Preparation and Characterization of Polyvinylidene Fluoride

The effect of water in the casting solution on the geometrical properties of the membrane is discussed and evaluated. .... Fabrication of hierarchical...
0 downloads 12 Views 164KB Size
5710

Ind. Eng. Chem. Res. 2001, 40, 5710-5718

Preparation and Characterization of Polyvinylidene Fluoride Membranes for Membrane Distillation Mohamed Khayet and Takeshi Matsuura* Industrial Membrane Research Institute, Department of Chemical Engineering, University of Ottawa, 161 Louis Pasteur Street, P.O. Box 450, Station A, Ottawa, Ontario K1N 6N5, Canada

Polyvinylidene fluoride (PVDF) flat-sheet membranes were prepared for membrane distillation (MD). Pure water was used as a pore-forming additive in the casting solution. Dimethylacetamide (DMAC) was used as the solvent. The polymer solutions were cast over a glass plate or over a nonwoven polyester backing material. The prepared supported and unsupported PVDF membranes were characterized in terms of their nonwettability, pore size and porosity. MD experiments were carried out using a vacuum membrane distillation (VMD) configuration and employing pure water or chloroform/water binary mixtures as the feed. The influence of some relevant parameters, such as the feed temperature, stirring rate, or downstream pressure on the MD flux was studied. The effect of water in the casting solution on the geometrical properties of the membrane is discussed and evaluated. The dependence of the MD flux and separation factor on the geometrical properties of the supported and unsupported membranes was also studied. 1. Introduction The separation process known as membrane distillation (MD) usually refers to the thermally driven transport of vapor through microporous and hydrophobic membranes.1 The liquid feed is always in contact with the membrane and cannot penetrate inside dried membrane pores unless a transmembrane hydrostatic pressure that exceeds the so-called “liquid entry pressure of water (LEPw)” is applied.1-6 When the hydrostatic pressure is lower than the LEPw the feed liquid cannot enter the pores, and a liquid-vapor interface is formed at the entrance of each membrane pore. This side of the pore is hereafter called the feed side. The vapor pressure at the interface is the saturation vapor pressure of the liquid. If a vapor pressure difference, which is the driving force in MD, is maintained by applying vacuum on the other end of the pore, hereafter called the permeate side, the MD process is termed vacuum membrane distillation (VMD).7,8 Generally, MD is carried out for aqueous solutions or pure water using microporous and hydrophobic membranes prepared especially for microfiltration purposes. Such membranes are commercially available in capillaries or flat sheets and are typically fabricated from poly(tetrafluoroethylene) (PTFE), polypropylene (PP), or polyvinylidene fluoride (PVDF) materials. Their pore sizes range from 100 Å (0.01 µm) to 1 µm. A historical review of the membranes used in MD was reported by Lawson and Lloyd.2 As far as we know, very few studies have been carried out on the preparation of membranes for MD. Cheng and Wiersma9,10 patented two types of composite MD membranes. One consisted of a hydrophobic layer made of PTFE or PVDF and a hydrophilic layer made of cellulose acetate, polysulfone, cellulose nitrate, or polyallylamine. In the other type of membranes, a * Author to whom correspondence should be addressed. Tel.: +1 (613) 5625800 ext. 6114. Fax: +1 (613) 5625172. E-mail address: [email protected].

hydrophobic layer was maintained between two hydrophilic layers. Wu et al.11 used hydrophilic porous partitions such as cellulose acetate whose surface was treated via radiation graft polymerization of styrene to achieve the required hydrophobicity. In the same way, Kong et al.12 employed a cellulose nitrate membrane modified via plasma polymerization of both vinyltrimethylsilicon/ carbon tetrafluoride and octafluorocyclobutane. Fujii et al.13,14 prepared tubular membranes from PVDF polymer dopes by using the dry-jet wet-spinning technique. Ortiz de Zarate et al.15 reported asymmetric PVDF flat membranes manufactured using the phase-inversion method from binary solutions of PVDF/dimethylacetamide (DMAC) or PVDF/dimethylformamide (DMF). Tomaszewska16 studied the effect of the LiCl concentration in the casting solution on the permeation flux through PVDF membranes prepared for MD. The same type of membranes were prepared by Bottino et al.17 for ultrafiltration applications. It was observed that the porosity and permeation flux of PVDF membranes increased with increasing content of LiCl in the casting solution. However, the mechanical strength of the membranes decreased. Despite all of the above work, detailed studies concerning the design of porous PVDF membranes for MD are still lacking. A good porous MD membrane should exhibit high permeability or low membrane resistance, high LEPw, and low thermal conductivity. To obtain a high permeability, the surface layer that governs the membrane transport must be as thin as possible, and its surface porosity and pore size must be as large as possible. However, a high LEPw is achieved by a membrane with high hydrophobicity and a small pore size at its surface. In addition, in MD, conductive heat loss is associated with thin membranes and low porosity because the conductive heat transfer coefficient of the vapor within the membrane pores is about an order of magnitude smaller than that of the membrane matrix. Moreover, MD membranes must have good thermal stability and excellent chemical resistance to feed

10.1021/ie010553y CCC: $20.00 © 2001 American Chemical Society Published on Web 10/24/2001

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001 5711

streams. PVDF is one polymer that is thermally stable and resistant to most of the corrosive chemicals and organic compounds. Hence, PVDF is chosen in this work for the membrane material. A series of studies has been carried out in an effort to improve the properties of PVDF membranes by introducing nonsolvent additives in the PVDF polymer dope. Deshmukh and Li18 and Wang et al.19 introduced poly(vinylpyrrolidone) (PVP) in the PVDF casting solution as an additive in order to obtain highly porous PVDF membranes. They expected that the complete rinsing of all PVP molecules from the prepared membrane would be very difficult, especially for PVP with high molecular weights. In addition, trace quantities of PVP in the membrane affect the hydrophobicity of the PVDF membrane. Uragami et al. tested the effect of the addition of polystyrene sulfonic acid20 in the PVDF dope, as well as the influence of poly(ethylene glycol) (PEG).21 The permeability of the prepared membranes was improved, as expected. In contrast, the strength of the membranes was reduced. Glycerol and phosphoric acid were employed by Benzinger and Robinson22 as poreforming agents in PVDF membranes to increase the membrane permeability. Shih et al.23 used ethanol to increase the gas permeability through dried PVDF membranes. An attempt is made in this paper to correlate the PVDF membrane preparation conditions to some characteristics of membranes required for MD. Pure water was used as a nonsolvent additive in the PVDF casting solution for pore making to improve the MD permeability of the PVDF membranes and to decrease the cost of the membranes, which must be taken into account as an important parameter. The membranes prepared in this study were characterized by their nonwettability, pore size, and porosity. MD experiments were carried out using a vacuum membrane distillation (VMD) configuration and employing pure water and chlorofom/water binary mixtures as the feed. The dependence of the MD flux and separation factor on the geometrical properties of the membrane is discussed and evaluated. 2. Experimental Section 2.1. Materials. Commercially available polyvinylidene fluoride (PVDF), Kynar grade 740 (Elf Atochem, Philadelphia, PA), was used for the membrane material, and N,N-dimethylacetamide (DMAC) (synthesis grade, Merck, >99%) was employed as the solvent to prepare the polymer solution. Pure water was used as a nonsolvent additive. Ethanol (GR grade, Merck) was used for the solvent-exchange method. Isopropyl alcohol (GR grade, Merck) was used as the wetting liquid of the prepared PVDF membranes. 2.2. Membrane Preparation. Flat-sheet polyvinylidene fluoride (PVDF) membranes for membrane distillation (MD) have been prepared by the phaseinversion method. In the preparation of casting solutions, water/DMAC mixtures (water content from 0 to 8 wt %) were first prepared, and then PVDF was added to the mixtures so that the PVDF concentration in the solution became 15 wt %. Thus, the water content in the casting solution changed from 0 to 6.8 wt %. The mixture was stirred at approximately 55 °C for about 12 h to ensure complete dissolution of polymer. The polymer dopes prepared were transparent and homogeneous at room tempera-

ture. It is worth noting that the casting solution corresponding to 6.8 wt % of water turned into a gel after the 12-h preparation. However, with increasing temperature, the gel became a clear solution again. This polymer solution was not used to cast membranes. All of the other polymer solutions were cast over a glass plate or over a nonwoven polyester backing material (Osmonics, Inc.) at room temperature using a casting blade. The cast films, together with the glass plates or with the backing material if used, were immersed immediately in distilled water at 21 °C for gelation and were maintained in the water for at least 1 h. During gelation, the membrane peeled from the glass plate spontaneously, resulting in an unsupported membrane, whereas the membrane stayed on the backing material during gelation. Hence, the latter membrane is called a supported membrane. Because of the significant shrinkage observed in the case of the unsupported PVDF membranes during drying process, the latter membranes were subjected to solvent exchange after gelation. First, the membranes were immersed for approximately 4 h in an aqueous ethanol solution (50 wt %) and then in pure ethanol for 24 h. In this process, water in the membrane pores was replaced by ethanol, which has a lower surface tension. These membranes were subsequently dried at room temperature before characterization tests. It should be pointed out that, before gas permeation tests, all of the membranes were dried first at the ambient conditions and then under vacuum for 24 h to ensure complete drying of the membrane pores. 2.3. Membrane Characterization. 2.3.1. Membrane Liquid Entry Pressure of Water (LEPw) Measurements. LEPw is the pressure that must be applied to pure water before it penetrates into dried membrane pores. This pressure depends on the pore size and the hydrophobicity of the membrane. The apparatus used for this measurement is shown in Figure 1a. The membrane was placed in a static stainless steel cell between the upper chamber, the feed side, which was filled with pure water, and the lower chamber, the permeate side, which was connected to a digital capillary flowmeter (Varian Optiflow 420). First, a slight pressure (∼0.3 × 105 Pa) was applied to the system for at least 10 min; then, the pressure was increased stepwise with an increment of 0.68 × 103 Pa. The pressure at which a continuous flow was observed in the permeate side is the membrane LEPw. This method has been described extensively by Smolder and Franken.24 2.3.2. Membrane Porosity Measurements. The membrane porosity is defined as the volume of the pores divided by the total volume of the membrane. It can be determined by measuring the density of the polymer material using isopropyl alcohol (IPA), which penetrates inside the pores of the membrane, and the density of the membrane using pure water, which does not enter the pores. In this method, a pycnometer and a balance were employed. The following equation can be used to determine the porosity as suggested by Smolder and Franken24

)1-

Fm Fpol

(1)

where Fm is the density of the membrane and Fpol is the density of the polymer material. More details are provided by Smolder and Franken.24

5712

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001

Figure 1. Schematics of the experimental systems used for the (a) membrane liquid entry pressure of water (LEPw) measurements and (b) bubble-point and gas permeation tests.

2.3.3. Gas Permeation Tests. The average pore size and the effective porosity of the prepared membranes were determined by employing the gas permeation method. The effective porosity is defined as the ratio of the porosity and the effective pore length, which takes into account the tortuosity of the membrane pores. The experimental apparatus used for these measurements is shown in Figure 1b. The permeation flux of air through the dried membranes was measured at various transmembrane pressures, in the range of (0.14-0.97) × 105 Pa, at room temperature using a soap-bubble flow meter when the gas flow rate was low and a wet flowmeter when the flow rate was high. The effective membrane area was 10.76 × 10-4 m2. In general, the gas permeance, B, for a porous medium contains both a diffusive term and a viscous term that depends on the pressure as expressed by Carman25

B)

Pm r2 0.5 r 4 2 + ) I0 + S0Pm 3 πMRT Lp 8µRT Lp

(

)

(2)

where R is the gas constant, T is the absolute temperature, M is the molecular weight of the gas, µ is the gas viscosity, Pm is the mean pressure within the membrane pore, r is the membrane pore radius,  is the porosity, and Lp is the effective pore length. Thus, /Lp is the effective porosity. By plotting a linear dependence between the permeance, B, and the mean pressure, Pm, the intercept (I0) and the slope (S0) can be determined, and consequently, the pore radius and the effective porosity can be calculated using the following two equations

r)

( )( )

16 S0 8RT 3 I0 πM

8µRT  ) S0 Lp r2

0.5

µ

(3) (4)

2.3.4. Bubble-Point and Gas Permeation Tests (Wet and Dry Flow Method). The bubble-point and gas permeation tests, sometimes called the wet and dry flow method, are usually employed to determine the maximum pore size, the mean pore size and the pore size distribution of a membrane. The system used is shown in Figure 1b. First, the gas permeation velocity, Fdry, is measured through a dried membrane and plotted as a function of the transmembrane pressure difference, ∆P, and generally, a straight line is obtained. Then, the membrane is wetted by isopropyl alcohol (IPA), and again the permeation velocity, Fwet, is measured for different values of ∆P. In this experiment, the gas permeation velocity was measured at room temperature with the downstream side maintained at atmospheric pressure. The range of pressures on the upstream side depended on the membrane. Because the pore radius, r, is related to ∆P by the Laplace equation26,27

r)

2σ ∆P

(5)

where σ is the surface tension of IPA, the ratio between the wet and dry flow rates can be plotted versus the pore radius, as discussed in more detail in section 3.4. 2.3.5. Vacuum Membrane Distillation (VMD). VMD experiments were conducted using the laboratory system shown schematically in Figure 2. The central part is a static stainless steel cell connected to a heating system through its jacket to control the temperature of the liquid feed. The membrane was placed between the upper chamber (feed side) and the lower chamber (permeate side) and had an effective area of about 9.84 × 10-4 m2. The total volume of the upper chamber was 381.7 mL. The temperature inside the cell was measured after steady state had been reached by a sensor connected to a digital meter with an accuracy of (0.1 °C. The liquid feed was stirred inside the cell using a

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001 5713

Figure 3. LEPw versus water concentration in the casting solution. Table 1. Porosity and Thickness of the Prepared PVDF Membranes mem- water content brane in casting porosity name solution (wt %)a 

Figure 2. Schematics of the experimental setup used for the VMD experiments: (FC) feed container, (HE) heat exchanger, (T) temperature sensor, (MS) magnetic stirrer, (V) valve, (P) vacuum pressure controller, (CT) cold trap, and (VP) vacuum pump.

graduated magnetic stirrer. A vacuum pump (Welch DuoSeal 1400), which was connected to a vacuum pressure controller (MKS type 651), was connected to the permeate side of the cell to remove the vapor. The downstream pressure, which was varied in the range 933-3000 Pa, was measured with a digital pressure transducer with an accuracy of about (2%. Two glass cold traps filled automatically with liquid nitrogen were installed to recover the permeate. The MD flux was calculated in every case by weighing the condensate collected in each trap for a predetermined period. The initial concentration of chloroform in the feed was varied from 1000 to 2000 mg/L. The concentration of chloroform in the feed and permeate solutions was analyzed by a Varian 3300 gas chromatograph equipped with an Alltech Chemipack C18, 80/100 column. Practically all VMD experiments were repeated at least two times to ensure reproducibility of the measurements. 3. Results and Discussion 3.1. Membrane Liquid Entry Pressure of Water (LEPw) Measurement. The LEPw values obtained experimentally, together with their standard deviations, are presented in Figure 3 as a function of the water content in the PVDF casting solution for both the supported and unsupported membranes. The LEPw decreases as the concentration of water in the PVDF casting solution increases. This is because of the increase in the pore size of the membranes as the amount of water is increased in the polymer solution. In addition, the pore sizes of the supported membranes are larger than those of the unsupported membranes when the water concentration in the PVDF casting solution is below 5.1 wt %. In contrast, when the water concentration in the casting solution is either 5.1 and 5.95 wt %, the LEPw values or the pore sizes are almost the same for both supported and unsupported membranes. 3.2. Membrane Porosity Measurements. The values of the membrane porosity obtained experimentally,

thickness δ (µm)b

pore effective radius r (× porosity 10-2 µm) /Lp (m-1)

WS0 WS1 WS2 WS3 WS4 WS5 WS6 WS7

0 0.85 1.70 2.55 3.40 4.25 5.10 5.95

supported membranes 30.1 ( 4.2 53.9 ( 6.2 41.2 ( 6.1 60.3 ( 9.8 38.3 ( 3.9 55.2 ( 8.1 44.3 ( 3.3 49.7 ( 7.6 44.9 ( 4.5 55.9 ( 7.1 56.7 ( 5.7 60.6 ( 6.3 67.3 ( 3.1 64.2 ( 9.1 79.6 ( 2.6 62.7 ( 8.7

2.95 4.59 5.64 7.42 10.14 15.65 23.78 34.08

2851.78 3676.66 4689.53 5540.16 5574.99 6000.45 6472.18 7216.45

W0 W1 W2 W3 W4 W5 W6 W7

0 0.85 1.70 2.55 3.40 4.25 5.10 5.95

unsupported membranes 26.8 ( 3.5 50.7 ( 7.1 31.6 ( 2.8 50.2 ( 5.2 40.1 ( 2.6 53.1 ( 4.7 33.6 ( 5.1 49.8 ( 8.0 48.3 ( 4.6 52.8 ( 6.7 52.6 ( 2.8 50.6 ( 5.8 70.5 ( 3.1 61.8 ( 6.1 74.9 ( 3.7 68.3 ( 4.9

1.09 1.98 2.99 4.17 5.92 9.95 22.03 30.80

4363.41 4559.44 5949.30 6191.80 6728.93 7878.14 7505.52 8181.63

a For all membranes, PVDF ) 15 wt %. b Thickness of the support ) 114.5 µm.

together with their standard deviations, are shown in Table 1 for the supported and unsupported membranes. The membrane thickness was measured with a digital micrometer, with a precision of (0.1 µm, at 10 locations for each membrane and the average value is also presented in Table 1, together with its standard deviation. In general, the membrane porosity increases as the water content in the casting solution is increased for both the supported and the unsupported membranes. Consequently, the effect of pure water as an additive in the PVDF casting solution is to increase both the pore size and the porosity of the prepared membranes. It is worth noting that the porosity of the Millipore PVDF membranes with the trade name Durapore that are frequently used in MD processes is about 75% as supplied by the manufacturer. The obtained experimental values are 70.1 ( 2.7% for membrane GVHP with a pore size of 0.22 µm and 71.3 ( 3.4% for membrane HVHP with a pore size 0.45 µm. As can be observed in Table 1, the porosity of the membranes corresponding to 5.95 wt % of water in the PVDF casting solution is higher than the porosity of the Millipore PVDF membranes. 3.3. Gas Permeation Tests. The results obtained for the pore radius and effective surface porosity using the gas permeation method are shown in Table 1. The pore radius and effective porosity increase with increasing water content in the PVDF casting solution

5714

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001

Figure 4. Ratio of dried membrane gas permeation velocity to wetted membrane gas permeation velocity versus pore radius for the (a) supported and (b) unsupported PVDF membranes.

for both the supported and unsupported membranes. The pore radii of the supported membranes are larger than those of the unsupported membranes for a given water content in the PVDF solution. This is due to the shrinkage of pores that occurs in the unsupported membranes during membrane drying even after solvent exchange. 3.4. Bubble-Point and Gas Permeation Tests. The ratios between the wet and dry flow rates are plotted versus the pore radius for supported and unsupported membranes in Figure 4a and b, respectively. For each wetted membrane, the pores are filled with IPA at low ∆P, and the gas permeation velocity, Fwet, is practically zero. At a certain value of ∆P corresponding to the bubble point of the membrane, the largest pores will be opened, and the gas permeation velocity, Fwet, will start to increase. Smaller pores will be opened as ∆P increases according to the Laplace equation. At the pressure corresponding to the minimum pore size, all of the pores become empty. Fwet/Fdry is unity when ∆P is higher than this pressure. From Figure 4a and b, one can determine the maximum pore radius by the bubble point, the mean pore radius as the radius corresponding to Fwet/Fdry equal to 0.5, and also the pore size distribution for each membrane, as suggested by Kesting.26 Figure 5a and b shows the pore size distribution of the supported and unsupported membranes, respectively. For both the supported and unsupported membranes, the pore size distribution becomes narrower around the mean pore radius as the water content in the PVDF casting solution decreases. As well, the pore

Figure 5. Pore size distributions of the (a) supported and (b) unsupported PVDF membranes. ∆F is the variation of the gas permeation velocity with the applied pressure.

radii calculated by the two different methods (i.e., gas permeation test and bubble-point and gas permeation test) approach each other. The ratio between the pore radius determined by the gas permeation test and the pore radius determined by the bubble-point and gas permeation test varies between 0.8 and 1.3. 3.5. Pure-Water VMD Experiments. It is wellknown that, in MD processes, the feed temperature is the operating variable that significantly affects the MD flux because of the exponential increase of the vapor pressure with temperature. This type of dependence has been well investigated in a number of studies.1-8 In addition, the MD flux increases with stirring because of the decrease of the temperature polarization effect, i.e., when the feed solution is stirred, the heat transfer through the boundary layer adjoining the membrane surface increases, and the temperature of the feed at the membrane surface approaches the temperature of the bulk phase. Consequently, the driving force increases.1-8 Furthermore, the effect of the downstream pressure was studied extensively by Bandini et al.7 The permeate pressure must be lower than the saturation pressure of the feed solution to drive the vapor through the membrane pores while preventing condensation inside the membrane pores. In this work, the first sets of experiments were carried out using pure water as the feed solution. The MD flux

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001 5715

Figure 6. MD flux versus water content in the casting solution of the supported and unsupported membranes. Downstream pressure, 1666.5 Pa; feed temperature, 25 °C; stirring rate, 53.3 rps.

was measured under different conditions, and the effects of the feed temperature, stirring rate, permeate pressure, and backing material on the MD flux were studied. It is worth noting that VMD experiments could not be carried out using the membrane prepared from the casting solution containing 5.95 wt % of water because of the observed wetting of the membrane pores. It was observed that the MD flux increased exponentially with the water content in the PVDF casting solution. This is due to the increase of the pore size and the effective porosity of the membranes, as shown previously. To study the effect of the backing material on the MD flux, similar sets of experiments were conducted using the supported and unsupported membranes. The feed temperature, downstream pressure, and stirring rate were maintained at 25 °C, 1666.5 Pa and 53.3 rps, respectively. Each membrane was tested at least two times and the discrepancy between the obtained MD fluxes was found to be less than 5%. Figure 6 shows the obtained MD flux versus the concentration of water in the PVDF casting solution for both the supported and unsupported membranes. As stated earlier, the MD flux through the supported and unsupported membranes increases exponentially with the concentration of water in the PVDF casting solution, and the MD flux through the supported membranes is higher than that through the unsupported ones for the casting solutions with water concentrations equal to or lower than 4.25 wt %. The MD flux depends on both the pore size and the effective porosity of the membranes. Therefore, the MD flux was plotted in Figure 7 versus the product of the pore radius and the effective porosity obtained by the gas permeation test. A linear dependence was observed for both the supported and unsupported membranes, with correlation coefficients higher than 0.99. This indicates that, in the VMD configuration, Knudsen-type flow is the mechanism responsible for transport through the membrane pores. This conclusion was also obtained in other VMD studies.7 In addition, under the conditions studied, the mean free path of water molecules is larger than the pore size of the used membranes because the mean free path of water molecules is inversely proportional to the pressure. From the kinetic theory of gases, the molecule-pore wall collisions are dominant compared to the molecule-molecule collisions when the mechanism of transport of water vapor through the membrane pores is Knudsen-type diffusion.28 As a consequence, the following explicit expression can be

Figure 7. MD flux versus r/Lp for the supported and unsupported membranes. Downstream pressure, 1666.5 Pa; feed temperature, 25 °C; stirring rate, 53.3 rps.

used to determine the net MD coefficient2,3,5,7

B′k )

(

4 2 3 πMRT

)( ) 0.5

r Lp

(6)

Moreover, in Figure 7, the surface diffusion flux through dense membranes can be evaluated from the intercept of the straight lines. The obtained values are 0.26 and 0.17 g/(m2 s) for the supported and unsupported membranes, respectively. In MD processes, this type of flow is considered negligible because the surface diffusion area of the membrane matrix is small in comparison to the pore area and the interaction between the molecules and the membrane matrix is low. In the case of water, this flow is very low because of the hydrophobicity of the membrane. Nevertheless, this mechanism could have a significant effect once other components are present in the feed solution. Considering Knudsen-type flow, the temperature polarization coefficient (TPC) can be evaluated if we assume that no vapor loss occurs on the permeate side. The TPC is the ratio between the global MD coefficient, which takes into account the heat transfer boundary layer near the membrane surface, and the net MD coefficient, B′k, which does not consider the boundary layer. The net MD coefficient is calculated using eq 6 and the global MD coefficient is determined experimentally as the ratio of the experimental MD flux to the transmembrane vapor pressure difference. Generally, it was observed that the temperature polarization effect becomes more significant as the permeability of the membrane is increased for both the supported and unsupported membranes. The obtained values of the TPC decrease from 0.89 to 0.61 and from 0.94 to 0.68 for the supported and unsupported membranes, respectively. 3.6. Chloroform/Water VMD Experiments. Experiments were performed to study the effect of the initial concentration in the feed solution. Those experiments were carried out using three different initial concentrations, i.e., 1000, 1500, and 2000 mg/L. The feed temperature, stirring rate, and permeate pressure were maintained at 25 °C, 53.3 rps, and 1666.5 Pa, respectively. Figure 8a and b shows the plot of ln(C0/C) versus time for the unsupported membranes W0 and W5. C0 and C are chloroform concentrations at times 0 and t, respectively. It is observed that the plot is linear and that the slope depends on the membrane regardless of the initial concentration. This indicates that, for each membrane, the dimensionless concentration of chloro-

5716

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001

Figure 9. Overall mass transfer coefficient for the supported and unsupported membranes as a function of the water content in the casting solution. Temperature, 25 °C; downstream pressure, 1666.5; stirring rate, 53.3 rps; initial chloroform concentration, ∼1000 mg/L.

Figure 8. ln(C0/C) versus time for (a) membrane W0 and (b) membrane W5. Temperature, 25 °C; permeate pressure, 1666.5 Pa; stirring rate, 53.3 rps.

form is not influenced by the initial chloroform concentration in the feed solution. In this case, the overall mass transfer coefficient K can be determined using the relation29,30

K)

( )

C0 V ln At C

(7)

where V is the initial volume of the liquid in the feed side and A is the membrane area. Experiments were carried out under the same conditions as in Figure 8a and b using an initial chloroform concentration of about 1000 mg/L for the supported and unsupported membranes. From the experimental data, the overall mass transfer coefficients were calculated and are plotted in Figure 9 as a function of the concentration of water in the casting solution. The overall mass transfer coefficients of the supported membranes are higher than those corresponding to the unsupported membranes when the water concentration is lower than 4.25 wt %. This is due to the higher permeability of the supported membranes, as can be observed in Figure 6. However, at the 4.25 and 5.1 wt % water concentrations in the casting solution, the overall mass transfer coefficients are lower for the supported membranes. This can be attributed to the resistance of the backing material, as was suggested by the resistances in series model in which the overall mass transfer coefficient is given by Gronda et al.29 as

1 1 1 1 1 ) + + + K Kfeed Kmembrane Ksupport Kpermeate

(8)

Figure 10. 1/K versus 1/(r/Lp) for the supported and unsupported membranes. Temperature, 25 °C; downstream pressure, 1666.5 Pa; stirring rate, 53.3 rps; initial chloroform concentration, ∼1000 mg/L; dashed line for the supported membrane; solid line for the unsupported membrane.

where Kfeed, Kmembrane, Ksupport, and Kpermeate are the mass transfer coefficients through the liquid feed boundary layer, the membrane, the backing material, and the permeate, respectively. It should be pointed out that Kmembrane is proportional to r/Lp. Figure 10 shows the plot of the reciprocal of the overall mass transfer coefficient, 1/K, versus 1/(r/Lp) for both the supported and unsupported membranes. As can be observed, the data could be fitted by straight lines with correlation coefficients of 0.97 and 0.94 for the supported and unsupported membranes, respectively. If we assume that the resistance in the permeate side is negligible at a downstream pressure as low as 1666.5 Pa, the mass transfer through the liquid feed boundary layer, Kfeed, can be evaluated from the intercept of the line for the unsupported membranes. The obtained value is Kfeed ) 3.295 × 10-5 m/s. This value depends on the hydrodynamic aspect in the feed side of the membrane cell as was determined by Gronda et al.,29 Urtiaga et al.,30,31 and Jou et al.32 On the other hand, the sum of the feed side resistance and the membrane support resistance, 1/Kfeed + 1/Ksupport, can be determined from the intercept of the line for the supported membranes. The obtained value, 29 706.34 s/m, is almost the same as the mass transfer in the feed, 1/Kfeed, obtained using the unsupported membranes (30 345.61 s/m). This indicates that the effect of the membrane support is negligible in our experiments.

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001 5717

Acknowledgment

Table 2. Results of the Chloroform/Water VMD Experiments Using the Supported and Unsupported Membranes membrane water content in permeate K name casting solution (wt %)a flux (g/m2 s) (10-5 m/s)

Rb

WS0 WS1 WS2 WS3 WS4 WS5 WS6

supported membranes 0 0.296 0.85 0.558 1.70 0.931 2.55 1.34 3.40 1.78 4.25 2.66 5.10 3.91

1.788 2.459 2.515 2.962 3.129 2.962 3.185

53.21 38.36 25.47 22.01 15.94 10.22 6.81

W0 W1 W2 W3 W4 W5 W6

unsupported membranes 0 0.178 0.85 0.336 1.70 0.634 2.55 0.862 3.40 1.385 4.25 2.493 5.10 4.46

1.453 1.676 2.403 2.738 2.740 3.241 3.297

72.45 45.23 34.91 29.63 18.85 12.93 7.89

a For all membranes, PVDF ) 15 wt %. b Mean value from the beginning to the end of the experiment.

In addition, the separation factor was calculated using the expression

R)

[xch/xw]p [xch/xw]f

(9)

where xch and xw are the mole fractions of chloroform and water, respectively. The superscripts p and f refer to permeate and feed, respectively. The results of the chloroform/water VMD experiments using the supported and unsupported membranes are presented in Table 2. The presented values of the separation factor refer to the mean values obtained from the beginning to the end of the experiment. As an example, for membranes W3 and WS6, the separation factors increased from the first 30 min to the last 30 min by about 3.2 and 2.5%, respectively. In the case of membranes WS3 and W6, decreases of about 1.8 and 3.7%, respectively, were observed. As can be seen in the Table 2, the separation factor decreases with increasing concentration of water in the PVDF casting solution for both the supported and supported membranes and is lower in the case of the supported membranes. The decrease of the separation factor is due to the increase of the permeation rate of water through the membrane pores, as was shown in Figures 6 and 7. 4. Conclusions Polyvinylidene fluoride (PVDF) flat-sheet membranes for membrane distillation (MD) with different pore sizes and porosities were prepared from casting solutions containing 15 wt % of PVDF, dimethylacetamide (DMAC), and different concentrations of pure water. The following observations were made: (1) The pore size and porosity of the prepared supported and unsupported membranes increase with increasing concentration of water in the PVDF casting solution. (2) The MD flux increases exponentially with the water content in the PVDF casting solution. (3) The MD flux and the overall mass transfer coefficient are higher for the supported membranes at concentrations of water in the PVDF casting solution below 4.3 wt %. (4) The separation factor is generally lower for the supported membranes.

The authors gratefully acknowledge the postdoctoral research grant provided by the University Complutense of Madrid to Mohamed Kahyet. The authors also thank Mr. Daniel Suk and Dr. Geeta Cowdhury for their help in membrane preparation experiments. Nomenclature Symbols A ) membrane area (m2) B ) permeance [mol/(m2 s Pa)] B′k ) net MD coefficient [mol/(m2 s Pa)] C ) chloroform concentration (mg/L) F ) gas permeation velocity (m/s) I0 ) intercept defined in eq 2 [mol/(m2 s Pa)] K ) overall mass transfer coefficient (m/s) Lp ) effective pore length (m) LEPw ) liquid entry pressure of water (Pa) M ) molecular weight (kg/kmol) Pm ) mean pressure (Pa) ∆P ) transmembrane pressure (Pa) R ) gas constant [J/(mol K)] r ) pore radius (µm) S0 ) slope defined in eq 2 [mol/(m2 s Pa2)] T ) absolute temperature (K) t ) time (min) V ) volume of liquid in the feed container (m3) X ) mole fraction Greek Letters R ) separation factor δ ) membrane thickness (m)  ) membrane porosity /Lp ) effective porosity (m-1) µ ) viscosity [kg/(m s)] F ) density (kg/m3) σ ) surface tension (Pa m) Superscripts f ) feed p ) permeate Subscripts ch ) chloroform dry ) dried membrane K ) Knudsen m ) membrane p ) pore pol ) polymer w ) water wet ) wetted membrane 0 ) initial (t ) 0)

Literature Cited (1) Mengual, J. I.; Pena, L. Membrane Distillation. Colloid Interface Sci. 1997, 1, 17. (2) Lawson, K. W.; Lloyd, D. R. Review: Membrane Distillation. J. Membr. Sci. 1997, 124, 1. (3) Khayet, M.; Godino, M. P.; Mengual, J. I. Modelling Transport Mechanism through a Porous Partition. J. Non-Equilb. Thermodyn. 2001, 26, 1. (4) Khayet, M.; Godino, M. P.; Mengual, J. I. Theory and Experiments on Sweeping Gas Membrane Distillation. J. Membr. Sci. 2000, 165, 261. (5) Khayet, M.; Godino, M. P.; Mengual, J. I. Nature of Flow on Sweeping Gas Membrane Distillation. J. Membr. Sci. 2000, 170, 243.

5718

Ind. Eng. Chem. Res., Vol. 40, No. 24, 2001

(6) Izquierdo-Gil, M. A.; Garcia-payo, M. C.; Fernandez-Pineda, C. Air Gap Membrane Distillation for Sucrose Aqueous Solutions. J. Membr. Sci. 1999, 155, 291. (7) Bandini, S.; Saavedra, A.; Sarti, G. C. Vacuum Membrane Distillation: Experiments and Modeling. AIChE J. 1997, 43, 2, 398. (8) Lawson, K. W.; Lloyd, D. R. Membrane Distillation: I. Module Design and Performance Evaluation using Vacuum Membrane Distillation. J. Membr. Sci. 1996, 120, 111. (9) Cheng, D. Y.; Wiersma, S. J. Composite Membrane for a Membrane Distillation System. U.S. Patents 4,316,772, 1982; 4,419,242, 1983. (10) Cheng, D. Y.; Wiersma, S. J. Apparatus and Method for Thermal Membrane Distillation. U.S. Patent 4,419,187, 1983. (11) Wu, Y.; Kong, Y.; Lin, X.; Liu, W.; Xu, J. Surface-Modified Hydrophilic Membranes in Membrane Distillation. J. Membr. Sci. 1992, 72, 189. (12) Kong, Y.; Lin, X.; Wu, Y.; Chen, J.; Xu, J. Plasma Polymerization of Octafluorocyclobutane and Hydrophobic Microporous Composite Membranes for Membrane Distillation. J. Appl. Polym. Sci. 1992, 46, 191. (13) Fujii, Y.; Kigoshi, S.; Iwatani, H.; Aoyama, M. Selectivity and Characteristics of Direct Contact Membrane Distillaiton Type Experiments: I. Permeability and Selectivity through Dried Hydrophobic Fine Porous Membranes. J. Membr. Sci. 1992, 72, 53. (14) Fujii, Y.; Kigoshi, S.; Iwatani, H.; Aoyama, M.; Fusaoka, Y. Selectivity and Characteristics of Direct Contact Membrane Distillaiton Type Experiments: II. Membrane Treatment and Selectivity Increase. J. Membr. Sci. 1992, 72, 73. (15) Ortiz de Zarate, J. M.; Pena, L.; Mengual, J. I. Characterization of Membrane Distillation Membranes Prepared by Phase Inversion. Desalination 1995, 100, 139. (16) Tomaszewska, M. Preparation and Properties of Flat-sheet Membranes from Polyvinylidene Fluoride for Membrane Distillation. Desalination 1996, 104, 1. (17) Bottino, A.; Capannelli, G.; Munari, S.; Turturro, A. High Performance Ultrafiltration Membranes Cast from LiCl Doped Solutions. Desalination 1988, 68, 167. (18) Deshmukh, S. P.; Li, K. Effect of Ethanol Composition in Water Coagulation Bath on Morphology of PVDF Hollow Fiber Membranes. J. Membr. Sci. 1998, 150, 75. (19) Wang, D., Li, K.; Teo, W. K. Preparation and Characterization of Polyvinylidene Fluoride (PVDF) Hollow Fiber Membranes. J. Membr. Sci. 1999, 163, 211. (20) Uragami, T.; Fujimoto, M.; Sugihara, M. Studies on Syntheses and Permeabilities of Special Polymer Membranes. 28.

Permeation Characteristics and Structure of Interpolymer Membranes from Poly(vinylidene fluoride) and Poly(styrene sulfonic acid). Desalination 1980, 34, 311. (21) Uragami, T.; Naito, Y.; Sugihara, M. Studies on Syntheses and Permeabilities of Special Polymer Membranes. 39. Permeation Characteristics and Structure of Polymer Blend Membranes from Poly(vinylidene fluoride) and Poly(ethylene glycol). Polym. Bull. 1981, 4, 617. (22) Benzinger, W. D.; Robinson, D. N. Porous Polyvinylidene Fluoride Membrane and Process for Its Preparation. U.S. Patent 4,384,047, 1982. (23) Shih, H. C.; Yeh, Y. S.; Yasuda, H. Morphology of Microporous Polyvinylidene Fluoride Membranes Prepared by Gas Permeation and Scanning Electron Microscopy. J. Membr. Sci. 1990, 50, 299. (24) Smolder, K.; Franken, A. D. M. Terminology for Membrane Distillation. Desalination 1989, 72, 249. (25) Carman, P. C. Flow of Gases Through Porous Media; Butterworth Publications: London, 1956. (26) Kesting, R. E. Synthetic Polymeric Membranes, 2nd ed.; John Wiley & Sons: New York, 1985. (27) Nakao, S. Review: Determination of Pore Size and Pore Size Distribution. 3. Filtration Membranes. J. Membr. Sci. 1994, 96, 131. (28) Mason, E. A.; Maulinaskas, A. P. Gas Transport in Porous Media: The Dusty Gas Model; Elsevier: Amsterdam, 1983. (29) Gronda, A. M.; Buechel, S.; Cussler, E. L. Mass Transfer in Corrugated Membranes. J. Membr. Sci. 2000, 165, 177. (30) Urtiaga, A. M.; Gorri, E. D.; Beasley, J. K.; Ortiz, I. Mass Transfer Analysis of the Pervaporative Separation of Chloroform from Aqueous Solutions in Hollow Fibers Devices. J. Membr. Sci. 1999, 156, 275. (31) Urtiaga, A. M.; Ruiz, G.; Ortiz, I. Kinetic Analysis of the Vacuum Membrane Distillation of Chloroform from Aqueous Solutions. J. Membr. Sci. 2000, 165, 99. (32) Jou, J.; Yoshida, W.; Cohen, Y. A Novel Ceramic-Supported Polymer Membrane for Pervaporation of Dilute Volatile Organic Compounds. J. Membr. Sci. 1999, 162, 269.

Received for review June 27, 2001 Revised manuscript received August 17, 2001 Accepted August 17, 2001 IE010553Y