- Email: [email protected]
I. Creaming Behavior
JOURNAL OF COLLOID AND INTERFACE SCIENCE ARTICLE NO.
207, 283–293 (1998)
CS985801
Characterization of a Depletion-Flocculated Polydisperse Emulsion I. Creaming Behavior Pretima Manoj, Annette J. Fillery-Travis,1 Andrew D. Watson, David J. Hibberd, and Margaret M. Robins Institute of Food Research, Norwich Research Park, Colney, Norwich, Norfolk, NR4 7UA, United Kingdom E-mail: [email protected] Received March 16, 1998; accepted August 11, 1998
We report an experimental investigation on the creaming behavior of flocculated, polydisperse, oil-in-water emulsions. Flocculation is by addition of a depletion flocculant, the polymer hydroxyethylcellulose (HEC), at a range of concentrations. The creaming behavior is dependent on the oil volume fraction and polymer concentration. At low concentrations of HEC, the droplets cream either individually or in two populations, a flocculated phase, and a coexistent phase of individual droplets. At higher HEC concentrations, the droplets appear to cream as a single entity, with a sharp lower boundary, separating the region with droplets from a clear serum at the base of the container. In these emulsions, and in some of the coexistent ones, there is a significant delay before creaming starts. Once started they cream at a constant rate. We have identified the continuous phase viscosity as a major factor. The aim of this work is to elucidate the mechanisms underlying the delay before creaming. We propose that as soon as they flocculate, the emulsions form space-filling structures, which slowly rearrange until channels are formed that allow the flow of bulk continuous phase to the base of the container. Scaling arguments are presented that suggest the delay could be related to the single-droplet diffusion rate. © 1998 Academic Press Key Words: depletion; flocculated network; creaming; delay period; induction phase; transient gelation.
INTRODUCTION
The stability of emulsions with respect to creaming is of great industrial and theoretical interest. This work concerns the rate of creaming and the time delay that is sometimes observed in systems prior to creaming. Here we call this the delay period, but other expressions used in the literature for this or related effects are induction phase and transient gelation. The creaming delay period has already been reported by this laboratory (1). Parker et al. reported that at certain concentrations of oil and added xanthan gum, oil-in-water emulsions exhibited a substantial delay period, after which rapid cream1
To whom correspondence should be addressed.
ing occurred with the boundary between flocculated emulsion and clear serum moving at a constant rate. The duration of the delay period was found to increase approximately with the cube of the xanthan concentration, but was relatively insensitive to the oil volume fraction. Delay periods were also observed in hydroxyethylcellulose (HEC) emulsions although at higher polymer concentrations than in xanthan systems. Pusey et al. (2) have studied the phase behavior of hardsphere colloids in the presence of nonadsorbing polymers. They used sterically stabilized monodisperse poly(methyl methacrylate) (PMMA) spheres dispersed in hydrocarbon solvents, selected to give refractive index matching, with added polystyrene polymer providing a weak depletion potential. In their monodisperse systems Pusey et al. observed the expected order–disorder transition at zero polymer, with a fluid– crystal coexistence region that was expanded by the introduction of polymer. Above a certain level of polymer, crystallization was suppressed, and nonequilibrium behavior appeared. For samples having sufficient polymer, a period of “latency” was displayed prior to settling. This latency period we identify with our delay period. The duration of the latency period was found to increase with polymer concentration and colloid volume fraction. Small angle light scattering measurements on samples having a composition just inside the nonequilibrium region of the phase diagram showed features indicative of spinodal decomposition or nucleation. Samples containing sufficient polymer to display transient gelation contained aggregates whose structure was consistent with a diffusion-limited aggregation process. Pusey et al. framed their discussion of transient gelation in terms of competition between diffusion-limited aggregation and thermal rearrangement. Verhaegh (3–5), Lekkerkerker (6), and co-workers observed transient gelation in their optical microscopy and light scattering studies of monodisperse silica spheres in cyclohexane in the presence of polydimethylsiloxane (PDMS) polymer. They proposed that transient gelation was driven by spinodal decomposition, which was responsible for forming a bicontinuous network of colloid-rich and colloid-poor domains. Aggregation in the colloid-rich phase was held responsible for the formation of a gel.
283
0021-9797/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.
284
MANOJ ET AL.
The reversible nature of the depletion-induced aggregation permitted network restructuring and eventual gel collapse. However, unlike Pusey and co-workers, who considered a system having a deeper interaction potential, these other authors observed no evidence of fractal clusters. The latter proposed that spinodal decomposition proceeds until a percolation threshold is reached, at which point the system assumes some rigidity. Restructuring leads to loss of connectivity in the network, allowing the phase separation to proceed, and the gel collapses. We note several other experiments reported in the literature which show delayed sedimentation. Smits et al. (5) studied silica in cyclohexane in the presence of polystyrene or PDMS, finding that certain concentrations of polymer-induced phase separation into colloid-rich and colloid-poor phases, followed after some days by crystallization and sedimentation in the colloid-rich phase. Partridge (6, 7) observed an induction time followed by sedimentation in sterically stabilized polystyrene lattices weakly flocculated by the addition of NaCl. There appeared to be a delay in creaming data presented by Dickinson for a triglyceride oil-in-water emulsion stabilized with guar gum (8), but this may have been a result of hazy boundaries that were not initially visible. Experiments by the same group on casein-stabilized emulsions also showed a characteristic induction time before separation (9). Here the casein, despite being adsorbed at the interface, was present in the continuous phase in sufficient quantity to act as a depletion flocculant. Mellor and Stavans observed what appeared to be delayed creaming in monodisperse emulsions of silicon oil emulsified by SDS in the presence of nonadsorbing PEO (10). Mellor and Stavans performed light scattering measurements at high polymer concentrations and found a broad peak commensurate with a correlation length of about 20 mm, or 20 drop diameters. Their corresponding visible flocs were about 25 mm in size. They concluded that the obstacle to creaming was the formation of a gel-like structure of interconnected flocs. Our current study continues our previous investigations of the phase behavior of a polydisperse alkane-in-water emulsion in the presence of the nonadsorbing polymer HEC (11). The main aim is to understand the mechanisms behind the delay period and the sudden collapse that ends in separation. Although it is possible that spinodal decomposition may be involved (4), the presence of polydispersity further complicates the phase behavior. Using visual observations in measuring cylinders, we have measured the duration of the delay period and the eventual creaming rate of oil-in-water emulsions. Data from the IFR creaming monitor, an ultrasonic technique, support our visual observations. We report here the effects of polymer concentration and oil volume fraction on the delay period, creaming rate, and final oil concentration in the cream layers in order to obtain insight into the mechanisms underlying the creaming behavior. The rheological behavior of these systems is discussed in a companion paper (12).
EXPERIMENTAL
Materials and Emulsion Preparation The emulsions were prepared from a premix. An oil phase was added to a solution of the nonionic surfactant Brij 35 (polyoxyethylene 23-lauryl ether, Sigma Chemical Company) and the emulsion prepared in a Waring blender using a predetermined shearing cycle. The resulting premix was stable to coalescence and disproportionation over the timescale of the experiments. Two oil phases were used within the experimental protocol: Premix 1 for visual characterization of creaming. Premix 1 contained n-hexadecane (Cetane–C16H34; density 773.4 kg m23 at 20°C; Sigma Chemical Company) in a surfactant solution. Premix 2 for ultrasonic monitoring of creaming. Premix 2 contained a mixture (density 693.2 kg m23 at 20°C) of nheptane (C7H16; density 683.8 kgm23 at 20°C; SLR, Fisons) and hexadecane in the volume ratio 9:1 in surfactant solution. A series of final emulsions were prepared by diluting the premixes with a polymer solution to obtain the desired concentrations of oil and polymer as in Table 1. The diluent was a mixture of an aqueous solution of the high molecular weight polymer HEC (Natrosol 250HR, Aqualon, mean r g (radius of gyration) 5 58 nm) and the preservative sodium azide (NaN3, Sigma Chemical Company). Prior to use the polymer was purified by dialysis and freeze-dried. The polymer was dispersed and hydrated by stirring together the ingredients while heating to 80°C, then allowing the solution to cool to room temperature. After preparation the emulsions were immediately transferred either to measuring cylinders for visual monitoring or to measurement cells for ultrasonic monitoring. Droplet Size Distribution The droplet size distribution of both premixes was measured using a Malvern Mastersizer laser diffraction sizer. These distributions were highly reproducible and did not change during the timescale of the creaming measurements. A typical size distribution is shown in Fig. 1 with a volume mean diameter (d 43 ) of 1.80 mm. Since the polymer was added after emulsification, it had no effect on the droplet size distributions. Density Densities were measured at 20.0°C using an Anton Paar DMA60 vibrating tube density meter with a DMA602 measuring cell. Rheological Measurements Viscosity (steady shear 2 shear rate) measurements were obtained using a controlled stress (CS) Bohlin Rheometer, using a double gap geometry at 25°C. Solvent traps were used to minimize the effects of drying out and dust settlement. The viscometry measurements were carried out on HEC solutions. Zero-shear viscosity values were obtained for each
285
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
TABLE 1 Emulsion Samples
Premix
Oil type
Oil volume fraction (%; (v/v))
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2
Hexc Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex HHd HH HH
10 10 10 10 25 25 25 25 34 34 34 34 40 40 40 20 20 20
Brij 35 (%; (w/w))a
Total HEC C av (%; (w/w))a,b
Estimated HEC in continuous phase C p (see text)
0.30 0.30 0.30 0.30 0.90 0.90 0.90 0.90 1.39 1.39 1.39 1.39 1.80 1.80 1.80 0.35 0.35 0.35
0.10 0.20 0.35 0.50 0.04 0.10 0.20 0.35 0.04 0.10 0.20 0.35 0.04 0.10 0.20 0 0.025 0.10
0.116 0.231 0.405 0.578 0.060 0.151 0.302 0.529 0.074 0.185 0.371 0.648 0.087 0.218 0.436 0 0.034 0.137
a
Expressed with respect to continuous phase. 0.3% (w/w) sodium azide was added, as a preservative, to all emulsions. c Hexadecane only. d Oil mixture of heptane and hexadecane. b
sample. The specific viscosity (hsp), which defines the fractional increase in viscosity due to the presence of the polymer, was calculated for each concentration
h sp 5
h 2 hs , hs
[1]
polymer concentration (Fig. 2) showed the critical polymer coil overlap concentration (C*) to be 0.275% (w/w) for HEC polymer. The behavior of the specific viscosity with concentration was as expected. Creaming Behavior of the Emulsions
where hs is the solvent viscosity. A log-log chart of hsp and
We have observed two distinct modes of creaming in emulsions containing nonabsorbing polysaccharides. In Type I
FIG. 1. Droplet size distribution of hexadecane premix from light diffraction particle sizer (Malvern Mastersizer).
FIG. 2. Specific viscosity and polymer concentration for estimation of critical polymer concentration.
286
MANOJ ET AL.
creaming, emulsions remain opaque at the base of the sample for a significant period of time, while a concentrated cream layer develops at the top of the sample. The thickness of the cream layer gradually increases with time, and only after most of the oil has reached the cream is there any visible sign of clearing from the base of the sample, which may remain turbid. The boundary between the serum and the creaming droplets is very hazy and indistinct. This behavior is characteristic of polydisperse emulsions in the presence of little or no polymer, where individual droplets or small aggregates move independently to the top of the container. Creaming of this type is very difficult to monitor visually but may be characterized using the IFR ultrasonic monitor (13). In Type II creaming, there is a clear boundary between the creaming emulsion and the serum layer at the base of the sample. Creaming occurs rapidly and the remaining serum can be clear or turbid. Previous investigations of this behavior (14) have suggested the presence of a high degree of flocculation within the creaming emulsion sufficient to form a space spanning structure. This type of creaming is clearly visible. Definition and Calculation of Delay Time Profiles of the creaming behavior of the sample emulsions were obtained by measuring the height of the boundary between the colloid-rich fraction and the colloid-poor fraction (type II creaming) as a function of time for each emulsion. Once started, the rate of creaming was constant (14) but a delay was sometimes evident before creaming began. The delay could be quantified by extrapolation of the boundary height to zero movement. Visual Assessment of Creaming Premix 1 dispersions and polymer diluents were poured into 100-ml measuring cylinders immediately after preparation. The cylinders were inverted 10 times to ensure thorough mixing. The sample emulsions were kept at 25°C and the movement of any creaming boundaries was followed with time. The turbidity of the serum layers below the developing creams was noted. The experiments were repeated between three and five times to obtain mean creaming rates and delay times. The standard deviation within each set of results was calculated to be less than 12%.
FIG. 3. (a) Ultrasonic creaming profiles during creaming of 20% emulsion in the absence of polymer. Each line shows the measured variation of oil volume fraction with height at the each sampling time. (b) The ultrasonic creaming data from (a) expressed as the height of contours of constant oil volume fraction with time. The data show the oil droplet started to move from time 5 0.
Ultrasonic Monitor Measurements
Emulsions Prepared without Polymer (Type I Creaming)
The IFR ultrasonic monitor (13) was used to characterize the creaming behavior of Premix 2 emulsions by measuring the speed of ultrasound in emulsions at a series of heights. The oil volume fraction (w) as a function of height was then calculated using a simple mixing theory (15). At each sampling time the validity of the calibration was verified by calculating the integrated oil volume fraction over sample height. These oil vol-
Visual monitoring. Accurate visual observations of Type I were not possible due to the absence of a well-defined boundary. Ultrasonic monitoring. The ultrasonic creaming profile for 20% (v/v) heptane/hexadecane emulsion in the absence of polymer is shown in Fig. 3a. At 0.28 day (;7 h) the oil had begun to move to the top of the cell and formed a thin cream
ume fraction profiles allowed early detection of creaming and gave detailed information on the creaming process. RESULTS
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
FIG. 4. Ultrasonic creaming profiles during creaming of 20% emulsion in the presence of 0.025% HEC, showing the coexistence of a rapidly rising flocculated phase and a fraction of slower-moving droplets.
layer of approximately 80% (v/v) oil. No sharp boundary was observed in the creaming emulsions. After 4 days, the oil had cleared from the base of the cell (no oil detected at ;13 mm height from the base). By interpolation, the height of contours of constant volume fraction, as a function of time, can be calculated, as shown in Fig. 3b. The linearity of these contours indicated that each oil fraction creamed at a constant velocity (constant gradient) until the droplets reached the closely packed cream layer. Comparison of the creaming velocities with the hindered Stokes velocity for the appropriate size fraction yielded an effective droplet size distribution that was similar to the Mastersizer result (11). The contour analysis also showed that there was no delay in the movement of droplets within the initial phase of creaming.
287
subcream serum containing no measurable oil. Since the serum was also visually clear, we interpreted the profiles as indicative of complete flocculation. Visual monitoring. The addition of HEC in the concentration range 0.04 – 0.35% (w/w) also gave a significant change in the visual creaming behavior of the emulsions, with the appearance of a rapidly creaming boundary at the base. At the onset of creaming, the boundary was characterized as either sharp or diffuse. The boundary regions in emulsions of 10% (v/v) oil volume fraction and low HEC concentration were initially diffuse but sharpened as creaming progressed. In emulsions with high oil volume fraction and high HEC concentration the boundary lines were sharp from the start and creamed rapidly, leaving a clear supernatant or subcream. An example of the creaming behavior is shown in Fig. 6 for 34% (v/v) hexadecane– 0.1% (w/w) HEC ([34:0.10]) and 34% (v/v) hexadecane– 0.35% (w/w) HEC ([34:0.35]) emulsions. A delay was observed before creaming for both emulsions, the delay time being shorter for 0.1% (w/w) HEC than for 0.35% (w/w) HEC. Once creaming was established, the rate of creaming was approximately constant. Figure 7 summarizes the variation of creaming rate with polymer concentration and oil volume fraction. For comparison Fig. 7 also shows the zeroshear viscosity of the continuous phases. After the onset of creaming, the appearance of the serums varied in their turbidity, consistent with the presence or absence of residual oil droplets. The emulsion with 0.35% (w/w) HEC left behind a clear layer from the start, whereas the 0.1% (w/w) HEC initially left behind a turbid serum layer which cleared slowly over a period of days. The initial turbidity (after the majority of the oil had creamed rapidly to the top) gave an indication of whether the system was fully flocculated or contained floc(s) in coexistence with unflocculated droplets.
Emulsions with Polymer (Types I and II Creaming) Ultrasonic monitoring. Figure 4 shows the creaming of a 20% (v/v) emulsion (Premix 2) in the presence of 0.025% (w/w) HEC, monitored using the ultrasonic technique. The profiles revealed two distinct fractions, a rapidly moving population which creamed within a day, leaving almost half the oil behind in a slower-moving fraction. Contour analysis indicated the emulsion contained two phases, a flocculated phase in coexistence with unflocculated droplets. By varying the HEC concentration in the range 0.015– 0.04% (w/w), the amount of flocculation varied in the range 1–98% (11). The unflocculated droplets creamed slowly, due to the increased viscosity of the polymer solution, and the serum was visually cloudy until all the oil had reached the cream layer. Figure 5 shows the ultrasonic profiles for 20% (v/v) oil (Premix 2) with 0.1% (w/w) HEC. The creaming profiles showed a sharp boundary rising rapidly with time, leaving a
FIG. 5. Ultrasonic creaming profiles during creaming of 20% emulsion in the presence of 0.10% HEC, showing formation of a sharp boundary as the flocculated droplets rise to the cream layer.
288
MANOJ ET AL.
of the flocculated structure, reducing the rearrangement on a macroscopic scale. The visual data shown in Fig. 9 were collected before the long-term rearrangements shown in Figs. 4 and 5 had occurred. DISCUSSION
The aim of this study was to investigate the origin of the delay time observed before the onset of creaming in emulsions containing nonadsorbing polymers. Before considering this question in detail we will outline the phase diagram of the system and hence the context of the observed behavior. Phase Behavior of the Emulsions
FIG. 6. Height of visual boundaries and time at 25°C for 34% (v/v) hexadecane emulsions: (■) 0.1% (w/w) HEC; (‚) 0.35% (w/w) HEC.
Plate 1 shows emulsions of 34% (v/v) hexadecane in the presence of 0.04, 0.10, and 0.35% (w/w) HEC, respectively, 16 days after preparation. At 0.04% (w/w) HEC there was a boundary at approximately 47 ml, although the emulsion below the boundary was opaque. At the higher HEC concentrations the serum layers became less opaque, and at 0.35% (w/w) HEC the serum was clear from the start of creaming. The turbid serum layers contained droplets which creamed very slowly due to the high viscosity of the HEC continuous phase. Close examination of the ultrasonic creaming data showed that the remaining fraction of oil was composed of individual droplets. However, at very low residual oil concentrations, visual observations of turbidity are more sensitive than the ultrasonic monitor to the presence of these suspended individual droplets. Summary of delay times. Figure 8 shows the variation of delay time with polymer concentration and oil volume fraction. The delay increased with increasing concentration of either component. It should be noted that for emulsions that initially creamed with diffuse boundary lines, there were significant uncertainties as to the instant that creaming started so the calculated delay times were taken as “upper limit delay times.” Figure 8 also shows the standard deviation of the observed delay times. Depths of cream layers. At the end of creaming the droplets formed a layer at the top of the container. Visual assessments showed that the thickness of the cream layers increased with increasing polymer concentration and oil volume fraction. To allow for the different initial oil volume fractions, the data were used to calculate an average volume fraction, wm, of the oil in the cream layers. Figure 9 shows that the packing density (wm) increased with oil volume fraction and decreased with polymer concentration. This implies that the polymer increases the compressive strength
In our previous study (11) we observed that increasing the HEC concentration caused the formation of a concentrated colloidal phase, in coexistence with a dispersed colloid-poor phase. These phases separated spatially under gravity, the colloid-rich fraction creaming rapidly in comparison with the colloid-poor fraction. Analysis of the ultrasonic creaming profiles of these emulsions showed the colloid-rich phase was flocculated and the colloid-poor phase was composed of individual droplets. We suggested the formation of these phases could be due to one or both of the following effects: ● size segregation of larger droplets due to preferential flocculation; ● phase separation which is independent of droplet size.
No significant difference was found between the size distributions of the droplets contained within each phase and we therefore concluded that phase separation was responsible, at least in part, for the observed creaming behavior. In this study we have investigated the flocculation behavior
FIG. 7. Creaming rates of hexadecane–HEC emulsions at 25°C as a function of polymer concentration as compared with the variation in the viscosity of the solution containing the same total concentration of HEC (---). (F) 10% (v/v) hexadecane; (■) 25% (v/v) hexadecane; (}) 34% (v/v) hexadecane; (Œ) 40% (v/v) hexadecane.
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
PLATE 1.
289
Appearance of cylinders containing 34% (v/v) hexadecane with HEC (from left to right) 0.04, 0.10, and 0.35% (w/w), after 16 days at 25°C.
of similar emulsions, using ultrasonic monitoring and visual observations. The results are summarized in Fig. 10, in the form of a phase diagram showing three regimes: nonflocculated, phase coexistent, and fully flocculated emulsions. The boundary between the nonflocculated and the phase coexistent region we have denoted as the primary critical concentration (PCC), and the boundary between the phase coexistent and fully flocculated region as the secondary critical concentration (SCC). No attempt was made in this study to delineate quantitatively the boundaries between the regions, and the upper and lower limits for SCC are shown. The data are shown as closed points where the serum subcream layers were initially turbid and open points where the serum was clear from the start. This phase diagram contains features consistent with the phase diagram of Ilett et al. (16) for r g/a , 0.3, i.e., two single phase regions separated by a two-phase region. In our study the polydispersed nature of the oil droplets would suggest a fluid– fluid coexistence region as opposed to the fluid–solid coexistence predicted.
Mechanism of Delay Figure 10 also shows the systems exhibiting a delay time significantly greater than zero. Delays were observed within both the phase coexistent and the fully flocculated regions. Within the phase coexistent region a prerequisite of delay was the presence of a sharp boundary upon creaming. We propose two hypotheses for the existence of the observed delay in creaming: ● The viscous nature of the continuous phase reduces the rate of flocculation and the delay is simply the timescale to form flocs which cream rapidly. ● There is the rapid formation of a space-spanning flocculated structure of sufficient “compressive strength” to withstand buoyancy forces. We speculate that the eventual collapse of such a network could be due to coarsening of the structure by droplet rearrangement.
We will consider each of these hypotheses in turn.
290
MANOJ ET AL.
FIG. 8. Delay times of hexadecane–HEC emulsions at 25°C as a function of HEC concentration. Symbols are as in Fig. 7. The error bars represent the standard deviation of the measurements on replicate samples.
Is the delay due to slow flocculation? The variation of delay time with HEC concentration for each oil volume fraction is shown in Fig. 8. For a particular HEC concentration the viscosity of the continuous phase is constant and the droplet encounter rate, and hence the rate of flocculation, would be expected to increase with volume fraction. Thus a decrease in delay time would be expected upon increasing the oil volume fraction if the observed delay time is the consequence of relatively slow flocculation. For all the HEC concentrations used in this study the delay time was found to increase with increasing oil volume fraction. This result contradicts the hypothesis of slow flocculation, particularly when the delay times observed within this study far exceed the characteristic timescale of droplet diffusion and depletion flocculation rates. The sharp boundary would require
FIG. 10. Phase diagram of hexadecane–HEC emulsions based on initial turbidity of serums left behind after creaming at 25°C (closed symbols, cloudy serum; open symbols, clear serum) in association with the existence of delay period (E). The HEC concentration is the total concentration in the continuous phase.
that the flocs all creamed at the same rate, and it seems unlikely that they would have a monodisperse effective size distribution. Rearrangement and collapse of a flocculated space-spanning structure. In our previous studies, and in common with other workers, we have identified the formation of a sharp creaming boundary to be symptomatic of a flocculated spacespanning structure creaming as a single entity (14). The form of the ultrasonic creaming profiles supports this interpretation. Upon creaming the oil droplets ascended the container at a single rate producing a sharp discontinuity in volume fraction. There was no broadening of the boundary that might suggest movement of separate aggregates with varying speed. Pursuing this approach we examined the creaming rates to determine their scaling behavior. In particular we aimed to separate the individual droplet interactions (controlled by the deletion potential and thus by the HEC concentration) from the effects of the initial volume fraction. Creaming Rate: Network Model
FIG. 9. Average oil volume fraction in the cream layers and total HEC concentration. Symbols are as in Fig. 7.
Emulsion creaming requires backflow of the continuous phase into the volume vacated by the rising oil droplets. Ilett et al. (16) considered that since nonadsorbing polymer is excluded from the immediate vicinity of the droplets, to a distance comparable with the polymer’s radius of gyration in solution, r g, then the effective volume of solvent available for the polymer is reduced, and the effective polymer concentration C p (outside the depleted zone) is higher than the overall concentration C av. Ilett et al. estimated the increase in polymer concentration a 5 C p/C av due to these excluded zones around droplets of radius a to be
291
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
a 5 ~1 2 w !exp@2A g 2 B g 2 2 C g 3#,
[2]
where
g5
w , A 5 3j 1 3j2 1 j3 , ~1 2 w ! B5
9 rg j 1 3 j 3, C 5 3 j 3, j 5 . 2 a
Applying the correction to the measured size distribution (Fig. 1) for each emulsion gave a significant increase in the effective concentration of HEC, as shown in Table 1. In the following analyses, we have used this estimated continuous phase concentration for all the HEC emulsions. We propose that the structures filled the available volume when they formed initially, but after a characteristic time (delay period) the structure was no longer space filling and separated physically from the surrounding continuous phase. The emulsions with w $ 25% (v/v) showed a sharp creaming boundary as soon as they began to cream, consistent with a space-filling structure. One model for this separation process is that the flocs form a porous network and the continuous phase flows through channels (pores) in the network. Michaels and Bolger (17) modeled the settling rate of kaolin flocs using the Kozeny–Carman equation applied to pore flow. They assumed the pores to be equivalent to smooth, straight channels of diameter d p, and obtained the maximum settling velocity V from g z D r z w z d 2p~1 2 w / w m! , V5 32 h
[3]
where Dr is the density difference between the dispersed particles and continuous phase, h is the viscosity of the continuous phase, g is the acceleration due to gravity, w is the volume fraction of particles, and wm is the maximum volume fraction within the flocs. Applying this expression to the measured creaming rates of the hexadecane emulsions, we obtained an estimate of the equivalent pore diameter d p in each emulsion, as shown in Fig. 11. The maximum oil concentration wm in the flocs was taken from the final concentration in the cream layers (Fig. 9). In the model, the initial w affected the overall voidage of the emulsions and therefore the total volume of channels, but it assumes that the size of the channels is determined by the flow properties of the continuous phase and geometry of the channels (roughness, tortuosity). Our data show that the effective pore size was indeed independent of initial w, despite a factor of up to 5 in creaming rate. The increase of effective pore diameter with HEC concentration, also observed by Michaels and Bolger, is attributed to the increased strength of the flocs under a higher depletion potential, which resisted the formation of
FIG. 11. Mean diameter of creaming channels in the flocculated emulsions, inferred from Michaels and Bolger (17). Symbols are as in Fig. 7. Closed symbols, cloudy serum; open symbols, clear serum.
smooth straight channels. The effective channel diameter, between 20 and 120 mm, was an appropriate order of magnitude for structures that cannot be resolved by the naked eye and consistent with the 25-mm characteristic length scale observed by Mellor and Stavans (10). Although at high HEC concentrations there was more scatter on the relationship, regression analysis on the data in Fig. 11 showed a correlation that was significant at better than the 99.9% level. Delay Period Figure 8 showed that the delay time was a strong function of both polymer concentration and droplet volume fraction, w. The polymer dependence is straightforward to rationalize, as the depth of the attractive interparticle potential is directly related to the concentration of free polymer (18). However, it is less easy to explain the strong w dependence. Full statistical mechanical treatment of the collective behavior of the droplets experiencing attractive interactions is outside the scope of this paper. However, it is interesting to examine the w dependence in terms of a mean-field approach, which is explored below. The porous model requires the formation of channels in the flocculated structure. It is reasonable to suppose that the delay time represents the time for the droplets to rearrange to a channeled structure. In this case, the delay time could be related to the diffusion of the droplets in and around the flocs. The diffusion coefficient of individual droplets, radius a, may be estimated using the Stokes–Einstein expression D5
k BT , 6 ph a
[4]
where h is the viscosity of the background medium. Considering the (flocculated) structure to hinder the diffusion of the
292
MANOJ ET AL.
than to be randomly dispersed (20). The rate of decomposition is also predicted to be dependent on the single-droplet diffusion rate. In dispersions flocculated by the addition of polymer, varying the interparticle potential by varying polymer concentration also changes the viscosity of the continuous phase. Thus it is not straightforward to separate the two mechanisms, flocculation and the process of spinodal decomposition. In principle, each process leads to spatial density variations that have a characteristic length scale (20). To determine which mechanism dominates in the droplet rearrangements prior to creaming, experiments are needed where the droplet interactions can be varied without changing the single-droplet diffusion rate. FIG. 12. Measured delay time and the estimated effective viscosity of the emulsions. Symbols are as in Fig. 7. Closed symbols, cloudy serum; open symbols, clear serum.
individual droplets, we can replace h by an effective viscosity, he, using the Ball/Richmond mean field model (19)
S
he 5 h0 1 2
w wm
D
25wm/ 2
.
[5]
If droplet diffusion is the controlling factor in the delay time, then we expect a relationship between the delay time and effective viscosity of the medium. Estimating the effective viscosity of the emulsions using Eq. [5] and the measured value of wm, we obtained the trends shown in Fig. 12. The analysis eliminated the effect of initial oil volume fraction, and there were two distinct trends. In the coexistent regime (closed data points) there was an approximately linear relationship (gradient (power) 1.06, significant at 99% level), and at high HEC concentrations when we observed complete flocculation (clear serum, open data points) the delay scaled with viscosity to a power 0.36 (significant at 99% level). The interesting point about this correlation is that an expression that is analogous to a single droplet diffusion rate allowed for the f dependence in the rearrangement time. So far we have discussed the rearrangement and subsequent creaming of the emulsion/polymer system in terms of flocculated structures that coarsened to produce channels through which continuous phase could flow. In flocculation, regions of the systems become colloid-rich (flocs) which grow from a nucleus analogous to a crystallization process. Flocculation is an instability that affects the rate at which equilibrium is achieved. However, the observed separation could also be due to slow spinodal decomposition, a demixing process. In this process, because the system has a lower energy when the droplets are in regions of high local droplet density, random fluctuations become asymmetric, so that over time a droplet is more likely to remain in a region of high droplet concentration
Additional Factors and Approximations Clearly the above analyses are tentative, as there are a number of factors that have not been included rigorously. The initial correction made to the effective continuous phase concentration (Eq. [2]) is accepted for low HEC concentrations, but must necessarily be flawed above C*, where the polymer chains are no longer able to adopt a random-coil configuration in solution. The droplet polydispersity is considered in the above analysis only because of its effect on the excluded volume, but it is likely to be a factor in the phase behavior of the droplet/polymer mixture, as well as the diffusion behavior of the droplets. Similarly, the polymer polydispersity is ignored, but it may well cause perturbations to the phase diagram. Modeling the creaming behavior as a porous network has been shown to rationalize the data for different oil volume fractions, but the model includes major assumptions on the channel geometry and floc structures. Equally, it is debatable whether the treatment of droplet concentration by means of an effective, mean-field viscosity is valid in the presence of attractive potentials. However, we offer the ideas above as a starting point for more detailed studies, which may provide an insight into the mechanisms underlying the gravitational collapse of flocculated structures for improved product design in the process industries. CONCLUSIONS
Experiments on a series of depletion-flocculated emulsions showed that the degree of flocculation, type of creaming, creaming rate, and delay before creaming were dependent on the oil volume fraction and polymer (HEC) concentration. With increasing HEC, the polydisperse emulsions displayed, in turn, creaming of individual droplets; creaming in two fractions corresponding to a flocculated phase in coexistence with a phase of individual droplets; and a fully flocculated state where the oil rose at a single, constant, rate. Delays before creaming were observed in the fully flocculated emulsions and in the partially flocculated emulsion at high oil volume fraction.
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
The phase behavior of the emulsions as a function of oil volume fraction and polymer concentration was similar in form to that observed by Ilett et al. (15) for monodisperse solid particles in the presence of nonadsorbing polymer. In the highly flocculated emulsions the dependence of creaming rate on oil volume fraction was consistent with the porous network model proposed by Michaels and Bolger (16), which inferred the existence of channels in the structure of size in the range 20 to 120 mm. We propose that the delay period represents the time required for the channels to form and show that the delay time for each emulsion scales with an effective continuous phase viscosity. Whether the mechanism of channel formation is controlled by the rate of spinodal decomposition or by diffusion-controlled flocculation is discussed. ACKNOWLEDGMENTS The authors acknowledge funding from BBSRC in the form of the Institute’s competitive strategic grant and a ROPA award. We are grateful to Wilson Poon for helpful discussions.
REFERENCES 1. Parker, A., Gunning, P. A., Ng, K., and Robins, M. M., Food Hydrocolloids 9, 333 (1995). 2. Pusey, P. N., Pirie, A. D., and Poon, W. C. K., Physica A 201, 322 (1993).
293
3. Verhaegh, N. A. M., Vanduijneveldt, J. S., Dhont, J. K. G., and Lekkerkerker, H. N. W., Physica A 230, 409 (1996). 4. Lekkerkerker, H. N. W., Poon, W. C. K., Pusey, P. N., Stroobants, A., and Warren, P. B., Europhys. Lett. 20, 559 (1992). 5. Smits, C., Vandermost, B., Dhont, J. K. G., and Lekkerkerker, H. N. W., Adv. Colloid Interface Sci. 42, 33 (1992). 6. Partridge, S. J., Ph.D. thesis, University of Bristol, UK, 1986. 7. Buscall, R., Colloids Surf. 43, 33 (1990). 8. Dickinson, E., in “Macromolecular Interactions in Food Technology.” Am. Chem. Soc., Washington, DC, 1996. 9. Dickinson, E., Golding, M., and Povey, M. J. W., J. Colloid Interface Sci. 185, 515 (1997). 10. Mellor, A., and Stavans, J., Langmuir 12, 301 (1996). 11. Fillery Travis, A. J., Gunning, P. A., Hibberd, D. J., and Robins, M. M., J. Colloid Interface Sci. 159, 189 (1993). 12. Manoj, P., Watson, A. D., Hibberd, D. J., Fillery-Travis, A. J., and Robins, M. M., J. Colloid Interface Sci., in press. 13. Howe, A. M., Mackie, A. R., and Robins, M. M., J. Dispersion Sci. Technol. 7, 231 (1986). 14. Robins, M. M., ACS Sympos. Ser. 448, 230 (1991). 15. Urick, R. J., J. Appl. Phys. 18, 983 (1947). 16. Ilett, S. M., Orrock, A., Poon, W. C. K., and Pusey, P. N., Phys. Rev. E 51, 1344 (1995). 17. Michaels, A. S., and Bolger, J. C., I&EC Fund. 1, 24 (1962). 18. Sperry, P. R., J. Colloid Interface Sci. 99, 97 (1984). 19. Ball, R. C., and Richmond, P., Phys. Chem. Liq. 9, 99 (1980). 20. Dhont, J. K. G., “An Introduction to Dynamics of Colloids.” Elsevier, Amsterdam 1996.
207, 283–293 (1998)
CS985801
Characterization of a Depletion-Flocculated Polydisperse Emulsion I. Creaming Behavior Pretima Manoj, Annette J. Fillery-Travis,1 Andrew D. Watson, David J. Hibberd, and Margaret M. Robins Institute of Food Research, Norwich Research Park, Colney, Norwich, Norfolk, NR4 7UA, United Kingdom E-mail: [email protected] Received March 16, 1998; accepted August 11, 1998
We report an experimental investigation on the creaming behavior of flocculated, polydisperse, oil-in-water emulsions. Flocculation is by addition of a depletion flocculant, the polymer hydroxyethylcellulose (HEC), at a range of concentrations. The creaming behavior is dependent on the oil volume fraction and polymer concentration. At low concentrations of HEC, the droplets cream either individually or in two populations, a flocculated phase, and a coexistent phase of individual droplets. At higher HEC concentrations, the droplets appear to cream as a single entity, with a sharp lower boundary, separating the region with droplets from a clear serum at the base of the container. In these emulsions, and in some of the coexistent ones, there is a significant delay before creaming starts. Once started they cream at a constant rate. We have identified the continuous phase viscosity as a major factor. The aim of this work is to elucidate the mechanisms underlying the delay before creaming. We propose that as soon as they flocculate, the emulsions form space-filling structures, which slowly rearrange until channels are formed that allow the flow of bulk continuous phase to the base of the container. Scaling arguments are presented that suggest the delay could be related to the single-droplet diffusion rate. © 1998 Academic Press Key Words: depletion; flocculated network; creaming; delay period; induction phase; transient gelation.
INTRODUCTION
The stability of emulsions with respect to creaming is of great industrial and theoretical interest. This work concerns the rate of creaming and the time delay that is sometimes observed in systems prior to creaming. Here we call this the delay period, but other expressions used in the literature for this or related effects are induction phase and transient gelation. The creaming delay period has already been reported by this laboratory (1). Parker et al. reported that at certain concentrations of oil and added xanthan gum, oil-in-water emulsions exhibited a substantial delay period, after which rapid cream1
To whom correspondence should be addressed.
ing occurred with the boundary between flocculated emulsion and clear serum moving at a constant rate. The duration of the delay period was found to increase approximately with the cube of the xanthan concentration, but was relatively insensitive to the oil volume fraction. Delay periods were also observed in hydroxyethylcellulose (HEC) emulsions although at higher polymer concentrations than in xanthan systems. Pusey et al. (2) have studied the phase behavior of hardsphere colloids in the presence of nonadsorbing polymers. They used sterically stabilized monodisperse poly(methyl methacrylate) (PMMA) spheres dispersed in hydrocarbon solvents, selected to give refractive index matching, with added polystyrene polymer providing a weak depletion potential. In their monodisperse systems Pusey et al. observed the expected order–disorder transition at zero polymer, with a fluid– crystal coexistence region that was expanded by the introduction of polymer. Above a certain level of polymer, crystallization was suppressed, and nonequilibrium behavior appeared. For samples having sufficient polymer, a period of “latency” was displayed prior to settling. This latency period we identify with our delay period. The duration of the latency period was found to increase with polymer concentration and colloid volume fraction. Small angle light scattering measurements on samples having a composition just inside the nonequilibrium region of the phase diagram showed features indicative of spinodal decomposition or nucleation. Samples containing sufficient polymer to display transient gelation contained aggregates whose structure was consistent with a diffusion-limited aggregation process. Pusey et al. framed their discussion of transient gelation in terms of competition between diffusion-limited aggregation and thermal rearrangement. Verhaegh (3–5), Lekkerkerker (6), and co-workers observed transient gelation in their optical microscopy and light scattering studies of monodisperse silica spheres in cyclohexane in the presence of polydimethylsiloxane (PDMS) polymer. They proposed that transient gelation was driven by spinodal decomposition, which was responsible for forming a bicontinuous network of colloid-rich and colloid-poor domains. Aggregation in the colloid-rich phase was held responsible for the formation of a gel.
283
0021-9797/98 $25.00 Copyright © 1998 by Academic Press All rights of reproduction in any form reserved.
284
MANOJ ET AL.
The reversible nature of the depletion-induced aggregation permitted network restructuring and eventual gel collapse. However, unlike Pusey and co-workers, who considered a system having a deeper interaction potential, these other authors observed no evidence of fractal clusters. The latter proposed that spinodal decomposition proceeds until a percolation threshold is reached, at which point the system assumes some rigidity. Restructuring leads to loss of connectivity in the network, allowing the phase separation to proceed, and the gel collapses. We note several other experiments reported in the literature which show delayed sedimentation. Smits et al. (5) studied silica in cyclohexane in the presence of polystyrene or PDMS, finding that certain concentrations of polymer-induced phase separation into colloid-rich and colloid-poor phases, followed after some days by crystallization and sedimentation in the colloid-rich phase. Partridge (6, 7) observed an induction time followed by sedimentation in sterically stabilized polystyrene lattices weakly flocculated by the addition of NaCl. There appeared to be a delay in creaming data presented by Dickinson for a triglyceride oil-in-water emulsion stabilized with guar gum (8), but this may have been a result of hazy boundaries that were not initially visible. Experiments by the same group on casein-stabilized emulsions also showed a characteristic induction time before separation (9). Here the casein, despite being adsorbed at the interface, was present in the continuous phase in sufficient quantity to act as a depletion flocculant. Mellor and Stavans observed what appeared to be delayed creaming in monodisperse emulsions of silicon oil emulsified by SDS in the presence of nonadsorbing PEO (10). Mellor and Stavans performed light scattering measurements at high polymer concentrations and found a broad peak commensurate with a correlation length of about 20 mm, or 20 drop diameters. Their corresponding visible flocs were about 25 mm in size. They concluded that the obstacle to creaming was the formation of a gel-like structure of interconnected flocs. Our current study continues our previous investigations of the phase behavior of a polydisperse alkane-in-water emulsion in the presence of the nonadsorbing polymer HEC (11). The main aim is to understand the mechanisms behind the delay period and the sudden collapse that ends in separation. Although it is possible that spinodal decomposition may be involved (4), the presence of polydispersity further complicates the phase behavior. Using visual observations in measuring cylinders, we have measured the duration of the delay period and the eventual creaming rate of oil-in-water emulsions. Data from the IFR creaming monitor, an ultrasonic technique, support our visual observations. We report here the effects of polymer concentration and oil volume fraction on the delay period, creaming rate, and final oil concentration in the cream layers in order to obtain insight into the mechanisms underlying the creaming behavior. The rheological behavior of these systems is discussed in a companion paper (12).
EXPERIMENTAL
Materials and Emulsion Preparation The emulsions were prepared from a premix. An oil phase was added to a solution of the nonionic surfactant Brij 35 (polyoxyethylene 23-lauryl ether, Sigma Chemical Company) and the emulsion prepared in a Waring blender using a predetermined shearing cycle. The resulting premix was stable to coalescence and disproportionation over the timescale of the experiments. Two oil phases were used within the experimental protocol: Premix 1 for visual characterization of creaming. Premix 1 contained n-hexadecane (Cetane–C16H34; density 773.4 kg m23 at 20°C; Sigma Chemical Company) in a surfactant solution. Premix 2 for ultrasonic monitoring of creaming. Premix 2 contained a mixture (density 693.2 kg m23 at 20°C) of nheptane (C7H16; density 683.8 kgm23 at 20°C; SLR, Fisons) and hexadecane in the volume ratio 9:1 in surfactant solution. A series of final emulsions were prepared by diluting the premixes with a polymer solution to obtain the desired concentrations of oil and polymer as in Table 1. The diluent was a mixture of an aqueous solution of the high molecular weight polymer HEC (Natrosol 250HR, Aqualon, mean r g (radius of gyration) 5 58 nm) and the preservative sodium azide (NaN3, Sigma Chemical Company). Prior to use the polymer was purified by dialysis and freeze-dried. The polymer was dispersed and hydrated by stirring together the ingredients while heating to 80°C, then allowing the solution to cool to room temperature. After preparation the emulsions were immediately transferred either to measuring cylinders for visual monitoring or to measurement cells for ultrasonic monitoring. Droplet Size Distribution The droplet size distribution of both premixes was measured using a Malvern Mastersizer laser diffraction sizer. These distributions were highly reproducible and did not change during the timescale of the creaming measurements. A typical size distribution is shown in Fig. 1 with a volume mean diameter (d 43 ) of 1.80 mm. Since the polymer was added after emulsification, it had no effect on the droplet size distributions. Density Densities were measured at 20.0°C using an Anton Paar DMA60 vibrating tube density meter with a DMA602 measuring cell. Rheological Measurements Viscosity (steady shear 2 shear rate) measurements were obtained using a controlled stress (CS) Bohlin Rheometer, using a double gap geometry at 25°C. Solvent traps were used to minimize the effects of drying out and dust settlement. The viscometry measurements were carried out on HEC solutions. Zero-shear viscosity values were obtained for each
285
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
TABLE 1 Emulsion Samples
Premix
Oil type
Oil volume fraction (%; (v/v))
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2
Hexc Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex Hex HHd HH HH
10 10 10 10 25 25 25 25 34 34 34 34 40 40 40 20 20 20
Brij 35 (%; (w/w))a
Total HEC C av (%; (w/w))a,b
Estimated HEC in continuous phase C p (see text)
0.30 0.30 0.30 0.30 0.90 0.90 0.90 0.90 1.39 1.39 1.39 1.39 1.80 1.80 1.80 0.35 0.35 0.35
0.10 0.20 0.35 0.50 0.04 0.10 0.20 0.35 0.04 0.10 0.20 0.35 0.04 0.10 0.20 0 0.025 0.10
0.116 0.231 0.405 0.578 0.060 0.151 0.302 0.529 0.074 0.185 0.371 0.648 0.087 0.218 0.436 0 0.034 0.137
a
Expressed with respect to continuous phase. 0.3% (w/w) sodium azide was added, as a preservative, to all emulsions. c Hexadecane only. d Oil mixture of heptane and hexadecane. b
sample. The specific viscosity (hsp), which defines the fractional increase in viscosity due to the presence of the polymer, was calculated for each concentration
h sp 5
h 2 hs , hs
[1]
polymer concentration (Fig. 2) showed the critical polymer coil overlap concentration (C*) to be 0.275% (w/w) for HEC polymer. The behavior of the specific viscosity with concentration was as expected. Creaming Behavior of the Emulsions
where hs is the solvent viscosity. A log-log chart of hsp and
We have observed two distinct modes of creaming in emulsions containing nonabsorbing polysaccharides. In Type I
FIG. 1. Droplet size distribution of hexadecane premix from light diffraction particle sizer (Malvern Mastersizer).
FIG. 2. Specific viscosity and polymer concentration for estimation of critical polymer concentration.
286
MANOJ ET AL.
creaming, emulsions remain opaque at the base of the sample for a significant period of time, while a concentrated cream layer develops at the top of the sample. The thickness of the cream layer gradually increases with time, and only after most of the oil has reached the cream is there any visible sign of clearing from the base of the sample, which may remain turbid. The boundary between the serum and the creaming droplets is very hazy and indistinct. This behavior is characteristic of polydisperse emulsions in the presence of little or no polymer, where individual droplets or small aggregates move independently to the top of the container. Creaming of this type is very difficult to monitor visually but may be characterized using the IFR ultrasonic monitor (13). In Type II creaming, there is a clear boundary between the creaming emulsion and the serum layer at the base of the sample. Creaming occurs rapidly and the remaining serum can be clear or turbid. Previous investigations of this behavior (14) have suggested the presence of a high degree of flocculation within the creaming emulsion sufficient to form a space spanning structure. This type of creaming is clearly visible. Definition and Calculation of Delay Time Profiles of the creaming behavior of the sample emulsions were obtained by measuring the height of the boundary between the colloid-rich fraction and the colloid-poor fraction (type II creaming) as a function of time for each emulsion. Once started, the rate of creaming was constant (14) but a delay was sometimes evident before creaming began. The delay could be quantified by extrapolation of the boundary height to zero movement. Visual Assessment of Creaming Premix 1 dispersions and polymer diluents were poured into 100-ml measuring cylinders immediately after preparation. The cylinders were inverted 10 times to ensure thorough mixing. The sample emulsions were kept at 25°C and the movement of any creaming boundaries was followed with time. The turbidity of the serum layers below the developing creams was noted. The experiments were repeated between three and five times to obtain mean creaming rates and delay times. The standard deviation within each set of results was calculated to be less than 12%.
FIG. 3. (a) Ultrasonic creaming profiles during creaming of 20% emulsion in the absence of polymer. Each line shows the measured variation of oil volume fraction with height at the each sampling time. (b) The ultrasonic creaming data from (a) expressed as the height of contours of constant oil volume fraction with time. The data show the oil droplet started to move from time 5 0.
Ultrasonic Monitor Measurements
Emulsions Prepared without Polymer (Type I Creaming)
The IFR ultrasonic monitor (13) was used to characterize the creaming behavior of Premix 2 emulsions by measuring the speed of ultrasound in emulsions at a series of heights. The oil volume fraction (w) as a function of height was then calculated using a simple mixing theory (15). At each sampling time the validity of the calibration was verified by calculating the integrated oil volume fraction over sample height. These oil vol-
Visual monitoring. Accurate visual observations of Type I were not possible due to the absence of a well-defined boundary. Ultrasonic monitoring. The ultrasonic creaming profile for 20% (v/v) heptane/hexadecane emulsion in the absence of polymer is shown in Fig. 3a. At 0.28 day (;7 h) the oil had begun to move to the top of the cell and formed a thin cream
ume fraction profiles allowed early detection of creaming and gave detailed information on the creaming process. RESULTS
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
FIG. 4. Ultrasonic creaming profiles during creaming of 20% emulsion in the presence of 0.025% HEC, showing the coexistence of a rapidly rising flocculated phase and a fraction of slower-moving droplets.
layer of approximately 80% (v/v) oil. No sharp boundary was observed in the creaming emulsions. After 4 days, the oil had cleared from the base of the cell (no oil detected at ;13 mm height from the base). By interpolation, the height of contours of constant volume fraction, as a function of time, can be calculated, as shown in Fig. 3b. The linearity of these contours indicated that each oil fraction creamed at a constant velocity (constant gradient) until the droplets reached the closely packed cream layer. Comparison of the creaming velocities with the hindered Stokes velocity for the appropriate size fraction yielded an effective droplet size distribution that was similar to the Mastersizer result (11). The contour analysis also showed that there was no delay in the movement of droplets within the initial phase of creaming.
287
subcream serum containing no measurable oil. Since the serum was also visually clear, we interpreted the profiles as indicative of complete flocculation. Visual monitoring. The addition of HEC in the concentration range 0.04 – 0.35% (w/w) also gave a significant change in the visual creaming behavior of the emulsions, with the appearance of a rapidly creaming boundary at the base. At the onset of creaming, the boundary was characterized as either sharp or diffuse. The boundary regions in emulsions of 10% (v/v) oil volume fraction and low HEC concentration were initially diffuse but sharpened as creaming progressed. In emulsions with high oil volume fraction and high HEC concentration the boundary lines were sharp from the start and creamed rapidly, leaving a clear supernatant or subcream. An example of the creaming behavior is shown in Fig. 6 for 34% (v/v) hexadecane– 0.1% (w/w) HEC ([34:0.10]) and 34% (v/v) hexadecane– 0.35% (w/w) HEC ([34:0.35]) emulsions. A delay was observed before creaming for both emulsions, the delay time being shorter for 0.1% (w/w) HEC than for 0.35% (w/w) HEC. Once creaming was established, the rate of creaming was approximately constant. Figure 7 summarizes the variation of creaming rate with polymer concentration and oil volume fraction. For comparison Fig. 7 also shows the zeroshear viscosity of the continuous phases. After the onset of creaming, the appearance of the serums varied in their turbidity, consistent with the presence or absence of residual oil droplets. The emulsion with 0.35% (w/w) HEC left behind a clear layer from the start, whereas the 0.1% (w/w) HEC initially left behind a turbid serum layer which cleared slowly over a period of days. The initial turbidity (after the majority of the oil had creamed rapidly to the top) gave an indication of whether the system was fully flocculated or contained floc(s) in coexistence with unflocculated droplets.
Emulsions with Polymer (Types I and II Creaming) Ultrasonic monitoring. Figure 4 shows the creaming of a 20% (v/v) emulsion (Premix 2) in the presence of 0.025% (w/w) HEC, monitored using the ultrasonic technique. The profiles revealed two distinct fractions, a rapidly moving population which creamed within a day, leaving almost half the oil behind in a slower-moving fraction. Contour analysis indicated the emulsion contained two phases, a flocculated phase in coexistence with unflocculated droplets. By varying the HEC concentration in the range 0.015– 0.04% (w/w), the amount of flocculation varied in the range 1–98% (11). The unflocculated droplets creamed slowly, due to the increased viscosity of the polymer solution, and the serum was visually cloudy until all the oil had reached the cream layer. Figure 5 shows the ultrasonic profiles for 20% (v/v) oil (Premix 2) with 0.1% (w/w) HEC. The creaming profiles showed a sharp boundary rising rapidly with time, leaving a
FIG. 5. Ultrasonic creaming profiles during creaming of 20% emulsion in the presence of 0.10% HEC, showing formation of a sharp boundary as the flocculated droplets rise to the cream layer.
288
MANOJ ET AL.
of the flocculated structure, reducing the rearrangement on a macroscopic scale. The visual data shown in Fig. 9 were collected before the long-term rearrangements shown in Figs. 4 and 5 had occurred. DISCUSSION
The aim of this study was to investigate the origin of the delay time observed before the onset of creaming in emulsions containing nonadsorbing polymers. Before considering this question in detail we will outline the phase diagram of the system and hence the context of the observed behavior. Phase Behavior of the Emulsions
FIG. 6. Height of visual boundaries and time at 25°C for 34% (v/v) hexadecane emulsions: (■) 0.1% (w/w) HEC; (‚) 0.35% (w/w) HEC.
Plate 1 shows emulsions of 34% (v/v) hexadecane in the presence of 0.04, 0.10, and 0.35% (w/w) HEC, respectively, 16 days after preparation. At 0.04% (w/w) HEC there was a boundary at approximately 47 ml, although the emulsion below the boundary was opaque. At the higher HEC concentrations the serum layers became less opaque, and at 0.35% (w/w) HEC the serum was clear from the start of creaming. The turbid serum layers contained droplets which creamed very slowly due to the high viscosity of the HEC continuous phase. Close examination of the ultrasonic creaming data showed that the remaining fraction of oil was composed of individual droplets. However, at very low residual oil concentrations, visual observations of turbidity are more sensitive than the ultrasonic monitor to the presence of these suspended individual droplets. Summary of delay times. Figure 8 shows the variation of delay time with polymer concentration and oil volume fraction. The delay increased with increasing concentration of either component. It should be noted that for emulsions that initially creamed with diffuse boundary lines, there were significant uncertainties as to the instant that creaming started so the calculated delay times were taken as “upper limit delay times.” Figure 8 also shows the standard deviation of the observed delay times. Depths of cream layers. At the end of creaming the droplets formed a layer at the top of the container. Visual assessments showed that the thickness of the cream layers increased with increasing polymer concentration and oil volume fraction. To allow for the different initial oil volume fractions, the data were used to calculate an average volume fraction, wm, of the oil in the cream layers. Figure 9 shows that the packing density (wm) increased with oil volume fraction and decreased with polymer concentration. This implies that the polymer increases the compressive strength
In our previous study (11) we observed that increasing the HEC concentration caused the formation of a concentrated colloidal phase, in coexistence with a dispersed colloid-poor phase. These phases separated spatially under gravity, the colloid-rich fraction creaming rapidly in comparison with the colloid-poor fraction. Analysis of the ultrasonic creaming profiles of these emulsions showed the colloid-rich phase was flocculated and the colloid-poor phase was composed of individual droplets. We suggested the formation of these phases could be due to one or both of the following effects: ● size segregation of larger droplets due to preferential flocculation; ● phase separation which is independent of droplet size.
No significant difference was found between the size distributions of the droplets contained within each phase and we therefore concluded that phase separation was responsible, at least in part, for the observed creaming behavior. In this study we have investigated the flocculation behavior
FIG. 7. Creaming rates of hexadecane–HEC emulsions at 25°C as a function of polymer concentration as compared with the variation in the viscosity of the solution containing the same total concentration of HEC (---). (F) 10% (v/v) hexadecane; (■) 25% (v/v) hexadecane; (}) 34% (v/v) hexadecane; (Œ) 40% (v/v) hexadecane.
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
PLATE 1.
289
Appearance of cylinders containing 34% (v/v) hexadecane with HEC (from left to right) 0.04, 0.10, and 0.35% (w/w), after 16 days at 25°C.
of similar emulsions, using ultrasonic monitoring and visual observations. The results are summarized in Fig. 10, in the form of a phase diagram showing three regimes: nonflocculated, phase coexistent, and fully flocculated emulsions. The boundary between the nonflocculated and the phase coexistent region we have denoted as the primary critical concentration (PCC), and the boundary between the phase coexistent and fully flocculated region as the secondary critical concentration (SCC). No attempt was made in this study to delineate quantitatively the boundaries between the regions, and the upper and lower limits for SCC are shown. The data are shown as closed points where the serum subcream layers were initially turbid and open points where the serum was clear from the start. This phase diagram contains features consistent with the phase diagram of Ilett et al. (16) for r g/a , 0.3, i.e., two single phase regions separated by a two-phase region. In our study the polydispersed nature of the oil droplets would suggest a fluid– fluid coexistence region as opposed to the fluid–solid coexistence predicted.
Mechanism of Delay Figure 10 also shows the systems exhibiting a delay time significantly greater than zero. Delays were observed within both the phase coexistent and the fully flocculated regions. Within the phase coexistent region a prerequisite of delay was the presence of a sharp boundary upon creaming. We propose two hypotheses for the existence of the observed delay in creaming: ● The viscous nature of the continuous phase reduces the rate of flocculation and the delay is simply the timescale to form flocs which cream rapidly. ● There is the rapid formation of a space-spanning flocculated structure of sufficient “compressive strength” to withstand buoyancy forces. We speculate that the eventual collapse of such a network could be due to coarsening of the structure by droplet rearrangement.
We will consider each of these hypotheses in turn.
290
MANOJ ET AL.
FIG. 8. Delay times of hexadecane–HEC emulsions at 25°C as a function of HEC concentration. Symbols are as in Fig. 7. The error bars represent the standard deviation of the measurements on replicate samples.
Is the delay due to slow flocculation? The variation of delay time with HEC concentration for each oil volume fraction is shown in Fig. 8. For a particular HEC concentration the viscosity of the continuous phase is constant and the droplet encounter rate, and hence the rate of flocculation, would be expected to increase with volume fraction. Thus a decrease in delay time would be expected upon increasing the oil volume fraction if the observed delay time is the consequence of relatively slow flocculation. For all the HEC concentrations used in this study the delay time was found to increase with increasing oil volume fraction. This result contradicts the hypothesis of slow flocculation, particularly when the delay times observed within this study far exceed the characteristic timescale of droplet diffusion and depletion flocculation rates. The sharp boundary would require
FIG. 10. Phase diagram of hexadecane–HEC emulsions based on initial turbidity of serums left behind after creaming at 25°C (closed symbols, cloudy serum; open symbols, clear serum) in association with the existence of delay period (E). The HEC concentration is the total concentration in the continuous phase.
that the flocs all creamed at the same rate, and it seems unlikely that they would have a monodisperse effective size distribution. Rearrangement and collapse of a flocculated space-spanning structure. In our previous studies, and in common with other workers, we have identified the formation of a sharp creaming boundary to be symptomatic of a flocculated spacespanning structure creaming as a single entity (14). The form of the ultrasonic creaming profiles supports this interpretation. Upon creaming the oil droplets ascended the container at a single rate producing a sharp discontinuity in volume fraction. There was no broadening of the boundary that might suggest movement of separate aggregates with varying speed. Pursuing this approach we examined the creaming rates to determine their scaling behavior. In particular we aimed to separate the individual droplet interactions (controlled by the deletion potential and thus by the HEC concentration) from the effects of the initial volume fraction. Creaming Rate: Network Model
FIG. 9. Average oil volume fraction in the cream layers and total HEC concentration. Symbols are as in Fig. 7.
Emulsion creaming requires backflow of the continuous phase into the volume vacated by the rising oil droplets. Ilett et al. (16) considered that since nonadsorbing polymer is excluded from the immediate vicinity of the droplets, to a distance comparable with the polymer’s radius of gyration in solution, r g, then the effective volume of solvent available for the polymer is reduced, and the effective polymer concentration C p (outside the depleted zone) is higher than the overall concentration C av. Ilett et al. estimated the increase in polymer concentration a 5 C p/C av due to these excluded zones around droplets of radius a to be
291
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
a 5 ~1 2 w !exp@2A g 2 B g 2 2 C g 3#,
[2]
where
g5
w , A 5 3j 1 3j2 1 j3 , ~1 2 w ! B5
9 rg j 1 3 j 3, C 5 3 j 3, j 5 . 2 a
Applying the correction to the measured size distribution (Fig. 1) for each emulsion gave a significant increase in the effective concentration of HEC, as shown in Table 1. In the following analyses, we have used this estimated continuous phase concentration for all the HEC emulsions. We propose that the structures filled the available volume when they formed initially, but after a characteristic time (delay period) the structure was no longer space filling and separated physically from the surrounding continuous phase. The emulsions with w $ 25% (v/v) showed a sharp creaming boundary as soon as they began to cream, consistent with a space-filling structure. One model for this separation process is that the flocs form a porous network and the continuous phase flows through channels (pores) in the network. Michaels and Bolger (17) modeled the settling rate of kaolin flocs using the Kozeny–Carman equation applied to pore flow. They assumed the pores to be equivalent to smooth, straight channels of diameter d p, and obtained the maximum settling velocity V from g z D r z w z d 2p~1 2 w / w m! , V5 32 h
[3]
where Dr is the density difference between the dispersed particles and continuous phase, h is the viscosity of the continuous phase, g is the acceleration due to gravity, w is the volume fraction of particles, and wm is the maximum volume fraction within the flocs. Applying this expression to the measured creaming rates of the hexadecane emulsions, we obtained an estimate of the equivalent pore diameter d p in each emulsion, as shown in Fig. 11. The maximum oil concentration wm in the flocs was taken from the final concentration in the cream layers (Fig. 9). In the model, the initial w affected the overall voidage of the emulsions and therefore the total volume of channels, but it assumes that the size of the channels is determined by the flow properties of the continuous phase and geometry of the channels (roughness, tortuosity). Our data show that the effective pore size was indeed independent of initial w, despite a factor of up to 5 in creaming rate. The increase of effective pore diameter with HEC concentration, also observed by Michaels and Bolger, is attributed to the increased strength of the flocs under a higher depletion potential, which resisted the formation of
FIG. 11. Mean diameter of creaming channels in the flocculated emulsions, inferred from Michaels and Bolger (17). Symbols are as in Fig. 7. Closed symbols, cloudy serum; open symbols, clear serum.
smooth straight channels. The effective channel diameter, between 20 and 120 mm, was an appropriate order of magnitude for structures that cannot be resolved by the naked eye and consistent with the 25-mm characteristic length scale observed by Mellor and Stavans (10). Although at high HEC concentrations there was more scatter on the relationship, regression analysis on the data in Fig. 11 showed a correlation that was significant at better than the 99.9% level. Delay Period Figure 8 showed that the delay time was a strong function of both polymer concentration and droplet volume fraction, w. The polymer dependence is straightforward to rationalize, as the depth of the attractive interparticle potential is directly related to the concentration of free polymer (18). However, it is less easy to explain the strong w dependence. Full statistical mechanical treatment of the collective behavior of the droplets experiencing attractive interactions is outside the scope of this paper. However, it is interesting to examine the w dependence in terms of a mean-field approach, which is explored below. The porous model requires the formation of channels in the flocculated structure. It is reasonable to suppose that the delay time represents the time for the droplets to rearrange to a channeled structure. In this case, the delay time could be related to the diffusion of the droplets in and around the flocs. The diffusion coefficient of individual droplets, radius a, may be estimated using the Stokes–Einstein expression D5
k BT , 6 ph a
[4]
where h is the viscosity of the background medium. Considering the (flocculated) structure to hinder the diffusion of the
292
MANOJ ET AL.
than to be randomly dispersed (20). The rate of decomposition is also predicted to be dependent on the single-droplet diffusion rate. In dispersions flocculated by the addition of polymer, varying the interparticle potential by varying polymer concentration also changes the viscosity of the continuous phase. Thus it is not straightforward to separate the two mechanisms, flocculation and the process of spinodal decomposition. In principle, each process leads to spatial density variations that have a characteristic length scale (20). To determine which mechanism dominates in the droplet rearrangements prior to creaming, experiments are needed where the droplet interactions can be varied without changing the single-droplet diffusion rate. FIG. 12. Measured delay time and the estimated effective viscosity of the emulsions. Symbols are as in Fig. 7. Closed symbols, cloudy serum; open symbols, clear serum.
individual droplets, we can replace h by an effective viscosity, he, using the Ball/Richmond mean field model (19)
S
he 5 h0 1 2
w wm
D
25wm/ 2
.
[5]
If droplet diffusion is the controlling factor in the delay time, then we expect a relationship between the delay time and effective viscosity of the medium. Estimating the effective viscosity of the emulsions using Eq. [5] and the measured value of wm, we obtained the trends shown in Fig. 12. The analysis eliminated the effect of initial oil volume fraction, and there were two distinct trends. In the coexistent regime (closed data points) there was an approximately linear relationship (gradient (power) 1.06, significant at 99% level), and at high HEC concentrations when we observed complete flocculation (clear serum, open data points) the delay scaled with viscosity to a power 0.36 (significant at 99% level). The interesting point about this correlation is that an expression that is analogous to a single droplet diffusion rate allowed for the f dependence in the rearrangement time. So far we have discussed the rearrangement and subsequent creaming of the emulsion/polymer system in terms of flocculated structures that coarsened to produce channels through which continuous phase could flow. In flocculation, regions of the systems become colloid-rich (flocs) which grow from a nucleus analogous to a crystallization process. Flocculation is an instability that affects the rate at which equilibrium is achieved. However, the observed separation could also be due to slow spinodal decomposition, a demixing process. In this process, because the system has a lower energy when the droplets are in regions of high local droplet density, random fluctuations become asymmetric, so that over time a droplet is more likely to remain in a region of high droplet concentration
Additional Factors and Approximations Clearly the above analyses are tentative, as there are a number of factors that have not been included rigorously. The initial correction made to the effective continuous phase concentration (Eq. [2]) is accepted for low HEC concentrations, but must necessarily be flawed above C*, where the polymer chains are no longer able to adopt a random-coil configuration in solution. The droplet polydispersity is considered in the above analysis only because of its effect on the excluded volume, but it is likely to be a factor in the phase behavior of the droplet/polymer mixture, as well as the diffusion behavior of the droplets. Similarly, the polymer polydispersity is ignored, but it may well cause perturbations to the phase diagram. Modeling the creaming behavior as a porous network has been shown to rationalize the data for different oil volume fractions, but the model includes major assumptions on the channel geometry and floc structures. Equally, it is debatable whether the treatment of droplet concentration by means of an effective, mean-field viscosity is valid in the presence of attractive potentials. However, we offer the ideas above as a starting point for more detailed studies, which may provide an insight into the mechanisms underlying the gravitational collapse of flocculated structures for improved product design in the process industries. CONCLUSIONS
Experiments on a series of depletion-flocculated emulsions showed that the degree of flocculation, type of creaming, creaming rate, and delay before creaming were dependent on the oil volume fraction and polymer (HEC) concentration. With increasing HEC, the polydisperse emulsions displayed, in turn, creaming of individual droplets; creaming in two fractions corresponding to a flocculated phase in coexistence with a phase of individual droplets; and a fully flocculated state where the oil rose at a single, constant, rate. Delays before creaming were observed in the fully flocculated emulsions and in the partially flocculated emulsion at high oil volume fraction.
CREAMING BEHAVIOR OF FLOCCULATED EMULSIONS
The phase behavior of the emulsions as a function of oil volume fraction and polymer concentration was similar in form to that observed by Ilett et al. (15) for monodisperse solid particles in the presence of nonadsorbing polymer. In the highly flocculated emulsions the dependence of creaming rate on oil volume fraction was consistent with the porous network model proposed by Michaels and Bolger (16), which inferred the existence of channels in the structure of size in the range 20 to 120 mm. We propose that the delay period represents the time required for the channels to form and show that the delay time for each emulsion scales with an effective continuous phase viscosity. Whether the mechanism of channel formation is controlled by the rate of spinodal decomposition or by diffusion-controlled flocculation is discussed. ACKNOWLEDGMENTS The authors acknowledge funding from BBSRC in the form of the Institute’s competitive strategic grant and a ROPA award. We are grateful to Wilson Poon for helpful discussions.
REFERENCES 1. Parker, A., Gunning, P. A., Ng, K., and Robins, M. M., Food Hydrocolloids 9, 333 (1995). 2. Pusey, P. N., Pirie, A. D., and Poon, W. C. K., Physica A 201, 322 (1993).
293
3. Verhaegh, N. A. M., Vanduijneveldt, J. S., Dhont, J. K. G., and Lekkerkerker, H. N. W., Physica A 230, 409 (1996). 4. Lekkerkerker, H. N. W., Poon, W. C. K., Pusey, P. N., Stroobants, A., and Warren, P. B., Europhys. Lett. 20, 559 (1992). 5. Smits, C., Vandermost, B., Dhont, J. K. G., and Lekkerkerker, H. N. W., Adv. Colloid Interface Sci. 42, 33 (1992). 6. Partridge, S. J., Ph.D. thesis, University of Bristol, UK, 1986. 7. Buscall, R., Colloids Surf. 43, 33 (1990). 8. Dickinson, E., in “Macromolecular Interactions in Food Technology.” Am. Chem. Soc., Washington, DC, 1996. 9. Dickinson, E., Golding, M., and Povey, M. J. W., J. Colloid Interface Sci. 185, 515 (1997). 10. Mellor, A., and Stavans, J., Langmuir 12, 301 (1996). 11. Fillery Travis, A. J., Gunning, P. A., Hibberd, D. J., and Robins, M. M., J. Colloid Interface Sci. 159, 189 (1993). 12. Manoj, P., Watson, A. D., Hibberd, D. J., Fillery-Travis, A. J., and Robins, M. M., J. Colloid Interface Sci., in press. 13. Howe, A. M., Mackie, A. R., and Robins, M. M., J. Dispersion Sci. Technol. 7, 231 (1986). 14. Robins, M. M., ACS Sympos. Ser. 448, 230 (1991). 15. Urick, R. J., J. Appl. Phys. 18, 983 (1947). 16. Ilett, S. M., Orrock, A., Poon, W. C. K., and Pusey, P. N., Phys. Rev. E 51, 1344 (1995). 17. Michaels, A. S., and Bolger, J. C., I&EC Fund. 1, 24 (1962). 18. Sperry, P. R., J. Colloid Interface Sci. 99, 97 (1984). 19. Ball, R. C., and Richmond, P., Phys. Chem. Liq. 9, 99 (1980). 20. Dhont, J. K. G., “An Introduction to Dynamics of Colloids.” Elsevier, Amsterdam 1996.