Double emulsions: how does release occur?

Journal of Controlled Release 79 (2002) 193–205 www.elsevier.com / locate / jconrel

Double emulsions: how does release occur? K. Pays a , J. Giermanska-Kahn a , B. Pouligny a , J. Bibette b , F. Leal-Calderon a , * b

a Centre de Recherche Paul Pascal, CNRS, Avenue Schweitzer, 33600 Pessac, France ¨ et Nanostructures’, UMR 7612, 10 Rue Vauquelin, 75005 Paris, France ESPCI, Laboratoire ‘ Colloıdes

Received 10 September 2001; accepted 27 November 2001

Abstract Water-in-oil-in-water double emulsions (W/ O / W) consist of dispersed oil globules containing smaller aqueous droplets. These materials offer interesting possibilities for the controlled release of chemical species initially entrapped in the internal droplets. A better understanding of the stability conditions and release properties in double emulsions requires the use of model systems with a well-defined droplet size. In this paper, we use quasi-monodisperse double emulsions made of calibrated water droplets and oil globules to investigate the two mechanisms that are responsible for the release of a chemical substance (NaCl). (i) One is due to the coalescence of the thin liquid film separating the internal droplets and the globule surfaces. (ii) The other mechanism termed as ‘compositional ripening’ occurs without film rupturing; instead it occurs by diffusion and / or permeation of the chemical substance across the oil phase. By varying the proportions and / or the chemical nature of the surface active species it is possible to shift from one mechanism to the other one. We therefore study separately both mechanisms and we establish some basic rules that govern the behavior of W/ O / W double emulsions.  2002 Elsevier Science B.V. All rights reserved. Keywords: Double emulsions; Coalescence; Compositional ripening

1. Introduction Water-in-oil-in-water double emulsions (W/ O / W) consist of dispersed oil globules containing smaller aqueous droplets. Taking advantage of their double (or multiple) compartment structure, an increasing interest has been devoted to double emulsions, as they can be considered as reservoirs of encapsulated substances to be released under variable conditions. Double emulsions are generally prepared with two surfactants of opposite solubility. To produce a W/ *Corresponding author. Tel.: 133-556-845-633; fax: 133-556845-600. E-mail address: [email protected] (F. Leal-Calderon).

O / W emulsion, a hydrophobic surfactant, i.e. a surfactant with low Hydrophilic to Lipophilic Balance (HLB,10) is first dissolved in oil. Then water is added and a W/ O emulsion is formed. The system is then emulsified again in an aqueous solution of surfactant with a high HLB number (.10) to produce a W/ O / W double emulsion. Both surfactants mix at the water / oil interfaces and the lifetime of the films as well as their permeation properties are governed by the composition of the binary surfactant mixture. The problems associated with the stability and release properties of double emulsions still raises interesting questions even if there has been a large literature published on the subject over the two past decades [1,2].

0168-3659 / 02 / $ – see front matter  2002 Elsevier Science B.V. All rights reserved. PII: S0168-3659( 01 )00535-1

194

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

A better understanding of the stability conditions and release properties in double emulsions should require the use of model systems with a well-defined droplet size. In this paper, we use quasi-monodisperse double emulsions made of calibrated water droplets and oil globules to investigate the two mechanisms that are responsible for the release of a chemical substance. (i) One is due to the coalescence of the thin liquid film separating the internal droplets and the globule surfaces. (ii) The other mechanism termed as ‘compositional ripening’ occurs without film rupturing; instead it occurs by diffusion and / or permeation of the chemical substance across the oil phase. By varying the proportions and / or the chemical nature of the surface active species, it is possible to shift from one mechanism to the other one. We therefore study separately both mechanisms and we establish some basic rules that govern the behavior of W/ O / W double emulsions.

2. Experimental section

2.1. System composition For the oil phase, we used either alkane oils (octane, dodecane, hexadecane from Aldrich) or commercial sunflower oil (Stora). To stabilize inverted W/ O emulsions, we used one of the three following oil soluble-surfactants: • Span 80 (Sorbitan monooleate) from Aldrich with HLB54. • Arlacel P135 (polyethylene-30 dipolyhydroxystearate) from ICI. MW ¯ 5000 g / mol. HLB5 5–6. • Admul Wol 1403 (Polyrycinoleate of polyglycerol) from QUEST International. MW ¯ 4400 g / mol. HLB54. To stabilize direct O / W emulsions, we choose different hydrophilic surfactants: • Sodium dodecyl sulphate (or SDS, HLB540) purchased from Aldrich. Its critical micellar concentration (CMC) in pure water is equal to 83 10 23 mol l 21 .

• Synperonic PE / F 68 provided by ICI. This surfactant is a tri-block copolymer of ethylene oxide (EO) and propylene oxide (PO). The average formula is the following: 75 EO / 30 PO / 75 EO. The average molecular weight is MW ¯ 8400 g / mol and HLB529. In some experiments, we added a polysaccharide, alginate HF120L, of molecular weight MW ¯ 54000 g / mol, provided by Disatec from Promova. This polymer consists of D-mannuronate and L-guluronate. It acts as a thickening agent and has only weak interfacial properties since it does not vary the interfacial tension between oil and water in the presence of the previously mentioned surface-active agents. Moreover, alginate molecules do not stabilize emulsions by themselves, i.e. in the absence of surface-active species. Other biopolymers like carboxymethyl cellulose (cmc, Akucell AF 0705 from Akzo Nobel) or hydroxyethyl cellulose (HEC, from Aldrich) were also probed.

2.2. Emulsion preparation 2.2.1. Preparation of the primary W/O emulsions We first prepare a monodisperse water-in-oil inverted emulsion, stabilized by an oil-soluble surfactant. Salt (NaCl) is added to the dispersed phase as a tracer to probe the release mechanisms and also to avoid the coarsening phenomena [3]. A Couette-type mixer consisting of two concentric cylinders was used for emulsion fabrication. A scheme of the set-up is available in Refs. [4,5]. After emulsification, the lipophilic surfactant concentration with respect to the continuous phase is always fixed at a given value Cl and the water droplet volume fraction is set to a given value f i0 . 2.2.2. Preparation of the double W/O /W double emulsion and characterization Two different methods were developed to produce quasi-monodisperse globules. In the first one [6], a turbulent flow is employed while in the second one [7,8] we apply a laminar shear flow to produce calibrated globules. Glucose is dissolved in the external water phase of the W/ O / W double emulsions to avoid any water transfer. The glucose concentration is always adjusted to match the os-

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

195

Fig. 1. (a) Microscopic image of a W/ O / W emulsion obtained following the procedure described in Ref. [6]. The double emulsion is stabilized by the surfactant mixture Span 80 / SDS, the oil phase is dodecane: d g 59 mm, d i 50.36 mm, fi 520%, fg 510%, Ch 50.1 CMC, Cl 52% (w / w). (b) Microscopic image of a W/ O / W emulsion obtained according the method described in Refs. [7,8]. The double emulsion is stabilized by the surfactant mixture Admul wol 1403-SynperonicPE / F68, the oil phase is dodecane. The continuous phase contained 2% (w / w) of Alginate HF 120L.

motic pressure induced by the presence of NaCl in the internal droplets. Fig. 1a,b are microscopic images of double emulsions fabricated according the two methods. Large oil globules very uniform in size are visible, and the smaller inverted water droplets are also distinguishable. The emulsions are observed with a phase contrast optic microscope (Zeiss, Axiovert X100). Since the double emulsions possess a very narrow size distribution, it is easy to determine the average globule diameter from microscopic images. We also used a Malvern Mastersizer granulometer to measure the size distribution of the emulsions. The collected scattered intensity as a function of the angle is transformed into the size distribution using the Mie theory. The polydispersity of the emulsion is characterized by a parameter termed as ‘uniformity’ and defined as: ] N D 3uD 2 Diu 1 i i i U 5] ] ]]]]] D Ni D 3i

O

O i

Where Ni is the total number of droplets with ] diameter Di and D is median diameter, i.e. the diameter for which the cumulative undersized volume fraction is equal to 50%. In the following, we shall characterize the obtained emulsions through their mean droplet size in volume and U. To ensure reliability of the data, three independent granulometric measurements were performed with samples collected in three distinct zones of the emulsion volume. The obtained results were reproducible to within 5%. Let us stress that in the experiments that will be described, globule–globule coalescence did not take place and therefore the globule monodispersity was preserved as revealed by measurements performed at different times of the experiments. For all the emulsions considered below (simple and double), the uniformity U was always lower than 30%. In comparison with previous studies, our systems possess the advantage to be well calibrated and reproducible. This is an important property since, as will be demonstrated in the following sections, the rate of release of double emulsions is

196

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

strongly dependent upon the colloidal size of the dispersed objects. In the following, we shall systematically use the following notations to characterize the double emulsion composition and properties: Cl 5concentration of the lipophilic surfactant in the oil globules; Ch 5 initial concentration of the hydrophilic surfactant in the external water phase; fi 5volume fraction of water droplets in the oil globules; fg 5volume fraction of oil globules in the external water phase; d i 5diameter of the water droplets; d g 5diameter of the oil globules.

In Eq. (1), rw is the density of the water phase, g is the acceleration of gravity (9.8 ms 22 ) and hc the continuous phase viscosity. It then becomes possible to determine the internal droplet volume fraction fi inside the globules according the relation:

rg 5 fi rwi 1 (1 2 fi )ro

(2)

Where ro is the oil density and rwi is the density of the internal water phase.

3. Results and discussion

2.3. Techniques to follow the kinetics of release 3.1. W/O /W surfactant-stabilized emulsions We used different techniques to study the kinetic evolution of the double emulsions. The concentration of salt (NaCl) present in the aqueous external phase is measured by means of an Ag /AgCl specific electrode (from Radiometer, France) which is sensitive to the chemical activity of chloride ions. The measured potential is transformed into salt concentration using a calibration curve. The emulsion is very gently stirred (maximum 60 rpm) in order to avoid the creaming of the oil globules and any non-homogeneous distribution of the salt concentration. We verified that this slow stirring does not perturb the release process. Indeed, as long as the stirring remains low, the rate of release does not vary and is the same as the one obtained with emulsions manually shaked at regular time intervals. We combined this technique with direct observations under microscope as well as repeated single globule creaming experiments using optical manipulation [9,10]. In our set-up, a unique non-Brownian globule (more than 10 mm in diameter) is illuminated by one or two moderately focused laser beams. The radiation pressure exerted by the lasers allows to capture and to displace a globule at any position in a transparent cell. When the lasers are switched off, the globule moves up because of buoyancy. We then measure under microscope the globule diameter d g as well as its creaming velocity Vcreaming in the stationary regime. From the stokes equation, we deduce the average density of the globule rg : ( rw 2 rg )gd g2 Vcreaming 5 ]]]] 18hc

(1)

3.1.1. General behavior We first consider a quasi-monodisperse double emulsion with moderate internal droplet and globule volume fractions (f 0i , 20%, fg , 20%. Span 80 is used for the stabilization of the primary W/ O emulsion at concentration Cl 52% (w / w) with respect to the continuous phase. The internal droplets have a diameter d i 50.36 mm with a uniformity of 30%. If the double globules are stabilized in water by SDS at CMC / 10 (CMC58310 23 mol l 21 ), the system does not exhibit any structural evolution after a few days of storage. However, if the SDS concentration is equal or larger to approximately 1 CMC, the double W/ O / W emulsion rapidly transforms into a simple W/ O emulsion. The characteristic time scale for the transformation becomes shorter when the SDS concentration increases, in perfect agreement with the pioneering experiments of Ficheux et al. [11]. For globules with diameter d g 54 mm, f i0 520% and Ch 510 CMC, it takes around 300 min for the transformation to occur. Repeated observations under the microscope reveal that the globules become progressively empty and that there is apparently no coarsening of the internal droplets. In order to elucidate the origin of this evolution, we produce double emulsions with large internal droplets that can be perfectly distinguished under the microscope. Within the same conditions, we observe that the internal droplets may spend some time in contact with the globule surface without exhibiting any structural change, and then suddenly disappear. From all the previous observations, it can be con-

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

cluded that the mechanism responsible for the transformation from double to simple emulsion is the coalescence of the internal droplets on the globule surface. A slightly distinct scenario of destruction occurs when the initial droplet volume fraction f 0i exceeds approximately 40%. The transition from double to simple emulsion is still observed but in this case there is some coarsening of the internal droplets during the process of destruction. Some large nuclei resulting from droplet–droplet coalescence are clearly distinguished under microscope. Once they reach the globule surface, they coalesce rapidly and disappear. In the last stages of the transformation, when the droplet concentration becomes small enough, we do not observe anymore the coarsening and the scenario becomes identical to the one observed for the systems with low initial droplet concentration. The phenomenology above described is general since it was reproduced using different ionic surfactants with high HLB values (.30), like alkyl sulfonates or alkyl quaternary ammonia.

3.1.2. Kinetics of release In this section, by varying the hydrophilic surfactant concentration, we identify and separate two release mechanisms in double emulsions. One occurs by diffusion and / or permeation of the salt across the oil globule. The second one involves coalescence, i.e. film rupturing between the internal droplets and the globule surface. The process of film rupturing is initiated by the spontaneous formation of a small hole. The nucleation frequency V of a hole that reaches a critical size, above which it becomes unstable and grows, determines the lifetime of the films with respect to coalescence. A mean field description [12] predicts that V varies with temperature T as an Arrhenius law: V 5 V 0 exp(2Ea /kT ). We study the kinetics of release in the regime dominated by coalescence and we propose an unambiguous method for the measurement of the microscopic parameters V 0 and Ea based on the use of monodisperse water-in-oil-in-water type (W/ O / W) double emulsions. Our method exploits the fact that the total number of internal droplets adsorbed on the globule surface governs the rate of release. An attractive London–Van der Waals interaction exists both between the small internal droplets and between

197

the droplets and the globule surface. However, since the globules are at least 10 times larger than the entrapped droplets, the attraction between the almost flat globule surface and the small droplets is nearly twice as large as that between the inner droplets [13]. This attraction is small enough for the small droplets to behave like a gas that reversibly adsorbs onto the globule surface. At low internal droplet volume fraction, by varying the concentration of the hydrophilic surfactant, we exploit the regime where the leakage is controlled only by the droplet / globule coalescence. In other words, the coalescence is the rate-determining mechanism because it occurs at much greater rate than diffusion and / or permeation. In such conditions, measuring the rate of release allows a direct determination of the average lifetime of the thin film that forms between the small internal droplet and the globule surface. We therefore deduce the activation energy and the natural frequency of the hole nucleation process by exploring the temperature dependence of the rate of release.

3.1.2.1. Limit of low SDS concentration and low internal droplet volume fraction. Because in the following experiments the internal water volume is about 100 times smaller than the external one, Ch fixes the chemical potential of the SDS molecules. The thin liquid film that forms between the internal droplets and the globule surface is comprised of two mixed monolayers covered by Span 80 and SDS molecules, separated by oil. Since SDS molecules migrate from the external to the internal water phase within very a very short period of time (1 min [14]), the film can be regarded as close to thermodynamic equilibrium with respect to surfactant adsorption a few minutes after preparation. Following the wellknown Bancroft rule, such inverted films possess a long-range stability when essentially covered by hydrophobic surfactant (Span 80) but become very unstable when a strong proportion of hydrophilic surfactant is adsorbed. From the previous experiments, we learned that the transition from long range to short range stability may be achieved by varying the concentration of the hydrophilic surfactant in the external water phase. Several emulsions are then prepared with 2% (w / w) of Span 80 as the emulsifier of the primary emulsion and SDS, at various concentrations Ch in

198

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

Fig. 2. Influence of SDS concentration in the external water phase on the kinetics of release. d g 54 mm, d i 50.36 mm, f 0i 520%, fg 510%, Cl 52% (w / w of Span 80).

the external continuous phase. Fig. 2 shows the quantity of released salt (expressed in relative percentage) as a function of time. The globule diameter d g is 4 mm, the initial droplet volume fraction f 0i is 20% and the globule volume fg fraction is 10%. From these curves, we can deduce that two limiting mechanisms control the salt release in double emulsions. Since they occur over time scales that are significantly different, they can be decoupled. For Ch # CMC, the rate of release is quite slow, occurring over a characteristic time scale of several days. The rate decreases with Ch , being minimal around 1 CMC. When the process is achieved (nearly 100% released), we observe under microscope that the water droplet concentration fi in the globules has apparently not varied. This is confirmed by the creaming technique, since we measure a constant rate of creaming of the globules after 3 days of storage. The fi value deduced from Eqs. (1) and (2) corresponds to the initial one. We therefore conclude that in this regime, the salt release occurs without film rupturing, instead it is produced by an entropically driven diffusion and / or permeation of the salt across the oil globule. For Ch . CMC, the release is quite fast and the rate increases with the hydrophilic surfactant concentration. Repeated observations under the microscope reveal a gradual decrease of the inner droplet concentration, being almost zero when 100% of release is attained. When Ch , 5 CMC, we do not observe any coarsening of the water-in-oil droplets. This definitely confirms that in

this SDS concentration range, the salt release is controlled by the coalescence of the internal droplets on the globule surface. We now examine, in more detail, the regime governed by coalescence. In the following experiments, Ch 53 CMC, d g 53.6 mm, fg 510% and we vary the initial internal droplet volume fraction between 5 and 35%. In the experimental conditions and within the time scales that are explored (less than 1000 min), the contribution of diffusion / permeation across the oil globule can be neglected. The data are represented in Fig. 3. A significant decay of the characteristic time of release as a function of the internal droplet volume fraction is observed. At this point, we make the assumption that a fraction of the internal droplets is adsorbed on the globule surface. This is a natural consequence of the Van der Waals attraction that exists between the internal droplets and the external water phase [13]. Let us define n i as the total internal droplet concentration within the globules, and n a as the concentration of adsorbed droplets (per unit volume of the globules). We assume that the number of coalescence events per unit time is simply proportional to the concentration of adsorbed droplets: dn i / dt 5 2 Vn a

(3)

where V is the characteristic frequency of coalescence between an adsorbed droplet and the globule surface. At any time t, n i is calculated from the ordinate of the curves (Fig. 3) and the number of

Fig. 3. Influence of the initial internal droplet volume fraction on the kinetics of release. d g 53.6 mm, d i 50.36 mm, fg 510%, Ch 53 CMC, Cl 52% (w / w of Span 80).

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

coalescence events dn i / dt may be deduced from the derivative of the curves. All the experimental points in Fig. 3 can be transformed and plotted again in (n i , dn i / dt) coordinates. We observe that all the data lie within a single curve, which means that the rate coalescence dn i / dt depends only on n i [14,15]. Following Eq. (3), this function is proportional to V and corresponds to the adsorption isotherm of the water droplets on the globule surface n a 5 f(n i ). We aim now to model the adsorption isotherm in order to deduce a numerical value for V. Following the model of Frumkin and Fowler [16], n a is given by the following set of equations:

S

u a 1 4ul Q Q ]] 5 K(n i 2 n a ) exp 2 ]]] 12Q kT na Q5] n0

D

(4) (5)

Where, n 0 is the total concentration (per unit volume) of available sites for adsorption easily deduced from geometrical considerations, u a is the adsorption

199

energy, ul is the lateral energy of interaction between the droplets and finally K is a constant calculated from the model [16]. From the average length of the surfactant tails (l ¯ 3 nm), we get an estimation of the Van der Waals interactions: Ad i Ad i ul ¯ 2 ] 5 2 1.75 kT and u a ¯ 2 ] 5 2 3.5 kT 24l 12l (for the evaluation of u a , the globule surface is assumed to be flat, and the Hamaker constant A between for the dodecane / water couple is taken to be 0.3310 220 J) [13]. The coalescence frequency is, therefore, the unique free parameter in our model and is determined from the best fit to the experimental curves. In Fig. 4a is plotted the kinetic evolution of n i at Ch 5 3 CMC and for a globules diameters of d g 53.6 mm. Using one and the same value of V, the theoretical points correctly fit the experimental data whatever the globule diameter (see [15]): V563 10 23 min 21 . From the obtained numerical value of V, it can be estimated that a droplet of diameter 0.36

Fig. 4. Number density of internal droplets in the globules as a function of time. d g 53.6 mm, d i 50.36 mm, f i0 520%, fg 512%, Cl 52% (w / w of Sapn 80). (a) Ch 53 CMC; (b) Ch 56 CMC; (c) Ch 510 CMC. In the inserts, the numbers correspond to the V values tested to fit the experimental data.

200

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

mm spends on average 3 h on the globule surface before a coalescence event to occur. The same type of measurements were performed in the same conditions (Ch 53 CMC) but varying the oil chemical nature: for octane globules we found V50 min 21 (no coalescence occurred) and V52.5310 22 min 21 for hexadecane. In order to determine the activation energy Ea and the hole natural frequency, we vary the temperature between 15 and 60 8C, all other parameters being constant. In Fig. 5, we show the evolution of Ln(V) as a function of 1 /kT (k being the Boltzmann constant) for the system SDS / Span 80 / Dodecane with d i 50.36 mm. From the best fit to our data (solid line), we get Ea 5 30 kT r , T r being the room temperature (25 8C). From the intercept, we obtain V 0 5 4310 10 min 21 . The ratio V(25 8C) / V 0 is roughly equal to 10 13 meaning that only one hole over 10 13 grows and ultimately produces a coalescence event.

3.1.2.2. Limit of high SDS concentration or high internal droplet volume fraction. In Fig. 4b,c, we illustrate the effect of varying the hydrophilic surfactant concentration from 3 to 10 CMC. It is clear from the graph that increasing the hydrophilic surfactant concentration has the effect to accelerate the salt release. We tried to fit the experimental curves with our adsorption-coalescence model, using V as unique free parameter. As can be observed in Fig. 4a, the agreement between theory and experiment is fairly

Fig. 5. Frequency of coalescence as a function of 1 /kT. d i 50.36 mm, f 0i 520%, fg 510%, Ch 53 CMC, Cl 52% (w / w).

good at 3 CMC but large deviations appear at higher SDS concentrations. In Fig. 4c, two different V values are probed to fit the experimental data but neither of them is satisfactory. From this, we conclude that the model valid at low SDS concentration is not correctly describing the experimental results at high SDS concentrations. In other words, the rate of release cannot be anymore described in terms of a unique coalescence frequency when SDS concentration is higher than about 5 CMC. We guess that at such high concentrations, droplet–droplet coalescence accelerates the rate of release. Indeed internal droplet coalescence produces large nuclei which are preferentially adsorbed on the globule surface due to their larger Van der Waals attraction. Moreover, when a nucleus coalesces at the globule surface, a larger amount of matter is released at a time. The inherent polydispersity resulting from droplet–droplet coalescence can, therefore, explain the complex behavior of highly concentrated SDS double emulsions and the fact that the process can not be described using a single coalescence frequency. The same type of conclusion can be established when the initial internal droplet concentration is large, even at low surfactant concentration. The general behavior for the release of double emulsions stabilized by short surfactants has been described. By appropriately choosing the surfactant concentrations as well as their chemical nature, we were able to dissociate the two mechanisms that are responsible for the release of chemical substances. This allowed us to study the stability towards coalescence of mixed surfactant films, i.e., films stabilized by both water and oil-soluble surfactants. From a practical point of view, we can conclude that long term encapsulation of small neutral or charged molecules is not really possible using short surfactants as stabilizing agents. Indeed, when the hydrophilic surfactant concentration is lower than a critical value C h* , the release occurs preferentially by diffusion and / or permeation across the oil phase, while above C h* it occurs preferentially by coalescence. In both cases, the characteristic period of release does not exceed several days, which is not sufficient for most of the practical applications. C h* is of the order of 1 CMC for highly hydrophilic surfactants (HLB. 30) and 100 CMC for lower HLB surfactants (10, HLB,30) [11]. Because of the inefficiency of short

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

201

surfactants, much effort is being spent in finding out formulations incorporating polymeric stabilizers [1] or associating short surfactants and polymers or proteins [2]. It is indeed expected that the presence of large molecules at the interfaces will reduce the inner droplet adsorption on the globule surface and increase the activation energy for the hole nucleation process. Promising results have already been empirically obtained [2] but there is still work to be done to perfectly control and understand the release properties of these new materials as well as to ensure their reliability as commercial products.

3.2. W/O /W polymer-stabilized emulsions The purpose of the next section is to explore the behavior of double emulsions stabilized by synthetic polymeric surfactants. We find a regime where the release in controlled by compositional ripening. We propose a model based on Fick’s theory to model the release kinetics. Finally, by associating polymeric surfactants and biopolymers, we fabricate new materials that ensure long term encapsulation of NaCl thus providing an interesting guide to formulate commercial products.

3.2.1. General behavior We used two different techniques previously mentioned to study the kinetic evolution of the double emulsions. The concentration of salt (NaCl) present in the aqueous external phase is measured by means of an Ag /AgCl specific electrode. We combined this technique with direct observations under microscope. In Fig. 6, we plot the relative percentage of salt released during the first 5 days following the preparation of the double emulsion stabilized by hydrophobic Arlacel P135 and hydrophilic Synperonic PE / F68 (curve at T525 8C). The detailed composition is reported in the caption of Fig. 6. The average globule diameter is 4 mm and the globule volume fraction is fixed at 18%. When the process is achieved (nearly 100% released), we observe under microscope that the water droplet concentration fi in the globules has apparently not varied. We therefore conclude than in this regime, the salt release occurs without film rupturing, instead it is produced by diffusion and / or permeation of the salt across the oil globule.

Fig. 6. Kinetics of leakage of an emulsion stabilized by Arlacel P 135 and Synperonic PE / F68 as a function of temperature. [NaCl] 0 50.4 mol / l, f i0 520%, fg 518%, d i 50.36 mm, d g 54 mm, Ch 55%, Cl 52% (w / w); oil:dodecane.

3.2.2. Phenomenological model for compositional ripening Assuming that there is no structural evolution of the double emulsion (absence of coalescence), the entropically driven molecular flow J of chloride ions across the oil globules is given by Fick’s law: dNi J 5 2 ] 5 PS(Ci 2 Ce ) dt where S is the total surface involved in the transfer process, Ni is the number of Cl 2 moles in the internal droplets, Ci and Ce are the concentration of chloride ions in the internal and external water phases respectively. P, termed as permeability coefficient, characterizes the rate of release across the oil membrane. P can be seen as a phenomenological constant that reflects the influence of all the microscopic parameters involved in the salt permeation process. Of course, it is expected that P depends, among others, on the chemical nature of the encapsulated substance, on the oil chemical nature, on the monolayer compositions, etc. The previous equation assumes a quasi-stationary process in which the permeation across the oil globule is the rate-determining factor. Integrating the previous differential equation, one gets: N 0i 11a Ni 5 ]] a 1 exp 2 PS ]] t 11a Vi

F

S

DG

(6)

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

202

where N 0i is the initial number of encapsulated moles and a 5Vi /Ve , Vi and Ve being the internal and external water volumes respectively, which are assumed to remain constant in time. Combining Eq. (6) and the conservation condition, we finally obtain the following expression for the relative percentage of release: Ce (t) %released(t) 5 100]] C`

F

S

DG

11a 5 100 1 2 exp 2 PS ]] t Vi with

fg fi C i0 N 0i C` 5 ]] 5 ]]]]] Vi 1Ve s1 2 fgd 1 fi fg where C 0i is the initial NaCl concentration in the internal droplets. The model predicts an exponential leakage of the encapsulated substance as a function of time. If S is assumed to be equal to the total globule surface, we can numerically deduce P from the initial slope p0 of the experimental curves:

fis1 2 fgd p0 d g P 5 ] ] ]]]]] 100 6 s1 2 fgd 1 fi fg

(7)

3.2.3. Microscopic approaches of the permeability—State of the art The compositional ripening in double emulsions is reminiscent of the so-called ‘passive’ leakage measured across phospholipid bilayers and which is partially responsible for chemical exchanges across biological membranes. At a microscopic level, several models to explain the permeation phenomenon have been proposed in the literature, all of them being in agreement with Fick’s law. Some models propose that the permeation across the membranes results from a solubilization process followed by diffusion across the hydrophobic part of the phospholipid membrane [17]. For hydrophilic substances, the rate-determining parameter is the so-called Born energy that represents the energy cost for the transfer of a hydrophilic species from a high dielectric to a low dielectric constant medium. Other models consider that the permeation is due to reversible and thermally activated holes that are permanently formed in the bilayers [17]. The characteristic hole

size may allow the passage of small hydrophilic substances with sufficiently high diffusion coefficient. The rate of transfer is controlled by the energy cost for hole formation, which includes line tension, surface tension and curvature effects. Although involving different microscopic mechanisms, the previously mentioned limiting models lead to quite similar expressions for the permeation coefficient. Indeed P is predicted to follow an effective Arrhenius law P 5 P 0 exp(2´a /kT ), where ´a is the activation energy, and P 0 a pre-factor which is not or only weakly temperature-dependant. These models have been established for simple and well-defined phospholipid membranes. It is clear that the composition of the membrane separating the internal and external water compartments in double emulsions is substantially more complex with two surfactant mixed monolayers separated by an oil membrane eventually containing micelles. However, as will be deduced from the next section, compositional ripening in double emulsions is affected by temperature and identically follow an Arrheniustype law.

3.2.4. Influence of temperature on compositional ripening We measure the rate of leakage of two distinct emulsions, one stabilized by short surfactants (Span 80, SDS) and the other one by amphiphilic polymers (Arlacel P135, Synperonic PE / F68). In both cases, the concentrations are chosen such as the leakage is ensured by compositional ripening only. This is confirmed by microscopic observations which reveal that both the internal droplet size as well as the droplet concentration inside the globules, apparently, does not vary in time, even when 100% of leakage is attained. The exact composition of the emulsions, as well as the colloidal diameters, are given in the captions of Figs. 6 and 7 representing the % released as a function of time at different temperatures. The profiles can be reasonably fitted by a mono-exponential function. Combining the initial slope of the experimental curves p0 and relation (7), we deduce the permeation coefficient P at a given temperature. In Fig. 8, we plot the evolution of log(P) as a function of the inverse thermal energy. Within experimental uncertainty, the variation is linear for both types of stabilizing agents in agreement with the

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

Fig. 7. Kinetics of leakage of an emulsion stabilized by Span 80 and SDS as a function of temperature: [NaCl] 0 50.4 mol / l, f 0i 520%, fg 512%, d i 50.36 mm, d g 53.6 mm, Cl 52% w, Ch 5 0.3 CMC, oil:dodecane.

previous theoretical models. From the Arrhenius plot, we deduce:

H

´a 5 20 kTr , (Tr 5 298K) P 0 5 2.8 3 10 23 ms 21

for the Span 80 / SDS stabilized system, and

H

´a 5 20 kTr , P 0 5 1.8 3 10 25 ms 21

for the Arlacel P135 / Synperonic PE / F68 stabilized system. Of course, the mere values can not, by themselves, elucidate the origin of the permeation mechanism (solubilization1diffusion or hole nucleation?) but it can be stated that the intrinsic permeation coefficient P 0 of the polymer-stabilized double

203

Fig. 9. Influence of the oil chemical nature on the kinetics of release for a double emulsion stabilised by Arlacel P135 and 0 0 Synperonic PE / F68. [NaCl] 50.4 mol / l, f i 516%, fg 518%, d i 50.36 mm, d g 55 mm, Cl 52% w, Ch 55% w, T555 8C.

emulsion is smaller resulting in a slower rate of leakage. This gives us a hint that polymers are more suitable surface-active species to ensure long term encapsulation than low molecular weight surfactants.

3.2.5. Influence of the oil chemical nature In Fig. 9, we plot the temporal evolution of the % released for two emulsions formulated with identical stabizing agents, with globules and droplets of the same colloidal size, but with oils of different chemical nature: in one case, the oil used is dodecane, while in the other one, it is sunflower oil. Within experimental uncertainty, we find that in both cases, the kinetic evolution is quite comparable. This means that for the stabilizing polymer association which

Fig. 8. Arrhenius plots deduced from Figs. 6 and 7.

204

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

was employed (Arlacel P135, synperonic PE / F68), the compositional ripening is occurring at comparable rates whatever the oil nature (alkane or vegetable oil).

3.2.6. Influence of the presence of biopolymers In Ref. [2], it was observed that the presence of natural proteins like Bovine Serum Albumin associated with classical short surfactants may considerably lower the rate of release in double emulsions. Following these pioneering observations, we decided to introduce in our double emulsions other biopolymers such as HEC, cmc and Alginate. Three emulsions were prepared according the protocol described in Refs. [7,8]. They all contain polyelectrolytes in both the internal and external water phases. The polyelectrolyte concentration in water is adjusted such as to fabricate globules with identical diameter. This was obtained by adjusting the composition in order to get the same viscosity of the water phase (300 MPa / s) during the emulsification process: • 2% w alginate HF 120L (Pronova). • 5% w of CMC Akucell AF 0705 (Akzo Nobel). • 4% w of HEC (Aldrich). The leakage properties of the three different emulsions stabilized by Arlacel P135 and Synperonic PE / F68 are compared in Fig. 10. In this figure, the reference refers to a double emulsion with the same

properties (diameters and concentrations) as the three other ones, but that does not contain polyelectrolyte. According to Eq. (7), the rate of release should depend on the globule diameter and this is why we fixed it at a value of 17 mm in order to obtain very low rates of leakage. It can be stressed that the entrapment yield right after the preparation step is rather elevated: indeed, we obtained at least 95% of the salt initially encapsulated, whatever the globule diameter was. This result can be regarded as a general feature of the fabrication method [7,8]. It can be deduced from Fig. 10, that the double emulsions containing alginate and HEC have similar leakage properties as the reference. In other words, the presence of such biopolymers does not affect much the rate of compositional ripening. However, a strong deviation with respect to the reference is observed for the system containing cmc. In this particular case, only 5% of the initial salt concentration have been released after 1 month. At this point, we can not give any interpretation to explain the particular role of cmc compared to the other biopolymers. Surface tension measurements at the oil–water interface in presence of the surface-active species did not reveal any significant difference between the three different systems (surface tension of the order of 0.1 mN / m). However, from a practical point of view, the result obtained in presence of cmc is quite encouraging in the perspective to perform long term encapsulation of active species. We should like to stress that the rate of release can be even lower if instead of NaCl which is quite a permeable species, we encapsulate larger ions or molecules with smaller permeation coefficients.

4. Conclusion

Fig. 10. Kinetics of release in presence of various natural polyelectrolytes: [NaCl] 0 50.4 mol / l, f 0i 540%, fg 560%, d i 51 mm, d g 517 mm, Cl 55% w; Ch 53% E; oil:Sunflower oil; T5 25 8C. The dashed lines are only guides to the eyes.

In the present paper, we have examined separately the two mechanisms that are responsible for the leakage of encapsulated substances in double emulsions. Our study was possible after finding experimental conditions where the two mechanisms occur over time scales that are sufficiently distinct to be decoupled. The methodology described here can be used to measure the microscopic parameters that control the rate of release in double emulsions. Finally, we hope this paper will provide some

K. Pays et al. / Journal of Controlled Release 79 (2002) 193 – 205

guidance to formulate double emulsions for longterm encapsulation. W/ O / W emulsions are potential vehicles for various hydrophilic drugs (vaccines, vitamins, enzymes and hormones) which would be then progressively released. Active substances may also migrate from the outer to the inner phase of multiple emulsion, providing in that case a kind of reservoir particularly suitable for detoxification (overdose treatment) or, in an again different domain, in the removal of toxic materials from wastewater. Anyway, the impact of double emulsions designed as drug delivery systems would be then of significant importance in the controlled release field, for oral, topical or parenteral administrations, provided that the stability and release mechanisms may be clearly understood and monitored.

Acknowledgements The authors gratefully acknowledge FOURNIER Company for financial support.

References [1] N. Garti, A. Aserin, Y. Cohen, Mechanistic considerations on the release of electrolytes from multiple emulsions stabilized by BSA and non-ionic surfactants, J. Controlled Release 29 (1994) 41–51. [2] N. Garti, A. Aserin, Double emulsions stabilized by macromolecular surfactants, Adv. Colloid Interface Sci. 65 (1996) 37–69. [3] M.P. Aronson, M.F. Petko, Highly concentrated water-in-oil emulsions: influence of electrolyte on their properties and stability, J. Colloid. Interf. Sci. 159 (1993) 134–149.

205

[4] C. Mabille, V. Schmitt, P. Gorria, F. Leal-Calderon, V. Faye, ` B. Deminiere, J. Bibette, Rheological and shearing conditions for the preparation of monodisperse emulsions, Langmuir 16 (2000) 422–429. [5] C. Mabille, Emulsion fragmentation in simple shear flow conditions, PhD thesis, University Bordeaux 1, October 2000. [6] K. Pays, F. Leal-Calderon, J. Bibette, Method for the preparation of monodisperse double emulsions, French Patent, CNRS, No. 00 05880, May 9, 2000. [7] J. Bibette, F. Leal-Calderon, P. Gorria, Polydisperse double emulsions, monodisperse double emulsions and preparation method, Patent WO 0121297, September 20, 1999. [8] C. Goubault, K. Pays, D. Olea, J. Bibette, V. Schmitt, F. Leal-Calderon, Shear rupturing of complex fluids: application to the preparation of quasi-monodisperse W/ O / W double emulsions, Langmuir 17 (2001) 5184–5188. [9] A. Ashkin, Acceleration and trapping of particles by radiation pressure, Phys. Rev. Lett. 24 (1970) 156–159. [10] M.I. Angelova, B. Pouligny, Trapping and levitation of a dielectric sphere with off-centred Gaussian beams, Pure Appl. Opt. 2 (1993) 261–276. [11] M.F. Ficheux, L. Bonakdar, F. Leal-Calderon, J. Bibette, Some stability criteria for double emulsions, Langmuir 14 (1998) 2702–2706. ¨ Macroemulsion stability: the [12] A. Kabalnov, H. Wennerstrom, oriented wedge theory revisited, Langmuir 12 (1996) 276– 292. [13] J.N. Israelachvili, Intermolecular and Surface Forces, Academic Press, London, 1992. [14] K. Pays, Double emulsions: coalescence and compositional ripening, PhD thesis, University Bordeaux 1, November 2000. [15] K. Pays, J. Kahn, B. Pouligny, J. Bibette, F. Leal-Calderon, Coalescence in surfactant-stabilized double emulsions, Langmuir 17 (2001) 7758–7769. [16] R.H. Fowler, A. Guggenheim, Statistical Thermodynamics, University Press, Cambridge, 1939. [17] R.T. Hamilton, E.W. Kaler, Alkali metal ion transport through thin bilayers, J. Phys. Chem. 94 (1990) 2560–2566.