Accepted Manuscript

Performance manipulations of a composite membrane of low thermal conduc-

tivity for seawater desalination

Li-Zhi Zhang, Qian-Wen Su

PII: S0009-2509(18)30487-1
DOI: https://doi.org/10.1016/j.ces.2018.07.019
Reference: CES 14373

To appear in: Chemical Engineering Science

Received Date: 4 May 2018

Revised Date: 3 July 2018
Accepted Date: 10 July 2018

Please cite this article as: L-Z. Zhang, Q-W. Su, Performance manipulations of a composite membrane of low thermal
conductivity for seawater desalination, Chemical Engineering Science (2018), doi: https://doi.org/10.1016/j.ces.
2018.07.019

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Performance manipulations of a composite membrane of low thermal

conductivity for seawater desalination

Li-Zhi Zhang1,2*, Qian-Wen Su1

1. Key Laboratory of Enhanced Heat Transfer and Energy Conservation of Education Ministry, School of

Chemistry and Chemical Engineering, South China University of Technology, Guangzhou 510640, China.

2. State Key Laboratory of Subtropical Building Science, South China University of Technology, Guangzhou

510640, China.

Abstract

Membranes of low thermal conductivity (thermal insulation) but high moisture permeability are highly

desired in membrane-based humidification-dehumidification (MHDD) seawater desalination technology. The

porous membranes with thermal insulation could decrease the heat losses in humidification, thus the system

energy performance can be improved. To satisfy this purpose, in this study, a PVDF/SiO 2 (Si aerogel)

composite bilayer membrane of low heat conductivity is prepared and analyzed. To solve the problem of

cracking in forming SiO2 layer, the casting solution (TEOS, H2O) is doped with a drying control chemical

additive (DCCA) N, N-dimethylformamide (DMF) to decrease the surface tensions in the gel. Then

performance analysis of the developed membranes are conducted and the membranes are characterized. The

results found that the heat conductivity is decreased from 0.08199 Wm-1K-1 to 0.04487 Wm-1K-1, with a

penalty of effective moisture diffusivity sacrificed from 4.93×10-6 m2/s to 2.42×10-6 m2/s. They are superior to

the available SiO2 blend insulation membranes. Further, the membranes can be optimized by facile structural

manipulations such as adjusting the SiO2 layer thickness to realize the desired performance while balancing

the paradox between the permeability and heat conductivity. This is the first time that heat conductivity of

membranes can be manipulated for seawater desalination.

Keywords: Seawater desalination; Thermal insulation; PVDF/SiO2 bilayer membrane; Low thermal

conductivity

*
Corresponding author, Tel/fax: 0086-20-87114268; Email: Lzzhang@scut.edu.cn
1
Nomenclature

dh hydrodynamic diameter (nm)

dp pore diameter (nm)

D moisture diffusivity (m2/s)

k mass transfer coefficient (m/s)

m1 substrate layer in membrane

m2 SiO2 aerogel layer in membrane

Nu Nusselt number

Pe′ permeability coefficient (g/(cm2 s cm Hg)

r mass transfer resistance; heat resistance (kW-1m2 K)

L the duct length

Lh the entry length

Re Reunolds number

Rq average roughness (m)

Sh Sherwood number

U dip-coating rate (m/s)

Greek letters

δ thickness (m)

ε porosity

η viscosity (Pa·s)

λ heat conductivity (Wm-1K-1)

ζ the electric potential

ρ density (kg/m3)

τ pore tortuosity

ω humidity ratio (kg moisture/kg air)

Subscripts

2
a air

i inlet

lm logarithmic

m mean, membrane

o outlet, ordinary

p pressure

t total

v vapor

1. Introduction

Sometimes we call our planet “Ocean” instead of “Earth” because over 70% of our earth’s surface is

covered by water. It seems that water is abundant, but the demand for the amount of fresh water available is

urgent (Yamali and Solmus, 2007). The majority of water on earth is seawater, leaving only 2.5% as fresh

water. Moreover, most of this fresh water exists in frozen areas such as Antarctica and Greenland. And there is

only about 1% fresh water accessible to direct human uses (Service, 2006). With the world’s population

growing rapidly, the quantity of water for humanity and other living beings is also increasing (Elimelech and

Phillip, 2011).

How can the growing demand for fresh water be met? Generally, desalination can be a solution to

overcome fresh water shortage. There are several technologies to achieve desalination (Banat et al., 2007;

El-Dessouky et al., 2004; Geng et al., 2014; Hassan et al., 1998; Matz and Zimerman, 1985; Mohamed et al.,

2006; Ortiz et al., 2006; Qtaishat and Banat, 2013): (1) Multi Stage Flash and Multi-Effect Distillation (MED).

(2) Reverse Osmosis (RO), Nano Filtration (NF) and Electro Dialysis (ED). (3) Membrane Distillation.

Among these technologies for seawater desalination, the recently announced membrane-based

humidification-dehumidification desalination (MHDD) is ideal because it can be operated at temperature

below 80 °C and atmospheric pressure. Therefore, it can be driven by low-grade energy such as industrial

waste heat and solar energy (Chen and Ho, 2010; Guillén-Burrieza et al., 2011; Zhang and Li, 2017). Based on

a humidification-dehumidification process, fresh water from seawater can be obtained (Li and Zhang, 2016).

However, a lot of thermal energy is lost during the membrane humidification processes, resulting in a low

energy COP of around 0.4 for MHDD (Zhang and Li, 2017). Then decreasing the energy losses in membrane

3
humidification processes becomes the key issue for performance improvement. The lost energy mostly comes

from the vapor through porous membrane by heat conduction, so membranes of low heat conductivity are

desired. Therefore, the issue of membranes with thermal insulation has come under the scientific spotlight (Tu

et al., 2010).

Generally, there are several characteristics for membranes to be suitable for MHDD process: (1) The

membranes must be hydrophobic (non-wetting) and only vapor exists in the pores. The membrane pore size

used in MHDD is between 5 nm and 1μm (El-Bourawi et al., 2006). (2) The membranes used in the MHDD

should have high mass transfer coefficient but low thermal conductivity to reduce heat losses as much as

possible. (3) The membranes should have good anti-fouling properties to prevent being blocked and reduced

permeability (Alkhudhiri et al., 2012).

Membranes of thermal insulation are also required in other membrane distillation operations. There are

varieties of membrane distillation. They are: (1) Direct Contact Membrane Distillation (DCMD). DCMD is

widely used in desalination processes (Alves and Coelhoso, 2006; Calabro et al., 1994; Godino et al., 1997;

Hsu et al., 2002; Tomaszewska et al., 1995). But its main disadvantages are the large heat losses by

conduction. (2) Air Gap Membrane Distillation (AGMD). AGMD is efficient to reduce heat losses in

desalination (Banat and Simandl, 1998). However, the mass transfer resistant is increased by static air layer. (3)

Sweeping Gas Membrane Distillation (SGMD). SGMD is often used to remove volatile organic compounds

from solution. But it is high energy-consuming to collect the distillate (Banat and Simandl, 1998). (4) Vacuum

membrane Distillation (VMD). VMD is used to separate aqueous volatile solutions. Although the conduction

heat loss can be neglected, it is only suitable for low-concentration solutions due to its complexity (Wirth and

Cabassud, 2002; Yun et al., 2006). Generally, thicker membranes may be an option to decrease heat

conduction. However, there are conflicts between low mass transfer resistance for thinner membranes and

high heat resistance for thicker membranes (Qtaishat and Banat, 2013). Therefore, membranes of low thermal

conductivity are the ultimate alternatives to meet these requirements.

Generally, there has existed one method to develop porous membranes of low thermal conductivity. They

are blend membranes doped with insulation materials. Chen et al. (Chen et al., 2014) fabricated membranes

from three different molecular weights PVDF via the phase inversion process to reduce thermal conductivity.

A porous blend membrane was prepared by Wang and Chung (Wang and Chung, 2015) using hydrophobic

clay particles as addititives. Efome et al. (Efome et al., 2015) fabricated hydrophobic PVDF-nanoparticles

4
membranes via adding superhydrophobic SiO2 nanoparticles into the dope. According to these methods, only

some parts of the matrixes were replaced by materials of low heat conductivity, while the structures of such

membranes were not changed. The mean free path of dry air (25 °C, 1.013×105 Pa) is 70 nm (Jennings, 1988),

so it is effective to decrease heat conduction by limiting the molecular oscillations if the pore sizes are less

than 70 nm. Regretfully the pore sizes of most of these membranes are above 100 nm, so the insulation effects

are quite limited.

So far, researches on the preparation of silica aerogel porous films for heat insulation have been rarely

reported. Considering this, this study adopts a dip-coating procedure to fabricate a composite membrane rather

than a blend membrane. First, a porous PVDF membrane is developed as the support layer. Then an insulation

layer made of silica aerogel film is directly fabricated on the support layer. The PVDF support layer is

mechanically strong but highly conductive. The silica aerogel layer is relatively fragile but it has good

thermally insulation properties. So the combination of the two layers would develop a new membrane of low

heat conductivity. The difficulty to make aerogel films lies in its easy cracking during ambient drying

processes. The first sol-gel coating procedure to prepare meso-porous silica films was reported by Brinker and

co-workers (Brinker and Scherer, 1990). However, the obtained films were mostly granular with

two-dimensional hexagonal structures. When the substrate layer is porous, intermixing of the two layers

would occur, leading to poor mass transfer. To solve this problem, Stucky and co-workers (Zhao et al., 1998)

used low-cost commercially available PEO-PPO-PEO triblock copolymers mixing with poly-(ethylene oxide)

non-ionic surfactants via dip-coat processing to fabricate continuous channel or ordered cage arrays films.

Here we introduce a drying control chemical additive (DCCA) N, N-dimethylformamide (DMF) into the sol

systems (TEOS, H2O) to solve the problem of cracking. Although several studies (Darmawan et al., 2016;

Lenza et al., 2015) have dealt with the influences of adding DMF on the structure of SiO2 films of micro pores,

few works investigate the effects on aerogel films which have nano pores. In this paper, the effects of DMF on

gel time, the morphology and the structures of the composite aerogel films are discussed. These films with the

addition of DMF have a less-defined pore structure than those prepared in the literature. They have a

three-dimensional network structure and a wider pore size distribution between 5 nm and 50 nm. This kind of

structure permits high property of thermal-insulation. They also permit the adjustment of pore sizes to

manipulate performances. When applied to seawater desalination, the thickness and structures of different

layers can be manipulated to balance the paradox between permeability and heat conductivity.

5
2. Experimental

2.1. Materials

Tetraethoxysilica (≥98%, TEOS), ethanol (≥99.7%, EtOH), hydrochloric acid (34.46wt%, HCl),

ammonia hydroxide (25wt%, NH4OH), dimethylformamide (DMF, reagent grade), trimethyl chlorosilane

(TMCS, reagent grade) and acetyl acetone (AcAc, reagent grade) were obtained from Guangzhou Qianhui Co.,

China. Poly (vinylidene fluoride) (PVDF, Mw=700000) and polyethylene glycol 2000 (PEG2000) were

purchased from Aladdin, Shanghai, China. SiO2 aerogel particles (pore size 5~40 nm, porosity ≥90%, grain

size 40~80 nm, hydrophobic) were bought from Xiamen Nameite New Materials Technology Co., Ltd, China.

Deionized water (DW) was obtained from a water purification system.

2.2. Membrane fabrication processes

2.2.1. Preparation of PVDF membrane

In this study PVDF hydrophobic membranes were prepared by a direct wet phase inversion method.

PVDF and PEG2000 were added into DMF with a specified mass ratio (PVDF/DMF/PEG2000, 9: 88.2: 2.8),

and the mixture was stirred at 55 °C for 8 h on a heating magnetic stirrer until fully dissolved. After the

solution was cooled down, the membrane preparation was the same as our previous publication (Zhang et al.,

2011). The prepared PVDF membrane was dried for later use.

2.2.2. Preparation of PVDF-SiO2 hybrid membrane

PVDF-SiO2 hybrid membranes were made by the following method. 2 g hydrophobic SiO2 aerogel

powder and 2.8 g PEG2000 were mixed in DMF and dispersed ultrasonically for 2 h. Subsequently, PVDF

was added to form the casting solution. After that, the membrane preparation was the same as the section

2.2.1.

2.2.3. Preparation of PVDF/SiO2 composite bilayer membrane

The PVDF/SiO2 composite bilayer membrane was prepared by two steps. Firstly, the fabrication of the

supporting PVDF layer was the same as that described in the section 2.2.1. Secondly, silica aerogel films were

dip-coated on PVDF surfaces.

6
 Dip-coating steps were shown in Fig. 1. They were: (1) the fabrication of silica sols (SS) by sol-gel

methods. TEOS, EtOH, DMF, deionized water and HCl were mixed (1: 5: 0.8: 1: 1×10-5, molar ratio) and

magnetically stirred for 60 min. Secondly, EtOH, H2O and NH4OH were added into the sol solution and

magnetically stirred for 30 min. The final molar ratio of TEOS: EtOH: H2O: DMF: HCl: NH4OH was 1: 7: 2:

0.8: 10-5: 3.57×10-5. (2) Preparation of SS with appropriate viscosity. When the viscosity of the silica sols

reached 10×10-3 Pa·s, EtOH was added with a volume ratio of SS: EtOH=1: 3. Then acetyl acetone (AcAc)

was added to maintain a constant viscosity of 5~7×10-3 Pa·s. Table 1 showed the effects of different viscosity

and dip-coating speed on film quality. (3) Fabrication of silica aerogel films. The PVDF layer was immersed

in SS at a dip-coating rate of 10 cm/min for different cycles. To ensure films only on one side, SS on the back

side of PVDF were wiped off. Then they were quickly dropped into saturated EtOH solutions. In order to

prevent abrupt solvent evaporation, the coated wet gels were first immersed in EtOH: H2O (7: 3, vol.) with a

moderate NH4OH for 24 h at 60 °C and then in EtOH: TEOS (7: 3, vol.) for 16 h at 60 °C to strengthen the gel

network. Continuously, the water and ethanol were exchanged with n-hexane repeatedly until solvents were

replaced completely. After surface modification for 8 h at 50 °C in TMCS: n-hexane (1: 9, vol.), the wet gel

films were washed in n-hexane repeatedly. Finally, the wet films were dried by slow heating at 60 °C, 80 °C,

120 °C, 180 °C for 2 h respectively. The resultant membranes were PVDF/SiO2 composite bilayer

membranes.

Table 1

The effects of different viscosity and dip-coating speed on film quality.

Sol viscosity Dip-coating rate Results from dip-coating in an EtOH

(Pa·s) (cm/min) atmosphere

5~7×10-3 5 breakage

5~7×10-3 10 Uniform and continuous film

5~7×10-3 20 breakage

>7×10-3 5 breakage

>7×10-3 10 breakage

>7×10-3 20 breakage

7
 Mixing of TEOS and EtOH using DMF as DCCA

HCl catalyst NH4OH catalyst

Dip-coating SiO2 sols of appropriate viscosity (5~7×10-3 Pa·s)

Aging of wet gel film to strengthen the structure

Solvent exchange with n-hexane

repeatedly

Surface modification using TMCS

Ambient drying by slow heating

Fig. 1. Fabrication of silica aerogel films.

Here, dip-coating rate referred to the speed of pulling out PVDF substrates from SS. Neither too fast nor

too slow dip-coating rates could lead to uniform films. As a result, to fabricate optimal silica films, the falling

speed was set as 8 cm/min and the pulling rate (dip-coating rate) was 10 cm/min. A constant viscosity of

5~7×10-3 Pa·s was applied in this study.

2.3. Characterization of the membranes

2.3.1. Membrane thickness

The PVDF membrane thickness was measured by an electronic digital indicator (0~25 mm, 0.01 mm,

Beijing Huafeng Science and Technology Co., Ltd, China). Each membrane was measured five times and the

average values was used.

Thicknesses of SiO2 aerogel films were given by the literature (Brinker et al., 1992):
η (1)
ρ
Where δ was film thickness (m), c was constant, η was sol viscosity (Pa·s), and U was dip-coating rate

(m/s). According to the equation, thickness of aerogel films was 100 nm at a dip-coating rate of 10 cm/min.

8
Under these conditions, thicknesses of silica thin films with multi-coating cycles were calculated. For

convenience, the different membranes were marked as M1, M2, M3, M4, M5, M6 and M7, respectively. They

were listed in Table 2. M5 was validated by SEM observations of the membrane cross sections. It was

consistent with the calculated thickness.

Table 2

The parameters of different films.

Membranes Composition Thickness Process

M1 PVDF 150 μm Direct wet phase inversion methods

M2 PVDF, SiO2 aerogel 150 μm Direct wet phase inversion methods

power

M3 PVDF, TEOS 152 μm (1) PVDF layer by direct wet phase inversion

methods; (2) SiO2 aerogel layer by sol-gel methods;

(3) 20 dip-coating cycles

M4 PVDF, TEOS 154 μm (1) (2) the same as above; (3) 40 dip-coating cycles

M5 PVDF, TEOS 156 μm (1) (2) the same as above; (3) 60 dip-coating cycles

M6 PVDF, TEOS 158 μm (1) (2) the same as above; (3) 80 dip-coating cycles

M7 PVDF, TEOS 160 μm (1) (2) the same as above; (3) 100 dip-coating cycles

2.3.2. Characterization techniques

The contact angle of these membranes was measured by a JC2000Cl contact angle system (Shanghai,

China) with a 4 μL water droplet. Each sample was measured three times and the average contact angle was

used. The surface topology of membrane was characterized by Atomic force microscopy (AFM, XE-100, Park,

Korea). The scan size was 5 μm×5 μm. Scanning electron microscopy (SEM, Merlin, LEO1530VP, Germany;

JEM-2100F, Japan) was used to observe the membrane morphologies. A polycrystall X-raydiffraction (XRD,

Bruker D8 advance, Germany) was performed to analyze the membrane crystalline microstructure. The θ

angular range was selected between 10° and 90°. Infrared spectroscopies of the membranes were obtained in

KBr pellets using a Bruker Hyperion FT-IR spectrometer. The parameters of SiO2 aerogel layer were

characterized by using a Micromeritics ASAP 2020 automated system. The parameters of PVDF layer were

9
characterized by using a mercury porosimeter (Micrometrics Autopore 9500) based on the Carman-Kozeny

theory (Kaviany, 1991). The thermal conductivities of membranes were measured by the thin film module of

the Hot Disk Thermal Constants Analyzer (hot disk, TPS 2500S, Sweden). The TPS 2500S system was based

on a specially designed Wheatstone bridge with the Hot Disk sensor in one of the arms.

2.3.3. Moisture exchange tests

The membranes were designed for seawater desalination. Air humidification were realized with these

porous membranes. The whole set of device had been set up and used for research in our previous paper

(Zhang et al., 2011). The mass exchange rates through membranes (isothermal moisture exchange between

two air flows) were measured by the apparatus. The schematic diagram is depicted in Fig. 2. The width and

length of the flow channel were 2.5 cm and 7 cm, respectively. Channel pitch was 2 mm. Therefore, the

exchange area through membrane was 17.5 cm2. As diagrammed in Fig. 3, stream 1 and 2 represented humid

air and dry air, respectively. Absolute moisture content difference between the two streams was about 8.14

g/kg, and the vapor pressure difference was 1264 Pa. Five flow rates from 0.05 m3 to 0.26 m3 were tested for

each membrane. In addition, the differences between the sum of moisture content to and from the exchanger

were checked to be within ±2%. The uncertainties were calculated in our previous paper (Zhang, 2006).

Membrane
Stream 2 out Stream 2 in

Stream 1 in Stream 1 out

Fig. 2. The schematic diagram of the membrane exchanger.

10
 Valve
Temperature and
By pass Humidity Sensor
Dehumidifier
Stream 2 in Stream 1 out

Pump

Flow Meter
and Controller

Stream 1 in Heat Mass

Exchanger Stream 2 out

Fig. 3. The test equipment for moisture exchange.

The total resistance in the test apparatus includes the membrane part itself and the convective resistance

on both sides of the membrane. The membrane part will not han e whether it’s in a liquid-gas system or in a

gas-gas system, as long as the correct boundary layer resistance is used. In the test, the gas –membrane

convective resistance is clarified and subtracted from the total resistance to obtain the membrane side

resistance. In future liquid-membrane-gas systems, like desalination, the overall resistance can be easily

calculated by the correct liquid side resistance, the previously obtained membrane resistance and the gas side

resistance.

The duct is a rectangular duct (2.57cm). The pitch is 2mm. When the duct length L>> Lh (the entry

length), the flow can be considered as fully developed. But in a short length channel, entrance effect should be

considered. For adequacy, care must be taken to ensure that fully developed flow prevails over most of the

channel. In the present design, the boundary layers on both sides in laminar flow will merge at a point

downstream. In rectangular ducts, the entrance length is given by (Muzychka and Yovanovich, 2009):
h . h (2)

Table 3
Changes of entrance length with Reynolds number.
V (m3/s) Re Lh (cm) Lh/L
0.05 68.06 0.28 3.96%
0.1 136.13 0.55 7.91%
0.16 217.81 0.89 12.66%
0.2 272.26 1.11 15.83%
0.26 353.94 1.44 20.58%

11
 From table 3, the range of Re varies from 68.04 to 353.94, while the entry length is from 0.28 cm to 1.44

cm. When the flow velocity is below 0.26 m3/h, the ratio of entry length to whole channel is below 21%.

Under flow rates of 0.05m3/s, the ratio is below 5%. Based on this situation, it is reasonable to neglect the

entry effect and calculate the mass transfer coefficient as fully developed flow.

In this study, the duct length is 7 cm, so it can be regarded as fully developed flow. For fully developed

laminar flow, Sh or Nu is a constant. In the measurement, to minimize uncertainties, isothermal operation is

performed. The two flows have the same temperature at inlets, so the influences from temperature can be

neglected. Further, the two flows are in a counter flow arrangement, under such conditions, a constant mass

flux boundary condition can be assumed (Pérez-Lombard et al., 2008).

3. Results and discussion

3.1 Effects of DMF on aerogel films

Effects of DMF on the gel time and viscosity of silicon sol

When the silica aerogel film is prepared by ambient drying, the most common problem is cracking. The

crack arises from the unbalanced capillary forces in pores, which are caused by the contraction of the network

structure of gels during drying. Several measures are taken to prevent cracking. By modifying the –OH groups

into non-active groups –CH3 with TMCS (Deshpande et al., 1992), the shrinkage of the gel network and

cracking are hindered effectively. Simultaneously, Using DMF as drying control chemical additives (DCCA)

is another method to further prevent cracking. In order to obtain appropriate molar ratios of DMF to TEOS,

the effects of DMF on the gel time and viscosity of the silicon sol are measured in Fig. 4.

12
 20 30
18 28
16 26
14 24

viscosity (×10-3 Pa·s)

gelation time (h)

12 22
10 20
8 18
6 16
4 14
2 12
0 10
0.0 0.2 0.4 0.6 0.8 1.0
nDMF/nTEOS

Fig. 4. Effects of molar ratios of DMF to TEOS on the gelation time and viscosity of silicon sol.

Fig. 4 shows the effects of molar ratio of DMF to TEOS on the gelation time and viscosity of silicon sol.

With an increase of DMF, the gelation time is obviously extended. It is resulted from hydrogen bonding

interaction with H+ in the sol. Combined with the hydrolyzed intermediate to reduce the H+ content absorbed

on the sol particles, the electric potential (ζ) of the micelles is increased. So the barrier for the sol particles to

agglomerate would increase (Adachi and Sakka, 1988). Hence, the gel reaction slows down and the stability

of the sol is improved. Simultaneously, after 20 h of aging, it is observed that the larger the molar ratios of

DMF to TEOS are, the smaller the sol viscosity is. As a polar solvent, it is easy for DMF to have an affinity to

EtOH. Therefore, strong hydration between ethanol and water is reduced and the sol system contains more

free water. Consequently, the condensation polymerization of the sol system is limited and the sol viscosity is

decreased. Too high or too low viscosity is bad for the coating. In order to fabricate a uniform film, the

content of DMF should be controlled within the appropriate range. When nDMF/nTEOS is up to 0.8, the gelation

time basically remains stable. Therefore, the molar ratio of DMF to TEOS is chosen as 0.8 in this study.

Effects of DMF on the structure of aerogel film

13
 (a) (b)

200 nm 200 nm

Fig. 5. The films before and after doping DMF. (a) before; (b) after.

To discuss the effects of DMF on the structure of the aerogel film, the films supported on Si wafers

before and after doping DMF were prepared. Fig. 5 shows the surface morphology of the films before and

after doping DMF. From Fig. 5 (a), the silica particles are stacked irregularly and the film surface is coarse.

After adding DMF, films of uniform pore sizes and high porosity are formed in Fig. 5 (b). This shows that

DMF can prevent gel breakage effectively and inhibit the formation of particle clusters. This is due to the

effects of DMF on silica colloidal particles. The mechanism is depicted in Fig. 6. In the gel process, when

combined with the hydrolyzed intermediate, the activities of the intermediate Si(OH)xOR4-x and the product

SixOy(OH)z are suppressed. The active participation of DMF in steric shielding around Si causes a decrease in

the colloid particle sizes. Therefore, it is effective on the promotion of polymerization and the inhibition of

coalescence. Meanwhile, because the growth of sol particles is inhibited, the pore diameter and pore volume

increase and the gel shrinkage during ambient drying decreases. Consequently aerogel films with a uniform

pore-size distribution is obtained.

Fig. 6. Mechanism of action of DMF on silicon sol.

3.2. Surface wettability

14
 The surface contact angles (SCA) were measured to describe the hydrophobicity of the modified PVDF

membranes. The values are listed in Table 4.

Table 4

The data of surface contact angle.

Samples M1 M2 M3 M4 M5 M6 M7

SCA(°) 76.2 92.4 81.2 94.5 115.7 93.9 73.6

The relationship between the contact angle of membranes and wettability is given by Young's equation

(Zisman, 1964). Wetting of the surface is very favorable (hydrophilic) when a contact angle is less than 90°;

wetting of the surface is unfavorable when a contact angle is greater than 90°.

For the PVDF membranes (M1), the formed macro-porous membrane is hydrophilic with a contact angle

of 76.2° due to -OH groups of PEG2000. However, the PVDF-SiO2 hybrid blend membrane (M2) becomes

hydrophobic with the addition of hydrophobic SiO2 nanoparticles. For the PVDF/SiO2 composite bilayer

membranes (M3-M7), it is observed that the SCA changes with different number of dip-coating cycles. For

M3, because the dipped silica sols are re-dissolved in EtOH, silica films are not formed. So M3 is hydrophilic.

Membranes M4~M6 are hydrophobic. The reason lies in the methylation reaction with TMCS. The more the

solution is coated on the surface, the more complete the silica films are formed. Therefore, M5 has the largest

contact angle of 115.7°, which helps to prevent the penetration of aqueous solution on both sides of the

membrane. After dip-coated 80 cycles, more coated solution would have adverse effects. For M7, because the

coating process is too long, during which a rapid solvent evaporation makes the film cracked. So it becomes

hydrophilic again.

3.3. Membrane morphology analysis

3.3.1. Surface morphological studies

AFM imaging is performed to assess the membrane uniformity on top surfaces. The scan size was 5

μm×5 μm. The mean roughness (Rq) is used to indicate the roughness of the membrane surfaces, which is

listed in Table 5.

15
Table 5

Surface roughness of membranes (s an size 5 μm×5 μm).

Samples M1 M2 M3 M4 M5 M6 M7

Rq(nm) 28.3 79.1 23.9 90.5 135 107 38.4

(a) (b) (c)

Fig. 7. The AFM images of different membranes: (a) M1; (b) M2; (c) M5.

The relationship between contact angle and surface morphology is described by the Wenzel equation

(Wenzel, 1936). According to this model, if the surface is chemically hydrophilic, it will become even more

hydrophilic when surface roughness is added. If the surface is chemically hydrophobic, it will become even

more hydrophobic when surface roughness is added. From Fig. 7 (a), M1 has good roughness, which is in

good agreement with the following discussed SEM image. Fig. 7 (b) shows that coarse SiO2 particles increase

the roughness of the PVDF-SiO2 hybrid membrane.

When dip-coating silica sols, the roughness of membranes increases first and decreases with the number

of dip-coating cycles. Since the PVDF membrane is coated by silica sols, it can be considered as a new

surface. For M3, the silica aerogel film is not formed and SiO2 nanoparticles adhere to the surface of PVDF,

causing pore blocking observed by SEM. Therefore, the roughness of M3 is lower than M1. The silica layer

contains many –OH groups and only a small amount of TMCS participates in the hydroxyl reaction with SiO2

nanoparticles. For M4~M6, the membrane surfaces are silica films with network structure as shown in SEM.

Therefore, the liquids can be well connected together on the surface protrusions so that the water droplets

adhere to the solid and air gaps explained by the Cassie-Baxter model (Cassie and Baxter, 1944). This means

that the increase in roughness reduces the contact area between the liquid and membrane surface. Therefore,

the roughness of M5 is the largest, with the largest contact angle as shown in SCA. The tapping-mode AFM

16
images of M5 is shown in Fig. 7 (c). For M7, the aerogel skeleton collapses due to fast solvent evaporation. A

large number of SiO2 nanoparticles stack together, so M7 has a larger roughness than M1. The silica layer

contains many –OH groups and it hardly participates in the hydroxyl reaction. All the results are consistent

with surface wettability.

3.3.2. Formation of PVDF/SiO2 composite bilayer membrane

Surface structures of the PVDF membranes are macro-porous with pore diameters from 200 nm ~ 1 μm

as indicated in Fig. 8 (a). However, in Fig. 8 (b) it can be seen that PVDF-SiO2 hybrid membrane has changed

from smooth to rough after 2.0% of hydrophobic SiO2 aerogel power is added. As verified by Fig. 8 (b), the

nanoparticles on the membrane surface are caused by agglomeration, which then results in pores reduction.

According to Fig. 8 (c)-(g), it is concluded that the PVDF/SiO2 composite bilayer membranes have different

pore structures. When the dip-coating cycles are limited, the silica sol on the surface is re-dissolved in EtOH.

So silica films are not formed, as shown in Fig. 8 (c). The more the solution is coated on the surface, the more

complete the silica films are formed. Especially in Fig. 8 (e), complete silica aerogel films with high porosity

are fabricated. The silica layer has network structure with an average pore diameter of about 10 nm. The

PVDF layer has macro-porous structure with a pore diameter of about 200 nm~1 μm. When performed with

more dip-coating cycles, the solvent evaporates rapidly. So silica aerogel film is cracked, as Fig. 8 (g) shows.

(a) (b) (c)

10 μm 2 μm 2 μm

(d-2) (e-2) (f-2)

(d-1) (e-1) (f-1)

2μm 2μm 2μm

为 为， 为
μ
，μ
500 nm 500 nm ，μ
500 nm

(g)

17
 1 μm

Fig. 8. The SEM graphs of membranes: (a~g) M1~M7. Subscripts 1 and 2 represent the same kind of

membranes but with different observation magnification, respectively.

SiO2 layer 6 μm

PVDF layer

20 μm

Fig. 9. SEM cross-sections photographs of M5 membrane with 20 μm amplification.

Fig. 9 shows cross-section of the membrane for M5. As seen, the meso-porous silica aerogel layer is

formed on the surface of PVDF layer. Continuous silica aerogel films are supported by PVDF layers to avoid

easy fragility. Silica aerogel is famous for its friability owing to porous nanostructure (Du et al., 2013). Herein,

the obtained PVDF/Si aerogel film has a moderate bend without cracking. The thickness of silica aerogel film

is 6 μm, which is consistent with that estimated by Eq. (1). It can be clearly seen that the PVDF layer is not

affected by the top layer and it has connected pores. The water vapor can permeate through these pores

efficiently.

3.4. Surface elemental composition analysis

Wide-angle X-ray diffractions in Fig. 10 are performed to study the membrane crystalline

microstructure. The crystal structure of two layers can be obtained.

18
 M1
12000 M2
M3
10000 M4
Intensity (counts) M5
8000 M6
M7
6000

4000

2000

10 20 30 40 50 60 70 80 90
2-theta
Fig. 10. The XRD patterns for different membranes (M1, M2, M3, M4, M5, M6 and M7).

It can be seen that M1 has the two dominant peaks at 18.3°, 20.08° assigned to α(0 1 0), α(1 1 0),

respectively. It is well known that these diffraction peaks are characteristic of α-phase structure (Kim et al.,

2002). This means that PVDF membrane contains mainly α-crystal. When adding SiO2, XRD analysis shows

that the crystalline structure of M2 is not much changed, indicating there is no interaction between PVDF and

SiO2.

When silica sols are dip-coated on the surface of PVDF membrane, the diffraction peaks are not changed

obviously. For M3~M7, no obvious diffraction peaks of SiO2 aerogel films are obtained and the peaks of M5

at 24.30° are amorphous. The main reason is that SiO2 aerogel films have porous network structures and hence

no complete crystal structures exist. This can be seen more clearly in the next Fourier transform infrared

spectroscopy (FTIR).

19
 3.6 M1
M2
3.2 M3
M4
2.8 M5
M6
Transmittance(%)

2915
840
2.4 M7
1232

2.0

1.6

1.2

0.8

3500 3000 2500 2000 1500 1000

-1
wavenumber/cm

Fig. 11. FTIR patterns of the membranes (M1, M2, M3, M4, M5, M6 and M7).

The infrared spectra is plotted in Fig. 11. We can see the changes of the chemical groups before and after

the dip-coating process of the silica films. Table 6 lists the wavenumbers of the FTIR peaks corresponded to

different groups and crystallites. There are main conclusions: (1) Confirmation of the crystalline structures

obtained by XRD; (2) Confirmation of the Si-O-Si and Si-CH3 groups (Yu et al., 2009). It can be obtained that

pure PVDF membrane contains mainly α-crystal. When adding SiO2, the PVDF-SiO2 hybrid membrane still

contains mainly α-crystal and it has no obvious asymmetrical Si-O-Si stretching. This may be explained that

drying in a oven leads to the collapse of SiO2 porous structure. When dip-coating with silica sols as listed in

Table 6, only M4~M6 has asymmetrical Si-O-Si stretching Si-CH3 stretching. It shows that only appropriate

cycles facilitate the formation of silica thin films.

Meanwhile, M1~M7 have β-phase characteristic peaks at 840 cm-1. It proves that the PVDF copolymer

possesses predominantly α-crystal crystalline structure and slightly β-crystal crystalline structure. Moreover, it

can be concluded that the crystalline structure of PVDF is not changed by dip-coating silica sols, indicating

that two layers have no chemical bonds and just stack together.

Table 6

FTIR peak assignments for membranes.

20
 Wavenumbers (cm-1) Functional groups and crystallites

3500-3000 -OH stretching

2915 (M4, M5, M6) Si-CH3 stretching

1640 H-O-H blending

1404 C-F deformation

1232 (M4, M5, M6) Asymmetrical Si-O-Si stretching

1170 C-F stretching

1072, 976, 876, 760, 614 (M1~ M7) α-Crystal of PVDF

840 (M1~M7) β-Crystal of PVDF

3.5. Distribution of pore diameters of PVDF layer and SiO2 aerogel layer

To analyze the effects of pore size distribution on moisture transfer and heat conductivity, the structures

of M5 are measured. The two layers of M5 are made separately to measure the parameters. The distribution of

pore sizes of PVDF layer and SiO2 aerogel layer are characterized and plotted in Fig. 12.

1.2 PVDF layer SiO2 aerogel layer

porosity: 77.76%
porosity: 90.18%
0.02
Pore volume dV/dlogdp (cm3/g)

1.0
Pore volume dV/dD (cm3/g/nm)

0.8

0.6
0.01
0.4

0.2

0.0
0.00
1 10 100 1000 10000 1 10 100
Pore diameter (nm) Pore diameter (nm)

(a) PVDF layer (b) SiO2 aerogel layer

Fig. 12. Distributions of pore diameters of (a) PVDF layer and (b) SiO2 aerogel layer.

Fig. 12 describes the distribution of pore sizes of the two layers. The PVDF layer is macro-porous with a

porosity of 77.76% and a mean pore size of 511.5 nm, allowing high vapor permeation. The SiO2 aerogel

layer is meso-porous with a porosity of 90.18% and a mean pore size of 6.6 nm. This small pore size can

effectively decrease heat conduction by limiting the molecular oscillations. The PVDF/Si aerogel composite

bilayer membrane can be manipulated to balance the paradox between permeability and heat conductivity.

21
3.6 Moisture transfer and heat conductivity

3.6.1 Vapor permeation

Heat and moisture transfer through membranes have been studied for several years in our laboratory and

they provide the theoretical basis for performance analysis (Zhang, 2006, 2008; Zhang and Li, 2017; Zhang et

al., 2011). Under constant mass flux boundary conditions, mass transfer coefficient ka is calculated by

equations in our previous study (Zhang et al., 2011).

a (3)
a
h

For laminar flow, Nu and Sh are constant values. So the air side mass transfer coefficient ka is calculated

as 0.049 m/s by Eq. (3).

The total moisture transfer coefficient kt, the mean exchanged humidity between the two air flows Δωm

(kg/kg), the logarithmic mean humidity difference Δωlm, and the membrane mass transfer coefficient km are

calculated by equations in our previous literature (Zhang et al., 2011).

Permeability coefficient Pe′ (g/(cm2 s cmHg)) ia a parameter to reflect moisture permeability. It is

calculated by

(4)
5
Moisture transfer resistance (m-1s) in the bilayer membrane composite membrane comprises two layers:

(5)
m m m

The PVDF membrane mass transfer coefficient km is 0.03289 m/s (the measurement from M1). When air

side resistance ka is constant, membrane side resistance km is the key factor for moisture transfer. For all the

composite membranes, the processes to prepare the porous PVDF membrane are the same. It can be assumed

that the mass transfer coefficient of PVDF layer km1 is fixed to 0.03289 m/s. Therefore, the mass transfer

coefficient of silica aerogel layer km2 can be calculated.

The equivalent diffusion efficient of moisture in the membrane Dvm (m2/s) is calculated by:
δ (6)
m m

where δi is membrane thickness in ith layer (m) and Dvmi (m2/s) is the diffusion efficient of moisture in ith

layer, i=1-2. Mass transfer coefficient and moisture diffusivity in each part are listed in the Table 7.

Table 7
22
 Mass transfer coefficient in each part and the total mass transfer coefficient for every membrane.

Sample δm1 km1 Dvm1 δm2 km2 Dvm2 ka km Dvm Pe'

(μm) (m/s) (m2/s) (μm) (m/s) (m2/s) (m/s) (m/s) (m2/s) (g/(cm2s cmHg))

M1 150 0.03289 4.93×10-6 − − − 0.049 0.03289 4.93×10-6 2.43×10-5

M2 150 0.01733 2.60×10-6 − − − 0.049 0.01733 2.60×10-6 1.28×10-5

M3 150 0.03289 4.93×10-6 2 0.02079 4.16×10-8 0.049 0.01274 1.94×10-6 0.94×10-5

M4 150 0.03289 4.93×10-6 4 0.02486 9.94×10-8 0.049 0.01416 2.18×10-6 1.05×10-5

M5 150 0.03289 4.93×10-6 6 0.02947 1.77×10-7 0.049 0.01554 2.42×10-6 1.15×10-5

M6 150 0.03289 4.93×10-6 8 0.01956 1.56×10-7 0.049 0.01227 1.94×10-6 0.91×10-5

M7 150 0.03289 4.93×10-6 10 0.00702 7.02×10-8 0.049 0.00579 0.93×10-6 0.43×10-5

3.5
M1
M2
3.0 M3
M4
M5
2.5 M6
km (10-2m/s)

M7
2.0

1.5

1.0

0.5
0.05 0.10 0.15 0.20 0.25
V (m3/h)
Fig. 13. Effects of dip coating cycles on mass transfer coefficients through membranes themselves.

The mass transfer coefficients for membranes themselves are shown in Fig. 13. As seen, the mass transfer

coefficient of the PVDF is the largest among all the membranes. In other words, the moisture transfer

resistance of the PVDF is the smallest. For M3~M7, the membrane mass transfer coefficients increase at first

from M3 to M5. Then they decrease with increased dip-coating cycles from M5 to M7. As seen, the moisture

23
transfer coefficients of M5 and M2 appear very close. This means the mass transfer resistance of the

well-filmed PVDF/SiO2 composite bilayer membrane is close to that of the PVDF-SiO2 hybrid membrane.

From Fig. 13, the permeability of the PVDF membrane with a 150 μm thickness is 2.43×10-5 g/(cm2 s cm

Hg). The permeability of the PVDF-SiO2 membrane with a 150 μm thickness is 1.28×10-5 g/(cm2 s cm Hg).

The performance of the PVDF/SiO2 bilayer membrane with a 156 μm thickness is 1.15×10-5 g/(cm2 s cm Hg).

These values are a bit higher than an asymmetric cellulose triacetate hollow fiber membrane (0.72×10-5 g/(cm2

s cm Hg)) (Pan et al., 1978) and two orders of magnitudes higher than the TES liquid membrane

(1.44×10-7~2.16×10-7 g/(cm2 s cm Hg)) (Ito, 2000). Therefore, the PVDF/SiO2 bilayer membrane fabricated in

this study has acceptable permeability available for seawater desalination.

3.6.2 Ordinary and Knudsen diffusion

In the porous media, the gas diffusion theory considers three mechanisms: Poisseuille flow, ordinary

molecular diffusion, and Knudsen diffusion. (Khayet and Matsuura, 2004; Zhang, 2006, 2008; Zhang et al.,

2008). In our study for humidification with PVDF membranes, Kn≥10, therefore only Knudsen and ordinary

diffusion is considered.

In the composite membrane, there are two layers moisture transfer resistance (m2s kg-1):

tot m m (7)

where
δm (8)
m
ρa m

Ordinary diffusion coefficient Do of water vapor molecule in air is calculated as 2.52×10-5 m2/s by Ref.

(Tomaszewska et al., 2000), Knudsen diffusion coefficient Dk is calculated by Ref.(Tomaszewska et al., 2000)

T (9)
p

where R is 8.314 J mol-1 K-1, dp (nm) is the pore diameter.

The effective diffusivity is calculated by Ref.(Tomaszewska et al., 2000)

(10)
o o

The relational expression between equivalent diffusion efficient of moisture Dvm and effective diffusivity

Dko is calculated by Ref.(Zhang, 2006)

24
 m o
(11)

The diffusion coefficients calculated by Knudsen model and by the experimental are listed in Table 8.

The parameter of M5 used in the calculation is from the 3.5 section.

Table 8

Parameter of M5 used in the analysis and diffusion coefficient calculated by Knudsen model.

Membrane dp ε τ V S Dv,cal Dv,exp Relative

(nm) (%) (ml/g) (m2/g) (m2/s) (m2/s) error (%)

PVDF layer 511.5 77.76 3.0 2.986 23.35 5.09×10-6 4.93×10-6 −3.26%

SiO2 aerogel layer 6.6 90.18 6.0 0.927 561.54 1.86×10-7 1.77×10-7 −5.18%

From Table 8, it can be clearly seen that the SiO2 aerogel layer has smaller pore sizes, higher porosity and

higher pore tortuosity, while the PVDF layer has larger pore sizes, lower porosity and lower pore tortuosity.

Compared the calculated and the experimental, the values of diffusion coefficient have a maximum difference

of 5.18%. The reason may be that pores are more zig-zagged. Therefore, the calculated value is relatively

larger than the experimental value.

3.6.3 Heat conductivity

There are two layer heat resistances in the composite membranes: resistance in the PVDF layer;

resistance in the silica layer. When the total membrane thermal conductivity is measured, the total heat

resistance (kW-1m2K) can be calculated:

δ (12)
λ
The heat resistance (kW-1m2K) in the bilayer membranes comprise two layers:

m m m (13)

For all the composite membranes, it can be assumed that the thermal conductivity of the PVDF layer is

equal to 0.08199 Wm-1K-1. Then the heat resistance of the PVDF layer is 1.8295 kW-1m2K. Therefore the heat

resistance of the silica aerogel layer can be calculated by Eq. (13). Thermal conductivity of each part and the

effective conductivity of the whole membrane are listed in Table 9. The ratios of heat resistance in each layer

25
to the total are plotted in Fig. 14.

Table 9

Thermal conductivity in each part and the heat resistance of the whole membrane.

sample δm1 λm1 rm1 δm2 λm2 rm2 λm rm

(μm) (Wm-1K-1) (kW-1m2K) (μm) (Wm-1K-1) (kW-1m2K) (Wm-1K-1) (kW-1m2K)

M1 150 0.08199 1.8295 — — — 0.08199 1.8295

M2 150 0.08078 1.8569 — — — 0.08078 1.8569

M3 150 0.08199 1.8295 2 0.01073 0.1864 0.07540 2.0159

M4 150 0.08199 1.8295 4 0.00498 0.8039 0.05848 2.6334

M5 150 0.08199 1.8295 6 0.00364 1.6472 0.04487 3.4767

M6 150 0.08199 1.8295 8 0.00740 1.0808 0.05429 2.9103

M7 150 0.08199 1.8295 10 0.00982 1.0180 0.05619 2.8475

3.5 rdm
rpm
3.0

2.5
rm (kW-1m2K)

2.0

1.5

1.0

0.5

0.0
M1 M2 M3 M4 M5 M6 M7
Membrane

Fig. 14. The ratios of heat resistance in each layer to the total heat resistance.

From Table 9, after doping with SiO2 aerogel powder, the thermal conductivity of the blend membrane

decreases from 0.08199 W m-1K-1 to 0.08078 Wm-1K-1 (a 1.5% decrease). After dip-coating with silica films,

the thermal conductivity of the membrane decreases from 0.08199 Wm-1K-1 to 0.04487 Wm-1K-1, significantly

(a 45.27% decrease). Therefore, the PVDF/SiO2 composite bilayer membranes are more promising than

PVDF-SiO2 hybrid membranes in thermal insulation. The heat conductivity of the PVDF/SiO2 bilayer

membranes (0.04487 Wm-1K-1) is also far lower than the PVDF/PDMS (0.071 Wm-1K-1) and
26
PVDF/PDMS-SiO2 (0.062 Wm-1K-1) membranes (Khayet and Matsuura, 2004) and it is close to the reported

hydrophobic/hydrophilic composite TF200 membrane (0.043 W m-1K-1) (Khayet et al., 2005).

When dip-coating silica sols for 60 cycles, the 150 μm PVDF support layer accounts for 52.62% of the

total resistance, and the silica thin film of 6 μm accounts for 47.38% of the total resistance, as shown in Fig.

14. The total heat resistance of PVDF/SiO2 composite bilayer membrane is up to 3.4767 kW-1m2K, which

increases by 87.23% upon M2 (only 1.8569 kW-1m2 K) and by 90.04% upon M1 (only 1.8295 kW-1m2K),

respectively. It can be obtained that the formed complete silica thin film has good thermal insulation.

Appropriate thickness of silica films can be designed and fabricated to balance the paradox between a lower

heat conductivity and a higher moisture permeability.

3.6.4 Effect of SiO2 aerogel layers on permeability and heat resistance

The mass transfer coefficient of PVDF layer is fixed at 0.03289 m/s. Substituting this value into Eqs.

(4)-(6), the relations between the SiO2 layer thickness δm2 and the membrane mass transfer coefficient km and

the permeability Pe′ (g/(cm2 s cmHg)) are obtained:

(14)
m
.4 5. 5δm
(15)
5 （ .4 5. 5δm ）
The heat resistance of the PVDF layer is 1.8295 kW-1m2K. Substituting this value into Eqs. (12)-(13), the

relation between the SiO2 layer thickness and membrane heat resistance is obtained:
δm (16)
m . 5
. 4
The effects of SiO2 layer thickness on mass transfer coefficient and heat resistance are plotted in Fig.15.

27
 16
2.5
3.5 km
rm 14
3.0
2.0
12
Pe' (10 g/(cm s cm Hg) 2.5
10
1.5

rm (kW m K)
2.0
km (10 m/s)
8
2

-1 2
-2

1.0 1.5
6
-5

1.0
(10.67,1.103) (10.67,4.76)
4
0.5
0.5 2

0.0 0.0 0
0 5 10 15 20 25 30 35 40
δm2 (μm)

Fig. 15. Effects of SiO2 layer thickness on mass transfer coefficient and heat resistance.

From Fig. 15, the greater the thickness is, the less the mass transfer coefficient is. The heat resistance

increases linearly with the thickness. It is worth noting that a larger thickness is not negatively related to a

lower permeability. Other parameters like pore diameter, porosity, tortuosity are also important parameters.

Therefore, in our study, other parameters are assumed to remain unchanged. When thickness is 10.67 μm, the

membrane mass transfer coefficient is 1.103×10-2 m/s, the permeability is 0.82×10-5 (g/(cm2 s cmHg)), and the

heat resistance is 4.76 kW-1m2K. This means a PVDF/SiO2 composite bilayer membrane with a silica layer

thickness of 10.67 μm has a larger heat resistance and an acceptable permeability.

Therefore, a PVDF/SiO2 composite bilayer membrane with thermal insulation and acceptable

permeability is prepared. Meanwhile, the membrane is hydrophobic and mechanically durable. All these

characteristics make it possible to be widely used in the seawater desalination. In the future, the membrane

with thermal insulation and high permeability can be fabricated by improving other parameters like pore

diameter, porosity, tortuosity, etc. In other words, performance can be manipulated.

4. Conclusions

A promising membrane is one that permits vapor permeation efficiently with low thermal conductivity.

For this respect, a dip-coating approach with the addition of DMF is taken to prepare PVDF/SiO2 composite

bilayer membrane with thermal insulation for seawater desalination. The optimum membrane is that

dip-coated silica sols with 60 cycles. With dip-coating silica sols on the surface of PVDF membranes, the

network structures are formed with high porosity. The results show that the heat conductivity decreases from
28
0.08199 Wm-1K-1 to 0.04487 Wm-1K-1, moisture permeability sacrifices from 4.93×10-6 m2/s to 2.42×10-6 m2/s.

The membranes made in this research are not only simple in method, easy accessible of raw materials, good in

low cost, low energy consumption, but also advantageous in membrane qualities like uniformity and

cracking-free. Although there is a slight drop in the mass transfer coefficient, the reduced heat losses is more

appreciable due to the tremendous decrease in thermal conductivity. This is the first time that heat

conductivity of membranes can be manipulated for seawater desalination. In the future, the PVDF/SiO2

composite bilayer membranes can also be alternatives for other membrane distillation processes, besides

membrane-based humidification-dehumidification (MHDD) applications.

Acknowledgement

This Project was supported by the National Science Fund for Distinguished Young Scholars of China, No.

51425601. It was also supported by the National Key Research and Development Program, No.

2016YFB0901404.

29
References

Adachi, T., Sakka, S., 1988. Sintering of silica gel derived from the alkoxysilane solution containing

N,N-dimethylformamide. Journal of Non-Crystalline Solids 100, 250-253.

Alkhudhiri, A., Darwish, N., Hilal, N., 2012. Membrane distillation: a comprehensive review. Desalination

287, 2-18.

Alves, V., Coelhoso, I., 2006. Orange juice concentration by osmotic evaporation and membrane distillation: a

comparative study. Journal of Food Engineering 74, 125-133.

Banat, F., Jwaied, N., Rommel, M., Koschikowski, J., Wieghaus, M., 2007. Performance evaluation of the

"large SMADES" autonomous desalination solar-driven membrane distillation plant in Aqaba, Jordan.

Desalination 217, 17-28.

Banat, F.A., Simandl, J., 1998. Desalination by membrane distillation: a parametric study. Separation Science

and Technology 33, 201-226.

Brinker, C., Hurd, A., Schunk, P., Frye, G., Ashley, C., 1992. Review of sol-gel thin film formation. Journal of

Non-Crystalline Solids 147, 424-436.

Brinker, C.J., Scherer, G.W., 1990. Sol-Gel Science: The Physics and Chemistry of Sol-Gel Processing.

Academic Press, Boston.

Calabro, V., Jiao, B.L., Drioli, E., 1994. Theoretical and experimental study on membrane distillation in the

concentration of orange juice. Industrial & Engineering Chemistry Research 33, 1803-1808.

Cassie, A., Baxter, S., 1944. Wettability of porous surfaces. Transactions of the Faraday Society 40, 546-551.

30
Chen, T.-C., Ho, C.-D., 2010. Immediate assisted solar direct contact membrane distillation in saline water

desalination. Journal of Membrane Science 358, 122-130.

Chen, Z., Rana, D., Matsuura, T., Yang, Y., Lan, C.Q., 2014. Study on the structure and vacuum membrane

distillation performance of PVDF composite membranes: I. Influence of blending. Separation and

Purification Technology 133, 303-312.

Darmawan, A., Karlina, L., Astuti, Y., Sriatun, Motuzas, J., Wang, D.K., da Costa, J.C.D., 2016. Structural

evolution of nickel oxide silica sol-gel for the preparation of interlayer-free membranes. Journal of

Non-Crystalline Solids 447, 9-15.

Deshpande, R., Hua, D.-W., Smith, D.M., Brinker, C.J., 1992. Pore structure evolution in silica gel during

aging/drying. III. Effects of surface tension. Journal of Non-Crystalline Solids 144, 32-44.

Du, A., Zhou, B., Zhang, Z., Shen, J., 2013. A special material or a new state of matter: a review and

reconsideration of the aerogel. Materials 6, 941-968.

Efome, J.E., Baghbanzadeh, M., Rana, D., Matsuura, T., Lan, C.Q., 2015. Effects of superhydrophobic SiO 2

nanoparticles on the performance of PVDF flat sheet membranes for vacuum membrane distillation.

Desalination 373, 47-57.

El-Bourawi, M., Ding, Z., Ma, R., Khayet, M., 2006. A framework for better understanding membrane

distillation separation process. Journal of Membrane Science 285, 4-29.

El-Dessouky, H., Ettouney, H., Al-Juwayhel, F., Al-Fulaij, H., 2004. Analysis of multistage flash desalination

flashing chambers. Chemical Engineering Research & Design 82, 967-978.

Elimelech, M., Phillip, W.A., 2011. The Future of Seawater Desalination: Energy, Technology, and the

Environment. Science 333, 712-717.

Geng, H., Wu, H., Li, P., He, Q., 2014. Study on a new air-gap membrane distillation module for desalination.

Desalination 334, 29-38.

Godino, M., Pena, L., Rincón, C., Mengual, J., 1997. Water production from brines by membrane distillation.

Desalination 108, 91-97.

Guillén-Burrieza, E., Blanco, J., Zaragoza, G., Alarcón, D.-C., Palenzuela, P., Ibarra, M., Gernjak, W., 2011.

Experimental analysis of an air gap membrane distillation solar desalination pilot system. Journal of

Membrane Science 379, 386-396.

Hassan, A.M., Al-Sofi, M.A.K., Al-Amoudi, A.S., Jamaluddin, A.T.M., Farooque, A.M., Rowaili, A., Dalvi,

31
 A.G.I., Kither, N.M., Mustafa, G.M., Al-Tisan, I.A.R., 1998. A new approach to membrane and thermal

seawater desalination processes using nanofiltration membranes (Part 1). Desalination 118, 35-51.

Hsu, S., Cheng, K., Chiou, J.-S., 2002. Seawater desalination by direct contact membrane distillation.

Desalination 143, 279-287.

Ito, A., 2000. Dehumidification of air by a hygroscopic liquid membrane supported on surface of a

hydrophobic microporous membrane. Journal of Membrane Science 175, 35-42.

Jennings, S., 1988. The mean free path in air. Journal of Aerosol Science 19, 159-166.

Kaviany, M., 1991. Principles of Heat Transfer in Porous Media, second ed. Springer, New York.

Khayet, M., Matsuura, T., 2004. Pervaporation and vacuum membrane distillation processes: modeling and

experiments. AIChE Journal 50, 1697-1712.

Khayet, M., Mengual, J.I., Matsuura, T., 2005. Porous hydrophobic/hydrophilic composite membranes.

Journal of Membrane Science 252, 101-113.

Kim, J.W., Cho, W.J., Ha, C.S., 2002. Morphology, crystalline structure, and properties of poly (vinylidene

fluoride)/silica hybrid composites. Journal of Polymer Science Part B: Polymer Physics 40, 19-30.

Lenza, R.F.S., Nunes, E.H.M., Vasconcelos, D.C.L., Vasconcelos, W.L., 2015. Preparation of sol-gel silica

samples modified with drying control chemical additives. Journal of Non-Crystalline Solids 423, 35-40.

Li, G.-P., Zhang, L.-Z., 2016. Investigation of a solar energy driven and hollow fiber membrane-based

humidification–dehumidification desalination system. Applied Energy 177, 393-408.

Matz, R., Zimerman, Z., 1985. Low-temperature vapour compression and multi-effect distillation of seawater.

Effects of design on operation and economics. Desalination 52, 201-216.

Mohamed, E.S., Papadakis, G., Mathioulakis, E., Belessiotis, V., 2006. An experimental comparative study of

the technical and economic performance of a small reverse osmosis desalination system equipped with an

hydraulic energy recovery unit. Desalination 194, 239-250.

Muzychka, Y.S., Yovanovich, M.M., 2009. Pressure Drop in Laminar Developing Flow in Noncircular Ducts: A

Scaling and Modeling Approach. Journal of Fluids Engineering 131, 111105.

Ortiz, J., Expósito, E., Gallud, F., García-García, V., Montiel, V., Aldaz, A., 2006. Photovoltaic electrodialysis

system for brackish water desalination: Modeling of global process. Journal of Membrane Science 274,

138-149.

Pan, C., Jensen, C., Bielech, C., Habgood, H., 1978. Permeation of water vapor through cellulose triacetate

32
 membranes in hollow fiber form. Journal of Applied Polymer Science 22, 2307-2323.

Pérez-Lombard, L., Ortiz, J., Pout, C., 2008. A review on buildings energy consumption information. Energy and

Buildings 40, 394-398.

Qtaishat, M.R., Banat, F., 2013. Desalination by solar powered membrane distillation systems. Desalination

308, 186-197.

Service, R.F., 2006. Desalination freshens up. Science 313, 1088-1090.

Tomaszewska, M., Gryta, M., Morawski, A., 1995. Study on the concentration of acids by membrane

distillation. Journal of Membrane Science 102, 113-122.

Tomaszewska, M., Gryta, M., Morawski, A., 2000. Mass transfer of HCl and H 2O across the hydrophobic

membrane during membrane distillation. Journal of Membrane Science 166, 149-157.

Tu, K.L., Nghiem, L.D., Chivas, A.R., 2010. Boron removal by reverse osmosis membranes in seawater

desalination applications. Separation and Purification Technology 75, 87-101.

Wang, P., Chung, T.-S., 2015. Recent advances in membrane distillation processes: Membrane development,

configuration design and application exploring. Journal of Membrane Science 474, 39-56.

Wenzel, R.N., 1936. Resistance of solid surfaces to wetting by water. Industrial & Engineering Chemistry 28,

988-994.

Wirth, D., Cabassud, C., 2002. Water desalination using membrane distillation: comparison between

inside/out and outside/in permeation. Desalination 147, 139-145.

Yamali, C., Solmus, I., 2007. Theoretical investigation of a humidification-dehumidification desalination

system configured by a double-pass flat plate solar air heater. Desalination 205, 163-177.

Yu, L.-Y., Xu, Z.-L., Shen, H.-M., Yang, H., 2009. Preparation and characterization of PVDF–SiO2 composite

hollow fiber UF membrane by sol–gel method. Journal of Membrane Science 337, 257-265.

Yun, Y., Ma, R., Zhang, W., Fane, A., Li, J., 2006. Direct contact membrane distillation mechanism for high

concentration NaCl solutions. Desalination 188, 251-262.

Zhang, L.-Z., 2006. Fabrication of a lithium chloride solution based composite supported liquid membrane

and its moisture permeation analysis. Journal of Membrane Science 276, 91-100.

Zhang, L.-Z., 2008. A fractal model for gas permeation through porous membranes. International Journal of

Heat and Mass Transfer 51, 5288-5295.

Zhang, L.-Z., Li, G.-P., 2017. Energy and economic analysis of a hollow fiber membrane-based desalination

33
 system driven by solar energy. Desalination 404, 200-214.

Zhang, L.-Z., Wang, Y.-Y., Wang, C.-L., Xiang, H., 2008. Synthesis and characterization of a PVA/LiCl blend

membrane for air dehumidification. Journal of Membrane Science 308, 198-206.

Zhang, L.Z., Niu, J.L., 2003. Laminar fluid flow and mass transfer in a standard field and laboratory emission

cell. International Journal of Heat and Mass Transfer 46, 91-100.

Zhang, X.-R., Zhang, L.-Z., Liu, H.-M., Pei, L.-X., 2011. One-step fabrication and analysis of an asymmetric

cellulose acetate membrane for heat and moisture recovery. Journal of Membrane Science 366, 158-165.

Zhao, D., Yang, P., Melosh, N., Feng, J., Chmelka, B.F., Stucky, G.D., 1998. Continuous mesoporous silica

films with highly ordered large pore structures. Advanced Materials 10, 1380-1385.

Zisman, W.A., 1964. Relation of the Equilibrium Contact Angle to Liquid and Solid Constitution, in: Fowkes

F.M., Contact Angle, Wettability, and Adhesion. VSP International Science Publishers, Los Angles, pp.

1-51.

Table Captions

Table 1. The effects of different viscosity and dip-coating speed on film quality.

Table 2. The parameters of different films.

Table 3. Changes of entrance length with Reynolds number.
Table 4. The data of surface contact angle.

Table 5. Surface roughness of membranes (s an size 5 μm×5 μm).

Table 6. FTIR peak assignments for membranes.

Table 7. Mass transfer coefficient in each part and the total mass transfer coefficient for every membrane.

Table 8. Parameter of M5 used in the analysis and diffusion coefficient calculated by Knudsen model.

Table 9. Thermal conductivity in each part and the heat resistance of the whole membrane.

Figure Captions

Fig. 1. Fabrication of silica aerogel films.

Fig. 2. The schematic diagram of the membrane exchanger.

Fig. 3. The test equipment for moisture exchange.

Fig. 4. Effects of molar ratios of DMF to TEOS on the gelation time and viscosity of silicon sol.
34
Fig. 5. The films before and after doping DMF. (a) before; (b) after.

Fig. 6. Mechanism of action of DMF on silicon sol.

Fig. 7. The AFM images of different membranes: (a) M1; (b) M2; (c) M5.

Fig. 8. The SEM graphs of membranes: (a~g) M1~M7. Subscripts 1 and 2 represent the same kind of

membranes but with different observation magnification, respectively.

Fig. 9. SEM cross-sections photographs of M5 membrane with 20 μm amplification.

Fig. 10. The XRD patterns for different membranes (M1, M2, M3, M4, M5, M6 and M7).

Fig. 11. FTIR patterns of the membranes (M1, M2, M3, M4, M5, M6 and M7).

Fig. 12. Distributions of pore diameters of (a) PVDF layer and (b) SiO2 aerogel layer.

Fig. 13. Effects of dip coating cycles on mass transfer coefficients through membranes themselves.

Fig. 14. The ratios of heat resistance in each layer to the total heat resistance.

Fig. 15. Effects of SiO2 layer thickness on mass transfer coefficient and heat resistance.

35
 SiO2 aerogel layer
SiO2 PVDF
porosity: 90.18%

Pore volume dV/dD (cm3/g/nm)

0.02

0.01

0.00
1 10 100
Pore diameter (nm)

1.2 PVDF layer

porosity: 77.76%

Pore volume dV/dlogdp (cm3/g)

1.0
PVDF/SiO2 composite membrane
0.8

0.6

0.4

0.2

0.0

1 10 100 1000 10000

Pore diameter (nm)

16
2.5
3.5 km
14
3.0
rm
2.0
12
2.5
Pe' (10 g/(cm s cm Hg)

10
1.5
rm (kW m K)

2.0
km (10 m/s)

8
2

-1 2
-2

1.0 1.5
6
-5

(10.67,1.103) (10.67,4.76)
1.0 4
0.5
0.5 2

0.0 0.0 0
0 5 10 15 20 25 30 35 40
δm2 (μm)
Effects of SiO2 layer thickness on mass transfer coefficient and heat resistance.
 A PVDF/SiO2 (Si aerogel) composite bilayer membrane of low heat conductivity is developed.

 Heat conductivity can be manipulated to satisfy the needs in desalination.

 Measures are taken to solve the problem of cracking in forming SiO 2 layer.

 A silica layer thickness of 10.67 μm has the largest heat resistance with an acceptable permeability.

 The bilayer composite membrane is superior to SiO 2 blend membrane in thermal insulation.

36