Effect of Emulsifier Type on Droplet Disruption in ... - ACS Publications

Further, homogenization is often followed by a considerable temperature elevation as a result of the poor energy utilization. Membrane emulsification ...
0 downloads 0 Views 440KB Size
Download PDF
4526

Langmuir 2006, 22, 4526-4533

Effect of Emulsifier Type on Droplet Disruption in Repeated Shirasu Porous Glass Membrane Homogenization Goran T. Vladisavljevic´,*,† Jeonghee Surh,‡ and Julian D. McClements‡ Institute of Food Technology and Biochemistry, Faculty of Agriculture, UniVersity of Belgrade, P.O. Box 127, YU-11081 Belgrade-Zemun, Serbia & Montenegro, and Department of Food Science, Biopolymer and Colloids Research Laboratory, UniVersity of Massachusetts, 100 Holdsworth Way, Amherst, Massachusetts 01003 ReceiVed December 16, 2005. In Final Form: March 7, 2006 The influence of various emulsifier types (anionic, nonionic, and zwitterionic) on the mean particle size, transmembrane flux, and membrane fouling in repeated membrane homogenization using a Shirasu porous glass (SPG) membrane has been investigated. Oil-in-water (O/W) emulsions (40 wt % corn oil stabilized by 0.06-2 wt % sodium dodecyl sulfate (SDS) or 0.1-2 wt % Tween 20 at pH 3 or 0.5-2 wt % β-lactoglobulin (β-Lg) at pH 7) were prepared by passing coarsely emulsified feed mixtures five times through the membrane with a mean pore size of 8.0 µm under the transmembrane pressure of 100 kPa. The flux increased as the number of passes increased, tending to a maximum limiting value. The maximum flux for the Tween 20-stabilized emulsions (5-47 m3‚m-2‚h-1) was smaller than that for the SDS-stabilized emulsions (29-60 m3‚m-2‚h-1) because less energy was needed for the disruption of a SDSstabilized droplet due to the lower interfacial tension. The mean particle size after five passes was 4.1-6.8 and 6.4-8.7 µm for 0.1-2 wt % SDS and Tween 20, respectively. The flux in the presence of β-Lg was much smaller than that in the presence of SDS and Tween 20, which was a consequence of more pronounced membrane fouling, due to the protein adsorption to the membrane surface. After five passes through the membrane, the fouling resistance in the presence of 2 wt % β-Lg (1.1 × 1010 1/m) was 2 orders of magnitude higher than that for 0.5 wt % Tween 20 and an order of magnitude higher than the membrane resistance. If a clean membrane was used in the fifth pass, a 2-fold reduction of the fouling resistance was observed.

Introduction Conventional emulsification devices such as high pressure valve homogenizers generally use inhomogeneous extensional and shear forces and high energy inputs of 106-108 J‚m-3 to rupture droplets.1,2 As a result, they generate emulsions with relatively small droplet sizes but wide particle size distributions. Further, homogenization is often followed by a considerable temperature elevation as a result of the poor energy utilization. Membrane emulsification is a relatively new emulsification technology, aimed at achieving precise control of the particle size distribution over a wide range of mean droplet sizes.3 This technique is particularly useful for producing multiple emulsions4 and monodisperse solid microparticles (microspheres and microcapsules)5 because of its effectiveness in preparing droplets with very narrow particle size distributions at low energy inputs. In “direct membrane emulsification”, a pure liquid (the disperse phase) is forced through the membrane pores into another immiscible liquid (the continuous phase), and the small droplets are formed in situ at the membrane-continuous phase interface.6-8 * Corresponding author. Tel: (+381) 11 2615 315/327. Fax: (+381) 11 199 711. [email protected]. † University of Belgrade. ‡ University of Massachusetts. (1) McClements, D. J. Food Emulsions: Principles, Practices, and Techniques, 2nd ed.; CRC Press: Boca Raton, FL, 2005; p 259. (2) Joscelyne, S. M.; Tra¨gårdh, G. J. Membr. Sci. 2000, 169, 107. (3) Nakashima, T.; Shimizu M.; Kukizaki, M. AdV. Drug DeliVery ReV. 2000, 45, 47. (4) van der Graaf, S.; Schroe¨n, C. G. P. H.; Boom, R. M. J. Membr. Sci. 2005, 251, 7. (5) Vladisavljevic´, G. T.; Williams, R. A. AdV. Colloid Interface Sci. 2005, 113, 1. (6) Vladisavljevic´, G. T.; Schubert, H. J. Membr. Sci. 2003, 225, 15. (7) Vladisavljevic´, G. T.; Schubert, H. Colloids Surf., A 2004, 232, 199. (8) Vladisavljevic´, G. T.; Schubert, H. J. Dispersion Sci. Technol. 2003, 24, 811.

In “premix membrane emulsification” (membrane homogenization), coarsely emulsified feeds are forced through the membrane, and the small droplets are formed by reducing the size of the large droplets in preexisting emulsions.9,10 The major advantages of this approach are that emulsions with higher droplet concentrations can more easily be produced, and higher transmembrane fluxes can be achieved, but at the expense of a higher extent of droplet polydispersity. The degree of monodispersity can be improved by passing the emulsion through the membrane a number of times.11-14 The repeated membrane homogenization was originally developed for the production of multilamellar lipid vesicles (liposomes) using track-etch polycarbonate filters, which contain almost identical cylindrical pores.15 In this process, the coarse liposome suspension is passed under moderate pressure repeatedly (usually 10 times) through filters with progressively smaller pore sizes, which leads to a gradual break up of the large vesicles into smaller ones.16 The most commonly used microporous membrane for emulsification is made of a special kind of CaO-Al2O3-B2O3SiO2-type porous glass called Shirasu porous glass (SPG).3 The major advantages of this membrane are that it can be fabricated with mean pore sizes in a wide interval between 0.1 and 20 µm, (9) Suzuki, K.; Fujiki, I.; Hagura, Y. Food Sci. Technol. Int. Tokyo 1998, 4, 164. (10) Shima, M.; Kobayashi, Y.; Fujii, T.; Tanaka, M.; Kimura, Y.; Adachi, S.; Matsuno, R. Food Hydrocolloids 2004, 18, 61. (11) Vladisavljevic´, G. T.; Shimizu, M.; Nakashima T. J. Membr. Sci. 2004, 244, 97. (12) Altenbach-Rehm, J.; Suzuki, K.; Schubert, H. Proceedings of the 3rd World Congress on Emulsions, Lyon, France, September 2002. (13) Park, S. H.; Yamaguchi, T.; Nakao, S. Chem. Eng. Sci. 2001, 56, 3539. (14) Ribeiro, H. S.; Rico, L. G.; Badolato, G. G.; Schubert, H. J. Food Sci. 2005, 70, E117. (15) Olson, F.; Hunt, C. A.; Szoka, F. C.; Vail, W. J.; Papahadjopoulos, D. Biochim. Biophys. Acta 1979, 557, 9. (16) Walde, P.; Ichikawa, S. Biomol. Eng. 2001, 18, 143.

10.1021/la053410f CCC: $33.50 © 2006 American Chemical Society Published on Web 04/04/2006

Emulsifier Type Effects in Membrane Homogenization

it contains very uniform pores with noncircular cross-sections facilitating droplet detachment from the pore openings, and it can easily be rendered hydrophobic by coating with a silicone resin.17 Hydrophobic SPG membranes are needed to produce water-in-oil (W/O) emulsions, whereas hydrophilic SPG membranes are needed to produce oil-in-water (O/W) emulsions. The adequate choice of emulsifier is of primary importance for the success of membrane emulsification. A chosen emulsifier should rapidly adsorb to the newly formed oil-water interface and reduce the interfacial tension to an optimum level. On the other hand, the emulsifier should not adsorb to the membrane surface by electrostatic or hydrophobic interactions because it can cause the alteration of membrane polarity from hydrophilic to hydrophobic or vice versa, nor should it accumulate inside the pores, which can lead to the pore plugging. The effect of the dynamic interfacial tension of emulsifiers on direct membrane emulsification has been studied by several authors.18-20 The general conclusion was that the faster the emulsifier molecules adsorbed at newly formed interfaces, the smaller the droplet sizes of the resultant emulsion. The effect of emulsifier charge on droplet formation in direct membrane emulsification has been investigated by Nakashima et al.21 for SPG membranes and by Kobayashi et al.22 for silicon microchannels. They have concluded that functional groups of the chosen emulsifiers must not carry the charge opposite to that of the membrane surface in order to avoid electrostatic attractions with the membrane surface. For example, an untreated SPG membrane has a negative surface potential of -15 to -35 mV within a pH range of 2-8 because of the dissociation of silanol groups on the surface (Si-OH f SiO- + H+). Hence, for the case of a hydrophilic (untreated) SPG membrane, the use of cationic emulsifiers such as alkylsubstituted quaternary ammonium salts, such as cetyltrimethylammonium bromide or tri-n-octylmethylammonium chloride, must be avoided. The effect of emulsifier on emulsification in pre-emulsified systems, however, has not yet been investigated. The aim of this work was to investigate the influence of different surfactants (nonionic, anionic, and zwitterionic) on the mean particle size, transmembrane flux, and membrane fouling in repeated membrane homogenization using an SPG membrane. Experimental Section Materials. Analytical-grade sodium dodecyl sulfate (SDS), polyoxyethylene sorbitan monolaureate (Tween 20), hydrochloric acid, sodium hydroxide, sodium azide, disodium phosphate, monosodium phosphate, and sodium chloride were purchased from the Sigma Chemical Company (St. Louis, MO). Acetic acid (glacial) was supplied from the Fisher Scientific Company (Fair Lawn, NJ). Powdered β-lactoglobulin (β-Lg) (pKa ≈ 5.2) was obtained from Davisco Foods International (LOT # JE 001-1-922, Le Sueur, MN). As stated by the manufacturer, the β-Lg content in the powder determined by electrophoresis was 98% (the remainder being mostly globulins). Corn oil was purchased from a local supermarket and used without further purification. Distilled, deionized water was used for the preparation of all solutions. Surfactant Solution Preparation. Solutions of Tween 20 (0.1-2 wt %) and SDS (0.06-2 wt %) were prepared by dispersing surfactant in a 100 mM acetic acid solution containing 10 mM NaCl and 0.02 (17) Vladisavljevic´, G. T.; Shimizu, M.; Nakashima T. J. Membr. Sci. 2005, 250, 69. (18) Schro¨der, V.; Behrend, O.; Schubert, H. J. Colloid Interface Sci. 1998, 202, 334. (19) van der Graaf, S.; Schroe¨n, C. G. P. H.; van der Sman, R. G. M.; Boom, R. M. J. Colloid Interface Sci. 2004, 277, 454. (20) Rayner, M.; Tra¨gårdh, G.; Tra¨gårdh, C. Colloids Surf., A 2005, 266, 1. (21) Nakashima, T.; Shimizu, M.; Kukizaki, M. Kagaku Kogaku Ronbunshu 1993, 19, 991. (22) Kobayashi, I.; Nakajima, M.; Mukataka, S. Colloids Surf., A 2003, 229, 33.

Langmuir, Vol. 22, No. 10, 2006 4527

Figure 1. Schematic diagram of the membrane homogenization apparatus and typical images of emulsion droplets before and after homogenization. wt % NaN3 (as an antibacterial agent). The pH of the solution was then adjusted to 3.0 with 1 M HCl. A solution of β-Lg (0.5 or 2 wt %) was prepared by dispersing β-Lg in 5 mM phosphate buffer (NaH2PO4/Na2HPO4 at pH 7) containing 10 mM NaCl and 0.02 wt % NaN3. Emulsion Preparation. O/W emulsions were prepared by homogenizing corn oil and an aqueous surfactant solution using the membrane homogenization apparatus shown in Figure 1. The oil and surfactant solution were first premixed for several minutes using a stirring bar followed by five passes through a membrane homogenizer (External pressure-type micro kit, MG-20-5, Kiyomoto Iron Works Ltd., Japan). The mean particle size of the premix was typically 100-170 µm, and the coefficient of variation (CV) was in the range of 30-90%. The mean particle size in the homogenized emulsions after five passes was 4-9 µm with a CV typically in the range between 12 and 30%. The pressure vessel was then filled up with 100 mL of the premix, and the required driving pressure was built up with compressed air using a precision pressure regulator (PRG101, Omega, Stamford, CT). The operating pressure was measured with an accuracy of ( 1 kPa using a digital pressure gauge (PG-200-103G-P, Copal Electronics, Tokyo, Japan). The fine emulsion that had passed through the membrane tube from outside to inside was collected into a beaker placed on an electronic balance (Accu-622, Fisher Scientific, Fair Lawn, NJ). The balance was interfaced to a personal computer (PC) to collect time and mass data every 2 s using data acquisition software (AccuSeries USB, version 1.2, Fisher Scientific, Fair Lawn, NJ). The experiments were carried out at 295 K. The membrane used was an SPG membrane (8.5 mm inner diameter × 0.8 mm wall thickness) supplied from SPG Technology Co., Ltd. (Sadowara, Japan). The mean pore size of the membrane was 8.0 µm, the effective membrane length was 12 mm (Figure 2), and the effective cross-sectional area was 3.75 cm2. The membrane was cleaned after use by immersing it for 2 days in ethanol plus 2 days in toluene, followed by heating at 500 °C for 30 min in an electric muffle furnace. The inherent membrane permeability to pure water was completely restored by this treatment. Particle Size Measurements. The particle size distribution of the emulsions was measured by static light scattering (Mastersizer X, Malvern Instruments Ltd., Malvern, U.K.). The computer software used to analyze the angular dependence of the scattered light intensity identified the particle size distribution that gave the best fit between the experimental measurements and the theoretical predictions made using the Mie theory. A refractive index ratio of 1.08 was assumed in the calculation of the particle size distributions, and the mean

4528 Langmuir, Vol. 22, No. 10, 2006

Figure 2. Schematic diagram of the membrane module (A) and a typical scanning electron microscopy (SEM) image of an SPG membrane showing interconnected pores with irregular cross-sections (B). particle size was reported as the volume median diameter, d50, which is the value for which 50 vol % of the particles are bigger and 50 vol % are smaller. Optical Microscopy. The photographs of prepared emulsion droplets were taken using conventional light microscopy (Nikon Eclipse E400 microscope, Nikon Corporation, Japan). Emulsions were gently agitated in a glass test tube before analysis to ensure that they were homogeneous. A drop of emulsion was then placed on a microscope slide and covered by a cover slip. The images were acquired using a CCD camera (CCD-300-RC, DAGE-MTI, Michigan City, IN) connected to Digital Image Processing software (Micro Video Instruments Inc., Avon, MA) installed on a PC. Interfacial Tension Measurements. The interfacial tension at the corn oil/surfactant solution interfaces was measured by processing the images of pendant drops acquired by the video camera using a drop shape analysis system (DSA 10, Kru¨ss, Hamburg, Germany). Viscosity Measurements. The viscosity of the surfactant solutions at 2.0 wt % and the viscosity of emulsions prepared with the surfactant solutions were measured using a dynamic shear rheometer with a temperature-controlled concentric cylinder measurement cell (Constant Stress Rheometer, CS-10, Bohlin Instruments, Cranbury, NJ). The diameter of the rotating inner cylinder was 25 mm, and the diameter of the static outer cylinder was 27.5 mm, so that the width of the gap between the two cylinders was 1.25 mm. Samples were placed in the temperature-controlled measurement cell with a thin layer of mineral oil on the top to prevent water evaporation, and allowed to equilibrate to the required temperature (25 °C) prior to measurements. Data are presented in the form of viscosity-shear stress curves over the shear stress region from 0.1 to 10.0 Pa.

Results and Discussion Viscosity of Emulsions. Figure 3 shows the effect of different surfactants on the apparent viscosity versus shear stress profiles for the emulsions prepared with a dispersed phase volume fraction of φ ) 0.4, using five passes through the membrane. As expected, all emulsions exhibited non-Newtonian pseudoplastic behavior, that is, the apparent viscosity, η, decreased with an increase in shear stress. The majority of viscosity change occurred over a shear stress range from 1.7 to 5 Pa. For shear stresses greater than 5 Pa, a constant η value of 4.1 ( 0.1 mPa‚s was established for all surfactants investigated. As shown in Figure 4, the relative viscosity, η/ηc, that is, the ratio of the viscosity of emulsion to that of the continuous phase (surfactant solution) decreased from 3.3 to 3.7 at τ ) 0.1 Pa to a constant value of η/ηc ≈ 1.58 at τ > 5 Pa.

VladisaVljeVic´ et al.

Figure 3. Apparent viscosity vs shear stress profiles for emulsions with φ ) 0.4 stabilized by different surfactants and prepared by repeated membrane homogenization (n ) 5, ∆ptm ) 100 kPa). The surfactant concentration in the continuous phase was 2 wt % for all cases.

Figure 4. Relative viscosity vs shear stress profiles for emulsions with φ ) 0.4 stabilized by different surfactants and prepared by repeated membrane homogenization (n ) 5, ∆ptm ) 100 kPa). The surfactant concentration in the continuous phase was fixed at 2 wt %.

In a first approximation, the wall shear stress generated inside the pores by continuous phase flow is given by11 τw,p ) 8ηcVp/dp, where dp is the mean pore diameter, and Vp is the mean velocity inside the pores, which can be calculated by the equation Vp ) Jξp/, where J is the transmembrane flux, ξp is the mean pore tortuosity, and  is the porosity of membrane. Hence,

τw,p ) 8ηcJξp/(dp)

(1)

For SPG membranes, ξp ) 1.28 and  ) 0.55, according to Vladisavljevic´ et al.17 In this study, dp ) 8 µm and J ) 2-60 m3/(m2‚h), so that one obtains τw,p ) 5.3-159 Pa from eq 1 and Figure 3. Equation 1 assumes that the continuous phase does not contain dispersed particles that might affect the flow. This is obviously not true for the case of emulsion being forced through pores. For a hydrophilic membrane, however, large oil droplets moving through the pores are separated from the pore walls by a lubrication layer of continuous phase, as suggested by Hunter and Frisken,23 who studied the flow of lipid vesicles in pores. As the flux through (23) Hunter, D. G.; Frisken, B. J. Biophys. J. 1998, 74, 2996.

Emulsifier Type Effects in Membrane Homogenization

Figure 5. Effect of the concentration of Tween 20 in the continuous phase and the number of passes of emulsion through the membrane on the transmembrane flux.

Figure 6. Effect of the concentration of SDS in the continuous phase and the number of passes through the membrane on the transmembrane flux.

the membrane increases, the thickness of the lubrication layer increases and the radius of oil cylinders decreases until they break up into smaller pieces. Rate of Permeation through the Membrane. Dependences of transmembrane fluxes on the number of passes for the emulsions with a disperse phase content of φ ) 0.4 stabilized by SDS and Tween 20 are shown in Figures 5 and 6. The error bars show the standard deviation (σ) of the J values obtained from at least three repeated experiments. Obviously, the flux increases as the number of passes increases, but at a decreasing rate until it reaches a constant limiting value. At a concentration of SDS greater than 0.08 wt %, the limiting flux was virtually established after three passes, which indicates that all of the large droplets in the feed emulsion then were completely disrupted and only fine droplets remained. Under the same experimental conditions, the limiting flux for the Tween 20-stabilized emulsions was substantially smaller that than for the SDS-stabilized emulsions and was not established after five passes because the ease at which a droplet can be disrupted increases as the interfacial tension decreases. The transmembrane flux in Figures 5 and 6 increases as the emulsifier concentration increases because the interfacial tension is a decreasing function of the emulsifier concentration, as shown in Figures 7 and 8. It is important to note that, using the drop shape analysis instrument, we could not

Langmuir, Vol. 22, No. 10, 2006 4529

Figure 7. Effect of the concentration of Tween 20 in the continuous phase on the dynamic interfacial tension at oil/water interfaces.

Figure 8. Effect of the concentration of β-Lg in the continuous phase on the dynamic interfacial tension at oil/water interfaces.

measure the interfacial tension from the very beginning of the droplet formation process because of the limitations of the instrument. Consequently, many of the short time scale affects related to droplet formation within the membrane homogenizer would not have been observed using this analytical technique. The driving pressure in membrane homogenization is used to overcome flow resistance forces inside the pores and interfacial tension forces associated with droplet disruption:11

where C is a proportionality constant independent of the number of passes, Ji and di are the transmembrane flux and the mean particle size corresponding to the ith pass, respectively, Rm is the hydraulic resistance of clean membrane, and Rf,i is the overall fouling resistance during the ith pass. The fouling resistance is a consequence of the accumulation of droplets on the membrane surface (external fouling) and inside the pores (internal fouling). The increase in flux in the subsequent passes can be explained by two main reasons: (i) As homogenization proceeds, less and less energy is needed for droplet disruption because the mean droplet size approaches a constant minimum value (di ≈ di-1), and hence both ∆pflow and Ji increase. (ii) As the mean droplet size decreases, the rejection of droplets by the membrane decreases, and the accumulation of droplets on the membrane

4530 Langmuir, Vol. 22, No. 10, 2006

Figure 9. Effect of emulsifier content and the number of passes on the permeate weight vs time profiles. A linear permeate weight vs time relationship is an indication that droplets were not accumulated on the membrane.

surface is less pronounced. Figure 9 shows that, except for 0.5 wt % SDS, the flux sharply decreased with time in the first pass through the membrane because the operating pressure was not sufficiently high to cause the breakup of all large droplets once they reached the membrane surface. Therefore, the amount of droplets deposited on the membrane surface increased with time, which increased the fouling resistance and decreased the flux. The flux decline was not observed in the second pass, because the smaller droplets can pass through the membrane more easily without being rejected by the membrane. In addition, the tendency for droplet creaming decreases with decreasing the droplet size, so the effective droplet concentration in the emulsion being homogenized is kept constant throughout the whole pass. According to eq 2, ∆pdisr is directly proportional to φ, and hence the flux is significantly affected by the effective droplet concentration. The higher transmembrane flux in the first pass for the SDS-stabilized emulsions (Figure 9) can also be attributed to repulsive forces between similarly charged droplets, preventing them from coming into close contact when they accumulate on the membrane surface. Therefore, the droplets cannot get as close together as uncharged Tween-stabilized droplets can, and thus the oil layer deposited on the membrane surface is more porous and thus more permeable. As shown in Figure 10, the transmembrane flux in the presence of 0.5-2 wt % β-Lg in the continuous phase was much smaller than that for the emulsions stabilized by SDS and Tween 20, which is a consequence of higher interfacial tension and more pronounced membrane fouling. At pH 3.0, the β-Lg-stabilized droplets have a positive charge and the membrane surface has a negative charge, which leads to attractive electrostatic interactions between the membrane surface and oppositely charged droplets. As a result of these interactions, the membrane pores become progressively blocked by the β-Lg macromolecules, and, after three passes, the transmembrane flux diminishes virtually to zero. Mean Particle Size in the Emulsions. The effect of emulsifier concentration and the number of passes on the mean particle size in the homogenized emulsions with φ ) 0.4 stabilized by SDS and Tween 20 is shown in Figures 11 and 12. The mean particle size decreases as the number of passes increases asymptotically, approaching a limiting minimum value. The main mechanisms of droplet breakup inside the pores are snap off due to localized shear in a constriction, snap off due to droplet deformation and steric hindrance between the droplets, breakup due to interfacial

VladisaVljeVic´ et al.

Figure 10. Effect of pH, concentration of β-Lg in the continuous phase, and the number of passes through the membrane on the transmembrane flux.

Figure 11. Influence of the concentration of Tween 20 in the continuous phase and the number of passes through the membrane on the mean particle size in product emulsions.

Figure 12. Effect of SDS concentration on mean particle size in the second peak for various numbers of passes through the membrane.

tension effects (Rayleigh and Laplace instabilities), and breakup due to pore branchings and junctions.24 Snap off due to droplet deformation at the pore entrance occurs at low flow rates, and the resultant droplets are larger than the pore size. This snap off mechanism dominates in the first pass through the membrane.

Emulsifier Type Effects in Membrane Homogenization

Langmuir, Vol. 22, No. 10, 2006 4531

Figure 14. Influence of emulsifier type on particle size distribution: (1) 0.5 wt % SDS, fine emulsion, n ) 5 (d50 ) 5.1 µm, CV ) 27%, span ) 0.733); (2) 0.5 wt % Tween 20, fine emulsion, n ) 5 (d50 ) 6.6 µm, CV ) 12%, span ) 0.326); (3) 2 wt % β-Lg, fine emulsion, n ) 5 (d50 ) 10.9 µm, CV ) 31%, span ) 0.73); (4) 0.5 wt % Tween 20, feed emulsion (d50 ) 119 µm, CV ) 31%, span ) 0.77).

Figure 13. Images of pendant drops acquired by the video camera of the drop shape analysis system in the presence of different emulsifiers in the continuous phase.

In the subsequent passes, droplet breakup due to collisions with the pore walls as a result of pore branching becomes increasingly important. A droplet breaks at the junction if a critical capillary number is exceeded, that is, if Ca > Cacr, where Ca ) ηcV/γ. Here, ηc is the continuous phase viscosity, and V is the droplet velocity through the pore. The critical capillary number in microfabricated channel devices with T junctions is given by25 Cacr ) Ro(1/o2/3 - 1)2, where R is a dimensionless constant, and o is the initial droplet extension defined as the ratio of its initial length to its circumference. Sufficiently large droplets (o > 1) always break at the junction, independent of the flow velocity. Smaller droplets do not break if they move too slowly down the channel or if the interfacial tension γ is too high. The breakup of drops in an SPG membrane is governed by the same principles, although the pore morphology substantially differs from that in microfluidic devices. As shown in Figures 11 and 12, the mean particle size in the SDS-stabilized emulsions is smaller than that in the Tweenstabilized emulsions, which is a consequence of both the higher flow velocity through the pores and the smaller interfacial tension. Therefore, under the same emulsifier concentration, the capillary number for the SDS-stabilized emulsions is higher than that for the Tween-stabilized emulsions, and a droplet stabilized by SDS breaks at smaller initial size. For 2% Tween 20, the mean particle size in the fifth pass was only slightly reduced (from 6.5 to 6.4 µm, as shown in Figure 11), so that Ca ≈ Cacr. Under these conditions, V ) Jξp/ ) 0.03 m/s (from Figure 5), ηc ) 2.6 mPa‚s (from Figures 3 and 4), and γ ) 5.7 mN/m (from Figure 7), from which Cacr ) 0.014. It is interesting to note that 2% Tween 20 and 0.1% SDS in the continuous phase give nearly the same mean particle size and transmembrane flux in the final (fifth) pass. Figure 13 shows the stages of droplet growth at the tip of a capillary tube as a function of emulsifier added in the surrounding (24) van der Zwan, E.; Schroe¨n, K.; van Dijke, K.; Boom, R. Colloids Surf., A 2006, 277, 223. (25) Link, D. R.; Anna, S. L.; Weitz, A.; Stone, H. A. Phys. ReV. Lett. 2004, 92, article no. 054503.

aqueous phase. The final droplet size was maximal for 0.1% β-Lg and minimal for 2% SDS, which reflects dynamic interfacial tension. In the case of SDS, a substantial necking before detachment was observed because of relatively short droplet formation time and small interfacial tension. An elongated shape is unfavorable from the surface energy point of view because of high surface area per unit volume, and this shape is only possible if the interfacial tension is sufficiently small. Particle Size Distribution. Figure 14 shows particle size distribution curves at φ ) 0.4 for a typical premix and for fine emulsions homogenized using five passes through the membrane. Particle diameters in the premix were distributed over a wide range of 10-204 µm, with a CV of 31% and a mean particle diameter of 119 µm. It is important to note that the minimum particle size in the premix was higher than the mean pore size of 8 µm. The narrowest particle size distribution in the fine emulsion with a CV of 12% and a mean diameter of 6.6 µm was achieved using 0.5 wt % Tween 20 as an emulsifier. The optimal conditions for SDS-stabilized emulsions with regard to particle size uniformity were five passes through the membrane at 0.2 wt % SDS, under which fine emulsions with CV ) 13%, span ) 0.35, and d50 ) 6.2 µm were obtained (the particle size distribution is not shown in Figure 14). The β-Lg-stabilized emulsions were polydisperse at φ ) 0.4 with a CV of 31%. Membrane Fouling. Figures 15 and 16 show the effect of membrane fouling on the mean particle size and transmembrane flux for the emulsions containing 0.5% Tween 20 in the continuous phase. The membrane resistance Rm can be calculated from the flux of pure continuous phase (surfactant solution): Rm ) (1/ Jc)(∆ptm/ηc) ) 1.49 × 109 1/m, where Jc ) 93.2 m3‚m-2‚h-1, as shown in Figure 15, and ηc ) 2.6 mPa‚s. In the absence of any fouling resistance and ignoring the second term (∆pdisr) in eq 2, the emulsion flux is given by J′ ) Jc/(η/ηc) ) 59 m3‚m-2‚h-1, where η/ηc ) 1.58 is deduced from Figure 4. However, the actual emulsion fluxes J shown in Figure 15 are smaller than J′, which is a consequence of membrane fouling and the expenditure of pressure for droplet disruption. Assuming that ∆pdisr ) 0 in the fifth pass since corresponding reduction of droplet size is very small, as shown in Figure 16, it is possible to calculate the fouling resistance in the fifth pass: Rf,5 ) (1/J5)(∆ptm/η) - Rm ) 5.78 × 108 1/m, which is a value 60% lower than Rm. If a clean membrane was used in the fifth pass, the fouling resistance was 3.78 × 108 1/m, which is a 35% reduction of Rf,5. However, since

4532 Langmuir, Vol. 22, No. 10, 2006

Figure 15. Influence of membrane fouling on the transmembrane flux for the emulsions stabilized by 0.5 wt % Tween 20. After four passes, a fouled membrane was replaced by a clean membrane, and the flux of emulsion through the clean membrane is shown by a dashed line. The flux of pure continuous phase (Jc) is shown by a dash-dot line.

Figure 16. Effect of membrane fouling on the mean particle size for the emulsions stabilized by 0.5 wt % Tween 20. The mean particle size obtained using a clean membrane after four passes is shown by the dashed line.

the membrane resistance was a main resistance to permeate flow, the flux was improved only by 11% (see dashed line in Figure 15). As a result of higher flux, the mean particle size obtained using a clean membrane was slightly lower than that obtained using a fouled membrane (dashed line in Figure 16). Figures 17 and 18 show the effect of membrane fouling on the mean particle size and transmembrane flux for the emulsions stabilized by 2% β-Lg. The fouling resistance in the fifth pass was 1.09 × 1010 1/m, which is 2 orders of magnitude higher than that for 0.5% Tween 20 and an order of magnitude higher than the membrane resistance. The membrane fouling was probably caused by the combination of pore blocking by oil particles due to protein adsorption on the pore walls and the deposition of a cake layer on the membrane surface. The adsorption of β-Lg on the membrane surface can be attributed to electrostatic interaction between anionic silanol groups (-Si-O-) on the SPG surface and cationic patches (e.g., -NH3+) on β-Lg molecules, although the net charge of β-Lg molecules was negative at pH 7. If a clean membrane was used in the fifth pass, the fouling resistance was 5.25 × 109 1/m, which is a 2-fold reduction when compared with the fouled membrane. Since the fouling resistance was a dominant

VladisaVljeVic´ et al.

Figure 17. Effect of membrane fouling on the transmembrane flux for the emulsions stabilized by 2 wt % β-Lg at pH 7. A fouled membrane was replaced by a clean membrane after four passes, and the flux using the clean membrane is shown by a dashed line.

Figure 18. Effect of membrane fouling on the mean particle size for the β-Lg-stabilized emulsions at pH 7. The mean particle size obtained using a clean membrane after four passes is shown by the dashed line.

resistance to permeate flow, a flux improvement by a factor of 2.4 was achieved using a clean membrane in the fifth pass (see the dashed line in Figure 17). As a result, the mean particle size obtained using a clean membrane was much lower than that obtained using a fouled membrane (the dashed line in Figure 18).

Conclusions We have examined the influence of various emulsifier types on the transmembrane flux, membrane fouling, and particle size distribution in repeated membrane homogenization using an SPG membrane. An adequate choice of emulsifier concentration and number of passes through the membrane in addition to pH control and possible membrane replacement between the passes enables one to obtain relatively uniform emulsion droplets, irrespective of the emulsifier type investigated. A minimum droplet size polydispersity of CV ) 12-13% was obtained for 0.5 wt % Tween 20 and 0.2 wt % SDS. The transmembrane flux for Tween 20-stabilized emulsions was substantially smaller than that for SDS-stabilized emulsions, as the ease at which a droplet can be disrupted increases as the interfacial tension decreases. The transmembrane flux for an emulsion containing β-Lg at pH 3

Emulsifier Type Effects in Membrane Homogenization

was much smaller than that in the presence of SDS and Tween 20, which was a consequence of protein adsorption to the membrane surface, presumably due to electrostatic attractions between the negatively charged membrane surface and positively charged patches on the protein molecules. The fouling resistance in the fifth pass for β-Lg-stabilized emulsions at pH 7 was 2 orders of magnitude higher than that for Tween 20 and an order of magnitude higher than the membrane resistance. To avoid the detrimental effect of membrane fouling due to protein adsorption, it is recommended to use a clean SPG membrane in the final

Langmuir, Vol. 22, No. 10, 2006 4533

pass. At a pH below the isoelectric point of β-Lg (pH ) 3), the membrane pores became progressively blocked by oil droplets, and, after several passes, the transmembrane flux diminished to zero. Acknowledgment. This material is based upon work supported by the Fulbright Scholar Program (G.T.V.) administrated by the Council for International Exchange of Scholars, Washington. LA053410F