Formulation and Preparation of Water-In-Oil-In-Water Emulsions Loaded with a Phenolic-Rich Inner Aqueous Phase by Application of High Energy Emulsification Methods

Currently, industry is requesting proven techniques that allow the use of encapsulated polyphenols, rather than free molecules, to improve their stability and bioavailability. Response surface methodology (RSM) was applied in this work to determine the optimal composition and operating conditions for preparation of water-in-oil-in-water (W/O/W) emulsions loaded with phenolic rich inner aqueous phase from olive mill wastewater. A rotor-stator mixer, an ultrasonic homogenizer and a microfluidizer processor were tested in this study as high-energy emulsification methods. Optimum results were obtained by means of microfluidizer with 148 MPa and seven cycles input levels yielding droplets of 105.3 ± 3.2 nm in average size and 0.233 ± 0.020 of polydispersity index. ζ-potential, chemical and physical stability of the optimal W/O/W emulsion were also evaluated after storage. No droplet size growth or changes in stability and ζ-potential were observed. Furthermore, a satisfactory level of phenolics retention (68.6%) and antioxidant activity (89.5%) after 35 days of storage at room temperature makes it suitable for application in the food industry.


Introduction
Emulsions are generally used for the encapsulation of bioactive compounds in aqueous solutions. They consist of at least two immiscible liquids (oil and water), one of them dispersed as small droplets in the other [1,2]. Typically, droplet diameters in food systems range from 0.1 to 100 µm [3]. Emulsions can be classified as oil-in-water (O/W) or water-in-oil (W/O) emulsions, depending on whether the dispersed phase is oil or water, respectively. Furthermore, there are several types of multiple emulsions, such as oil-in-water-in-oil (O/W/O) or water-in-oil-in-water (W/O/W) emulsions [4,5]. Emulsifiers are commonly added as stabilizers to obtain a kinetically stable system [6].
The formation, stability, and properties of emulsions depend on the characteristics (polarity, water solubility, viscosity, density, etc.) of the oil phase and the type and concentration of components present in the aqueous phase [2,[7][8][9][10][11]. The choice of surfactant used as emulsifier is very important in the food industry because it must be able not only to create and stabilize the dispersed phase droplets, but also be biodegradable and nontoxic [12].

Preparation of Primary Emulsion (W/O)
W/O emulsions were prepared by mixing Miglyol 812 and Span 80 in amounts described by response surface methodology (RSM) and shown in Table 1. Then the aqueous dispersed phase was added dropwise and the mixture was stirred for 10 min at 500 rpm. The prepared W/O emulsions were immediately used as the dispersed phase in the preparation of W/O/W double emulsions.

Preparation of Double Emulsion (W/O/W)
W/O/W emulsions were prepared by the dropwise addition of the dispersed phase (W/O emulsion) to a continuous phase formed by Milli-Q water and the Tween 80 surfactant.
The primary goal for preparation of double emulsion was optimization of formulation. For this purpose, emulsification process was performed with 26 different compositions based on RSM experimental design ( Table 1). The emulsification was performed by using a high intensity ultrasonic homogenizer, described below, for 6 min effective time, in 5 s pulses (5 s off and 5 s on) and 50% amplitude.
The second goal was emulsification of optimized formulation with different high energy methods in order to achieve the optimal operating conditions. A high speed blender (Miccra D9 equipped with a DS-5/K-1 rotor-stator, ART Prozess & Labortechnik, Mülheim, Germany), a high intensity ultrasonic homogenizer (Sonics VCX 500, 500 W, 20 kHz, Newtown, CT, USA) with a titanium alloy microtip probe of 3 mm diameter, and a microfluidizer high shear fluid processor (LM20, Microfluidics, Westwood, MA, USA) were the equipment used for the formation of W/O/W emulsions.

Droplet Size Analysis of the Emulsions
Droplet size distribution, mean droplet diameter and polydispersity index (PDI) of samples were measured by dynamic light scattering (DLS) using a Zetasizer Nano ZS apparatus (Malvern Instruments Ltd., Malvern, UK). Measurements were performed by triplicate at 25 • C. The PDI is a dimensionless measure of the width of the size distribution ranging from 0 to 1, a higher value being indicative of a broader distribution of particle size.
2.6. ζ-Potential ζ-potential was also measured with the aforementioned Zetasizer Nano ZS apparatus, using the laser Doppler velocimetry technique. The measurement was conducted six times for each diluted sample at 25 • C. The ζ-potential was calculated using Henry's equation and the Smoluchowski approximation.

Total Phenolic Content
Total phenolic content (TPC) of samples was measured using Folin-Ciocalteu standard method [42] with some modifications. Briefly, a standard calibration curve (R 2 = 0.998) was prepared using gallic acid solution. 100 µL of sample was added to a test tube. Then 2.8 mL of distilled water, 2 mL of 7.5% sodium carbonate solution and 100 µL of Folin-Ciocalteu reagent were added to the test tube and mixed. After 60 min of incubation in darkness at room temperature, the absorbance was measured spectrophotometrically at 750 nm wavelength (U-2000 spectrophotometer, Hitachi, Tokyo, Japan). TPC was expressed as milligrams of gallic acid equivalents per liter of sample (mg GA/L).

Antioxidant Activity
Antioxidant activity (AA) of the samples was measured by the DPPH free radical scavenging activity method described by Shen et al. [43]. Briefly, 1 mg of DPPH reagent (2,2-diphenyl-1picrylhydrazyl) was dissolved in 50 mL of methanol and stored in darkness at 4 • C for 4 h. In order to perform DPPH assay, 60 µL of sample were mixed with 2940 µL of DPPH reagent solution. For the control sample 60 µL of the solvent (in this case Milli-Q water) were used. The mixtures were allowed to stand in darkness at room temperature for 60 min. Then the absorbance was measured at 517 nm wavelength using the U-2000 spectrophotometer. A standard calibration curve (R 2 = 0.995) was prepared using Trolox solution, and antioxidant activity was expressed as milligrams of Trolox equivalents per liter of sample (mg Trolox/L).

Chemical Stability Measurements
The percentage of phenolic compounds held within the aqueous phase after 35 days of storage at room temperature was measured following the method proposed by Regan and Mulvihill [44]. Thus, 3 g of optimal double emulsions were mixed with 3 g of phosphate buffer solution (pH 7) and centrifuged (Eppendorf 5804 centrifuge) at 4500 rpm for 90 min. Then, the lower phase was collected carefully for total phenolic content (TPC) and antioxidant activity (AA) analysis. The percentage of encapsulated compounds (E) was identified by using Equation (1): where C 2 is the concentration of phenolic compounds found in the aqueous phase after centrifugation and C 1 is the initial concentration of phenolic compounds in the inner aqueous phase [22,44].

Physical Stability Measurements
Stability of W/O/W emulsions was measured in terms of their droplet growth ratio. Since emulsions tend to aggregate during storage, the droplet size of the emulsions was measured after 1 day and also 35 days after preparation. Two different storage conditions were evaluated: 4 • C and room temperature in darkness. Furthermore, optical characterization of the optimal double emulsion was done by static multiple light scattering (S-MLS) using a Turbiscan Lab Expert equipment (Formulaction Co., L'Union, France). The apparatus send a light beam from an electroluminescent diode (λ = 880 nm) through a cylindrical glass cell containing the sample. The emulsion sample (20 mL) without dilution was placed in a cylindrical glass cell and two synchronous optical sensors received the light transmitted through the sample (180 • from the incident light) and the light backscattered by the droplets in the sample (45 • Table 1. The following second-degree polynomial equation (Equation (2)) was used to express the predicted response (Y) as a function of the independent variables (X 1 , X 2 , X 3 and X 4 ): Y = a 0 + a 1 X 1 + a 2 X 2 + a 3 X 3 + a 4 X 4 + a 11 X 2 1 + a 12 X 1 X 2 + a 13 X 1 X 3 + a 14 X 1 X 4 + a 22 X 2 2 + a 23 X 2 X 3 + a 24 X 2 X 4 + a 33 X 2 3 + a 34 X 3 X 4 + a 44 X 2 where Y represents the response variable (droplet size in this case), a 0 is a constant, and a i , a ii and a ij are the linear, quadratic and interactive coefficients, respectively. After optimization of the W/O/W formulation, three experimental designs based on RSM were applied in order to determine the optimal emulsification method. The three aforementioned apparatus were tested: a rotor-stator mixer, an ultrasonic homogenizer and a microfluidizer processor. For this purpose, the CCD model generated 10 experimental runs with two replicates of the central point for each emulsification method, as shown in Table 2. The factors selected were rotation speed (X 1 : 11,000-29,000 rpm) and time (X 2 : 5-15 min) for the rotor-stator mixer, sonication time (X 1 : 5-15 min) and amplitude (X 2 : 20-60%) for ultrasonic homogenizer (while temperature (30 • C) and pulses (5 s off and 5 s on) were kept constant), and pressure (X 1 : 50-150 MPa) and number of cycles (X 2 : 1-11) for microfluidizer processor. The related data can be seen in Table 2. The following polynomial equation (Equation (3)) was used to express predicted responses (Y) as a function of the independent variables (X 1 , X 2 ) under study: Y = a 0 + a 1 X 1 + a 2 X 2 + a 3 X 3 + a 11 X 2 1 + a 22 X 2 2 + a 12 X 1 X 2 where Y represents the response variable (droplet size), a 0 is a constant, and a i , a ii and a ij are the linear, quadratic and interactive coefficients, respectively. The value of the factors and their effect on the response was determined by analysis of variance (ANOVA) and LSD (lesser significant difference) test. The model was adjusted by means of multiple linear regressions (MLR) and its validity was determined by ANOVA. The level of significance of each coefficient was evaluated through the values of the statistical parameters F and p (probability), with a confidence level of 95% [47].

Model Fitting of Formulation
One response variable, the droplet size, and four experimental factors were used on the CCD for the optimization of formulation, as shown in Table 1. PDI of the LSD measurements is also shown in Table 1. The default model is quadratic with 15 coefficients, and it has been fit to the response variable. The R-squared statistic indicates that the model as fitted explains 80.57% of the variability in particle size. Three of the coefficients (a 3 , a 4 , and a 34 ) of the quadratic polynomial model, Equation (2), have p-values less than 0.05, indicating that they are significantly different from zero at the 95% confidence level (Table 3). F-ratio values indicate that, for the range of studied variables, W/O content in W/O/W composition (X 3 ) had stronger influence on the droplet size of the emulsions than the other independent variables. F-ratio values also indicate that the interaction with the highest incidence was the one occurring between the W/O concentration in W/O/W (X 3 ) and surfactant content in W/O/W emulsion (X 4 ). Table 3. Analysis of variance of the regression coefficients of the quadratic model (Equation (2)

Response Surface Analysis
In order to study the effect of the independent variables on the droplet size, surface responses of the quadratic polynomial model were generated by varying two of the independent variables within the experimental range while holding the other two constant at the central points. It shows that increasing X 4 in the lower levels of X 2 causes an increase in emulsion particle size which is unfavorable, while for higher levels of X 2 no considerable variations in particle size were observed. The effect of X 1 and X 4 on the particle size of the emulsion at a fixed content of X 2 and X 3 in their central values can be seen in Figure 1b. This figure shows that the increase in X 1 and X 4 hardly affects the particle size, except at the higher levels for both factors. The effect of X 1 and X 2 changing on particle size at the central values of X 3 and X 4 is depicted in Figure 1c. It shows that changes of X 1 and X 2 hardly affect the particle size. The effects of X 3 and X 4 variations on particle size at central values of X 1 and X 2 are shown in Figure 1d: a particle size increase is observed as X 3 increases, being this effect greater at higher levels of X 4 . Figure 1e depicts the effect of X 2 and X 3 on particle size at central values of X 1 and X 4 . Somehow, X 2 variation hardly modifies the particle size, while increasing Foods 2020, 9, 1411 8 of 19 X 3 has a significant effect on particle size growth. Figure 1f shows the effect of X 1 and X 3 variation on particle size at central levels of X 2 and X 4 . A significant growth in particle size with the increase in X 1 is only observed at the highest levels of X 3 factor. It can be concluded that these figures prove the significance of X 3 and X 4 over other factors on particle size, as well as the interaction between both factors represented in Figure 1d.

Optimization of Double Emulsion Formulation
Following Stoke's law, the stability of emulsion would increase as the droplet size decreases. Furthermore, the emulsion with higher resistance and control to creaming should also be homogenously distributed in particle size [48]. Numerical optimization of W/O/W emulsion formulation loaded with phenolic rich inner aqueous phase was carried out through design expert software, using desirability function. The W/O/W optimal formulation is expected to be those leading to a stable emulsion with minimum droplet size. Optimum formulation with 98.8% desirability was predicted for 20% (w/w) of aqueous phase content (X1), 4% (w/w) Span 80 surfactant content (X2), 10% (w/w) of aqueous phase (W/O) content (X3), and 4.6% (w/w) of Tween 80 surfactant content (X4). Because some of these optimal values are in the lower range of those selected in the CCD matrix, data extrapolation was performed. It was carried out through expert design software to expand factor input levels and to evaluate the possibility of achieving a better response beyond the levels (e) X 2 vs. X 3 ; (b) X 1 vs. X 3 .

Optimization of Double Emulsion Formulation
Following Stoke's law, the stability of emulsion would increase as the droplet size decreases. Furthermore, the emulsion with higher resistance and control to creaming should also be homogenously distributed in particle size [48]. Numerical optimization of W/O/W emulsion formulation loaded with phenolic rich inner aqueous phase was carried out through design expert software, using desirability function. The W/O/W optimal formulation is expected to be those leading to a stable emulsion with minimum droplet size. Optimum formulation with 98.8% desirability was predicted for 20% (w/w) of aqueous phase content (X 1 ), 4% (w/w) Span 80 surfactant content (X 2 ), 10% (w/w) of aqueous phase (W/O) content (X 3 ), and 4.6% (w/w) of Tween 80 surfactant content (X 4 ). Because some of these optimal values are in the lower range of those selected in the CCD matrix, data extrapolation was performed. It was carried out through expert design software to expand factor input levels and to evaluate the possibility of achieving a better response beyond the levels considered. Formulation optimization process for main response (minimum particle size) yields 100% desirability for 20.3% (w/w) of aqueous phase content (X 1 ), 3.7% (w/w) Span 80 surfactant content (X 2 ), 9.8% (w/w) of aqueous phase (W/O) content (X 3 ), and 4.1% (w/w) of Tween 80 surfactant content (X 4 ). The predicted optimal response is very close to that obtained without extrapolation, which reveals the validity of the experimental design carried out. Three experimental replicates of the optimal formulation were made (20.3% X 1 , 3.7% X 2 , 9.8% X 3 and 4.1% X 4 ). The average droplet size of these samples was 232.5 ± 2.9 nm and the PDI was 0.274 ± 0.013, indicating a narrow distribution in droplet size. As expected, the results were very close to those obtained in run 16 of Table 1.

Effect of the High Energy Emulsification Method on Double Emulsion Droplet Size
Under suitable homogenization conditions, the final droplet size depends strongly on the characteristics of the oil and emulsifier used [34,[49][50][51][52][53][54][55][56][57][58][59][60]. In general, small droplet sizes can be obtained more effectively by low-energy approaches than by high-energy approaches, but the former are more limited in the types of oils and emulsifiers that can be used [11].
Microfluidizers generate intense disruptive forces when two fast-moving emulsion streams impinge upon each other within an interaction chamber, leading to highly efficient droplet disruption [11]. Ultrasonic homogenizers use high-intensity ultrasonic waves to create the intense disruptive forces needed to fracture oil and water phases into very small droplets [53,[61][62][63]. Ultrasound requires less energy expenditure than other high-energy methods, but sonicator probe-induced contamination is an important drawback. For scale-up applicability, commercial homogenizers based on sonication have been developed in which nanoemulsion is made to flow through a special column capable of producing ultrasonic waves [64]. Rotor-stator mixers are mechanical homogenizers especially used for high viscosity and high disperse phase volume fraction dispersions, e.g., in pharmaceuticals, cosmetics and food processing [65][66][67].
After obtaining the optimal W/O/W double emulsion formulation, the second aim of this work was to achieve the optimal operating conditions for its preparation. For this purpose, the following methods and factors were studied: ultrasonic homogenizer by varying time and amplitude, rotor-stator mixer by varying time and rotor speed, and microfluidizer processor by varying pressure and cycle numbers.

Model Fitting of Emulsification Method
Three experimental designs based on RSM were prepared by the application of rotor-stator mixer, ultrasonic homogenizer and microfluidizer processor. For this purpose, the CCD model generated 10 experimental runs with two replicates of central point for each emulsification apparatus. The average droplet size and PDI of the experiments corresponding to the CCD design is given in Table 2.
The default model is quadratic and statistical models have been fit to the response variables. The R-squared statistic indicates that the fitted model explains 84.23%, 56.13% and 86.18% of the variability in particle size for rotor-stator mixer, ultrasonic homogenizer and microfluidizer processor, respectively. Regarding the rotor-stator mixer, the F-ratio values indicate that the time factor (X 2 ) has a stronger influence on the droplet size of the emulsions than the rotational speed factor (X 1 ). In the ultrasonic homogenizer, the F-ratio values indicate that time (X 1 ) has a slightly stronger influence on the droplet size of double emulsions than the amplitude (X 2 ), whereas for microfluidizer processor pressure (X 1 ) has considerable stronger influence on the droplet size of the emulsions than the number of cycles (X 2 ), as shown in Table 4.
ANOVA showed the significance of the coefficients of the quadratic polynomial models (Equation (3)). Regarding the high speed mixer and microfluidizer, only one coefficient in each model has a p-value less than 0.05, indicating that it is significantly different to zero with 95% confidence level. Moreover, no coefficient has a p-value less than 0.05 for the ultrasonic homogenizer model (Table 4).

Response Surface Analysis
In order to study the effect of the emulsification methods and operating conditions on the droplet size, surface response plots of the quadratic polynomial model were generated. Figure 2a shows the effects of time and rotation speed in rotor-stator mixer on the emulsion droplet size. This figure shows that regardless of time level, 20,000-23,000 rpm is the optimal range for rotation speed. However, increasing time causes a continuous decrease in particle size which is favorable.

Response Surface Analysis
In order to study the effect of the emulsification methods and operating conditions on the droplet size, surface response plots of the quadratic polynomial model were generated. Figure 2a shows the effects of time and rotation speed in rotor-stator mixer on the emulsion droplet size. This figure shows that regardless of time level, 20000-23000 rpm is the optimal range for rotation speed. However, increasing time causes a continuous decrease in particle size which is favorable.  Figure 2b shows the interaction of time and amplitude on the achieved particle size in ultrasonic homogenizer; as it can be observed, amplitude hardly affects the particle size except for lower time levels. Furthermore, increasing time is significant on reducing particle size [68], as it causes an increase in temperature and cavitation intensity which accelerates the breakdown of droplets [57,69]. Somehow, the minimum particle size would be achieved for high time and lower amplitude levels.
In emulsions prepared by microfluidization the mean droplet diameter decreased with increasing homogenization pressure. It is in accordance with the study performed by Bai et al. [49], in which the mean droplet diameter decreased from around 213 to 150 nm as the homogenization pressure increased from 4 to 14 kbar. The decrease in droplet size with increasing pressure can be attributed to the increase in the magnitude of the disruptive forces generated within the homogenization chamber. Therefore double emulsions containing small droplets with a narrow particle size distribution can be produced by using microfluidizer processor. Figure 2c shows the interaction between pressure and cycles and their effect on particle size in microfluidizer emulsification tests. It can be observed that increasing pressure is significantly effective in decreasing particle size at middle cycle levels. Somehow, the minimum particle size can be achieved at 140-150 MPa pressure levels and 6-7 cycles.

Optimization of W/O/W Emulsion Preparation Conditions
The optimal conditions for the emulsification of the phenolic rich olive mill aqueous phase used in this work would be those leading to a stable double emulsion with minimum droplet size. Numerical optimization was performed through design expert software, using desirability function method. Regarding the W/O/W emulsification by rotor-stator mixer, the combined optimum ingredient levels for average particle size with 100% desirability were predicted to be achieved by emulsification at 20000 rpm and 10 min. The predicted response at optimal value was 169.4 nm for particle size and polydispersity index (PDI) of 0.546, while observed results were 301.7 ± 12.1 nm and 0.439 ± 0.030 for particle size and PDI, respectively ( Table 2). This large difference between predicted and observed results is due to imprecision in the average size for heterogeneous populations, as revealed the high PDI values In the case of optimal double emulsion prepared by ultrasonic homogenizer, response of 316.15 nm for particle size and PDI of 0.453 with 100% desirability was predicted while emulsification for 10 min at 20% amplitude was the combined optimum factor levels. The observed results, as shown in Table 2, were 338.5 ± 16.5 nm and 0.468 ± 0.021 for particle size and PDI, respectively. The narrow difference between the predicted and observed responses can verify the reliability of the ultrasonic assisted experimental design.
Best results were obtained with microfluidizer processor (Table 2). Numerical optimization performed by the design expert software predicted the optimal factor levels as 148 MPa of pressure and 7 cycles. The predicted response was 96.03 nm and 0.330 for particle size and PDI, respectively, whereas the observed responses for the mentioned factor levels were 105.3 ± 3.2 nm and 0.233 ± 0.020 (c) Figure 2. Response surface plots of interaction between emulsification input factors on particle size: (a) Rotor-stator mixer; (b) Ultrasonic homogenizer; (c) Microfluidizer processor. Figure 2b shows the interaction of time and amplitude on the achieved particle size in ultrasonic homogenizer; as it can be observed, amplitude hardly affects the particle size except for lower time levels. Furthermore, increasing time is significant on reducing particle size [68], as it causes an increase in temperature and cavitation intensity which accelerates the breakdown of droplets [57,69]. Somehow, the minimum particle size would be achieved for high time and lower amplitude levels.
In emulsions prepared by microfluidization the mean droplet diameter decreased with increasing homogenization pressure. It is in accordance with the study performed by Bai et al. [49], in which the mean droplet diameter decreased from around 213 to 150 nm as the homogenization pressure increased from 4 to 14 kbar. The decrease in droplet size with increasing pressure can be attributed to the increase in the magnitude of the disruptive forces generated within the homogenization chamber. Therefore double emulsions containing small droplets with a narrow particle size distribution can be produced by using microfluidizer processor. Figure 2c shows the interaction between pressure and cycles and their effect on particle size in microfluidizer emulsification tests. It can be observed that increasing pressure is significantly effective in decreasing particle size at middle cycle levels. Somehow, the minimum particle size can be achieved at 140-150 MPa pressure levels and 6-7 cycles.

Optimization of W/O/W Emulsion Preparation Conditions
The optimal conditions for the emulsification of the phenolic rich olive mill aqueous phase used in this work would be those leading to a stable double emulsion with minimum droplet size. Numerical optimization was performed through design expert software, using desirability function method. Regarding the W/O/W emulsification by rotor-stator mixer, the combined optimum ingredient levels for average particle size with 100% desirability were predicted to be achieved by emulsification at 20,000 rpm and 10 min. The predicted response at optimal value was 169.4 nm for particle size and polydispersity index (PDI) of 0.546, while observed results were 301.7 ± 12.1 nm and 0.439 ± 0.030 for particle size and PDI, respectively (Table 2). This large difference between predicted and observed results is due to imprecision in the average size for heterogeneous populations, as revealed the high PDI values.
In the case of optimal double emulsion prepared by ultrasonic homogenizer, response of 316.15 nm for particle size and PDI of 0.453 with 100% desirability was predicted while emulsification for 10 min at 20% amplitude was the combined optimum factor levels. The observed results, as shown in Table 2, were 338.5 ± 16.5 nm and 0.468 ± 0.021 for particle size and PDI, respectively. The narrow difference between the predicted and observed responses can verify the reliability of the ultrasonic assisted experimental design.
Best results were obtained with microfluidizer processor (Table 2). Numerical optimization performed by the design expert software predicted the optimal factor levels as 148 MPa of pressure and 7 cycles. The predicted response was 96.03 nm and 0.330 for particle size and PDI, respectively, whereas the observed responses for the mentioned factor levels were 105.3 ± 3.2 nm and 0.233 ± 0.020 Foods 2020, 9, 1411 12 of 19 for particle size and PDI, respectively. The narrow difference between the predicted and observed responses can also verify the reliability of the microfluidizer experimental design. Therefore, it can be concluded from the observed responses of optimal points from three emulsification methods that microfluidizer processor can be considered as the optimal method to achieve double emulsions with lower particle size and PDI.

Stability Evaluations During Storage
The stability of the optimized double emulsions during their storage was evaluated by two methods. The first one was the comparative evaluation of droplet size (at room temperature and 4 • C) after 1 day and 35 days from emulsification in order to study the possible droplet growth. The second method was the backscattering (BS) evolution over time using Turbiscan Lab Expert apparatus to detect possible creaming, sedimentation, coalescence, flocculation or Ostwald ripening effect. Sedimentation, creaming and flocculation phenomena are unexpected, while those of coalescence and Ostwald ripening are more likely in the behavior of nanoemulsions [55]. Figure 3 shows the changes in droplet size between 1 day and 35 days of storage at 4 • C and room temperature. As it can be seen from Figure 3a, there is no considerable change in droplet size distribution of rotor-stator optimal W/O/W emulsion during the storage period while the intensity increased regardless of storage temperature condition. According to Figure 3b, which shows the droplet size distribution of emulsion prepared in ultrasonic homogenizer, average droplet size did not vary considerably but PDI was increased during the storage period. It is already mentioned the advantage of optimal W/O/W prepared by microfluidizer processor over other devices due to significantly better average droplet size. Furthermore, it showed the best droplet growth stability, as shown in Figure 3c: neither average droplet size nor PDI was changed after 35 days of storage at both 4 • C and room temperature. ζ-potential was also measured for this sample and the results indicated the stable value of −2.7 mV after 1 day and 35 days of emulsification.
Foods 2020, 9, x 12 of 19 for particle size and PDI, respectively. The narrow difference between the predicted and observed responses can also verify the reliability of the microfluidizer experimental design. Therefore, it can be concluded from the observed responses of optimal points from three emulsification methods that microfluidizer processor can be considered as the optimal method to achieve double emulsions with lower particle size and PDI.

Stability Evaluations During Storage
The stability of the optimized double emulsions during their storage was evaluated by two methods. The first one was the comparative evaluation of droplet size (at room temperature and 4 °C) after 1 day and 35 days from emulsification in order to study the possible droplet growth. The second method was the backscattering (BS) evolution over time using Turbiscan Lab Expert apparatus to detect possible creaming, sedimentation, coalescence, flocculation or Ostwald ripening effect. Sedimentation, creaming and flocculation phenomena are unexpected, while those of coalescence and Ostwald ripening are more likely in the behavior of nanoemulsions [55]. Figure 3 shows the changes in droplet size between 1 day and 35 days of storage at 4 °C and room temperature. As it can be seen from Figure 3a, there is no considerable change in droplet size distribution of rotor-stator optimal W/O/W emulsion during the storage period while the intensity increased regardless of storage temperature condition. According to Figure 3b, which shows the droplet size distribution of emulsion prepared in ultrasonic homogenizer, average droplet size did not vary considerably but PDI was increased during the storage period. It is already mentioned the advantage of optimal W/O/W prepared by microfluidizer processor over other devices due to significantly better average droplet size. Furthermore, it showed the best droplet growth stability, as shown in Figure 3c: neither average droplet size nor PDI was changed after 35 days of storage at both 4 °C and room temperature. ζ-potential was also measured for this sample and the results indicated the stable value of −2.7 mV after 1 day and 35 days of emulsification.
(a)  There is a direct relationship between the increased emulsion droplet diameter and creaming, which is the formation of a lipid rich cream layer on the liquid surface as a result of the generation of big lipid droplets from smaller ones due to weak steric repulsions [18,33]. The coalescence process mainly takes place after double emulsion production and is thus stronger influenced by geometrical parameters, like inner and outer droplet sizes and dispersed phase concentrations, than by process parameters [59]. Both phenomena negatively affect emulsion stability, a very important factor determining its shelf life in commercial food and beverage applications [33,70].
BS profiles monitored during 35 days at 25 °C are depicted in Figure 4. As it is shown in Figures  4a and 4b, creaming instability occurs in emulsification by rotor-stator mixer and ultrasonic homogenizer. In contrast, the optimal W/O/W emulsion prepared by microfluidizer processor, Figure  4c, showed no considerable destabilization during the storage period and the sample was also visually stable, maintaining the same bluish color and semitransparent appearance after 35 days.  There is a direct relationship between the increased emulsion droplet diameter and creaming, which is the formation of a lipid rich cream layer on the liquid surface as a result of the generation of big lipid droplets from smaller ones due to weak steric repulsions [18,33]. The coalescence process mainly takes place after double emulsion production and is thus stronger influenced by geometrical parameters, like inner and outer droplet sizes and dispersed phase concentrations, than by process parameters [59]. Both phenomena negatively affect emulsion stability, a very important factor determining its shelf life in commercial food and beverage applications [33,70].
BS profiles monitored during 35 days at 25 • C are depicted in Figure 4. As it is shown in Figure 4a,b, creaming instability occurs in emulsification by rotor-stator mixer and ultrasonic homogenizer. In contrast, the optimal W/O/W emulsion prepared by microfluidizer processor, Figure 4c, showed no considerable destabilization during the storage period and the sample was also visually stable, maintaining the same bluish color and semitransparent appearance after 35 days.
determining its shelf life in commercial food and beverage applications [33,70].
BS profiles monitored during 35 days at 25 °C are depicted in Figure 4. As it is shown in Figures  4a and 4b, creaming instability occurs in emulsification by rotor-stator mixer and ultrasonic homogenizer. In contrast, the optimal W/O/W emulsion prepared by microfluidizer processor, Figure  4c, showed no considerable destabilization during the storage period and the sample was also visually stable, maintaining the same bluish color and semitransparent appearance after 35 days.  The rotor-stator mixer sample happened to exhibit visual creaming instability after a few hours of storage. Nevertheless, double emulsion obtained by using ultrasonic homogenizer showed a slightly larger droplet size but visual instability was not observed after the same storage period. Similar results were obtained by Einhorn-Stoll et al. [71], who observed a rapid destabilization of emulsions prepared by a single step with the Ultra-Turrax homogenizer. However, both chemical analysis and physical appearance indicated slight levels of creaming after few days of storage ( Figure  4b). The shifting of the curves in Figures 4b and 4c could be attributed to the presence of air bubbles in cells or being a little shaken during each analysis.

Retention Properties of Nanoemulsions
As it was aforementioned the OMW used in this study was acidified to pH = 1.8 prior to membrane treatment in order to increase the content of selective bioactive compounds. The effect of pH on physical stability of different emulsions has been previously determined by several authors. Recently, the effect of pH on curcumin emulsions was investigated [72]. The authors found that more than 85% of curcumin was present after one month of storage at 37 °C when acidic conditions were employed. However, emulsions at pH 7.0, 7.4 and 8.0 contained only 62, 60, and 53% of the initial curcumin, respectively, thus demonstrating low stability. The effect of pH on the stability of emulsions stabilized by pectin-zein complexes has also been studied by Juttulapa et al. [73]: they found a greater cross-linking polymer network at pH 4 than pH 7, providing thus a smaller droplet size distribution.  The rotor-stator mixer sample happened to exhibit visual creaming instability after a few hours of storage. Nevertheless, double emulsion obtained by using ultrasonic homogenizer showed a slightly larger droplet size but visual instability was not observed after the same storage period. Similar results were obtained by Einhorn-Stoll et al. [71], who observed a rapid destabilization of emulsions prepared by a single step with the Ultra-Turrax homogenizer. However, both chemical analysis and physical appearance indicated slight levels of creaming after few days of storage (Figure 4b). The shifting of the curves in Figure 4b,c could be attributed to the presence of air bubbles in cells or being a little shaken during each analysis.

Retention Properties of Nanoemulsions
As it was aforementioned the OMW used in this study was acidified to pH = 1.8 prior to membrane treatment in order to increase the content of selective bioactive compounds. The effect of pH on physical stability of different emulsions has been previously determined by several authors. Recently, the effect of pH on curcumin emulsions was investigated [72]. The authors found that more than 85% of curcumin was present after one month of storage at 37 • C when acidic conditions were employed. However, emulsions at pH 7.0, 7.4 and 8.0 contained only 62, 60, and 53% of the initial curcumin, respectively, thus demonstrating low stability. The effect of pH on the stability of emulsions stabilized by pectin-zein complexes has also been studied by Juttulapa et al. [73]: they found a greater cross-linking polymer network at pH 4 than pH 7, providing thus a smaller droplet size distribution.
Chemical stability of W/O/W emulsion was evaluated by measurements of retention levels of total phenolic content (TPC) and antioxidant activity (AA) after 35 days of storage at room temperature. TPC and AA measurements were done on the W/O/W double emulsion using the optimal formulation prepared by microfluidizer processor at 148 MPa and 7 cycles. Polyphenols are very sensitive compounds that can be easily degraded over time; therefore, their encapsulation in suitable double emulsions is an effective solution to prevent the degradation of phenolic compounds at a satisfactory level. The phenolic rich inner aqueous phase used in this study (after membrane treatments and before emulsification) had a TPC of 1399.8 ± 17.9 mg GA/L and 286.5 ± 0.3 mg Trolox/L of AA. After 35 days of storage at room temperature, 68.6% and 89.5% retentions of TPC and AA, respectively, were preserved.
In related studies, Akhtar et al. [74] reported 72% rutin and anthocyanin flavonoids retention after 10 days of storage in a W/O/W nanoemulsion system, but a polymodal droplet size distribution and inner phase leakage were observed. Gomes et al. [75] studied the retention of gallic acid in W/O and O/W emulsion systems with soybean oil, obtaining a 15% reduction in gallic acid content after 7 days of storage. Mohammadi et al. [22] reported 22% of phenolic compounds release by preparing double emulsions stabilized only with whey protein concentrate (WPC) containing olive leaf extract after 20 days. Furthermore, in the study performed by Gadkari et al. [18] it was noted that when the emulsions were stored at temperatures of 277 K, 300 K and 310 K, the green tea polyphenols present in emulsions were degraded by 4.25%, 15.97% and 22.78%, respectively. It can be concluded that the results obtained in the present study are in the usual retention range of phenols encapsulated in emulsions with applications for the food industry.

Conclusions
Response surface methodology was applied in this work to determine both the optimal composition and operating conditions for preparation of W/O/W nanoemulsion loaded with phenolic rich inner aqueous phase from olive mill wastewater. The optimal formulation for primary W/O emulsion was 20.3% (w/w) of phenolic rich aqueous solution as dispersed phase and 3.7% (w/w) of Span 80 in MCT oil as continuous phase. The optimal composition for W/O/W emulsion was 9.8% (w/w) of W/O as dispersed phase and 4.1% (w/w) of Tween 80 in Milli-Q water as external phase. Three methods were tested to obtain the optimal emulsification conditions: mechanical homogenization (rotor-stator mixer), ultrasonic homogenization and microfluidization. Optimum results were achieved by microfluidization at 148 MPa and 7 cycles input levels, obtaining a W/O/W nanoemulsion with an average droplet diameter of 105.3 ± 3.2 nm and a polydispersity index of 0.233 ± 0.02. Samples obtained by rotor-stator mixer and ultrasonic homogenizer showed creaming instability after few days of emulsification, while emulsion obtained by microfluidization showed no droplet size growth or changes in stability and ζ-potential after 35 days of storage at 25 • C. Furthermore, it showed a satisfactory level of phenolics retention (68.6%) and antioxidant activity (89.5%) after 35 days of storage, which are suitable for application in the food or pharmaceutical industry.