- Email: [email protected]
Effects of organic solvents on ultrafiltration polyamide membranes for the preparation of oil-in-water emulsions
Journal of Colloid and Interface Science 287 (2005) 612–623 www.elsevier.com/locate/jcis
Effects of organic solvents on ultrafiltration polyamide membranes for the preparation of oil-in-water emulsions L. Giorno a,∗ , R. Mazzei a , M. Oriolo a , G. De Luca a , M. Davoli b , E. Drioli a a Institute on Membrane Technology, CNR-ITM, University of Calabria, Via P. Bucci 17/C, 87036 Rende (CS), Italy b Department of Earth Science, University of Calabria, Via P. Bucci 14/B, 87036 Rende (CS), Italy
Received 20 July 2004; accepted 11 February 2005 Available online 31 March 2005
Abstract Hydrophilic ultrafiltration membranes made of polyamide with molecular weight cutoff 10 and 50 kDa have been studied for the preparation of oil-in-water emulsions by a cross-flow membrane emulsification technique. Isooctane and phosphate buffer were used as disperse and continuous phase, respectively. The permeation of apolar isooctane through the polar hydrophilic membrane was achieved by pretreatment of membranes with a gradient of miscible solvents of decreasing polarity to remove water from the pores and replace it with isooctane. Four different procedures were investigated, based on the solvent mixture percentage and contact time with membranes. After pretreatment, the performance of the membranes in terms of pure isooctane permeate flux and emulsion preparation was evaluated. The influence of organic solvents on polyamide (PA) membranes has been studied by SEM analysis, which showed a clear change in the structure and morphology of the thin selective layers. The effects proved stronger for PA 10 kDa than for 50 kDa. In fact, similar pretreatment procedures caused larger pore size and pore size distribution for PA 10 kDa than for 50 kDa. The properties of emulsions in terms of droplet size distribution reflected the membrane pore sizes obtained after pretreatment. The correlation between pore size and droplet size, for the physicochemical and fluid dynamic conditions used, has been evaluated. 2005 Elsevier Inc. All rights reserved. Keywords: Membrane emulsification; O/W emulsion; Polyamide membrane; Solvent effect; Membrane pretreatment
1. Introduction Emulsions are mixtures that consist of a dispersed phase in a finely divided state uniformly distributed in a continuous phase. In the case of oil-in-water emulsions, finely divided oil droplets are uniformly dispersed in water [1,2]. During the last two years more than 100 publications dealing with microemulsions have appeared in the literature [3,4]. Microemulsions, or emulsions with droplet size in the range 20–80 nm, are thermodynamically stable mixtures of oil in water or water in oil. When droplets are larger, in order to have a stable emulsion, the interface between the oil and the water phase must be stabilized by surfactant mole* Corresponding author. Fax +39 0984 402103.
E-mail address: [email protected] (L. Giorno). 0021-9797/$ – see front matter 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.jcis.2005.02.015
cules, which prevent immediate aggregation or coalescence and whose properties largely determine the behavior of the emulsion. Emulsions destabilize in several ways. Sufficiently large droplets suspended in an aqueous medium of low viscosity or yield stress will cream. Creaming in itself does not destabilize an emulsion, but the high concentration of oil droplets in the creamed layer promotes interactions that lead to flocculation, aggregation, or coalescence [5]. It is well known that, since emulsions are thermodynamically unstable, flocculation and coalescence occur immediately after emulsification. Generally, the instability of an emulsion depends upon the emulsifying agent, droplet size, net charge, and mechanical and physical properties of the adsorbed film. In particular, the distribution of droplets is the most important parameter characterizing any emulsion. Stability and resistance to creaming, rheology, chemical re-
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
activity, and physiological efficiency are influenced by both the relative size and the size distribution [6]. With membrane emulsification, in principle, monodisperse emulsions can be produced, requiring a relatively low energy input, which implies that the shear stress exerted on the ingredients is low. This technology is therefore particularly suitable for emulsions containing labile compounds such as biomolecules. In cross-flow membrane emulsification the dispersed phase is forced to pass through the membrane, where it forms droplets at the pore mouth on the other side of the membrane. Here the droplets are detached by the action of a drag force provided by the tangential flow of the continuous phase flowing along the membrane surface. If the membrane pore has an adequately narrow size distribution, uniform droplets can be obtained by carefully controlling the pressure driving the disperse phase [7]. A limiting factor for emulsion production on a commercial scale is the low disperse phase flux. Advantages of this process are low energy consumption, control of droplet size and droplet size distribution, and the low shear stress that is needed [1]. This is determined by the balance between the drag force on the droplets from the flowing continuous phase, the buoyancy of the droplets, the interfacial tension forces, and the driving pressure. The final droplet size and size distribution are determined not only by the pore size and size distribution of the membrane but also by the degree of coalescence, both at the membrane surface and in bulk solution [8]. Membrane emulsification was developed by Nakashima et al. [9,10] for making monodispersed emulsions over a wide spectrum of mean droplet sizes, ranging from about 0.5 µm to several tens of micrometers [10]. At the present time, membrane emulsification is primarily used for the production of special “high-technology” products, where uniform droplets of controlled mean diameters are needed. These applications include preparation of core particles for toner application [11], preparation of liposome [12] spacers for liquid crystal displays [13] preparation of uniform silica hydrogel particles [14], and synthesis of monodisperse polymer microspheres. Such polymer spheres have been used as packings for HPLC columns [13], as immobilizing carriers of enzyme [15], and as biodegradable drug delivery systems [16]. Membranes used in emulsification process are mainly inorganic microfiltration membranes. Polymeric membranes are essentially confined to the treatment of aqueous solutions. This is because of materials difficulties; polymeric membranes in contact with organic solvents tend to swell and lose their separation capabilities [17]. Several commercial polymeric ultrafiltration membranes in contact with solvent can swell or dissolve, leading to unacceptable changes in solvent flux and solute separation [18]. Rapid solvent exchange between water and a high concentration of alcohol disrupts the polymer matrix [18]. A similar phenomenon was observed with alcohols such as ethanol, isopropanol, and isobutanol at concentrations higher than 20%. Permeability of alcohols and hydrocarbons
613
through different polymeric membranes has been measured [18]. Permeability was found to be influenced by molecular size, solubility, viscosity, etc. Although polymeric membranes are in general less stable in organic solvents and are primarily used for separation of aqueous solutions, the possibility of using polymeric membranes with organic solvents for various applications has been reported. Polymeric membranes are widely used in two separate phase membrane reactors for phase transfer catalysis [19] or in membrane contactors for membrane-based solvent extraction [20]. The performance of nanofiltration membranes with organic solvents has been recently discussed by Van der Bruggen [21]. Ultrafiltration hydrophilic membranes have been used to prepare oil-in-water emulsions by membrane emulsification [22]. The hydrophilic membranes were conditioned with organic solvents to allow permeation of nonpolar solvents through the membrane. In the present work, the influence of organic solvents on polymeric membrane properties has been clarified. Several pretreatment procedures based on different contact time with solvents and solvent mixtures of decreasing polarity have been investigated. The effects of organic solvents on asymmetric polyamide membranes with different cutoffs (10 and 50 kDa) have been studied by scanning electron microscopy (SEM) and permeability measurements to pure organic solvents. The stability and reproducibility of pretreated membranes have been evaluated. The conditioned membranes have been used to prepare oil-in-water emulsions, which have been characterized in terms of droplet size distribution and stability. Although a theoretical treatment of solvent influence on the properties of the polymeric membrane used was not the aim of this work, basic theoretical considerations were necessary to qualitatively interpret some fundamental empirical observations. In particular, the morphology of the polyamide membrane used admits a Darcy model to describe the disperse phase (oil) flux in first approximation. Consequently, for the oil to start flowing and producing droplets, it is necessary that the applied transmembrane pressure be higher than the pore critical pressure P c , which depends on the membrane pore diameters and oil–water interfacial tension. The critical pressure can be calculated with the Young–Laplace equation, i.e., Pc = γo,w /Rc , in which R c and γ o,w are the droplet curvature radius and the oil–water interfacial tension, respectively. More precisely, when the height of the growing droplet equals the radius of the pore, the P c , i.e., the maximum pressure resistance, is reached [23].
2. Materials and methods Isooctane (2,2,4-trimethylpentane, from Sigma Aldrich, Italy) was the organic solvent used as dispersed phase. Ultrapure water (bidistilled water, filtered with a membrane of pore size 0.2 µm), anhydrous sodium dihydrogen phosphate
614
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Table 1 Pretreatment procedures Method A
Method B
Method C
Method D
Solvents
Conc. (v/v)
Contact time (h)
Solvents
Conc. (v/v)
Contact time (h)
Solvents
Conc. (v/v)
Contact time (h)
Solvents
Conc. (v/v)
Contact time (h)
H2 O H2 O–IPA ISO
100 50:50 100
∼4 3 12a
H2 O H2 O–IPA IPA–ISO IPA–ISO ISO
100 80:20 50:50 20:80 100
∼4 12 12 12 12a
H2 O H2 O–IPA IPA–ISO IPA–ISO ISO
100 80:20 50:50 20:80 100
∼4 12 3 12 3a
H2 O IPA–ISO Pure ISO
100 50:50 100
∼4 1.5 3a
a After this time, the module was washed with isooctane and used to prepare the emulsion.
(NaH2 PO4 ), and disodium hydrogen phosphate anhydrous (Na2 HPO4 ), from Fluka (Germany) or Sigma Aldrich (Italy) were used to prepare phosphate buffer solutions. A 50 mM buffer phosphate at pH 7 containing sodium dodecyl sulfate (SDS) and polyvinyl alcohol (PVA) was used as the continuous phase. SDS and PVA (molecular weight 22,000 and 97.5–99% degree of hydrolysis) from Fluka were used as emulsifier and stabilizer (0.2 and 0.8%), respectively. Isopropanol (IPA) of analytical grade from Fluka was used with water or isooctane for membrane pretreatment, as indicated in the following. Polyamide capillary membranes with nominal molecular weight cutoff (NMWCO) of 10 and 50 kDa, having inner/outer diameter of 1.5/2.2 and 1.2/2.4 mm, respectively, were used. The membranes were kindly provided by Forschstung Institut Berghof, Germany. The structure of this kind of membrane is asymmetric, with the selective layer on the lumen side and the sponge layer on the shell side. The lab-made membrane modules were prepared by assembling six capillary membranes inside a Pyrex glass cylinder of inner diameter 1.2 cm, 20 cm long. The internal/external membrane surface area were 4.9 × 10−3 m2 /7.2 × 10−3 m2 for 10 kDa and 3.6 × 10−3 m2 /7.9 × 10−3 m2 for 50 kDa.
brane, it was necessary to pretreat the membrane to remove the internal liquid phase (water) and replace it with the organic solvent. The pretreatment consisted in washing the membrane with a gradient of miscible solvents and solvent solutions of decreasing polarity, to shift the internal phase from polar to nonpolar and to allow the permeation of the isooctane. Four different procedures were used (see also Table 1). Method A: the pretreatment consisted in the subsequent permeation of pure water, water–isopropanol (50:50), and pure isooctane. Methods B and C used pure water, water–isopropanol (80:20), isopropanol–isooctane (50:50 and 20:80), and then pure isooctane; the difference between B and C is the different contact time of membranes with the solutions (Table 1). The last method (D) consisted of the use of water, isopropanol–isooctane (50:50), and then pure isooctane. After all the pretreatments the membranes were washed with isooctane and the pure isooctane flux was measured in order to monitor membrane stability. Afterward, the membranes were used to prepare oil-in-water emulsions (isooctane in aqueous phase constituted by phosphate buffer pH 7 containing 0.2 and 0.8% SDS and PVA, respectively). During the preparation of emulsions, the permeate flux of the organic phase was about 2 l/(h m2 ).
2.1. Membrane pretreatment The original membranes were both asymmetric in the structure, with a thin selective layer on the lumen side. Since the size of the pores on the selective layer is very low (ultrafiltration range) and not easily detectable by scanning electron microscopy, the concept of nominal molecular weight cutoff was used instead. The cutoff gives an indirect measure of the pore dimension; it is determined on the basis of the size of molecules that are retained by the membrane at 90%. The membranes used originally had 10 or 50 kDa cutoff. We have studied the structure of the thin selective layer of these membranes and investigated how the structure and pore size changed after contact with organic solvents. The membranes were first washed with ultra pure water to remove the water-soluble residues, and then the pure water permeance (l/(h m2 bar)) was tested. Afterward, to allow the permeation of the apolar organic solvent (isooctane, log P 4.5) through the hydrophilic mem-
2.2. SEM analysis To understand the effects of solvent on the structure and morphology of the asymmetric polyamide membranes with different cutoffs, new and pretreated membranes were analyzed by scanning electron microscopy (SEM, Mod. Stereoscan 360, Cambridge Instrument). The analysis was carried out after the overall pretreatments and, to identify the influence of each solvent, the membranes were also analyzed after contact with each solvent (isooctane and isopropanol). In addition, characterization of membranes after long-term contact with isooctane was carried out. Cross sections of new and treated fibers were obtained in liquid nitrogen. Cross-section enlargements were by up to 40,000 times. The analysis of SEM pictures made it possible to determine the original structure of the thin selective layer and to measure the pore size and size distribution of membranes after treatment with the solvents.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
615
Fig. 1. Schematic of membrane emulsifier: capillary membrane module with inlet and outlet of the two separate circuits (lumen and shell); the enlargement shows the longitudinal section of the asymmetric capillary membrane.
2.3. Droplet measurement Droplets were measured by optical microscopy. Droplet samples were placed on a cell counter having nine squares, each one subdivided into 16 smaller well-defined squares having a side of 200 µm and a volume of 0.1 mm3 . The images were acquired by means of a video camera connected to a personal computer and then were processed by the “Scion Image” program. This software automatically counts the particles and measures geometrical parameters (e.g., diameter, area) of identified particles. The droplet size distribution was calculated on the basis of at least 10 processed samples. 2.4. Membrane emulsification equipment Oil-in-water emulsions were prepared using a cross-flow membrane emulsification system. A schematic of the membrane emulsifier is shown in Fig. 1, where the module with capillary membranes is represented; the apparatus where the module is located includes two control panels fixed up with valves and pressure gauges to control the fluid dynamic conditions of the two circuits, separately. The experiments have been carried out at room temperature (23 ± 2 ◦ C). The disperse phase (organic phase) was circulated in the shell side of membrane module and forced to permeate through the membrane by applying a transmembrane pressure of about 0.1 bar. The disperse phase was then collected in the continuous phase (aqueous phase), which was recirculated on the lumen side of the membrane, where emulsions were formed. The axial velocity of the aqueous phase was 5.2 (±0.6) × 10−5 m/s, unless otherwise specified in the Experimental section. The initial volumes of the aqueous and organic phases recirculated in the two separate circuits were 150 ml each. In general, a percentage of 15% organic phase in the aqueous phase was obtained.
3. Results The aim of this work was to study the preparation of oilin-water emulsions by means of polyamide membranes with
cutoff 10 and 50 kDa and to investigate the influence of organic solvents on the membrane morphology and structure, which would affect the emulsion properties. The most suitable pretreatment was estimated in terms of pure isooctane flux stability after pretreatment and of droplet size distribution. The results of methods B–C and A–D are reported together as they gave similar results. 3.1. Effects of solvents on PA 10 kDa The thin layer morphology of new and pretreated PA 10kDa membranes is shown in the cross-section pictures depicted in Fig. 2. The new membranes present a dense thin layer with a compact organization (pores not detectable) (Fig. 2a). The morphology of the membrane appeared the same before and after washing with water, which means that the membrane was stored without additives. The pure water permeance was 180 (±20) l/(h m2 bar). After pretreatments B and C the thin layer is characterized by an open pore structure with pores like tortuous channels (Fig. 2b). The modifications after pretreatments A and D resulted in a different morphology; there are also, in this case, alterations of the thin layer structure and formation of an open pore structure, but pores appear more like a network rather than channels (Fig. 2c). As reported in Table 1, in all pretreatments the solvents used were isooctane and isopropanol, the difference being in the composition of mixtures and contact time. To understand the effect of each solvent, membranes were kept in contact with pure solvents for periods comparable with the ones used in the pretreatment procedures. Cross sections of 10-kDa polyamide membranes after treatment with pure isopropanol and pure isooctane are shown in Fig. 3. After the contact with isopropanol (Fig. 3a) the membranes look like a granular sphere structure; pores are clearly evident with respect to the initial situation shown in Fig. 2a. Membranes after contact with isooctane (Fig. 3b) are characterized by a homogeneous composition, with a more dense and packed morphology, where it is possible to ob-
616 L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Fig. 2. Scanning electron microscopic pictures of 10-kDa hollow fibers, cross section at lumen side. (a) New fibre; (b) after pretreatment B–C.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
617
figure it is shown that isooctane flux under a transmembrane pressure of 0.1 bar was more or less constant, with a decrease of 13% over a period of 5 days. Between sets of measurements the module was kept in isooctane with no applied pressure. 3.3. Preparation of oil-in-water emulsion
(c) Fig. 2. Continued. (c) After pretreatment A–D.
serve irregular pore channels that are smaller than those for isopropanol (Fig. 3a). An example of flux as a function of time for PA 10-kDa pretreated with method C is reported in (Fig. 4). This is an average value over three experiments with different PA 10 kDa modules. The figure shows the flux remained constant over a period of 3 days with an applied pressure of 0.1 bar. Between sets of flux measurements the membrane module was immersed in isooctane at room temperature. 3.2. Effects of solvents on PA 50 kDa The initial situation of PA with 50-kDa cross section is characterized by a nodular structure [24] with spherical units and by a homogeneous thin layer (Fig. 5a1). After washing with water, the big spherical units are not observed any more and the porous structure of the membrane becomes visible, as shown in Fig. 5a2. The spherical units previously observed were probably due to the presence of additives (e.g., glycerine) used to preserve the membrane during storage. This effect, as stated in the previous section, was not observed for PA 10 kDa. The pure water permeance of PA 50 kDa was about 320 (±40) l/(h m2 bar). After pretreatment of PA 50 kDa with methods B and C the pores also became more visible in the thin layer (Fig. 5b). After pretreatment with methods A and D (Fig. 5c) the situation is about the same as observed in B and C, but the structure remains more compact and the pores are almost not visible. Fibers of 50 kDa after contact with pure isopropanol are represented by an open pore structure and nodular distribution (Fig. 6a). After treatment with pure isooctane the organization of polymer is compact and dense and few pores are observed (Fig. 6b). The steady isooctane flux through the PA 50-kDa membrane after pretreatment C is illustrated in Fig. 4, in this
In Fig. 7 the pore size distribution of PA 10 and 50 kDa after each pretreatment used is summarized. Results of methods A and D and of B and C are represented together, as they gave similar results. For both type of membranes, treatments A and D, which used fewer steps and a lower contact time of solvents with membrane, gave a narrow pore size distribution compared to treatments B and C. A comparison with the original pore size distribution is not applicable since only the cutoff of the original membrane is known. Therefore, a relationship between original cutoff and pore size after pretreatment has been reported (Fig. 7a). PA 10 kDa membranes after pretreatment with method A showed a pore size distribution with a maximum at 0.038 µm. These membranes were used to prepare oil-inwater emulsion. During the process, the transmembrane pressure from oil to water was about 0.5 bar and the axial velocity of the continuous aqueous phase was 2.65 × 10−3 m/min (4.42 × 10−5 m/s). Polydisperse emulsions reflecting the pore size distribution were obtained. Droplet size ranged from 1.6 (±0.3) to 4.4 (±0.4) µm with a maximum at 2.2 (±0.2) µm. Similar results were reached using D pretreatment, where pore size distribution had a maximum at 0.038 µm, and droplet size ranged from 1.5 (±0.2) µm to 4.2 (±0.3) µm with a maximum at about 2.4 (±0.3) µm. As an example, the results from treatment D are reported in Fig. 8, which shows the number of droplets and the corresponding volume as a function of droplet size. Their size and density was monitored by optical microscopy. Emulsions resulted stable for at least 3 months. After pretreatment B and C pore size distributions showed higher values, with a maximum of about 0.075 µm; corresponding values for droplet size were estimated about 4.1 (±0.4) µm, with smaller droplets of about 2 (±0.4) µm and larger ones of about 6.3 (±0.5) µm. Using PA 50 kDa pretreated with methods B and C the pore size distribution showed a maximum of about 0.04 µm. These membranes were used to prepare emulsions at two different aqueous-phase axial velocities. The droplet size obtained using 3.45 × 10−3 m/min (5.75 × 10−5 m/s) showed a maximum around 2.8 (±0.3) µm (smaller droplets were about 1.5 (±0.3) µm and larger ones were about 4.2 (±0.5) µm). These conditions of axial velocity can be considered, as a reasonable approximation, to be comparable with the ones used for the experiments carried out with the other type of membrane (10 kDa). The same membranes were used to prepare emulsions at lower aqueous phase axial velocity, e.g., 6.13 × 10−4 m/min (1.02 × 10−5 m/s). In this case, as expected, the droplet size increased, with a maximum at 3.5 (±0.3) µm.
618 L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Fig. 3. Scanning electron microscopic pictures of 10-kDa hollow fibers, cross section at lumen side. (a) After treatment with isopropanol; (b) after treatment with isooctane.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
619
Fig. 4. Pure isooctane flux through PA 10 and 50 kDa after pretreatment C.
(a1)
(a2) Fig. 5. Scanning electron microscope pictures of 50-kDa hollow fibers, cross section at lumen side. (a1) New fibers, before washing with water; (a2) new fibers after washing with water.
620
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
(b)
(c) Fig. 5. Continued. (b) Fibers after pretreatment B–C; (c) fibers after pretreatment A–D.
PA 50 kDa pretreated with methods A and D showed a pore size distribution with a maximum at 0.02 µm. The droplet size obtained using an axial flow rate of 3.40 × 10−3 m/min (5.75 × 10−5 m/s) showed a maximum about 1.6 (±0.3) µm with a range from <0.5 to 3.7 µm. The relationship between pore and droplet size concerning PA 10 and 50 kDa is reported in Fig 9. When the pore size is plotted as a function of the droplet size, a linear relationship is obtained, regardless of the pretreatment procedure (A–D or B–C) and membrane type (PA 10 kDa or PA 50 kDa), with a proportional coefficient of about 50.
4. Discussions When membranes were subjected to pretreatment with solvents, it was expected that the effects of solvents on the final size of the pores for the two membrane types (PA 10 and 50 kDa) would have been proportional to the original cutoff.
On the contrary, an inversion relationship was observed between original cutoff and final pore size: for the same type of pretreatment method, larger pore dimensions were obtained with the originally lower cutoff and vice versa (see Fig. 7 and part a). For example, for the treatment AD, the original 10-kDa membranes had a peak around 0.038 µm, while the original 50-kDa membrane had 0.02 µm, with treatment BC, the original 10-kDa membranes got 0.075 and the original 50-kDa got 0.04 µm pore size. This effect is probably due to different mass transfer properties and swelling capacities of the dense (10-kDa membrane) and porous (50-kDa membrane) structures in the thin layer. The isopropanol caused swelling of the polymer on the thin dense layer with formation of an open pore sponge structure. The isooctane caused melting of the swollen polymer on the thin dense layer with formation of tortuous pore channels. In general, increase of contact time with solvents caused increased pore size and size distribution. Although
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
621
(a)
(b) Fig. 6. Scanning electron microscopic pictures of 50-kDa hollow fibers, cross-section at the lumen side. (a) After treatment with isopropanol; (b) after treatment with isooctane.
Fig. 7. Pore size distribution for PA 10 and 50 kDa after various pretreatments. (a) Relationship between original membrane cutoff and pore size after pretreatments.
622
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Fig. 8. Percentage of droplets and volume as a function of droplet size for emulsion prepared using 10 kDa PA membrane after pretreatment D.
Fig. 9. Relationship between pore size and droplet size for emulsion prepared using PA 10 kDa and PA 50 kDa after pretreatment with A–D and B–C methods.
these effects were significant for both types of membranes (PA 10 and 50 kDa) they were more severe for PA 10 kDa. The pure isooctane flux through membranes after pretreatment was more or less constant for the observed period. This makes it possible to state that after the initial effect, the membranes were stable in contact with the solvent. The results were reproducible with an average error of about 14%. Some information concerning the disperse phase flow model can be obtained, comparing applied transmembrane pressure (0.1 bar) and membrane mean pore sizes, e.g., 10 kDa after pretreatment B or C (Fig. 7). In particular, for an isooctane interfacial tension of 18 mN/m and pore size 0.08 µm, the critical pressure (maximum pressure resistance), evaluated from the Young–Laplace equation (see Section 1), is almost 8.95 bar. The Young–Laplace relation takes into account the evolution of the droplet curvature radius at the pore mouth, which reaches a minimum when the height of the growing droplet equals the pore radius, and hence gives the maximum pressure resistance to the dispersed phase flow [25]. Although an equivalent form of
Young–Laplace equation could be used to evaluate this critical pressure, e.g., 2γo,w /Rp cos θ , in which θ represents the contact angle, the expression presented in Section 1 and used here can equally be employed to interpret the above experimental data. As previously mentioned, the calculated critical pressure is higher than the applied experimental transmembrane pressure. To interpret this result, according to the Young–Laplace relation, two possible effects can be evoked. (i) The first one concerns the equilibrium γo,w value (i.e., the interfacial tension corresponding to the oil–water interface, in which the emulsifier is completely adsorbed), which is smaller than the initial 18 mN/m interfacial tension; (ii) the second considers the membrane pretreatment effects on pore sizes. More precisely, it could be admitted that chemically active species, which can be generated on the membrane surface during pretreatment, yield less “hydrophilic” pore borders. In other words, at the beginning of the process, the membrane pores are filled with oil, as well as the borders of these pores. Thus, this effect leads to an increase of the droplet sizes and consequently permits obtaining relatively high fluxes with low transmembrane pressure, lower than the theoretical value evaluated from the Y–L equation. Moreover, the large proportional coefficient (about 50) obtained for the droplet size versus pore size relationship may also be due to this effect. Various proportionality coefficients are reported in the literature [8], which are related to the physical chemical properties of fluid and membrane components as well as to the fluid dynamic conditions of dispersed and continuous phases.
5. Conclusions This study clarifies the behavior of polymeric ultrafiltration membranes made of polyamide used in organic solvents, and in particular, for the preparation of oil-in-water emulsions. In order to permeate the apolar organic phase through the hydrophilic membranes, these have been conditioned with solvents of decreasing polarity. Several procedures were investigated based on different contact time with solvents and different solvent mixtures. The effect of solvents on the thin layer of asymmetric polyamide membranes showed a clear change in the membrane morphology and pore structure. The effects were greater for 10 kDa than for 50 kDa. The isopropanol caused a pore open structure of nodular type, while the isooctane caused a pore open structure with pores like tortuous channels. The overall effect of pretreatments gave microporous structure with both type of pores present, with the tortuous channels being the most frequent. Although the pretreatments caused the described modifications, the obtained structures made it possible to obtain stable and reproducible performance in terms of permeate isooctane flux and emulsion droplet size distribution.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
The obtained pore size distribution was related to the membrane type and conditioning methodology. PA 10 kDa membranes pretreated with method D have been successfully used to produce O/W emulsions with average droplet size 2.4 (±0.3) stable for at least 3 months. The pretreatment methodology made it possible to obtain very high fluxes with low transmembrane pressure for a direct oil-in-water emulsification.
Acknowledgment The work was carried out with the support of Ministero degli Affari Esteri, Direzione Generale per la Promozione e la Cooperazione Culturale.
References [1] [2] [3] [4]
A. Hong, A.G. Fane, J. Membr. Sci. 222 (2003) 19–39. G.T. Vladisavjevi´c, H. Schubert, J. Membr. Sci. 225 (2003) 15–25. T. Hellweg, Curr. Opin. Colloid Interface Sci. 7 (2002) 50–56. I. Kobayashi, M. Yasuno, S. Iwamoto, Colloids Surf. 207 (2002) 185– 196. [5] D.G. Dalgleish, Trends Food Sci. Technol. 8 (1997) 1–8. [6] Y. Mine, M. Shimizu, Colloids Surf. B Biointerfaces 6 (1996) 261– 268.
623
[7] N. Yamazaki, H. Yuyama, J. Dispers. Sci. Technol. 23 (1–3) (2002) 279–292. [8] S.M. Joscelyne, G. Tragardh, J. Membr. Sci. 169 (2000) 107–117. [9] T. Nakashima, M. Shimizu, M. Kukizaki, Presented at the 2nd International Conference on Inorganic Membranes of Montpellier, France, 1991. [10] G.T. Vladisavljevi´c, H. Schubert, J. Membr. Sci. 225 (2003) 15–23. [11] Y.K. Ha, H.S. Song, H. Lee, J.H. Kim, Colloids Surf. A 162 (2000) 289. [12] L. Zhang, J. Hu, Z. Lu, J. Colloid Interface Sci. 190 (1997) 76–80. [13] T. Nakashima, M. Shimizu, M. Kukizaki, Adv. Drug Deliv. Rev. 45 (2000) 47. [14] S. Nagashima, S. Ando, T. Tsukamoto, H. Ohshima, K. Makino, Colloids Surf. B Biointerfaces 11 (1998) 47–56. [15] S. Omi, K. Kaneka, A. Nakayama, K. Katami, T. Taguchi, M. Iso, M. Nagai, M.G.H. Ma, J. Appl. Polym. Sci. 65 (1997) 2655. [16] G.T. Vladisavljevi´c, Chem. Eng. Process. 41 (2002) 231–238. [17] D.R. Machado, D. Hasson, R. Semiat, J. Membr. Sci. 163 (1999) 93– 102. [18] R. Shukla, M. Cheryan, J. Membr. Sci. 198 (2002) 75–85. [19] E. Drioli, L. Giorno, Application in Biotechnology and the Pharmaceutical Industry: Biocatalytic Membrane Reactors, Taylor & Francis, London, 1999. [20] L. Kubišová, E. Sabolová, Š. Schlosser, J. Marták, R. Kertész, Desalination 148 (2002) 205–211. [21] B. Van der Bruggen, J. Geens, Chem. Eng. Sci. 57 (2002) 2511–2518. [22] L. Giorno, N. Li, J. Membr. Sci. 217 (2003) 173–180. [23] S.J. Peng, R.A. Williams, Trans. IChemE A 76 (1998) 894–901. [24] R.M. Boom, I.M. Wienk, J. Membr. Sci. 73 (1992) 277–292. [25] G. De Luca, A. Sindona, L. Giorno, E. Drioli, J. Membr. Sci. 229 (2004) 199–209.
Effects of organic solvents on ultrafiltration polyamide membranes for the preparation of oil-in-water emulsions L. Giorno a,∗ , R. Mazzei a , M. Oriolo a , G. De Luca a , M. Davoli b , E. Drioli a a Institute on Membrane Technology, CNR-ITM, University of Calabria, Via P. Bucci 17/C, 87036 Rende (CS), Italy b Department of Earth Science, University of Calabria, Via P. Bucci 14/B, 87036 Rende (CS), Italy
Received 20 July 2004; accepted 11 February 2005 Available online 31 March 2005
Abstract Hydrophilic ultrafiltration membranes made of polyamide with molecular weight cutoff 10 and 50 kDa have been studied for the preparation of oil-in-water emulsions by a cross-flow membrane emulsification technique. Isooctane and phosphate buffer were used as disperse and continuous phase, respectively. The permeation of apolar isooctane through the polar hydrophilic membrane was achieved by pretreatment of membranes with a gradient of miscible solvents of decreasing polarity to remove water from the pores and replace it with isooctane. Four different procedures were investigated, based on the solvent mixture percentage and contact time with membranes. After pretreatment, the performance of the membranes in terms of pure isooctane permeate flux and emulsion preparation was evaluated. The influence of organic solvents on polyamide (PA) membranes has been studied by SEM analysis, which showed a clear change in the structure and morphology of the thin selective layers. The effects proved stronger for PA 10 kDa than for 50 kDa. In fact, similar pretreatment procedures caused larger pore size and pore size distribution for PA 10 kDa than for 50 kDa. The properties of emulsions in terms of droplet size distribution reflected the membrane pore sizes obtained after pretreatment. The correlation between pore size and droplet size, for the physicochemical and fluid dynamic conditions used, has been evaluated. 2005 Elsevier Inc. All rights reserved. Keywords: Membrane emulsification; O/W emulsion; Polyamide membrane; Solvent effect; Membrane pretreatment
1. Introduction Emulsions are mixtures that consist of a dispersed phase in a finely divided state uniformly distributed in a continuous phase. In the case of oil-in-water emulsions, finely divided oil droplets are uniformly dispersed in water [1,2]. During the last two years more than 100 publications dealing with microemulsions have appeared in the literature [3,4]. Microemulsions, or emulsions with droplet size in the range 20–80 nm, are thermodynamically stable mixtures of oil in water or water in oil. When droplets are larger, in order to have a stable emulsion, the interface between the oil and the water phase must be stabilized by surfactant mole* Corresponding author. Fax +39 0984 402103.
E-mail address: [email protected] (L. Giorno). 0021-9797/$ – see front matter 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.jcis.2005.02.015
cules, which prevent immediate aggregation or coalescence and whose properties largely determine the behavior of the emulsion. Emulsions destabilize in several ways. Sufficiently large droplets suspended in an aqueous medium of low viscosity or yield stress will cream. Creaming in itself does not destabilize an emulsion, but the high concentration of oil droplets in the creamed layer promotes interactions that lead to flocculation, aggregation, or coalescence [5]. It is well known that, since emulsions are thermodynamically unstable, flocculation and coalescence occur immediately after emulsification. Generally, the instability of an emulsion depends upon the emulsifying agent, droplet size, net charge, and mechanical and physical properties of the adsorbed film. In particular, the distribution of droplets is the most important parameter characterizing any emulsion. Stability and resistance to creaming, rheology, chemical re-
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
activity, and physiological efficiency are influenced by both the relative size and the size distribution [6]. With membrane emulsification, in principle, monodisperse emulsions can be produced, requiring a relatively low energy input, which implies that the shear stress exerted on the ingredients is low. This technology is therefore particularly suitable for emulsions containing labile compounds such as biomolecules. In cross-flow membrane emulsification the dispersed phase is forced to pass through the membrane, where it forms droplets at the pore mouth on the other side of the membrane. Here the droplets are detached by the action of a drag force provided by the tangential flow of the continuous phase flowing along the membrane surface. If the membrane pore has an adequately narrow size distribution, uniform droplets can be obtained by carefully controlling the pressure driving the disperse phase [7]. A limiting factor for emulsion production on a commercial scale is the low disperse phase flux. Advantages of this process are low energy consumption, control of droplet size and droplet size distribution, and the low shear stress that is needed [1]. This is determined by the balance between the drag force on the droplets from the flowing continuous phase, the buoyancy of the droplets, the interfacial tension forces, and the driving pressure. The final droplet size and size distribution are determined not only by the pore size and size distribution of the membrane but also by the degree of coalescence, both at the membrane surface and in bulk solution [8]. Membrane emulsification was developed by Nakashima et al. [9,10] for making monodispersed emulsions over a wide spectrum of mean droplet sizes, ranging from about 0.5 µm to several tens of micrometers [10]. At the present time, membrane emulsification is primarily used for the production of special “high-technology” products, where uniform droplets of controlled mean diameters are needed. These applications include preparation of core particles for toner application [11], preparation of liposome [12] spacers for liquid crystal displays [13] preparation of uniform silica hydrogel particles [14], and synthesis of monodisperse polymer microspheres. Such polymer spheres have been used as packings for HPLC columns [13], as immobilizing carriers of enzyme [15], and as biodegradable drug delivery systems [16]. Membranes used in emulsification process are mainly inorganic microfiltration membranes. Polymeric membranes are essentially confined to the treatment of aqueous solutions. This is because of materials difficulties; polymeric membranes in contact with organic solvents tend to swell and lose their separation capabilities [17]. Several commercial polymeric ultrafiltration membranes in contact with solvent can swell or dissolve, leading to unacceptable changes in solvent flux and solute separation [18]. Rapid solvent exchange between water and a high concentration of alcohol disrupts the polymer matrix [18]. A similar phenomenon was observed with alcohols such as ethanol, isopropanol, and isobutanol at concentrations higher than 20%. Permeability of alcohols and hydrocarbons
613
through different polymeric membranes has been measured [18]. Permeability was found to be influenced by molecular size, solubility, viscosity, etc. Although polymeric membranes are in general less stable in organic solvents and are primarily used for separation of aqueous solutions, the possibility of using polymeric membranes with organic solvents for various applications has been reported. Polymeric membranes are widely used in two separate phase membrane reactors for phase transfer catalysis [19] or in membrane contactors for membrane-based solvent extraction [20]. The performance of nanofiltration membranes with organic solvents has been recently discussed by Van der Bruggen [21]. Ultrafiltration hydrophilic membranes have been used to prepare oil-in-water emulsions by membrane emulsification [22]. The hydrophilic membranes were conditioned with organic solvents to allow permeation of nonpolar solvents through the membrane. In the present work, the influence of organic solvents on polymeric membrane properties has been clarified. Several pretreatment procedures based on different contact time with solvents and solvent mixtures of decreasing polarity have been investigated. The effects of organic solvents on asymmetric polyamide membranes with different cutoffs (10 and 50 kDa) have been studied by scanning electron microscopy (SEM) and permeability measurements to pure organic solvents. The stability and reproducibility of pretreated membranes have been evaluated. The conditioned membranes have been used to prepare oil-in-water emulsions, which have been characterized in terms of droplet size distribution and stability. Although a theoretical treatment of solvent influence on the properties of the polymeric membrane used was not the aim of this work, basic theoretical considerations were necessary to qualitatively interpret some fundamental empirical observations. In particular, the morphology of the polyamide membrane used admits a Darcy model to describe the disperse phase (oil) flux in first approximation. Consequently, for the oil to start flowing and producing droplets, it is necessary that the applied transmembrane pressure be higher than the pore critical pressure P c , which depends on the membrane pore diameters and oil–water interfacial tension. The critical pressure can be calculated with the Young–Laplace equation, i.e., Pc = γo,w /Rc , in which R c and γ o,w are the droplet curvature radius and the oil–water interfacial tension, respectively. More precisely, when the height of the growing droplet equals the radius of the pore, the P c , i.e., the maximum pressure resistance, is reached [23].
2. Materials and methods Isooctane (2,2,4-trimethylpentane, from Sigma Aldrich, Italy) was the organic solvent used as dispersed phase. Ultrapure water (bidistilled water, filtered with a membrane of pore size 0.2 µm), anhydrous sodium dihydrogen phosphate
614
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Table 1 Pretreatment procedures Method A
Method B
Method C
Method D
Solvents
Conc. (v/v)
Contact time (h)
Solvents
Conc. (v/v)
Contact time (h)
Solvents
Conc. (v/v)
Contact time (h)
Solvents
Conc. (v/v)
Contact time (h)
H2 O H2 O–IPA ISO
100 50:50 100
∼4 3 12a
H2 O H2 O–IPA IPA–ISO IPA–ISO ISO
100 80:20 50:50 20:80 100
∼4 12 12 12 12a
H2 O H2 O–IPA IPA–ISO IPA–ISO ISO
100 80:20 50:50 20:80 100
∼4 12 3 12 3a
H2 O IPA–ISO Pure ISO
100 50:50 100
∼4 1.5 3a
a After this time, the module was washed with isooctane and used to prepare the emulsion.
(NaH2 PO4 ), and disodium hydrogen phosphate anhydrous (Na2 HPO4 ), from Fluka (Germany) or Sigma Aldrich (Italy) were used to prepare phosphate buffer solutions. A 50 mM buffer phosphate at pH 7 containing sodium dodecyl sulfate (SDS) and polyvinyl alcohol (PVA) was used as the continuous phase. SDS and PVA (molecular weight 22,000 and 97.5–99% degree of hydrolysis) from Fluka were used as emulsifier and stabilizer (0.2 and 0.8%), respectively. Isopropanol (IPA) of analytical grade from Fluka was used with water or isooctane for membrane pretreatment, as indicated in the following. Polyamide capillary membranes with nominal molecular weight cutoff (NMWCO) of 10 and 50 kDa, having inner/outer diameter of 1.5/2.2 and 1.2/2.4 mm, respectively, were used. The membranes were kindly provided by Forschstung Institut Berghof, Germany. The structure of this kind of membrane is asymmetric, with the selective layer on the lumen side and the sponge layer on the shell side. The lab-made membrane modules were prepared by assembling six capillary membranes inside a Pyrex glass cylinder of inner diameter 1.2 cm, 20 cm long. The internal/external membrane surface area were 4.9 × 10−3 m2 /7.2 × 10−3 m2 for 10 kDa and 3.6 × 10−3 m2 /7.9 × 10−3 m2 for 50 kDa.
brane, it was necessary to pretreat the membrane to remove the internal liquid phase (water) and replace it with the organic solvent. The pretreatment consisted in washing the membrane with a gradient of miscible solvents and solvent solutions of decreasing polarity, to shift the internal phase from polar to nonpolar and to allow the permeation of the isooctane. Four different procedures were used (see also Table 1). Method A: the pretreatment consisted in the subsequent permeation of pure water, water–isopropanol (50:50), and pure isooctane. Methods B and C used pure water, water–isopropanol (80:20), isopropanol–isooctane (50:50 and 20:80), and then pure isooctane; the difference between B and C is the different contact time of membranes with the solutions (Table 1). The last method (D) consisted of the use of water, isopropanol–isooctane (50:50), and then pure isooctane. After all the pretreatments the membranes were washed with isooctane and the pure isooctane flux was measured in order to monitor membrane stability. Afterward, the membranes were used to prepare oil-in-water emulsions (isooctane in aqueous phase constituted by phosphate buffer pH 7 containing 0.2 and 0.8% SDS and PVA, respectively). During the preparation of emulsions, the permeate flux of the organic phase was about 2 l/(h m2 ).
2.1. Membrane pretreatment The original membranes were both asymmetric in the structure, with a thin selective layer on the lumen side. Since the size of the pores on the selective layer is very low (ultrafiltration range) and not easily detectable by scanning electron microscopy, the concept of nominal molecular weight cutoff was used instead. The cutoff gives an indirect measure of the pore dimension; it is determined on the basis of the size of molecules that are retained by the membrane at 90%. The membranes used originally had 10 or 50 kDa cutoff. We have studied the structure of the thin selective layer of these membranes and investigated how the structure and pore size changed after contact with organic solvents. The membranes were first washed with ultra pure water to remove the water-soluble residues, and then the pure water permeance (l/(h m2 bar)) was tested. Afterward, to allow the permeation of the apolar organic solvent (isooctane, log P 4.5) through the hydrophilic mem-
2.2. SEM analysis To understand the effects of solvent on the structure and morphology of the asymmetric polyamide membranes with different cutoffs, new and pretreated membranes were analyzed by scanning electron microscopy (SEM, Mod. Stereoscan 360, Cambridge Instrument). The analysis was carried out after the overall pretreatments and, to identify the influence of each solvent, the membranes were also analyzed after contact with each solvent (isooctane and isopropanol). In addition, characterization of membranes after long-term contact with isooctane was carried out. Cross sections of new and treated fibers were obtained in liquid nitrogen. Cross-section enlargements were by up to 40,000 times. The analysis of SEM pictures made it possible to determine the original structure of the thin selective layer and to measure the pore size and size distribution of membranes after treatment with the solvents.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
615
Fig. 1. Schematic of membrane emulsifier: capillary membrane module with inlet and outlet of the two separate circuits (lumen and shell); the enlargement shows the longitudinal section of the asymmetric capillary membrane.
2.3. Droplet measurement Droplets were measured by optical microscopy. Droplet samples were placed on a cell counter having nine squares, each one subdivided into 16 smaller well-defined squares having a side of 200 µm and a volume of 0.1 mm3 . The images were acquired by means of a video camera connected to a personal computer and then were processed by the “Scion Image” program. This software automatically counts the particles and measures geometrical parameters (e.g., diameter, area) of identified particles. The droplet size distribution was calculated on the basis of at least 10 processed samples. 2.4. Membrane emulsification equipment Oil-in-water emulsions were prepared using a cross-flow membrane emulsification system. A schematic of the membrane emulsifier is shown in Fig. 1, where the module with capillary membranes is represented; the apparatus where the module is located includes two control panels fixed up with valves and pressure gauges to control the fluid dynamic conditions of the two circuits, separately. The experiments have been carried out at room temperature (23 ± 2 ◦ C). The disperse phase (organic phase) was circulated in the shell side of membrane module and forced to permeate through the membrane by applying a transmembrane pressure of about 0.1 bar. The disperse phase was then collected in the continuous phase (aqueous phase), which was recirculated on the lumen side of the membrane, where emulsions were formed. The axial velocity of the aqueous phase was 5.2 (±0.6) × 10−5 m/s, unless otherwise specified in the Experimental section. The initial volumes of the aqueous and organic phases recirculated in the two separate circuits were 150 ml each. In general, a percentage of 15% organic phase in the aqueous phase was obtained.
3. Results The aim of this work was to study the preparation of oilin-water emulsions by means of polyamide membranes with
cutoff 10 and 50 kDa and to investigate the influence of organic solvents on the membrane morphology and structure, which would affect the emulsion properties. The most suitable pretreatment was estimated in terms of pure isooctane flux stability after pretreatment and of droplet size distribution. The results of methods B–C and A–D are reported together as they gave similar results. 3.1. Effects of solvents on PA 10 kDa The thin layer morphology of new and pretreated PA 10kDa membranes is shown in the cross-section pictures depicted in Fig. 2. The new membranes present a dense thin layer with a compact organization (pores not detectable) (Fig. 2a). The morphology of the membrane appeared the same before and after washing with water, which means that the membrane was stored without additives. The pure water permeance was 180 (±20) l/(h m2 bar). After pretreatments B and C the thin layer is characterized by an open pore structure with pores like tortuous channels (Fig. 2b). The modifications after pretreatments A and D resulted in a different morphology; there are also, in this case, alterations of the thin layer structure and formation of an open pore structure, but pores appear more like a network rather than channels (Fig. 2c). As reported in Table 1, in all pretreatments the solvents used were isooctane and isopropanol, the difference being in the composition of mixtures and contact time. To understand the effect of each solvent, membranes were kept in contact with pure solvents for periods comparable with the ones used in the pretreatment procedures. Cross sections of 10-kDa polyamide membranes after treatment with pure isopropanol and pure isooctane are shown in Fig. 3. After the contact with isopropanol (Fig. 3a) the membranes look like a granular sphere structure; pores are clearly evident with respect to the initial situation shown in Fig. 2a. Membranes after contact with isooctane (Fig. 3b) are characterized by a homogeneous composition, with a more dense and packed morphology, where it is possible to ob-
616 L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Fig. 2. Scanning electron microscopic pictures of 10-kDa hollow fibers, cross section at lumen side. (a) New fibre; (b) after pretreatment B–C.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
617
figure it is shown that isooctane flux under a transmembrane pressure of 0.1 bar was more or less constant, with a decrease of 13% over a period of 5 days. Between sets of measurements the module was kept in isooctane with no applied pressure. 3.3. Preparation of oil-in-water emulsion
(c) Fig. 2. Continued. (c) After pretreatment A–D.
serve irregular pore channels that are smaller than those for isopropanol (Fig. 3a). An example of flux as a function of time for PA 10-kDa pretreated with method C is reported in (Fig. 4). This is an average value over three experiments with different PA 10 kDa modules. The figure shows the flux remained constant over a period of 3 days with an applied pressure of 0.1 bar. Between sets of flux measurements the membrane module was immersed in isooctane at room temperature. 3.2. Effects of solvents on PA 50 kDa The initial situation of PA with 50-kDa cross section is characterized by a nodular structure [24] with spherical units and by a homogeneous thin layer (Fig. 5a1). After washing with water, the big spherical units are not observed any more and the porous structure of the membrane becomes visible, as shown in Fig. 5a2. The spherical units previously observed were probably due to the presence of additives (e.g., glycerine) used to preserve the membrane during storage. This effect, as stated in the previous section, was not observed for PA 10 kDa. The pure water permeance of PA 50 kDa was about 320 (±40) l/(h m2 bar). After pretreatment of PA 50 kDa with methods B and C the pores also became more visible in the thin layer (Fig. 5b). After pretreatment with methods A and D (Fig. 5c) the situation is about the same as observed in B and C, but the structure remains more compact and the pores are almost not visible. Fibers of 50 kDa after contact with pure isopropanol are represented by an open pore structure and nodular distribution (Fig. 6a). After treatment with pure isooctane the organization of polymer is compact and dense and few pores are observed (Fig. 6b). The steady isooctane flux through the PA 50-kDa membrane after pretreatment C is illustrated in Fig. 4, in this
In Fig. 7 the pore size distribution of PA 10 and 50 kDa after each pretreatment used is summarized. Results of methods A and D and of B and C are represented together, as they gave similar results. For both type of membranes, treatments A and D, which used fewer steps and a lower contact time of solvents with membrane, gave a narrow pore size distribution compared to treatments B and C. A comparison with the original pore size distribution is not applicable since only the cutoff of the original membrane is known. Therefore, a relationship between original cutoff and pore size after pretreatment has been reported (Fig. 7a). PA 10 kDa membranes after pretreatment with method A showed a pore size distribution with a maximum at 0.038 µm. These membranes were used to prepare oil-inwater emulsion. During the process, the transmembrane pressure from oil to water was about 0.5 bar and the axial velocity of the continuous aqueous phase was 2.65 × 10−3 m/min (4.42 × 10−5 m/s). Polydisperse emulsions reflecting the pore size distribution were obtained. Droplet size ranged from 1.6 (±0.3) to 4.4 (±0.4) µm with a maximum at 2.2 (±0.2) µm. Similar results were reached using D pretreatment, where pore size distribution had a maximum at 0.038 µm, and droplet size ranged from 1.5 (±0.2) µm to 4.2 (±0.3) µm with a maximum at about 2.4 (±0.3) µm. As an example, the results from treatment D are reported in Fig. 8, which shows the number of droplets and the corresponding volume as a function of droplet size. Their size and density was monitored by optical microscopy. Emulsions resulted stable for at least 3 months. After pretreatment B and C pore size distributions showed higher values, with a maximum of about 0.075 µm; corresponding values for droplet size were estimated about 4.1 (±0.4) µm, with smaller droplets of about 2 (±0.4) µm and larger ones of about 6.3 (±0.5) µm. Using PA 50 kDa pretreated with methods B and C the pore size distribution showed a maximum of about 0.04 µm. These membranes were used to prepare emulsions at two different aqueous-phase axial velocities. The droplet size obtained using 3.45 × 10−3 m/min (5.75 × 10−5 m/s) showed a maximum around 2.8 (±0.3) µm (smaller droplets were about 1.5 (±0.3) µm and larger ones were about 4.2 (±0.5) µm). These conditions of axial velocity can be considered, as a reasonable approximation, to be comparable with the ones used for the experiments carried out with the other type of membrane (10 kDa). The same membranes were used to prepare emulsions at lower aqueous phase axial velocity, e.g., 6.13 × 10−4 m/min (1.02 × 10−5 m/s). In this case, as expected, the droplet size increased, with a maximum at 3.5 (±0.3) µm.
618 L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Fig. 3. Scanning electron microscopic pictures of 10-kDa hollow fibers, cross section at lumen side. (a) After treatment with isopropanol; (b) after treatment with isooctane.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
619
Fig. 4. Pure isooctane flux through PA 10 and 50 kDa after pretreatment C.
(a1)
(a2) Fig. 5. Scanning electron microscope pictures of 50-kDa hollow fibers, cross section at lumen side. (a1) New fibers, before washing with water; (a2) new fibers after washing with water.
620
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
(b)
(c) Fig. 5. Continued. (b) Fibers after pretreatment B–C; (c) fibers after pretreatment A–D.
PA 50 kDa pretreated with methods A and D showed a pore size distribution with a maximum at 0.02 µm. The droplet size obtained using an axial flow rate of 3.40 × 10−3 m/min (5.75 × 10−5 m/s) showed a maximum about 1.6 (±0.3) µm with a range from <0.5 to 3.7 µm. The relationship between pore and droplet size concerning PA 10 and 50 kDa is reported in Fig 9. When the pore size is plotted as a function of the droplet size, a linear relationship is obtained, regardless of the pretreatment procedure (A–D or B–C) and membrane type (PA 10 kDa or PA 50 kDa), with a proportional coefficient of about 50.
4. Discussions When membranes were subjected to pretreatment with solvents, it was expected that the effects of solvents on the final size of the pores for the two membrane types (PA 10 and 50 kDa) would have been proportional to the original cutoff.
On the contrary, an inversion relationship was observed between original cutoff and final pore size: for the same type of pretreatment method, larger pore dimensions were obtained with the originally lower cutoff and vice versa (see Fig. 7 and part a). For example, for the treatment AD, the original 10-kDa membranes had a peak around 0.038 µm, while the original 50-kDa membrane had 0.02 µm, with treatment BC, the original 10-kDa membranes got 0.075 and the original 50-kDa got 0.04 µm pore size. This effect is probably due to different mass transfer properties and swelling capacities of the dense (10-kDa membrane) and porous (50-kDa membrane) structures in the thin layer. The isopropanol caused swelling of the polymer on the thin dense layer with formation of an open pore sponge structure. The isooctane caused melting of the swollen polymer on the thin dense layer with formation of tortuous pore channels. In general, increase of contact time with solvents caused increased pore size and size distribution. Although
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
621
(a)
(b) Fig. 6. Scanning electron microscopic pictures of 50-kDa hollow fibers, cross-section at the lumen side. (a) After treatment with isopropanol; (b) after treatment with isooctane.
Fig. 7. Pore size distribution for PA 10 and 50 kDa after various pretreatments. (a) Relationship between original membrane cutoff and pore size after pretreatments.
622
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
Fig. 8. Percentage of droplets and volume as a function of droplet size for emulsion prepared using 10 kDa PA membrane after pretreatment D.
Fig. 9. Relationship between pore size and droplet size for emulsion prepared using PA 10 kDa and PA 50 kDa after pretreatment with A–D and B–C methods.
these effects were significant for both types of membranes (PA 10 and 50 kDa) they were more severe for PA 10 kDa. The pure isooctane flux through membranes after pretreatment was more or less constant for the observed period. This makes it possible to state that after the initial effect, the membranes were stable in contact with the solvent. The results were reproducible with an average error of about 14%. Some information concerning the disperse phase flow model can be obtained, comparing applied transmembrane pressure (0.1 bar) and membrane mean pore sizes, e.g., 10 kDa after pretreatment B or C (Fig. 7). In particular, for an isooctane interfacial tension of 18 mN/m and pore size 0.08 µm, the critical pressure (maximum pressure resistance), evaluated from the Young–Laplace equation (see Section 1), is almost 8.95 bar. The Young–Laplace relation takes into account the evolution of the droplet curvature radius at the pore mouth, which reaches a minimum when the height of the growing droplet equals the pore radius, and hence gives the maximum pressure resistance to the dispersed phase flow [25]. Although an equivalent form of
Young–Laplace equation could be used to evaluate this critical pressure, e.g., 2γo,w /Rp cos θ , in which θ represents the contact angle, the expression presented in Section 1 and used here can equally be employed to interpret the above experimental data. As previously mentioned, the calculated critical pressure is higher than the applied experimental transmembrane pressure. To interpret this result, according to the Young–Laplace relation, two possible effects can be evoked. (i) The first one concerns the equilibrium γo,w value (i.e., the interfacial tension corresponding to the oil–water interface, in which the emulsifier is completely adsorbed), which is smaller than the initial 18 mN/m interfacial tension; (ii) the second considers the membrane pretreatment effects on pore sizes. More precisely, it could be admitted that chemically active species, which can be generated on the membrane surface during pretreatment, yield less “hydrophilic” pore borders. In other words, at the beginning of the process, the membrane pores are filled with oil, as well as the borders of these pores. Thus, this effect leads to an increase of the droplet sizes and consequently permits obtaining relatively high fluxes with low transmembrane pressure, lower than the theoretical value evaluated from the Y–L equation. Moreover, the large proportional coefficient (about 50) obtained for the droplet size versus pore size relationship may also be due to this effect. Various proportionality coefficients are reported in the literature [8], which are related to the physical chemical properties of fluid and membrane components as well as to the fluid dynamic conditions of dispersed and continuous phases.
5. Conclusions This study clarifies the behavior of polymeric ultrafiltration membranes made of polyamide used in organic solvents, and in particular, for the preparation of oil-in-water emulsions. In order to permeate the apolar organic phase through the hydrophilic membranes, these have been conditioned with solvents of decreasing polarity. Several procedures were investigated based on different contact time with solvents and different solvent mixtures. The effect of solvents on the thin layer of asymmetric polyamide membranes showed a clear change in the membrane morphology and pore structure. The effects were greater for 10 kDa than for 50 kDa. The isopropanol caused a pore open structure of nodular type, while the isooctane caused a pore open structure with pores like tortuous channels. The overall effect of pretreatments gave microporous structure with both type of pores present, with the tortuous channels being the most frequent. Although the pretreatments caused the described modifications, the obtained structures made it possible to obtain stable and reproducible performance in terms of permeate isooctane flux and emulsion droplet size distribution.
L. Giorno et al. / Journal of Colloid and Interface Science 287 (2005) 612–623
The obtained pore size distribution was related to the membrane type and conditioning methodology. PA 10 kDa membranes pretreated with method D have been successfully used to produce O/W emulsions with average droplet size 2.4 (±0.3) stable for at least 3 months. The pretreatment methodology made it possible to obtain very high fluxes with low transmembrane pressure for a direct oil-in-water emulsification.
Acknowledgment The work was carried out with the support of Ministero degli Affari Esteri, Direzione Generale per la Promozione e la Cooperazione Culturale.
References [1] [2] [3] [4]
A. Hong, A.G. Fane, J. Membr. Sci. 222 (2003) 19–39. G.T. Vladisavjevi´c, H. Schubert, J. Membr. Sci. 225 (2003) 15–25. T. Hellweg, Curr. Opin. Colloid Interface Sci. 7 (2002) 50–56. I. Kobayashi, M. Yasuno, S. Iwamoto, Colloids Surf. 207 (2002) 185– 196. [5] D.G. Dalgleish, Trends Food Sci. Technol. 8 (1997) 1–8. [6] Y. Mine, M. Shimizu, Colloids Surf. B Biointerfaces 6 (1996) 261– 268.
623
[7] N. Yamazaki, H. Yuyama, J. Dispers. Sci. Technol. 23 (1–3) (2002) 279–292. [8] S.M. Joscelyne, G. Tragardh, J. Membr. Sci. 169 (2000) 107–117. [9] T. Nakashima, M. Shimizu, M. Kukizaki, Presented at the 2nd International Conference on Inorganic Membranes of Montpellier, France, 1991. [10] G.T. Vladisavljevi´c, H. Schubert, J. Membr. Sci. 225 (2003) 15–23. [11] Y.K. Ha, H.S. Song, H. Lee, J.H. Kim, Colloids Surf. A 162 (2000) 289. [12] L. Zhang, J. Hu, Z. Lu, J. Colloid Interface Sci. 190 (1997) 76–80. [13] T. Nakashima, M. Shimizu, M. Kukizaki, Adv. Drug Deliv. Rev. 45 (2000) 47. [14] S. Nagashima, S. Ando, T. Tsukamoto, H. Ohshima, K. Makino, Colloids Surf. B Biointerfaces 11 (1998) 47–56. [15] S. Omi, K. Kaneka, A. Nakayama, K. Katami, T. Taguchi, M. Iso, M. Nagai, M.G.H. Ma, J. Appl. Polym. Sci. 65 (1997) 2655. [16] G.T. Vladisavljevi´c, Chem. Eng. Process. 41 (2002) 231–238. [17] D.R. Machado, D. Hasson, R. Semiat, J. Membr. Sci. 163 (1999) 93– 102. [18] R. Shukla, M. Cheryan, J. Membr. Sci. 198 (2002) 75–85. [19] E. Drioli, L. Giorno, Application in Biotechnology and the Pharmaceutical Industry: Biocatalytic Membrane Reactors, Taylor & Francis, London, 1999. [20] L. Kubišová, E. Sabolová, Š. Schlosser, J. Marták, R. Kertész, Desalination 148 (2002) 205–211. [21] B. Van der Bruggen, J. Geens, Chem. Eng. Sci. 57 (2002) 2511–2518. [22] L. Giorno, N. Li, J. Membr. Sci. 217 (2003) 173–180. [23] S.J. Peng, R.A. Williams, Trans. IChemE A 76 (1998) 894–901. [24] R.M. Boom, I.M. Wienk, J. Membr. Sci. 73 (1992) 277–292. [25] G. De Luca, A. Sindona, L. Giorno, E. Drioli, J. Membr. Sci. 229 (2004) 199–209.