m = Tf - (J x ΔHv + km/δm x (Tf,m - Tp,m))/ hf (52)

Tp,m = Tp - (J x ΔHv + km/δm x (Tf,m - Tp,m))/ hp (53)

The value of Hv should be evaluated at average membrane temperature. However,

the model was evaluated at logarithmic average membrane temperature.

The surface membrane temperature in terms of temperature polarisation

coefficient, ψ, for pure water and very diluted solution,

Tf,m - Tp,m = 1/ (1 + H/ hf +H/ hp) x (Tf- Tp) = ψ x (Tf - Tp) (54)

where hm is equal to

hm = (Cm x dP/dT) x ΔΗv + km/δ (55)

The (Tf,m - Tp,m) is about 0.1 °C at low flux and does not exceed 0.5 °C at high flux.

Concerning the presence of free and force convection in laminar flow in DCMD, the

following equation to calculate the heat transfer coefficient is suggested:

Νu = 0,74 x Re0,2 x (Gr x Pr)0,1 x Pr0,2 (56)

31

For the AGMD configuration, the heat transfer through the AGMD could be

represented as in DCMD, except for the heat transfer across the air gap, which

occurs by conduction and vapour (mass transfer)

hf x (Tf - Tf,m) = J x ΔHv + km/δ x (Tf,m - Tp,m) = J x ΔHv + kg/l x (p,m - Tfilm) = hd

x (Tfilm - T5) (57)

In addition, we can use the following equation to calculate the heat transfer

coefficient for the condensate film (pure vapour) on a vertical wall:

hd = 2/3 x x [(kfilm3 x ρ2 x g x ΔHv)/ (μ x L x (Tfilm - T5)]1/4 (58)

The sensible heat for the MD system can be neglected, because it has a very small

magnitude compared to the heat of vaporization

Q = J x ΔHv (59)

Free convection heat transfer between two vertical plates is also used to describe

the heat transfer phenomenon in the air gap region, when the air gap distance is

over 5 mm

Nu = c x (Pr x Gr)n x (l/L)1/9 (60)

where

105 < Gr < 107, c = 0,07 and n = 1/3

104 < Gr < 105, c = 0,2 and n = 1/4

For VMD configuration, heat transfer by convection in the feed boundary layer can

be expressed as:

Qf = hf x (Tf - Tf,m) (61)

However, the heat transfer by conduction through the membrane is ignored, so the

heat transfer across the VMD can be written as:

hf x (Tf - Tf,m) = J x ΔHv (62)

For SGMD, the heat transfer equations, which describe the DCMD can be used.

The heat transfer coefficients of the boundary layers can be estimated by the Nusselt

correlation (see Table 6). Its empirical form is:

Nu = Constant x Rea x Prb (63)

32

Consequently, the heat transfer coefficient h can be calculated using Reynolds and

Prandtl numbers(Re and Pr), i.e.

Reynolds number = Re = (ν x d x ρ)/ μ (64)

Prandtl number = Pr = (cp x μ)/ k (65)

Grashoff number = Gr = (g x β x ΔΤ x L3 x ρ2)/ μ3 (66)

where v, ρ, μ, cp, g, β, L and k are fluid velocity, density, viscosity, heat capacity,

gravity acceleration, thermal expansion coefficient, height and thermal conductivity.

The mass transfer and the heat transfer can be related, by:

Sh x Sc-1/3 = Nu x Pr-1/3 (67)

33

9. Thermal efficiency and energy consumption

The thermal efficiency Π in MD can be specified as the ratio of latent heat of

vaporization to the total (latent and conduction) heat. The thermal efficiency can be

improved by increasing the feed temperature, feed flow rate and membrane

thickness. In contrast, it decreases when the concentration for salt solution

increases.

For DCMD, the thermal efficiency Π can be expressed as:

Π = (J x ΔΗv)/ (J x ΔHv + km/δ x (Tf,m - Tp,m)) (68)

For pure water, the characteristics of the membrane, such as porosity and tortuosity,

determine the thermal efficiency, with no dependence on membrane thickness.

Around 50–80% of the total heat flux across the membrane is considered to be

latent heat; whereas 20–40% of heat is lost by conduction through the membrane.

The heat lost by mass flux can be estimated by:

Qlost/J = km/Cm x (Tf,m - Tp,m)/ (P2-P3) (69)

For a very dilute solution, and low membrane temperature we can use the following:

Qlost/J = km/Cm x 1/ (dP/dT)Tm (70)

Working at a high temperature and flow rate reduces the heat loss.

There are three forms for heat transfer to be lost in the DCMD system. The first form

is due to the presence of air within the membrane. Secondly, heat loss through the

membrane by conduction, and finally by temperature polarization. Solutions to

minimize heat loss in the DCMD, are: de-aeration of the feed solution, increasing the

membrane thickness, creating an air gap between the membrane and the

condensation surface, and operating within a turbulent flow regime.

In terms of AGMD thermal efficiency, suggested that the thermal efficiency is

proportional to the membrane distillation temperature difference. They introduced

two parameters α and β, which can be determined experimentally for an air gap less

than 5 mm, and average membrane distillation temperature, Ta, between 30 °C and

80 °C by:

η = 1 - (α x Tα-2.1)/λ x ((Tf-Tp) x Cp)/( α x Tα

-2.1 + β) + kα/l) (71)

cp and ka are specific heat and air gap thermal conductivity.

34

It is observed that by increasing the feed temperature from 40 °C to 80 °C, the

thermal efficiency increased by 12%, whereas the salt concentration has a marginal

effect on the thermal efficiency.

With regard to energy consumption, if we use a simple energy balance to compute

the energy consumption of hot and cold streams for DCMD and VMD using different

flow configurations,

Q = m x cp x ΔΤ (72)

We can found that the cross-flow configuration is the best, in terms of high flux and

energy consumption. Moreover, hybrid RO/MD becomes the best choice when an

external energy source is available. In addition, heat transfer to the cooling side by

heat conduction, and by heat of condensation can be used (recovered) to preheat

the feed solution, which minimizes the heat requirement and improves the

operation cost. The percentage of heat recovery depends on the heat exchanger

area. It is pointed out that the MD performance rises by 8% when heat recovery is

used. The heat exchanger capacity should be optimized with membrane area, in

order to get high production flux for a solar powered membrane distillation system.

From the economic point of view, the capital cost is very sensitive to heat recovery,

because the heat exchanger is the most expensive item in a solar-powered MD plant.

We must optimized the solar collector area, membrane area and heat recovery to

achieve low capital cost and high flux.

10. Temperature polarization and concentration polarization

Since the vaporization phenomenon occurs at the membrane hot surface and

condensation at the other side of membrane, thermal boundary layers are

established on both sides of the membrane. The temperature difference between

the liquid-vapour interface and the bulk is called temperature polarization, ψ , which

is defined as:

ψ = (Tf,m - Tp,m)/ (Tf - Tp) (73)

Lawson represented ψ with slight difference for VMD as:

ψ = (Tf - Tm,f)/ (Tf - Tp) (74)

The effect of heat transfer boundary layer to total heat transfer resistance of the

system is measured by temperature polarization.

When the thermal boundary layer resistances are reduced, the temperature

difference between the liquid-vapour interface and the bulk temperature becomes

close to each other and, consequently, ψ approaches 1, which means a typical

35

system. On the other side, zero ψ means a high degree of concentration polarization

is taking place, and the system is controlled by large boundary layer resistance.

Usually, the value of ψ lies between 0.4–0.7 for DCMD. It is pointed out that

temperature polarization becomes important at high concentration, high

temperature and low feed velocity.

Concentration polarization, Φ is defined as the increase of solute concentration on

the membrane surface (cm) to the bulk solute concentration(cf):

Φ = cm / cf (75)

In order to estimate the concentration of the solute (mole fraction) on the

membrane surface, the following relation is suggested:

cm = cf x exp(j/(ρ x K)) (76)

where ρ is the liquid density and K is mass transfer coefficient.

Concerning the influence of high concentration on mass transfer coefficient and

distilled flux, the viscosity, density of the feed, solute diffusion coefficient, and the

convective heat transfer coefficient are directly related to the concentration and

temperature. The concentration polarization and fouling must be considered in

modelling, and the permeate flux cannot be predicted by Knudsen, molecular and

Poiseuille flow, because the properties of the boundary layer at the membrane

surface vary from the bulk solution.

11. Fouling

Membrane fouling is a major obstacle in the application of membrane technologies,

as it causes flux to decline. The foulant, e.g., bio-film, precipitations of organic and

inorganic matter, can reduce the permeability of a membrane by clogging the

membrane surface and/or pores. Although membrane distillation is more resistant

to fouling than conventional thermal processes, dosing of anti-scalants can be used

to control scaling. Lower feed temperatures can substantially reduce the influence of

fouling in DCMD.

Since the hydrophobic MD membrane is the barrier between the feed and permeate,

membrane wetting will reduce the rejection of the non-volatiles. Membrane wetting

can occur under the following conditions:

• The hydraulic pressure applied on the surface of the membrane is greater than the

LEP;

• The foulant depositing on the membrane surface can effectively reduce the

hydrophobicity of the membrane, which was generally found in a long-term

36

operation or in treating high-concentration feeds such as for brine crystallisation;

and

• In the presence of high organic content or surfactant in the feed, which can lower

the surface tension of feed solution and/or reduce the hydrophobicity of the

membrane via adsorption and lead to membrane wetting.

The fouling problem is significantly lower than that encountered in conventional

pressure-driven membrane separation. It is pointed out that membrane fouling by

inorganic salt depends on the membrane properties, module geometry, feed

solution characteristic and operating conditions. There are several types of fouling,

which may block the membrane pores. Biological fouling is growth on the surface of

the membrane (by bacteria), and scaling (for the high concentration solution), which

will create an additional layer on the membrane surface, composed of the particles

present in the liquid.

Fouling and scaling lead to blocking the membrane pores, which reduces the

effective membrane, and therefore the permeate flux obviously decreases. These

may also cause a pressure drop, and higher temperature polarization effect. The

deposits formed on the membrane surface leads to the adjacent pores being filled

with feed solution (partial membrane wetting). Moreover, additional thermal

resistance will be created by the fouling layer, which is deposited on the membrane

surface. As a result, the overall heat transfer coefficient is changed. For DCMD at

steady state:

hfx(Tf - Tf,fouling) = kfouling/δfoulingx(Tf,fouling - Tf,m) = km/δx(Tf,m - Tp,m) + JxΔHv =

hpx(Tp,m - Tp) (77)

where kfouling, δfouling and Tf, fouling are the fouling layer thermal conductivity, thickness,

and fouling layer temperature, respectively.

Concerning the effect of high concentration of NaCl and Na2SO4 on the permeate

flux, the flux gradually decreases during the MD process, until the feed

concentration reaches the super saturation point, and then the flux decrease sharply

to zero. Afterwards, the membrane was completely covered by crystal deposits.

When the membrane surface concentration reaches saturation, the properties of the

boundary layer will differ from the bulk solution properties. Currently, pre-treatment

and membrane cleaning are the main techniques to control fouling. Pre-treatment

process increases the product flux by 25%, which means that the pre-treatment

process is important, in order to enhance the permeate flux. Fouling intensity can be

limited by operating at low temperature (feed temperature), and increasing the feed

flow rate.

37

Notes: θf is the angle

between spacer fibres in

the flow direction; lm is the

distance between parallel

spacer fibres; hsp is the

height of the spacer and df

is the diameter of a single

spacer fibre.

12. Operating parameters

In this section, the influence of feed temperature, concentration and air gap will be

reviewed and major findings will be cited and discussed.

12.1 Parameters to Reducing Temperature Polarization

To maximise flux, it is necessary to increase the vapor pressure difference across the

membrane or to reduce temperature polarization. Therefore, it is necessary to

improve the convective heat transfer coefficient for the purpose of producing more

flux. The convective heat transfer coefficient can be expressed as:

αf = - λf/ (Tf - T1) x (dT/dy)boundary (78)

where λf is thermal conductivity of the feed, and (dT/dy)boundary is the temperature

gradient in the thermal boundary layer of the feed. From Equation (10b), it can be

seen that the convective heat transfer coefficient can be improved effectively by

reducing the thickness of the thermal boundary layer. As the thickness of the

thermal boundary layer can be reduced by enhancing the stream turbulence,

increasing flow rate can effectively improve the flux. However, the hydrodynamic

pressure has a square relationship to the flow rate, and the increased pressure will

diminish the effect of increasing turbulence if the membrane is compressible.

The presence of turbulence promoters, e.g., net-like spacers or zigzag spacers shown

schematically in Figure 5 can effectively reduce the thickness of the thermal

boundary layer and improve αf. It is also important that high heat transfer rates are

achieved with a low pressure drop in the channels where the feed solution and

cooling liquid are flowing.

38

From reported data, it is found that the temperature polarization coefficient of

spacer-filled channels falls in the range of 0,9–0,97, in comparison with a

temperature polarization coefficient 0,57–0,76 for flowing channels without spacers.

It is also noticed that the influence of turbulence on flux becomes less at higher

turbulence levels. Therefore, it is necessary to control turbulence within an adequate

range to reduce the energy cost associated with pumping.

12.2 Feed temperature

As MD is driven by vapor pressures which vary exponentially with the stream

temperature, the flux is affected greatly by the feed temperature. Furthermore,

since the heat loss through thermal conduction is linear to the temperature

difference across the membrane as according to Equation (3), the proportion of

energy used for evaporation will increase as the feed temperature increases.

However, an increase of temperature polarization due to the high flux and greater

heat and mass transfer was also observed with rising temperature, but this can be

reduced by using turbulence promotors such as spacers.

As can be seen in Table 7, the feed temperature has a strong influence on the

distilled flux. According to the Antoine equation, the vapour pressure increases

exponentially with temperature. Therefore, the operating temperature has an

exponential effect on the permeate flux.

At constant temperature difference between the hot and the cold fluid, the

permeate flux increases when the temperature of the hot fluid rises, which means

the permeate flux is more dependent of the hot fluid temperature. It is pointed out

that increasing the temperature gradient between the membrane surfaces will affect

the diffusion coefficient positively, which leads to increased vapour flux. Similarly, it

is believed that there is a direct relation between diffusivity and temperature, so

that working at high temperature will increase the mass transfer coefficient across

the membrane. Moreover, temperature polarization decreases with increasing feed

temperature. In terms of coolant temperature, a noticeable change takes place in

39

the permeate flux when the cold side temperature decreases. In addition, more than

double permeate flux can be achieved compared to a solution, at the same

temperature difference. However it was found that the effect of the cold side

temperature on the permeate flux is neglected at fixed hot side temperature,

because of low variation of vapour pressure at low temperatures.

12.3 The concentration and solution feature

There is a significant fall in the flux product when feed concentration increases due

to decreasing vapour pressure and increasing temperature polarization and the

reduction in product flux is linear with time. Furthermore, there is a reduction in the

permeate when the acid concentration increase. About 12% reduction in permeate

flux happened when the feed (NaCl) increased from 0 to 2 Molar concentration. This

decrease in the permeate flux amount is due to the reduction in the water vapour

pressure. Lawson and Lloyd studied the reasons for decreasing product flux when

the concentration of NaCl increases.

They found three reasons for this reduction;

1) water activity, which is a function of temperature, decreases when the

concentration increases

2) the mass transfer coefficient of the boundary layer at the feed side decreases due

to increased influence of concentration polarization, and

3) the heat transfer coefficient decreases as well at the boundary layer, because of

the reduction in the surface membrane temperature.

Therefore, the vapour pressure of the feed declines, which leads to reduced

performance of MD. The viscosity is an important factor in flux reduction. The heat

transfer coefficient decreases due to the reduced Reynolds number. The effect of

thermal conductivity and heat capacity on the flux reduction is negligible.

Furthermore, the impact of density on the flux production is important for salt

solutions. There is variation in the permeate flux with time (see Table 8), and that it

is difficult to calculate the permeate flux using existing models.

40

12.4 Recirculation rate

Table 9 summarizes the effect of recirculation rate. Working at a high recirculation

rate minimizes the boundary layer resistance and maximizes the heat transfer

coefficient.

As a result, higher flux can be achieved. It is indicated that the increasing volumetric

flow rate will enhance the permeate flux. The fluid velocity rises when the

volumetric flow rates increases, so that the convective heat transfer coefficient

develops and the thermal boundary layer thickness decreases. As a result, the

temperature polarization effect reduces. Moreover, there is a significant change in

temperature polarization, when the rate of recirculation changes. This is because the

recirculation rate enhances the heat transfer, which leads to rise in the product flux

and temperature polarization. The effect of the cold side flow rate on the permeate

flux has yet to be decided.

12.5 The air gap

It is suggested that the flux declines linearly with 1/l . Reducing the air gap width will

increase the temperature gradient within the gap, which leads to increased

permeate flux. Table 10 summarizes the air gap effect on the permeate flux.

41

12.6 Membrane type

The membrane permeation flux is proportional to the porosity, and inversely

proportional to the membrane thickness and tortuosity. For a larger pore size

membrane, higher permeate flux is obtained. In addition, higher flux is achieved

using a membrane without support, compared to the same membrane pore size

with support. For a more efficient MD process, low thermal conductivity material

should be used (unsupported membrane).

13. Future Developments and Conclusions

MD has been explored since the early 1960s, but only in the last decade has the

interest grown substantially such that commercial systems are readily available,

backed with pilot trial experience. Various MD providers offer solutions that are

primarily focused on minimizing thermal energy demand, but there are also

possibilities to reduce electrical energy demand.

MD has been used mainly trailed for removing salt for sea water and brackish water.

It has also proven to be a suitable technology for removal of other contaminants,

such as heavy metals, radionuclides, and organics from brackish, produced, industrial

and other impaired water. While it is capable of treating many kinds of water, its

ability to compete with established technologies such as RO, ED, MED and MSF is

currently limited due to its high energy use. Consequently it is likely to find

application where current established technologies are unable to operate or in

applications that substantially favour its use. For instance, the treatment of brine

streams that reverse osmosis finds difficult to treat may be a possible application,

and integration of MD with RO to treat RO brine may be a suitable application where

brine disposal is problematic. Treatment of CSG water brine is one such potential

application, where reduction of brine pond areas has substantial capital cost

benefits.

Similarly, application in industries that have significant low grade waste heat

sources, such as power stations and chemical plants, would also seem to be strong

candidates for application of MD. The high quality of MD permeate compared to RO

permeate may also provide advantages in these applications, particularly if purified

water is required as boiler feed.

Finding suitable applications for MD currently seems to be the major impediment to

its wider commercial use. The theory of its operation is well known, and models are

available to allow design and scale up of MD systems using local heat sources. The

ability to design MD processes using site specific heat flows is critical for its

application, as it is dependent upon waste heat sources to achieve economic

advantages, and the quality and available heat flows from such heat sources will vary

42

from site to site. Efficient designs will be required to take this variability in available

heat in to account.

Low fluxes and wetting have also been limitations for MD implementation. Having

highly permeable membranes and suitable modules with improved hydrodynamics

will allow increased permeate flux and overall performance of the MD process.

Membrane hydrophobicity and pore geometry are critical parameters in reducing

MD membrane wetting, and surface coatings are enabling reduced wetting to be

achieved. For example, oleophobic coatings can reduce wetting and fouling from oily

feeds. Membrane hydrophobicity also determines the largest possible membrane

pore size for scale up as do process parameters such as feed water temperature,

operating pressure, flow rate, and liquid composition.

A large variety of materials has been tested and investigated as MD membranes. It

appears that although morphological features are critical to achieve high flux,

improving the membrane performance is a complex issue involving a number of

parameters. The variety of the MD configurations, membrane morphologies, module

shape and size, as well as the testing conditions cause the large scatter.

As defined by theory, pore size, porosity and thickness of the active layer matter,

and the characteristics of the support are critical to achieving high flux. Controlling

the thermal transfers across the different strata of the membrane is also critical and

more efforts should be focused on improving the interfaces between the active layer

and the supporting layer in order to reduce temperature polarization effects.

Thermal conductivity measurements are often difficult to perform due to the

difficulty in controlling the interfacial contact, but this should be an area of focus for

researchers in order to better understand their structures. Although a few studies

did investigate the long term performance of their membranes, it would also be

interesting to investigate the long term flux and rejection stability of these novel

membranes as very few groups investigated the impact of contaminants, such as

chlorine, or chemical and thermal degradation on the process. MD induces strong

temperature gradients across the membranes, and thermal degradation could occur

over time depending on the composition of the feed. In addition, the compressibility

of the membrane when stressed in the module under the pressure difference will

likely affect flux and energy efficiency, particularly for DCMD.

To date commercial large pore size PTFE flat sheet membranes still show higher

permeance than laboratory fabricated membranes when tested under similar

conditions. A number of routes are open for researchers to improve the

performance of membranes for MD. These routes include fabricating smaller pore

size, but thinner active membrane layers with more hydrophilic materials. The

smaller pore size will then lead to a larger LEP reducing the risk of liquid water

penetration into pores while hydrophilic surfaces may reduce fouling. Research

43

could also, on the other hand, be driven towards the processing of larger pore size

hydrophobic membranes to achieve higher water vapor permeability in order to

become more competitive with commercially available structures. Tuning the

surface energy of the membrane is also critical, and novel approaches combining

hydrophilic and hydrophobic materials have shown highly promising results. Other

routes include the control of the support morphology by introducing large macro

cavities to maximize the liquid water or vapor transport and reducing possible heat

and concentration polarization effects.

Ceramic membranes are possible candidates in place of polymeric membranes in MD

applications due to higher thermal resistance, mechanical strength, chemical

stability and oxidant tolerance. Additional research is required to find optimal

chemical modification candidates as well as optimal procedures to change the

hydrophilic inorganic membranes to hydrophobic membranes without compromising

the performance and permeate flux of the MD process. Nanoparticles are important

emerging candidates to be used in the manufacturing of membranes for MD. They

allow for control of membrane wetting and fouling. Graphene and carbon nanotubes

are the most promising candidates due to their physico-chemical properties, which

help engineering of desired structures and selectivity of the membrane separation

process. Electro-spun webs, which are manufactured as affinity membranes for the

study and growth of biological cells, may open opportunities for research in the area

of membranes for MD.

MD appears to be poised for commercial implementation, and identification of

opportunities that maximise the advantages of MD over competing technologies is

emerging. In developing these opportunities, the energy consumption and desalted

unit cost will decrease; therefore, competitive values with those of other

desalination processes can be reached.

44

Membrane Distillation (MD)

Reverse osmosis (RO)

Ultrafiltration (UF)

Direct Contact Membrane Distillation (DCMD)

Air Gap Membrane Distillation (AGMD)

Sweeping Gas Membrane Distillation (SGMD)

Thermostatic sweeping gas membrane distillation (TSGMD)

Vacuum Membrane Distillation (VMD)

Polytetrafluoroethylene (PTFE)

Polypropylene (PP)

Polyvinylidene fluoride (PVDF)

Liquid entry pressure (LEP)

Scanning Electron Microscopy (SEM)

Atomic Force Microscopy (AFM)

Direct Contact Membrane Distillation (DCMD)

45