Novel PVDF hollow fiber membranes for vacuum and direct contact membrane distillation applications

Accepted Manuscript Novel PVDF Hollow Fiber Membranes for Vacuum and Direct Contact Mem‐ brane Distillation Applications E. Drioli, A. Ali, S. Simone, F. Macedonio, S.A. AL-Jlil, F.S. Al Shabonah, H.S. Al-Romaih, O. Al-Harbi, A. Figoli, A. Criscuoli PII: DOI: Reference:

S1383-5866(13)00264-5 http://dx.doi.org/10.1016/j.seppur.2013.04.040 SEPPUR 11176

To appear in:

Separation and Purification Technology

Received Date: Revised Date: Accepted Date:

19 December 2012 19 April 2013 22 April 2013

Please cite this article as: E. Drioli, A. Ali, S. Simone, F. Macedonio, S.A. AL-Jlil, F.S. Al Shabonah, H.S. AlRomaih, O. Al-Harbi, A. Figoli, A. Criscuoli, Novel PVDF Hollow Fiber Membranes for Vacuum and Direct Contact Membrane Distillation Applications, Separation and Purification Technology (2013), doi: http://dx.doi.org/10.1016/ j.seppur.2013.04.040

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.

Novel PVDF Hollow Fiber Membranes for Vacuum and Direct Contact Membrane Distillation Applications E. Drioli1, A. Ali2, S. Simone1, F. Macedonio1, S. A. AL-Jlil3, F. S. Al Shabonah3, H. S. AlRomaih3, O. Al-Harbi3, A. Figoli1*, A. Criscuoli1* 1 2

Institute on Membrane Technology (ITM-CNR), via P. Bucci 17/C, 87030 Rende (CS) Italy Dept. of Chemical Engineering and Materials, University of Calabria, via P. Bucci 44/A,

87030 Rende (CS) Italy 3

National Center for Water Technology, King Abdulaziz City for Science and Technology

(KACST), King Addullah Road (West), Riyad, Kingdom of Saudi Arabia * corresponding authors ([email protected]; [email protected])

Abstract

Microporous, hydrophobic hollow fiber membranes were prepared from polyvinylidene fluoride

(PVDF) under different processing conditions and by changing the composition of the polymeric

dope. In particular, different additives were incorporated into the dope (Polyvinyl Pyrrolidone

and Maleic Anhydride) and the molecular weight and the concentration of PVDF polymer into

the dope solution (medium and high Mw, 15-18 wt.%) were varied, in order to obtain fibers with

different morphologies. The prepared membranes were characterized in terms of morphology,

mechanical properties, bubble point, overall porosity, pore size distribution and thickness.

Comparison between their performance in membrane distillation (MD), working under vacuum

and direct contact configurations was also performed. The trans-membrane flux obtained was

related to the membrane properties, which are function of the operating spinning conditions and

the polymeric dope composition. This systematic investigation is an attempt to establish a

correlation between the MD performance (in terms of trans-membrane flux) and the hollow fiber preparation conditions (polymeric dope composition and spinning conditions).

Keywords: PVDF Solef® 6012, PVDF Solef® 1015, hollow fiber membrane, membrane morphology, membrane distillation.

1

1. Introduction Membrane distillation (MD) can be applied to various fields involving separation and purification of the effluent streams and crystallization from various solutions [1-6]. MD is a process in which vapors diffuse through a microporous hydrophobic membrane thanks to a difference of partial pressures, and are collected at the distillate or permeate side. Various configurations can be used in this process, including direct contact membrane distillation (DCMD), vacuum membrane distillation (VMD), sweeping gas membrane distillation (SGMD) and air gap membrane distillation (AGMD). DCMD is the most studied type of membrane distillation due to the simplicity of the set-up used, while VMD is the configuration that usually leads to the highest trans-membrane fluxes [7]. The process possesses some unique advantages over the conventional processes, the worth mentioning being the theoretically complete rejection of the non-volatile species, the potential to use the waste grade heat or energy, operations carried out at low pressures and less strict requirements in terms of mechanical strength of the membranes used [8]. In MD, the feed is in direct contact with the membrane. The vapor pressure gradient across the membrane can be created through different schemes depending upon the configuration applied. The membrane must allow only the passage of vapor phase and must strictly retain the liquid, which implies that the membranes must exhibit excellent hydrophobic character. Thermal conductivity of the membrane must be reduced as much as possible to avoid the heat losses through conduction. High porosity of the membrane is required to ensure high molecular diffusion across the membrane. The presence of even a few pores with size exceeding those suggested by water liquid entry pressure (LEPw) can affect the quality of the product obtained. Therefore, the maximum pore size and the pore size distribution must lie within a narrow range. Other requirements for membranes for MD include excellent chemical-resistance, appropriate mechanical stability and low price material of construction. The detailed theoretical design of membranes for MD has been provided in various studies [9-11]. Historically, membranes prepared for other applications, mainly for microfiltration and ultrafiltration, were utilized for MD. Various studies pointed out that the further development and commercialization of MD is strongly associated with the availability of suitable membranes. Some serious attempts for preparation of membranes for MD can be found mostly during last decades. In particular, many researches deal with the use of PVDF membranes, due to their good

2

hydrophobicity and chemical stability, together with the possibility of using the phase inversion technique, both thermal induced (TIPS) and diffusion induced (DIPS or NIPS), for their preparation. For example, Prince et al. [12] prepared electrospun nanofiber membranes from PVDF-nano clay composites. They claimed that such membranes show superior hydrophobicity, improved heat resistance and better performance in direct contact membrane distillation.

Edwie et al. [13] prepared dual layer hydrophobic-hydrophilic membranes for membrane

distillation. Methanol was used as non-solvent additive while self-synthesized silica particles

were used to improve the hydrophobicity of the membranes produced.

Teoh and Chung [14] prepared hydrophobic polyvinylidene fluoride–polytetrafluoroethylene

(PVDF–PTFE) hollow fibers showing improved hydrophobicity and macrovoid-free structure.

The prepared membranes show improved hydrophobic character, mechanical strength and better

performance in membrane distillation.

Wang et al [15] used clay particles to produce mixed matrix PVDF hollow fiber membranes.

Hydrophobic cloisite clay particles were added into the polymeric dope. The obtained fibers

show high porosity, high flux, low thermal insulation and reduced risk of membrane pore

wetting.

Tang et al [16] studied the effect of various parameters on microporous hydrophobic PVDF

membrane characteristics prepared through dry-jet wet phase inversion technique. The effect of

bore fluid temperature, the composition of the dope, length of air gap, take up speed and extrusion rate on membrane properties has been investigated. The membrane performance was

tested in DCMD and VMD configurations.

Fan and Peng [17] investigated the effect of different operating variables on performance of

PVDF flat sheet membranes for VMD and DCMD. Dusty gas model was successfully applied to

model the flux in both configurations. The authors concluded that thin membranes with good

hydrophobicity and sponge like cross section are more suitable for VMD than DCMD. The results achieved in various studies are reported in Table 1. In general, all the membranes show good overall porosity, however, the contact angle, mean pore size and the thickness vary considerably. Concerning their performance in membrane distillation, it is quite difficult to make a fair comparison, as the different tests were carried out by working at different operating conditions and with different membrane modules (that means different fluidodynamics and

3

mass/heat transport resistances). Nevertheless, Table 1 gives an overview of the main results obtained in literature with PVDF-based membranes and can be considered as a “reference point” for future developments in the field. In a previous work, it was shown how the synergistic effect of water and low molecular weight poly(vinyl pyrrolidone) K-17 (PVP K-17, Mw 7–11 kDa) as pore forming additives can lead to production of PVDF microporous hydrophobic hollow fibers [18]. Even if PVP is hydrophilic, the use of low molecular weight additive, coupled to treatment with sodium hypochlorite, ensured complete leaching of PVP from the polymeric matrix; therefore, the produced fibers were hydrophobic and suitable for VMD application. It was also evidenced how fibers morphology was dependent on PVP concentration. In particular, while low percentages of PVP enhanced macrovoids formation (due to hydrophilic nature of PVP and thermodynamic instability of dope solution), higher PVP concentration can hinder macrovoid formation, due to increased dope viscosity (kinetic effect), thus ensuring high flux coupled to good mechanical properties. On the basis of the positive results previously obtained, the focus of the present study was to continue the investigation of the effect of processing conditions and different additives on PVDF hollow fiber membrane properties (morphology, pore size distribution, porosity, mechanical strength), with the aim of further improving their performance for membrane distillation applications. In particular, air gap, bore fluid, coagulation bath and dope composition effect was further investigated. The prepared membranes were used for carrying out both DCMD and VMD tests, to make a comparison of their performance in the two MD configurations. The obtained results were also compared to those obtained using commercial polypropylene (PP) hollow fiber membranes tested both in DCMD and VMD under the same conditions.

2. General theory on mass and heat transfer across the membrane 2.1 Mass transfer The transport of vapor across the membrane can occur through different mechanisms: Knudsen diffusion, molecular diffusion and viscous flow. In DCMD, the feed and the permeate streams are usually at atmospheric pressure and the viscous flow can be neglected, while in VMD, due to the removal of the air entrapped into pores by

4

means of the vacuum, the molecular diffusion does not take place. Moreover, in many cases, the viscous flow is also neglected. All types of mechanisms are strongly dependent on the membrane geometry, as clearly evidenced by the expressions of the relative coefficients: Dv (viscous coefficient) = εr2/8τ Dk (Knudsen coefficient) = 2εr/3τ Dm (molecular coefficient) = ε/τ All coefficients increase with the membrane pore size and porosity and decrease with the pore tortuosity. Furthermore, the trans-membrane flux is inversely proportional to the membrane thickness.

2.2 Heat transfer The heat is transferred across the membrane by two main mechanisms: the conduction through the membrane material and the gas-filled pores, and the latent heat due to the water evaporation. The conduction (Qc) is considered as a “heat loss mechanism”, because no corresponding mass transfer takes place. It depends on the thermal conductivity of the membrane (km), the temperatures at the feed/membrane (Tfm) and permeate/membrane (Tpm) sides and on the membrane thickness:

Qc= km (Tfm-Tpm)/δ

km is function of the gas (kg) and of the membrane material (ks) thermal conductivities: km = εkg + (1-ε)ks

Due to the high vacuum at the permeate side and in the membrane pores, the heat lost by conduction is often negligible in VMD, while is always present in DCMD. However, this loss can be reduced by using materials with low thermal conductivity and by increasing the membrane porosity, because, usually, the water vapor has a thermal conductivity one order of magnitude lower than that of membrane materials. Although an increase of the membrane

5

thickness also leads to a reduction of the heat loss, its value has to be optimized in DCMD, in order to do not increase too much the mass transport resistance offered by the membrane.

3. Experimental 3.1. Hollow fiber preparation For preparation of polymeric dopes, the polymer poly(vinylidene fluoride) (PVDF Solef® 6012 or 1015 from Solvay chemical company) was dissolved in N-methyl pyrrolidone (NMP) or N,Ndimethylformamide (DMF) in concentration ranging from 15 to 18 wt.%. Double distilled water, Polyvinyl Pyrrolidone (PVP K-17) and Maleic Anhydride (AMAL) were also used as pore forming. The solution was stirred at selected temperature (60-85°C) until a homogeneous dope was obtained. The dope was allowed to degas overnight before being transferred to the pressurized reservoir of the spinning set-up. Hollow fibers were spun by the dry/wet technique described elsewhere [18, 25]. Most significant spinning conditions used in this work are reported in Table 2, while Table 3 summarizes the main parameters used in ref. 18 for the preparation of fibers having a similar dope composition of M1 and M2. The choice of the experimental conditions used in this work, with respect to ref. 18, can be explained as follows. 1) For Fiber M1 and M2 dope composition: the polymer type is the same (PVDF Solef 6012). Polymer concentration was slightly reduced (from 20 to 18%) attempting to improve porosity and, hence, VMD flux without affecting too much fibers mechanical resistance. Solvent was changed from DMF to NMP. This choice was made taking into account the different solvent miscibility with water. In particular, NMP has less affinity to water with respect to DMF; the use of this solvent should delay solvent outflux and nonsolvent influx during coagulation and, hence, help to avoid the formation of macrovoids. Dope temperature and spinning rate are the same. 2) For Fiber M1: the coagulation bath composition was changed from pure water to ethanol 30%. The use of this mixture instead of pure water as outer coagulant should delay L/L demixing during fiber coagulation and promote the formation of spongy structures. Because of the use of a different coagulation bath in the hollow fiber spinning set-up, it was necessary to change also the air gap (from 25 to 14 cm). The choice of the bore fluid (NMP 30%) can be explained as delaying demixing, avoid macrovoids and promote formation of spongy-type structures. In ref.

6

18 it was observed that use of aqueous mixtures of solvent (25-35%) as bore fluids avoids the formation of macrovoids. Therefore, the strategy is the same except that NMP was used instead of DMF in this work. 3) For Fiber M2: The dope composition and spinning conditions, including air gap and coagulation bath composition, are the same to that used in ref. 18. In this case, the effect of H2O 100% as bore fluid was investigated. In ref. 18, it was observed that the bore fluid composition affects macrovoid formation at the fibres inner layer; but, at high PVP concentration (15%), this effect is less evident. However, in the previous work, water 100% was not tested as bore fluid. Therefore, the aim here was to use pure water as bore fluid for overcoming the effect of PVP and promote transition of fibers morphology from symmetric to asymmetric. 4) For Fiber M3: A further additive, maleic anhydride, as pore former, was added in the dope composition. This additive was already used, in combination with high Mw PVP (PVP K-90, Mw 1300 KDa), for preparation of microfiltration PVDF hollow fibre membranes in literature [19]. High Mw PVP was replaced by PVP K-17 (Mw 7-11 KDa) (with the same concentration used in ref. 18) in order to facilitate its removal by sodium hypochlorite, as suggested in literature [20-24]. The PVDF type Solef 1015 and its concentration (15%) were chosen on the basis of what reported in literature, as well as the dope temperature [19]; while, air gap, spinning rate and coagulation bath composition are the same used in ref. 18. For these fibers isopropanol 30% (injection rate 17 ml/min) was tested as bore fluid. This mixture can be considered a “soft” coagulant, and was chosen attempting to delay demixing. The high injection rate was chosen to tailor fiber thickness and, in particular, to obtain thinner fibers with respect to M1 and M2.

3.2. Hollow fiber post treatment The produced PVDF hollow fibers were rinsed completely with abundant water to ensure complete coagulation and total removal of residual solvent. PVP was leached out as suggested in literature [20-24, 26-29] by washing the fibers with a solution of sodium hypochlorite 4000 ppm buffered to pH 7. Fibers were washed again and soaked in an aqueous solution of glycerol (30%) for 3-4 hours before drying to avoid collapse of their porous structure.

7

3.3. Hollow fiber modules Hollow fiber modules were prepared in the lab, by coaxially inserting three hollow fibers in a glass device (length 20 cm). Both sides were potted by epoxy glue. Fibers part covered with glue was not considered in the calculation of active membrane area. In order to remove glycerol, prior to use, modules were washed using double distilled water at 50°C and trans-membrane pressure of ~1 bar for at least 2 hours.

3.3. Scanning electron microscopy (SEM) characterization The morphology of the PVDF hollow fibers prepared in this work was analyzed by SEM (Quanta FENG 200, FEI Company). For the cross sectional study, the samples were prepared by freeze fracturing the selected fibers into liquid nitrogen.

3.4. Mechanical properties The mechanical properties of the produced fibers were measured by using ZWICK/ROELL Z 2.5 test unit. The stretching rate was adjusted at 5mm/min while maintaining the unidirectional stretching. The initial distance between the clamps was adjusted at 50 mm. For each type of fiber, 5 tests were performed and the results were reported as the average values of these tests.

3.5. Bubble point and pore size distribution Fibers bubble points, largest pore size and pore size distribution were determined using a PMI Capillary Flow Porometer (CFP 1500 AEXL, Porous Materials Inc., USA). By this technique, only active pores are measured, that means pores interconnected and open on both surfaces. For each test, fiber samples not treated with glycerol were fully wetted using Porewick (16 dyne/cm). Samples were connected to the instrument and tests were performed according to the wet-up/dryup mode using the software Capwin. The measurement of bubble point, largest pore size and pore size distribution is based on the Laplace’s equation: dP = 4τ cosθ/P where dP is the pore diameter, τ is the surface tension of the liquid, θ is the contact angle of the liquid (assumed to be 0 in case of full wetting, which means cosθ = 1) and P is the external pressure.

8

The results of each test were imported as an excel file using the software Caprep for further processing.

3.6. Porosity measurement The porosity of the fibers was measured according to the procedure described elsewhere [18]. The gravimetric method was used, which is based on measuring the weight of the liquid entrapped within the membrane pores. The overall porosity was calculated according to the following formula: εm(%) = [(w1 −w2)/Dk]/[(w1 −w2)/Dk + w2/Dpol]×100 Where w1

Weight of the wet membrane

w2

Weight of the dry membrane

Dk

Density of kerosene oil (0.82 g/cm3)

Dpol

Density of PVDF (1.78 g/cm3)

For each fiber type, five measurements were carried out; then, average and standard deviation were calculated.

3.7. Trans-membrane flux Membrane characterization in terms of trans-membrane flux was carried out in vacuum and direct contact configuration.

3.7.1 Vacuum membrane distillation The procedure and the set-up used for vacuum membrane distillation are described elsewhere [18] and shown in Figure 1a. Briefly, the feed is introduced in the lumen side of the fibers by using a peristaltic pump after heating. Vacuum is applied on the outer side (module shell side) by using a vacuum pump. The vapors transported through the membrane are condensed in the condensate tank by using liquid nitrogen. Double distilled water was used as feed. The feed temperature and flow-rate were adjusted at 50°C and 6 L/h, respectively, while the vacuum pressure applied was 40 mbar. The distillate collected (m) after specific interval of time was weighted and the flux was calculated according to the following relation.

9

J=

m A.t

Where J is the trans-membrane flux, A is the active membrane surface area calculated on the basis of membrane inner diameter, length of each fiber and the total number of fiber, and t is the time interval.

3.7.2 Direct contact membrane distillation The set-up used for DCMD is shown in Figure 1b. Double distilled water was used on both feed and permeate sides. The feed is heated before entering into the membrane module and is introduced on the lumen side of the fibers while the distillate stream is sent at the shell side. The feed and permeate flow rates are controlled by using peristaltic pumps. The distillate tank was placed on a balance that registered the mass of permeate produced during the experiments. The trans-membrane flux was calculated as for the VMD operation. DCMD tests were performed by varying temperature conditions on the feed side, while keeping the permeate temperature constant. The permeate temperature and flow-rate were fixed at 25°C and 6 L/h, respectively, while the feed temperature was changed from 50°C to 70°C with an interval of 10°C each. For each temperature condition, four different flow rates were applied, ranging from 6 L/h to 11.4 L/h, with an interval of 1.8 L/h each.

3.8. Commercial hollow fiber membranes Commercial poly propylene hollow fiber membranes type ACCUREL® PP S6/2 were purchased from Membrana GmbH. The commercial hollow fibers were characterized using the same methods described for lab-produced PVDF hollow fibers. The results are reported in Table 4. The commercial PP fibers were assembled in modules and tested in both VMD and DCMD setups described in section 2.7.

4. Results and discussion 4.1. Membrane morphology The scanning electron microscopy pictures of the fibers produced in the present work are shown in Figures 2-4. Figures 2 reveals that the morphology of the fibers type M1 is dominated by sponge-like structure. In Figure 3, it can be noticed that fibers type M2 show finger like

10

macrovoids at both the inner and outer surface. Fibers type M3 shows tear-drop macrovoids at the inner surface and a thin sponge-like layer at the outer surface, as shown in Figure 4. The morphology, properties and, subsequently, MD performance of hollow fiber membranes arise from interplay between different factors. For fibers produced by Non-solvent or Diffusion Induced Phase Separation (NIPS or DIPS) the main factors are: polymer molecular weight and concentration [30-32]; type of solvent [30, 31]; temperature and viscosity of polymeric dope [30, 31]; temperature and composition of outer coagulation bath [32-35]; eventual use of additives such as polymers (PVP, PEG), small molecules (Glycerol, Ethanol, Water) or salts (LiCl, LiClO4) [31, 36-37]; mutual affinity between solvent and nonsolvent (Hildebrand solubility parameters) [38]; composition and injection rate of inner coagulant [23, 24, 39-40]; typical spinning parameters such as dope extrusion rate, air gap and fiber take-up speed [39-40]. In this study, the attention has been focused on the combined effect of dope composition (PVDF Mw and concentration, use of pore forming additives) and operating conditions used in the spinning experiments, i.e. the composition of the used bore fluid, the air gap and the composition of the outer coagulant. Indeed, coagulation media are among the most important factors affecting the final morphology of hollow fiber membranes prepared by NIPS, since they directly influence the coagulation rate. In general, while large finger-like macrovoids and cavities-like structures are usually produced by fast coagulation rate, slower coagulation gives rise to porous sponge-like structures, as reported in literature [31-32]. For fibers type M1, the presence of water and NMP, in both solvent and bore fluid, decreases the difference between the solubility parameters of solvent and coagulant and, thus, attributes a “soft” character to the inner coagulant. Consequently, de-mixing between polymer rich and polymer lean phases is delayed and a dominant sponge-like structure is obtained. Formation of sponge-like structure is also promoted by the use of ethanol 30% as outer coagulant instead of tap water, in agreement to what reported in literature. Deshmukh and Li [41] examined the effect of coagulation bath composition on morphology of polyvinylidene fluoride (PVDF) hollow fiber membranes using ethanol and water mixtures with different ratio (from 10/90 to 50/50). They found that fibers morphology near the outer wall was changed from long finger-like structure, to short finger-like structure and, finally, to a sponge-like structure, by increasing ethanol concentration in outer coagulation medium. Fibers morphology also depends on the pore forming additive used in the dope solution (PVP K-17). It has been observed in the previous studies [18]

11

that the addition of PVP as pore forming additive interferes with thermodynamic stability of the solution and influences the de-mixing behavior of the solution during phase separation. At low concentration, the de-mixing is accelerated in presence of PVP. However, due to the compatibility of PVP with PVDF, dope viscosity increases and de-mixing is delayed at higher concentration of PVP into the solution; this leads towards the formation of a dominant macrovoids free, sponge like structure, as observed in case of fiber type M1. The morphology of fiber type M2 is strongly affected by the use of water as both internal and external coagulant, which generates finger like macrovoids on both surfaces of the fiber. As widely accepted in literature [33, 42-43], water is a harsh non-solvent for PVDF. It causes instantaneous liquid-liquid de-mixing and leaves a structure containing finger-like macrovoids. The morphology of the inner surface seems to be more sensitive towards the composition of the bore fluid as compared to the dependence of outer surface on the composition of coagulant. In fact, as it can be noticed in Figure 2, finger-like structures are much more evident at the fiber inner wall. This is due to the different coagulation dynamics of fibers inner and outer walls. Coagulation of the inner layer starts immediately, due to the contact of the nascent fiber inner wall with the bore fluid, while the nascent fiber passes through an air gap of 25 cm before entering in the coagulation bath. Therefore, coagulation of the outer surface is delayed with respect to the inner one, thus resulting in reduction of macrovoids. Fibers type M3 were produced using a different PVDF type (Solef 1015, which has higher Mw with respect to Solef 6012), lower polymer concentration (15 wt.%), and different additives (PVP K-17 and AMAL), with respect to fibers M1 and M2, as reported in Table 2. However, we added also this type of fibers in our analysis, to discuss the effect of different membrane structures on the membrane distillation performance. Their morphology arises from an interplay between different factors, particularly the dope composition and the air gap. Regarding the fibers outer surface, the formation of sponge-like layer can be connected to delayed demixing due to the used air gap distance of 25 cm. Regarding fibers inner surface, even if IPA 30% is a softer coagulant for PVDF with respect to pure water [43], tear-drop macrovoids can be noticed. The development of this structure can be connected to the low polymer concentration in the dope (15%), in agreement to what reported by Smolders et al. [44]. Other factors, such as combination of two additives (PVP and maleic anhydride) and different solvent type (DMF instead of NMP) could be also involved in the development of M3 features. Particularly, the higher DMF miscibility with water,

12

with respect to NMP [45] could be also a reason for the development of macrovoids, due to faster coagulation rate.

4.2. Mechanical properties The mechanical properties of the PVDF hollow fibers produced are reported in Table 4. The fiber type M1 shows the best mechanical properties among the tested ones, in terms of both Young’s modulus/tensile strength and elongation at break. Fibers type M2 show the lowest Young’s Modulus; while, fibers type M3 are characterized by lower tensile strength and elongation at break. It is well known that the mechanical properties are strongly related with the microstructure developed during the fiber formation. Therefore, fibers mechanical properties can be explained taking into account the dope composition and coagulations conditions employed during spinning experiments. Fibers M1 are characterized by sponge-like structure, which is consequence of dope and coagulants composition, as explained in Section 3.1. This structure imparts superior mechanical properties to the fibers. On the other hand, finger-like macrovoids represent weak points in the structure and, hence, the corresponding fibers show mediocre mechanical properties. The increased porosity also generates the fibers with weaker mechanical properties. Relatively substandard mechanical properties exhibited by M2 can be related with the application of water as both inner and outer coagulant, which produces a structure containing finger-like macrovoids at both the faces due to instantaneous de-mixing. The low graded mechanical character of M3 can be associated with the lower polymeric composition in the dope and, consequently, the higher porosity achieved.

4.3. Porosity As shown in Table 4, the porosity of all the membranes obtained in current work is quite high, and ranges from 79% to 83%. As also observed in section 3.2, fibers properties can be explained taking into account dope composition and coagulation conditions, which influence de-mixing rate during coagulation and, hence, final fibers morphology. The porosity of fibers types M1 and M2 is similar. Therefore, in this case, fibers void fraction is mostly affected by dope composition, which is the same in the two cases; while, the bore fluid composition has less

13

effect. Surprisingly, the finger-like macrovoids observed in fibers M2 structure do not contribute to increase the overall fibers void fraction. This let us to conclude that the sponge-like structure sandwiched between the two finger-like layers in fiber M2 (Figure 3) should be tighter, with respect to fibers type M1. The higher porosity observed for M3 can be related to the reduced amount of polymer and the additional pore forming additive into the dope solution, which affect fibers morphology as explained in Section 3.1. Fibers porosity is enhanced by the presence of finger-like macrovoids, which are predominant with respect to the thin sponge-like layer in fibers structure, as shown in Figure 4.

4.4. Bubble point and pore size distribution The results of the characterization procedures are reported in Table 4. In general, fibers pore size distributions are sharp, with bubble points ranging from 1.3 (Fiber M3) to 2.8 (Fiber M2) bar, corresponding to largest pore size of 0.36 and 0.16 µm respectively. For fibers M1, the average pore size is about 0.2 µm. Fibers type M2 shows lower average pore size (about 0.12 µm). The observed average pore size values of fibers M1 and M2 can be explained taking into account the preparation procedure. For fiber M1, the use of “soft” coagulants, both as bore fluid and coagulation bath, delayed liquid/liquid demixing during coagulation and promoted the formation of larger pores with respect to M2. In agreement with what observed by Liu et al. [46], when only water is used as bore liquid, as for M2, fast solvent/nonsolvent exchange could result in small pores. Also in ref. 39 it is observed that the rapid phase inversion and polymer solidification in the case of strong nonsolvents, such as water, does not allow a significant pore growth. Fibers type M3 show larger average pore size, with respect to M1 and M2, and pore size distribution located between 0.36 and 0.32 µm. The use of lower polymer concentration in the dope, together with two pore forming additives (PVP and Maleic anhydride), and, also, use of “soft” inner coagulant (IPA 30%) resulted in the development of porous structure. As widely accepted, lower polymer concentration results, in general, in enhanced coalescence of the polymer lean phase during phase inversion. Moreover, Liu et al. [46] observed that, when solvent/non solvent exchange takes place with slower rate, the initiated nuclei get the chance to grow before phase separation occurs. Therefore, the dope

14

composition and the coagulation conditions used during fibers M3 spinning resulted in larger pore size.

4.5. Membrane distillation The PVDF hollow fibers produced in this study were assembled in modules to be tested in MD working both under vacuum and direct contact mode. In this section, the results obtained in both configurations will be illustrated and discussed. Before any VMD test, fibers were checked to verify that, after vacuum application, no liquid water passed from the fibers lumen (feed side) to the shell side of the module. For all the tested modules, no liquid droplets were evidenced on fibers outer surface, therefore all fibers were used for VMD experiments.

4.5.1 VMD results Looking at the results reported in Table 5, it can be noticed that fibers type M1 shows lower trans-membrane water vapor flux with respect to fibers M2 (15.06 kg/m2 h vs. 22.37 kg/m2 h) and that fibers type M3 shows the highest fluxes 41.78 kg/m2h. As widely accepted in literature, the MD trans-membrane water vapor flux is strongly associated with the membranes porosity and thickness [16, 47]. The amount of vapor transported through the membrane increases by increasing the membrane porosity. Similarly, reduced membrane thickness offers less resistance to the vapor transport through the membrane. The morphology of fibers M1 is dominated by the sponge-like structure with an average pore size of 0.19 µm. The limited flux is connected to both its thickness (0.38 mm) and structure. Fibers type M2 have an average pore size of 0.12 µm and are thicker (0.45 mm), but present a sponge-like structure only 0.1 mm thick, and have finger-like layers (which can be noticed in Figure 3) that offer a reduced resistance to the transport. Therefore, these fibers show higher VMD flux. Fibers M3 show typical dual layer structure, with reduced thickness (0.23 mm) and a very thin sponge-like layer (about 0.055 mm) at the outer surface, and finger-like elements close to the inner surface, as evidenced in Figure 4. As pointed out in Section 3.3, this morphology results in higher porosity with respect to the other fibers, due to the presence of finger-like structures, that,

15

together with a reduced thickness and higher average pore size (of about 0.32 µm), lead to a high trans-membrane flux.

4.5.2 DCMD results The results obtained in the DCMD tests performed using fibers type M1-M3 are shown in Figure 5 a-c, respectively. In all cases, it can be noticed that the trans-membrane flux increases by increasing feed flow rates for all the temperatures investigated. The increase in flow rate decreases the boundary layer thickness and thus the flux increases due to more effective heat transfer from the bulk to the membrane surface. For each flow rate, trans-membrane flux increases when increasing the temperature at the feed side. This behavior can be related to the higher vapor pressure at the feed side that leads to higher partial pressure gradient between the two phases (higher driving force). As discussed in Section 3.1, the structure of fibers M1 is dominated by the sponge-like morphology, which offers more resistance to the mass transfer and, hence, limits the vapor transport across the membrane. Additionally, Table 4 indicates that fiber type M1 possesses high thickness, which also contributes in decreasing the value of the trans-membrane flux. Moreover, the presence of some fingers at the outer surface could lead to a partial wetting by the distillate stream that flows at the shell side. This would imply that not all the membrane cross section is available for the vapor transport and that the overall membrane resistance increases. The maximum flux attained reached the value of 14 kg/m2 h. Fibers type M2 show trans-membrane water vapor fluxes higher than that for M1. As discussed in Section 3.1, and evidenced in MD tests carried out under vacuum, these fibers have finger-like structures on both inner and outer surface. The presence of finger-like structure reduces the resistance to vapor transport and effective tortuosity along the membrane thickness, as well as the heat loss by conduction. However, significant increase of the trans-membrane flux with respect to fibers dominated by the sponge-like structure was not observed when working under the direct contact mode, on the contrary of what observed during VMD tests. This result could be probably due to the partial wetting of the finger-like structure at the outer surface of the fibers, present in both membrane types (M1 and M2). As reported above, in partially wetted membranes, the cross section dry and available for the vapor transport is reduced and the overall resistance offered by the membrane increases. Therefore, the gain in membrane flux, that could

16

be obtained with fibers type M2 thanks to the presence of the finger-like structure (characterized by a lower membrane resistance), is, in fact, reduced because of the higher resistance offered by the wetted portion of the membrane. Similarly to what observed in VMD tests, the obtained flux is the maximum for fibers type M3 among all the tested ones. As already explained above, this can be related with the higher porosity and pore size, presence of the fingers and the lowest thickness of these fibers among all the used ones. In this case, a spongy structure is observed on the outer surface and, thus, no wetting phenomena by the distillate stream can occur. The maximum flux achieved approaches the value of 22 kg/m2 h, as shown in Figure 5c. Again, the gain in flux is not so high as for VMD, due to the increase of the heat loss by conduction at lower membrane thickness. As discussed in literature [16], sponge-like structure narrows down the difference between DCMD and VMD flux. As the structure changes from spongy to the one containing macrovoids, the difference in flux becomes more and more significant. The results obtained for all the investigated fibers and that for commercial polypropylene (PP Accurel® S6/2) are compared in Figure 6, which shows the average flux obtained for each fiber at different feed temperatures. The flux obtained for each type of fiber at various flow rates and at a specific temperature was averaged and the results are shown in the figure. It can be noticed that the maximum flux is obtained with the fibers type M3 at all the tested temperatures.

4.5.3 Comparison between VMD and DCMD Looking at the results obtained in VMD and DCMD experiments, it can be noticed that the trend of trans-membrane flux for various types of fibers is same in both configurations i.e. the highest flux is exhibited by fiber type M3 while M1 shows the limited values of flux, flux for M2 lays between M3 and M1. The heat transferred through the membrane is governed by the following equation

Qm = k m / δ (TFm − TPm ) + J λ Where km is the thermal conductivity of the membrane, δ is the membrane thickness and λ is the latent heat of vaporization. Hot and cold phases separated by thin membranes in DCMD cause the thermal losses through the membrane. Moreover, the vapors transported from the hot to the cold side produce the local increase in temperature at the distillate side. Consequently, the effective vapor pressure gradient between the two phases decreases. In VMD, the conductive

17

losses can be neglected due to the high vacuum applied at the other side of the membrane. Thus, no temperature profile is present at the permeate side. As already reported, the mass transfer flux through the membrane is proportional to the vapor pressure gradient across the membrane, which is a function of the membrane surface temperatures.   B  B  ∆Pm = exp  A −  − exp  A −  TFm − C  TPm − C   

(1)

In case of vacuum membrane distillation, the second term in above equation equals the vacuum pressure applied and therefore, the effective pressure gradient across the membrane increases. Due to the lower driving force and the heat loss by conduction, trans-membrane water vapor fluxes obtained while working in the direct contact mode are lower with respect to those obtained working under vacuum for the same feed temperature and flow-rate.

The comparison of the fluxes obtained for the two configurations at the same feed inlet temperatures and flow rates (50°C and 6L/h) is given in Figure 7. The general trend observed for trans-membrane water vapor flux among the tested fibers is the same working under both MD configuration, i.e. M3>M2>M1. However, the difference between the VMD and DCMD performance becomes much more evident going from M1 to M3. As already discussed, this could be attributed to the different fibers morphology and properties (M1 mainly characterized by a sponge-like structure; M2 and M3 asymmetric type: M2, sponge structure “sandwiched” between finger-type structures at inner and outer surfaces; M3, thin sponge layer at the outer surface and finger elements at the inner surface, together with higher porosity and pore size and lower thickness than M1 and M2). In the case of M1, the high thickness of the dominant spongelike structure increases the resistance to the vapor transport while reducing the heat loss across the membrane. Therefore, the difference between VMD and DCMD fluxes is reduced. Due to the same reasons, when reducing the membrane thickness and increasing the percentage of macrovoids in the fibers structure, i.e. going to M2 and, especially, to M3, the difference between DCMD and VMD performance becomes much more evident. The highest difference observed between the two MD configurations, achieved for fibers type M3, can be explained

18

taking into account that, due to the fibers reduced thickness, the heat losses by conduction increase and affect the DCMD flux, whereas the VMD flux benefits of the reduced resistance to the transport.

4.5.4. Comparison of membrane properties and performance As described in the Introduction, the aim of the present work was to investigate the behaviour of different fiber morphologies in membrane distillation. In particular, PVDF hollow fibers were produced by changing some of the operating parameters used in a previous work [18]. The comparison between properties and VMD performance of fibers produced in this and in the previous work is shown in Table 6 and Fig. 8, respectively. 1) Comparison M1-previous work: The different air gap (25 vs. 14 cm) accounts for the different thickness. In fact, thicker fibers are expected to be produced with shorter air gap due to less stretch stress on the nascent membrane by the gravity effect. Slightly higher porosity of M1 could be justified due to the lower polymer percentage (18 wt% instead of 20 wt%) in the dope. The similar dope composition explains the similar morphology. As a result, the fibers type M1 produced in this work show similar performance with respect to those produced under similar conditions in Ref. 18. The plot reported in Fig. 8 highlights that similar VMD fluxes are obtained with symmetric spongy fibers. 2) Comparison M2-previous work: M2 morphology and higher thickness with respect to fibers produced in ref. 18 are due to the bore fluid composition. In fact, pure water, used as bore fluid, promotes the formation of fingerlike macrovoids as already explained in Section 4.1. Moreover, fibers produced in ref. 18 with higher solvent percentage in the bore fluid show reduced thickness, due to enhanced inflation of the inner lumen by the bore fluid pressure, in agreement with literature [39, 50]. Fibers porosities are similar. It can be concluded that fibers M2 show higher flux due to their asymmetric morphology, as highlighted in Fig. 8. 3) Comparison M3-previous work:

19

The comparison is difficult due to different dope compositions, i.e. polymer molecular weight and concetration, operating temperature and presence of further additive. Thickness and porosity of fiber M3 are similar to that of fibers of ref. 18. The asymmetric morphology and the larger pore size are the reasons for the higher VMD flux (as shown in Fig. 8) even if comparison with fibers spun with same bore fluid (isopropanol 30%) was not applicable since data are not available in ref. 18.

20

5. Conclusions In this work, microporous hydrophobic PVDF hollow fibers were produced by the phase inversion method, varying polymer Mw and concentration, type and concentration of different additives (H2O, PVP and AMAL) in the polymeric dope, composition of inner and outer coagulants. The effect of all these parameters on the morphology of the produced hollow fiber membranes was analyzed. A correlation between the developed morphology, fibers characteristics in terms of thickness and porosity, and the trans-membrane water vapor flux in MD was observed. In general, fibers showing macrovoids in the structure possess higher porosity, reduced mechanical strength and enhanced flux. The best mechanical properties are obtained when a weak non-solvent, like NMP or EtOH solution, is used as the bore fluid and external coagulant. It has been shown that the introduction of further pore forming additive (AMAL), combined with the relatively lower composition of polymer into the dope solution (Fiber type M3), leads to the structures exhibiting maximum trans-membrane flux working under both MD configurations (JVMD (50°C)= 41.78 kg/m2 h, JDCMD (70°C)= 21.78 kg/m2 h). Furthermore, the fluxes obtained are higher or comparable to that mentioned in literature. In particular, the VMD flux is 90% higher than that obtained with the commercial membranes (PP Accurel) when using the M3 membrane. This interesting result can be attributed to the asymmetric structure of the produced (M3) membrane where the spongy layer has a reduced thickness and is supported by a very open finger type structure.

21

Acknowledgements The authors acknowledge the financial support of the King Abdulaziz City for Science and Technology (KACST) Kingdom of Saudi Arabia. The authors are grateful to Dr. M. Davoli (Department of Earth Science, University of Calabria, Via P. Bucci, 87030 Rende (CS), Italy) for the use of SEM microscope.

22

References

[1] Membrane Distillation: Basic aspects and applications - A Journal of Membrane Science

Virtual Special Issue. Edited by Enrico Drioli, Francesca Macedonio, Aamer Ali. Journal of

Membrane

Science.

Elsevier.

2012.

Link

al

Virtual

Special

Issue:

http://www.journals.elsevier.com/journal-of-membrane-science/virtual-special issues/membrane-

distillation-basic-aspects-and-applications/

[2] M. Tomaszewska, M. Gryta, A. W. Morawski, The influence of salt in solutions

on hydrochloric acid recovery by membrane distillation Sep. Purif. Technol., 14 (1998) 183–188.

[3] E. Curcio, E. Drioli, Membrane Distillation and Related Operations - A Review, Sep.Purif.

Rev., 34 (2005) 35 – 86.

[4] M. Tomaszewska, Membrane Distillation - Examples of Applications in Technology and

Environmental Protection, Pol. J. Environ. Stud., 9 (2000) 27-36.

[5] X. Ji, E. Curcio, S. Al Obaidani, G. Di Profio, E. Fontananova, E. Drioli, Membrane

distillation-crystallization of seawater reverse osmosis brines, Sep. Purif. Technol., 71 (2010) 76-

82.

[6] M. Gryta, K. Karakulski, M. Tomaszewska, A. Morawski, Treatment of effluents from

the regeneration of ion exchangers using the MD process, Desalination, 180 (2005) 173–180.

[7] E. Drioli, A. Criscuoli, E. Curcio, Membrane Contactors: Fundamentals, Applications and

Potentialities, Membr. Sci. Technol. Ser. 11, Amsterdam; Boston: Elsevier, 2006.

[8] M. K. Souhaimi, T. Matsuura., Membrane distillation: principles and applications,

Amsterdam, Elsevier, 2011.

[9] M. Khayet, Membranes and theoretical modeling of membrane distillation: A review, Adv.

Colloid Interface Sci., 164 (2011) 56–88.

[10] M. Khayet,

T. Matsuura, J. I. Mengual, M. Qtaishat, Design of novel direct contact

membrane distillation membranes, Desalination, 192 (2006) 105–111.

[11] M.S. El-Bourawi, Z. Ding, R. Ma, M. Khayet, A framework for better understanding

membrane distillation separation process, J. of Membr. Sci., 285 (2006) 4–29.

[12] J. A. Prince, G. Singh, D. Rana, T. Matsuura, V. Anbharasi, T.S. Shanmugasundaram,

Preparation and characterization of highly

hydrophobic

poly(vinylidene fluoride)Clay

nanocomposite nanofiber membranes (PVDF–clayNNMs) for desalination using direct contact

membrane distillation, J. of Membr. Sci., 397–398 (2012) 80–86.

23

[13] F. Edwie, M. M. Teoh, T.-S. Chung, Effects of additives on dual-layer hydrophobic–

hydrophilic

PVDF hollow

fiber

membranes for membrane

distillation and

continuous

performance, Chem. Eng. Sci., 68 (2012) 567–57.

[14] M. Teoh, T. Chung, Membrane distillation with hydrophobic macrovoid-free PVDF–PTFE

hollow fiber membranes, Sep. Purif. Technol., 66 (2009) 229–236.

[15] K.Y. Wang, S.W. Foo, T. Chung, Mixed matrix PVDF hollow fibre membranes with

nanoscale pores for desalination through direct contact membrane distillation, Ind. Eng. Chem.

Res., 48 (2009) 4474–4483

[16] Y. Tang, Na Li , A. Liu, S. Ding, C. Yi, H. Liu, Effect of spinning conditions on the

structure and performance of hydrophobic PVDF hollow fiber membranes for membrane

distillation, Desalination, 287 (2012) 326–339.

[17] H. Fan, Y. Peng, Application of PVDF membranes in desalination and comparison of the

VMD and DCMD processes, Chem. Eng. Sci., 79 (2012) 94–102.

[18] S. Simone, A. Figoli, A. Criscuoli, M.C. Carnevale, A. Rosselli, E. Drioli, Preparation of

hollow fibre membranes from PVDF/PVP blends and their application in VMD, J. Membr. Sci.,

364 (2010) 219-232

[19] J. Ji, C. Di Cecca, E. M. Schulz, M. Mehta, D. Stead, D. McKinley, D. H. Koch, Hollow

fibre microfiltration membranes and a method of making these membranes, Patent N° US

6,890.435 B2, (2005)

[20] J.J. Qin, Y.M. Cao, Y.Q. Li, Y. Li, M.H. Oo, H. Lee, Hollow fibre ultrafiltration

membranes made from blends of PAN and PVP, Sep. Purif. Technol., 36 (2004) 149–155.

[21] B. Jung, J.K. Yoon, B. Kim, H.W. Rhee, Effect of molecular weight of polymeric additives

on formation, permeation properties and hypochlorite treatment of asymmetric polyacrylonitrile

membranes, J. Membr. Sci., 243 (2004) 45–57.

[22] J.J Qin, M.H. Oo, Y. Li, Development of high flux polyethersulfone hollow fibre

ultrafiltration membranes from a low critical solution temperature dope via hypochlorite

treatment, J. Membr. Sci., 247 (2005) 137–142. [23] L.S. Wan, Z.K. Xu, Z.G. Wang, Leaching of PVP from polyacrylonitrile/PVP blending membranes: a comparative study of asymmetric and dense membranes, J Polym. Sci. Pol. Phys.,

44 (2006) 1490–1498.

24

[24] S. Rouaix, C. Causserand, P. Aimar, Experimental study of the effects of hypochlorite on

polysulfone membrane properties, J. Membr. Sci., 277 (2006) 137–147.

[25] F. Tasselli, J.C. Jansen, F. Sidari, E. Drioli, Morphology and transport property control of

modified poly(ether ether ketone) (PEEKWC) hollow fiber membranes prepared from

PEEKWC/PVP blends: influence of the relative humidity in the air gap, J. Membr. Sci., 255

(2005) 13–22.

[26] L. Wu, J. Sun, C. He, Effects of solvent sort, PES and PVP concentration on the properties

and morphology of PVDF/PES blend hollow fiber membranes, J. Appl. Polym. Sci., 116 (2010)

1566-1573

[27] R. Liikanen, J. Yli-Kuivila, R. Laukkanen, Efficiency of various chemical cleanings for

nanofiltration membrane fouled by conventionally-treated surface water, J. Membr. Sci., 195

(2002) 265-276

[28] S.H. Wolff, A.L. Zydney, Effect of bleach on the transport characteristics of polysulfone

hemodialysers, J. Membr. Sci., 243 (2004) 389-399

[29] I.M. Wienk, E.E.B. Meuleman, Z. Borneman, T. van den Boomgaard, C.A. Smolders,

Chemical treatment of a polymer blend: mechanism of the reaction of hypochlorite with

poly(vinyl pyrrolidone), J. Polym. Sci. A1, 33 (1995) 49–54.

[30] Q. Li, Z. Xu, L. Yu, “Effects of mixed solvents and PVDF types on performances of PVDF

microporous membranes”, J. Appl. Polym. Sci., 115 (2010) 2277-2287.

[31] M. L. Yeow, Y. T. Liu, K. Li, “Morphological Study of Poly(vinylidene fluoride)

Asymmetric Membranes: Effects of the Solvent, Additive, and Dope Temperature”, J.Appl.

Polym. Sci., 92 (2004) 1782–1789.

[32] A. Akbari, M. Hamadanian, V. Jabbari, A. Y. Lehi, M. Bojaran, Influence of PVDF

concentration on the morphology, surface roughness, crystalline structure, and filtration

separation properties of semicrystalline phase inversion polymeric membranes, Desalination and

Water Treatment, 46 (2012) 96-106

[33] M. G. Buonomenna, P. Macchi, M. Davoli, E. Drioli, Poly(vinylidene fluoride) membranes

by phase inversion: the role the casting and coagulation conditions play in their morphology,

crystalline structure and properties, Eur. Polym. J., 43 (2007) 1557–1572.

25

[34] X. Wang, . Zhang, D. Sun, Q. Ana, H. Chen, Formation mechanism and crystallization of

poly(vinylidene fluoride) membrane via immersion precipitation method, Desalination 236

(2009) 170–178.

[35] D. Lin, C. Chang, C. Chang, T. Chen, L. Cheng, Fine Structure of Poly(vinylidene fluoride)

Membranes Prepared by Phase Inversion from a Water/N-Methyl-2-pyrollidone/Poly(vinylidene

fluoride) System, J. Polym. Sci. Pol. Phys., 42 (2004) 830–842.

[36] E. Fontananova, J. C. Jansen, A. Cristiano, E. Curcio, E. Drioli, Effect of additives in the

casting solution on the formation of PVDF membranes, Desalination 192 (2006) 190–197.

[37] J.-H. Kim, K.H. Lee, Effect of PEG additive on membrane formation by phase inversion, J.

Appl. Polym. Sci., 138 (1998) 153–163.

[38] M. G. Buonomenna, A. Figoli, J. C. Jansen, E. Drioli, Preparation of Asymmetric PEEKWC

Flat Membranes with Different Microstructures by Wet Phase Inversion, J.Appl. Polym. Sci., 92

(2004) 576—591.

[39] F. Tasselli, J. C. Jansen, E. Drioli, PEEKWC Ultrafiltration Hollow-Fiber Membranes:

Preparation, Morphology, and Transport Properties, J. Appl. Polym. Sci., 91 (2004) 841–853.

[40] F. Tasselli, E. Drioli, Tuning of hollow fiber membrane properties using different bore

fluids, J. Membr. Sci., 301 (2007) 11–18.

[41] S.P. Deshmukh, K. Li, Effect of ethanol composition in water coagulation bath on

morphology of PVDF hollow fiber membranes, J. Membr. Sci., 150 (1998) 75-85.

[42] A. Xu, A. Yang, S. Young, D. deMontigny, P. Tontiwachwuthikul, Effect of internal

coagulant on effectiveness of polyvinylidene fluoride membrane for carbon dioxide separation

and absorption, J. Membr. Sci., 311 (2008) 153–158.

[43] P. Sukitpaneenit, T. S. Chung, Molecular elucidation of morphology and mechanical

properties of PVDF hollow fiber membranes from aspects of phase inversion, crystallization and

reology, J. Membr. Sci., 340 (2009) 192-205.

[44] C.A. Smolders, A.J. Reuvers, R.M Boom, I.M. Wienk, Microstructures in phase-inversion

membranes. Part I. Formation of macrovoids, J. Membr. Sci., 73 (1992) 259-275.

[45] Y.H. Teow, A.L. Ahmad, J.K. Lim, B.S. Ooi, Preparation and characterization of

PVDF/TiO2 mixed matrix membrane via in situ colloidal precipitation method, Desal., 295

(2012) 61–69.

26

[46] Y. Liu, G.H. Koops, H. Strathmann, Characterization of morphology controlled

polyethersulfone hollow fiber membranes by the addition of polyethylene glycol to the dope and

bore liquid solution, Journal of Membrane Science 223 (2003) 187–199

[47] S. Al-Obaidani, E. Curcio, F. Macedonio, G. Di Profio, H. Al-Hinai, E. Drioli, Potential of

membrane distillation in seawater desalination: thermal efficiency, sensitivity study and cost

estimation, J. Membr. Sci. 323 (2008) 85–98.

[48] S. Bonyadi, T.S. Chung, Flux enhancement in membrane distillation by fabrication of dual

layer hydrophilic hydrophobic hollow fiber membranes, Journal of Membrane Science 306

(2007), 134-146

[49] M. Khayet, T. Matsuura, Preparation and characterization of polyvinylidene fluoride

membranes for membrane distillation, Ind. Eng. Chem. Res., 40 (2001) 5710–5718 [50] Q. F. Alsalhy, K. T. Rashid, W. A. Noori, S. Simone, A. Figoli, E. Drioli, Poly(vinyl chloride) hollow-fiber membranes for ultrafiltration applications: Effects of the internal coagulant composition, Journal of Applied Polymer Science, 2012, Vol. 124 pp. 2087-2099. [51] M. Khayet, T. Matsuura, Preparation and characterization of polyvinylidene fluoride membranes for membrane distillation, Ind. Eng. Chem. Res., 40 (2001) 5710–5718.

Captions: Table 1: Main literature data for PVDF membranes Table 2: The compositions and operating conditions applied in spinning experiments Table 3: Spinning conditions used in ref. 18 for the preparation of selected fibers Table 4: Properties of the PVDF hollow fibers produced in this work and of the commercial PP membrane used Table 5: Results of the VMD tests performed using double distilled water as feed (Pvacuum 40 mbar, Tfin 50°C, Qfeed 6L/h) Table 6: Comparison of properties of fibers produced in this work and in ref. 18 Figure 1: a) VMD set-up and b) DCMD set-up used in this work. Figure 2: SEM pictures of fibers M1. a) Cross section (54X); b) Enlarged cross section (300X); c) Inner surface (10800X). Figure 3: SEM pictures of fiber M2. a) Cross section (54X); b) Enlarged cross section (300X); c) Inner surface (10800X). 27

Figure 4: SEM pictures of fiber M3. a) Cross section (54X); b) Enlarged cross section (300X); c) Inner surface (10800X). Figure 5: Asymptotic flux obtained at various temperatures and flow-rates for fibers type (a) M1, (b) M2, (c) M3. Figure 6: Average asymptotic flux obtained for various types of fibers at different temperatures. Figure 7: Comparison of flux obtained at the same feed conditions (50°C, 6 L/h) for different types of membranes used in DCMD and VMD tests. Figure 8: Comparison between VMD fluxes of fibers produced in this work and in ref. 18. Fibers morphologies are highlighted.

28

Table 1: Main literature data for PVDF membranes Tfin Feed inlet temperature

Membrane type

Membranes properties

Operating parameters

Ref.

d (µm)

MPS (µm)

CA (o)

Feed solution

DCMD

VMD

PVDF-FS

∂ (%) 78

82

0.49

145

35g/L NaCl

Tfin=73oC Tpin=25oC

-

18.9

-

[17]

PVDF-HF

79.17

199.5

0.1434

100.81

9.09% NaCl

Tfin=65-65.2oC Tpin=20.3-21.4oC

65±0.2oC, P=900 mbar

~16.5

~23.5

[16]

PVDF-HF

75.42±0.34

119.80±2.45

-

136.13±0.04

83.4±3.66

-

[13]

80±5

110

0.248±0.007

108±5

Tfin=80 oC, Tpin=17±0.5oC Tfin=80±0.5oC, Tpin=17.5±0.5oC

-

PVDF–PTFE PVDF–PTFEHF PVDF-Clay nanocompositeNF PVDF-FS

3.5% NaCl 3.5% NaCl

-

40.4

-

[14]

81±3

300±25

0.64±0.22

154.2±3.0

3.5% NaCl

Tfin=80oC Tpin=17±2 oC

-

~5.7

-

[12]

70.5±3.1

61.8± 6.1

0.22

-

16.05

[51]

80.95± 0.4

203±26

0.16

Tfin=25oC, P=16.66 mbar Tfin=50oC, P=20mbar

-

PVDF-HF

Chloroform/ Water mixture Double distilled water

18.5

[18]

HF FS NF MPS CA Tfin Tpin P

Hollow fiber Flat sheet Nanofiber Mean pore size Contact angle Feed inlet temperature Permeate inlet temperature Vacuum pressure

1  

Flux (kg/m2.h) DCMD VMD

Tfin=73oC Tpin=25oC

Table 2: The compositions and operating conditions applied in spinning experiments Fiber

Polymer

Additives

Solvent

PVDF

T

Air gap

Spinning rate

(°C)

(cm)

(g/min)

Bore fluid (50°C)

M1

6012-18%

PVP(K-17) 15%, H2O 6%

NMP 62%

85

14

12

NMP 30%, 10 ml/min

M2

6012-18%

PVP(K-17) 15%, H2O 6%

NMP 62%

85

25

12

H2O 100%, 10 ml/min

M3

1015-15%

PVP(K-17) 15%, AMAL 10%

DMF 60%

60

25

12

IPA 30%, 17 ml/min

*Note: Fibers type M1 was prepared using Ethanol 30% as coagulation bath. In the other cases, tap water was used as other coagulant. Abbreviations list: PVDF = poly(vinylidene fluoride); PVP = poly (vinyl pyrrolidone); AMAL = maleic anhydride; NMP = N-methyl pyrrolidone; DMF = N,N-dimethyl formamide; IPA = isopropanol.

2

Table 3: Spinning conditions used in ref. 18 for the preparation of selected fibers Fiber

Polymer

Additives

Solvent

PVDF

T

Air gap

Spinning rate

(°C)

(cm)

(g/min)

Ref. 18 fiber 4b Ref. 18 fiber 4d

Bore fluid (50°C)

EtOH 30% 14 ml/min 6012-20%

PVP(K-17) 15%, H2O 6%

DMF 58%

Ref. 18 fiber 4g

85

25

12

DMF 25% 14 ml/min DMF 35% 14 ml/min

*Note: All fibers were prepared using tap water as coagulation bath. Abbreviations list: PVDF = poly(vinylidene fluoride); PVP = poly (vinyl pyrrolidone); DMF = N,N-dimethyl formamide; EtOH = Ethanol.

3

Table 4: Properties of the PVDF hollow fibers produced in this work and of the commercial PP membrane used Largest

Average

pore size

pore size

(bar)

(µm)

(µm)

(%)

0.60

2.093

0.219

0.191

80.90

192.88

0.30

2.789

0.164

0.121

79.11

2.62

163.82

0.19

1.267

0.362

0.318

83.39

4.16

174.40

1.11

0.672

0.682

0.2

70.00

O.D.

I.D.

Thickness

Emod

Rm

break

W

Bubble point

(mm)

(mm)

(mm)

N/mm²

N/mm²

%

Nm

M1

1.80

1.04

0.38

70.34

3.43

251.97

M2

1.59

0.70

0.45

57.15

2.84

M3

1.60

1.15

0.23

63.71

PP Accurel®

2.70

1.80

0.45

103.75

Fiber type

Note: The relative standard error is less than 5% in all cases.

4

Porosity

Table 5: Results of the VMD tests performed using double distilled water as feed (Pvacuum 40 mbar, Tfin 50°C, Qfeed 6 L/h) Fiber

VMD flux (H2O) (kg/h·m2)

M1

15.06

M2

22.37

M3

41.78

5

Table 6: Comparison of properties of fibers produced in this work and in ref. 18 Thickness

Porosity

(mm)

(%)

Spongy with some fingers at outer surface

0.174

81.23

Ref. 18 fiber 4d

Spongy with some fingers at outer surface

0.202

78.09

Ref. 18 fiber 4g

Spongy with some fingers at outer surface

0.139

78.45

M1

Spongy with some fingers at outer surface

0.38

80.9

M2

Asymmetric with fingers at inner and outer surface

0.45

79.11

M3

Asymmetric with fingers towards inner surface

0.23

83.39

Fiber

Morphology

Ref. 18 fiber 4b

6

Fig. 1 a-b

T

P Membrane module

Water trap

Heater

Peristaltic feed pump

Vacuum pump

Feed tank

(a)

T

P Membrane module

T

P

Feed tank

Cooler Heater Permeate tank 8. 2

Feed pump

g

Electric balance

Permeate pump

(b)

Figure 1: a) VMD set-up and b) DCMD set-up used in this work.

Figure 2

M1

M1

M1

Figure 2: SEM pictures of fibers M1. a) Cross section (54X); b) Enlarged cross section (300X); c) Inner surface (10800X).

Figure 3

M2

M2

M2

Figure 3: SEM pictures of fibers M2. a) Cross section (54X); b) Enlarged cross section (300X); c) Inner surface (10800X).

Figure 4

M3

M3

M3

Figure 4: SEM pictures of fibers M3. a) Cross section (54X); b) Enlarged cross section (300X); c) Inner surface (10800X).

Figure 5

14

2.

Flux (kg/m h)

12

10

8 6 L/h 7.8 L/h 9.6 L/h 11.4 L/h

6 50

60

70

Temperature (°C) (a)

18

14

2

Flux (kg/m .h)

16

12 10 6 L/h 7.8 L/h 9.6 L/h 11.4 L/h

8 6 50

60

Temperature (°C) (b)

70

21

2

Flux (kg/m .h)

18

15

12 6 L/h 7.8 L/h 9.6 L/h 11.4 L/h

9

6 50

60

70

Temperature (°C) (c)

Figure 5: Asymptotic flux obtained at various temperatures and flow-rates for fibers type (a) M1, (b) M2, (c) M3.

Figure 6

21

15

2

Flux (kg/m .h)

18

12 9

M1 M2 M3 PP

6 50

60

70

Temperature (°C) Figure 6: Average asymptotic flux obtained for various types of fibers at different temperatures

Figure 7

Figure 7: Comparison of flux obtained at the same feed conditions (50°C, 6 L/h) for different types of membranes used in DCMD and VMD tests.

Fig. 8

50 asymmetric

45 40

VMD flux (Kg/m2h)

35 30 25 fully sponge-like

20 15 10 5 0 Ref. 18 fiber 4d

Ref. 18 fiber 4g

M1

M2

M3

Figure 8: Comparison between VMD fluxes of fibers produced in this work and in ref. 18. Fibers morphologies are highlighted

Research highlights:  PVDF hollow fiber membranes were prepared under different processing conditions.  Different PVDF types and additives were used into the polymeric dope.  Comparison between fiber performances in direct contact and vacuum membrane distillation configurations was performed.  Correlation between preparation conditions, morphology and MD performance was established.