of Chemical Engineering HOW TO FORMULATE A STABLE AND MONODISPERSE WATER-IN-OIL NANOEMULSION CONTAINING PUMPKIN SEED OIL : THE USE OF MULTIOBJECTIVE OPTIMIZATION

The multiobjective optimization method was applied in order to improve the droplet size distribution and stability of water-in-oil emulsions composed of sunflower and pumpkin seed oils as continuous phase, polyglycerol polyricinoleate as emulsifier, water as dispersed phase and sodium chloride as co-stabilizer (lipophobe). Three composition factors were varied based on the three level Box-Behnken design and three characteristics of the obtained emulsions were measured for each experimental run. The mean volume diameter of water droplets and the span of the droplet size distribution, both determined immediately upon preparation of the emulsion, as well as the stability index over a three-month period were interrelated by regression functions with the surfactant concentration, oil composition and the salt content in the water phase of the emulsion. Also, the fourth objective function based on a difference in the prices of pumpkin seed and sunflower oils was considered for optimization. The multiobjective optimum was calculated by using the minimal loss method with weight factors. Additionally, effects of the continuous phase composition and the salt content on the equilibrium interfacial tension of water-oil systems and the changes of the droplet size distribution over time were studied.


INTRODUCTION
Tasty, healthy, nutritious and more convenient food products with enhanced stability and shelf life are imperative for today's increasingly demanding markets.The production of water-in-oil (W/O) and water-in-oil-in-water (W/O/W) emulsions is one possible step towards employing the novel idea that biologically active substances and active ingredients in the food industry should be entrapped in some carrier material to form microcapsules or nanoparti-cles in order to achieve the controlled release of active ingredients and flavour retention, to mask the bad taste or smell of some components, stabilize food ingredients, prevent their oxidation or hydrolysis, and adjust their properties and/or increase their bioavailability (Nikolovski et al., 2011).
Different dispersing methods have been employed to generate emulsions such as conventional simple agitation, colloidal mills, static mixers and high shear mixers, as well as novel methods like membrane emulsification (Dragosavac et al., 2012, Brazilian Journal of Chemical Engineering Vladisavljević and Williams, 2005) and ultrasound cavitation (Sivakumar et al., 2014, Tang et al., 2013, Tang et al., 2012).For production of nanoemulsions intense shear should be applied in order to overcome the Laplace pressure and break up droplets into smaller (nanometre scale) dimensions (Sivakumar et al., 2014).The developed high-energy input techniques adequate for production of nanoemulsions include the use of high-pressure homogenizers, ultrasonicators and microfluidizers (Sivakumar et al., 2014, Landfester 2006).Also, low-energy input techniques adequate for the production of nanoemulsions have been developed such as phase inversion temperature, solvent-diffusion and spontaneous emulsification (Sivakumar et al., 2014).However, low-energy input techniques have their own limitations (Sivakumar et al., 2014) including the use of a large quantity of surfactant, usually not of the food grade type, and instability after long-term storage, which can be improved when the droplet disruption is provided predominantly by high-energy input techniques (Santana et al., 2013, Sivakumar et al., 2014, Tang et al., 2013).
The stability problems in food emulsions cannot be improved by increasing the concentration of the emulsifying agent (like in cosmetic and pharmaceutical emulsions) due to limitations in the permitted dose for human consumption (Dickinson, 2011, Jiménez-Colmenero, 2013).The polyglycerol ester of polyricinoleic acid (PGPR) (low HLB value) is a synthetic non-ionic and the most effective hydrophobic emulsifier, commonly used in the food industry as a chocolate thickening agent with excellent waterbinding characteristics (Gülseren andCorredig, 2012, Wilson et al., 1998a).Used as a food additive, it is recognised as generally safe with a maximum per capita mean daily intake of 2.64 mg/kg body weight/day (Wilson et al., 1998b).
Although PGPR reduces the interfacial tension very well, thereby facilitating droplet break-up, and prevents coalescence of newly formed water droplets via the Gibbs-Marangoni effect (Walstra, 1993) and by steric stabilisation (Landfester, 2006), diffusional degradation remains the destabilising mechanism, which has to be precluded by osmotic pressure regulation.There are claims that the presence of salt is crucial for emulsion formation and the stability of primary W/O emulsions (Aronson and Petko, 1993).Despite the fact that some authors claim that a stable emulsion can be obtained without electrolytes in the emulsion and/or that the addition of salt (NaCl) or sodium phosphate buffer destabilizes the emulsion (Su et al., 2008), it is well known that the stability rating of primary W/O emulsions is markedly affected by the addition of electrolytes in the inner aqueous phase (Moguet et al., 2001, Srinivasan et al., 2000).Actually, salt is considered to be a costabilizer, a lipophobe that builds-up the osmotic pressure to counterbalance the Laplace pressure, and consequentially stabilizes emulsions against diffusional degradation known as Ostwald ripening (Capek, 2010, Colmán et al., 2014, Landfester, 2006).Therefore, the addition of an osmotic agent that cannot interdiffuse between two droplets, and the use of an appropriate hydrophobic surfactant, could both be essential for preparation of stable and monodisperse water-in oil emulsions, with droplet sizes ranging from 50 to 500 nm, well known as inverse miniemulsions (Landfester, 2000, Landfester, 2003).For inverse miniemulsions, relations between surfactant content and droplet size, as well as particle size and the coverage of the particles by surfactant would additionally depend on the amount of the osmotic agent (Landfester, 2006).
The chemical structure of the oil phase, i.e., the chain length of the fatty acids, molecular configuration and the number of unsaturated bonds, is crucial for the stability of the emulsion.Particularly, the polarity of the oil phase can affect the interfacial tension of the W/O interface and the allocation of the components at the interface (Ushikubo and Cunha, 2014).W/O and W/O/W emulsions were usually prepared with sunflower, corn, soybean oil, canola, olive and rapeseed oils (Jiménez-Colmenero, 2013), with olein and miglyol (Bonnet et al., 2009), and sometimes with specialty oils like Moringa oleifera oil (Khalid et al., 2013).
On the one hand, this work aimed to give an insight into an inverse miniemulsion system composed of two oils with different physicochemical characteristics (refined sunflower and unrefined pumpkin seed oil) as continuous phase, polyglycerol polyricinoleate as emulsifier, water as dispersed phase and sodium chloride as co-stabilizer.The influence of three composition factors on three characteristics of the obtained W/O emulsions was studied.On the other hand, optimization of the formulation of this system was attained by the use of a multiobjective optimization technique involving the Box-Behnken experimental design, which is considerably cheaper than the three-level full factorial designs and is considered to be very efficient, when efficiency is estimated as the number of coefficients in the estimated model divided by the number of experiments (Ferreira et al., 2007).

MULTIOBJECTIVE OPTIMIZATION
Since we needed to take into account and eventually simultaneously optimize three different quality parameters of emulsions: the mean volume diameter of water droplets, the span value of the droplet size distribution and the stability index of the emulsions that were studied over a three-month period, it was necessary to resort to some type of multiobjective optimization technique.Our aims had different dimensions, different orders of magnitude and might potentially be of different importance in the decision making.
Multiobjective optimization was performed in three steps.Firstly, each aim had to be described mathematically by its unique objective function f i = (X 1 , X 2 , X 3 ) (i=1, 2 and 3).However, the three aforementioned aims depended on the same independent variables/ factors -in our case polyglycerol polyricinoleate (X 1 ) and pumpkin seed oil (X 2 ) contents in the oil phase and sodium chloride (X 3 ) content in the water phase of the emulsion.In order to calculate the multiobjective optimum by taking into account all the aims simultaneously, some compromises had to be made.Therefore, in the second step, the importance of the objectives was considered by defining the weighting factors (w i ).The weighting factors influence the minimum of the loss function and the set of compromise optimum values of the independent variables X 1,opt , X 2,opt , and X 3,opt .As the third step, the multicriterion optimization method had to be chosen.For our purposes we applied the loss-minimization method (Osyczka, 1984, Gergely et al., 2003).This method calculates the minimum of the sum of the weighted relative deviations: where L is the loss function, and * i f individual optimum value of the i objective function.
A constrained minimization of the multivariate scalar function (L) was done by using the Scipy.optimizemodule of The Scientific Computing Tools for Python (ScyPy 0.14.0.).The algorithms that gave good results were: L-BFGS-B (Limited-memory Broyden-Fletcher-Goldfarb-Shanno Bound-constrained) algorithm, the Constrained Optimization BY Linear Approximation (COBYLA) method, and a Truncated Newton Conjugate-gradient (TNC) algorithm.

Emulsion Preparation
Water-in-oil (W/O) emulsions were prepared by homogenization using a high speed homogenizer (UltraturraxT-25, IKA, Germany) at 24000 rpm for 10 min.Continuous phases were prepared by dissolving a certain amount of PGPR (1, 3 and 5% (w/w)) in an oil phase (the sunflower oil, the pumpkin seed oil and a mixture of the oils (mass ratio 1:1)) at 50 °C, by mixing on a magnetic stirrer for 30 minutes.Final 20% (v/v) W/O emulsions were prepared by careful addition (drop by drop) of the dispersed phase (water and water solutions of NaCl (0.15 and 0.3 M) into the continuous oil phase stirred by the high-speed homogenizer.The emulsification temperature was maintained at 25 °C by means of a water bath.After preparation all emulsions were initially analysed and then stored in a refrigerator at 4 °C.

Sedimentation Stability
For the stability test, the prepared emulsions were transferred into 10ml graduated glass cylinders and stored at room temperature for three months.During storage, the emulsions separated into an opaque layer of emulsion and a transparent serum layer consisting of oil at the top or water at the bottom of the cylinders.The stability index (the creaming stability) was measured by the height of the serum layer (H S ) with the storage time.The stability index (SI) (Perrechil et al., 2014) was reported as: where H E represents the initial height of the emulsion.

Droplet Size Analysis
The mean droplet size and droplet size distribution of the primary W/O emulsions were measured immediately after the formation using a laser light scattering instrument ZetasizerNano ZS (Malvern Instruments, U.K.).To avoid multiple light scattering, samples of the emulsions were diluted with sunflower oil at a dilution ratio of 1:100, and analysed in triplicate.The optical parameters selected were: dispersed phase refractive index 1.33, dispersant liquid viscosity (sunflower oil) 51.32 mP•s and refractive index 1.4723.The stability of W/O emulsions was determined weekly during a month of storage.
The results of the measurements are shown as the droplet size distribution and the volume-weighted mean droplet diameter, d 4,3 , given by Eq. ( 3): where n i is the number of droplets with diameter of d i .
The width of the droplet size distribution and polydispersity are expressed through the span value, defined by Eq. ( 4): ( ) where d 10 , d 50 and d 90 are standard percentile readings from the cumulative droplet volume distribution curve, meaning the droplet diameters below which 10%, 50% and 90% of the sample lies, respectively (Ushikubo and Cuncha, 2014).

Surface/Interfacial Tension
A digital tensiometer KSV -Sigma 703D (Finland) was used and the Du Noüy ring method was employed for interfacial tension measurements between the water and oil phases.Before the measurement, the ring was immersed in the water phase, the oil phase was added slowly and the surface was left for 10 min to equilibrate.The reported values of the interfacial tension were the average values of at least three measurements.All measurements were performed at 25 °C.

Viscosity Measurements
An RV20 rotational viscometer (cone plate geometry) with a SVI measuring sensor (Haake, Germany) was used for viscosity measurements of the oil phase.The samples were transferred to the instrument and allowed to equilibrate to 25 °C for 5 min prior to measurement.Shear stress τ (Pa) was determined with continually changing shear rates D (s -1 ) from zero to 500 s -1 and the reverse.The apparent viscosity was calculated as:

Statistical Analysis
Multivariate optimization schemes involve designs for which the levels of all the variables (the significant factors) are changed simultaneously.The optimum operational conditions are attained by using more complex experimental designs such as the three-level Box-Behnken design, which has proven to be slightly more efficient than other experimental designs in use (Ferreira et al., 2007).
The Box-Behnken designs are a class of rotatable or nearly rotatable second-order designs based on three-level incomplete factorial designs.The required number of experiments (N) is defined as where k is number of factors and C 0 is the number of central points) (Ferreira et al., 2007).The Box-Behnken designs give the possibility to estimate all linear effects, all quadratic effects, and all linear 2-way interactions between factors.
In order to investigate the effect of emulsion composition parameters on the droplet size distribution and the sedimentation stability, the Box-Behnken design for three factors was applied.The three input variables were the content of PGPR in the continuous phase (1-5% (w/w), X 1 ), pumpkin seed oil content in the continuous phase (0-100% (w/w), X 2 ) and NaCl concentration in water phase (0-0.3M, X 3 ), where three levels were chosen (−1, 0, +1) as shown in Table 1.The experimental design consisted of 15 runs as shown in Table 2.The experimental values were expressed as the means of three determinations and the standard deviation.All statistical analyses were performed using STATISTICA 12 software (StatSoft, Inc., 2012).

RESULTS AND DISCUSSION
Varying PGPR and PO content in the oil phase and NaCl content in the water phase, thirteen W/O emulsions of different formulations were prepared and investigated in a such a way as to determine the mean volume diameter (d 4,3 ) and span value of the freshly prepared emulsions and the stability index over a three-month period.All emulsions in the experimental design contained 20% (v/v) of water and the results obtained are shown in Table 2.The stability index determined after 90 days of storage was labelled as SI90. the water molecules associated with the polar parts of phospholipids and PGPR, changing the aggregation state of phospholipids, modifying the interactions between phospholipids and PGPR and thus affecting the properties of the adsorbed layer (Dedinaite and Campbell, 2000).The addition of PGPR in a range between 4% and 5% did not produce a significant change of interfacial tension.Table 5 indicates that the addition of NaCl slightly reduces interfacial tension between the water and oil phases in the presence of PGPR from 5.3±0.12 to 4.4±0.22mN•m -1 ; nevertheless, it provides better stability of the emulsion during storage.Very stable droplets could be formed shortly after preparation of inverse miniemulsions due to the fact that the added salt enables the occurrence of a real zero-effective-pressure situation (i.e., the osmotic pressure counterbalances the Laplace pressure) (Landfester, 2006).This study confirms the hypothesis of Pawlik et al. (2010) that the addition of salt in the water phase strengthens the interaction between adsorbed molecules, and provides better packing of the PGPR in the interfacial layer.Therefore, increasing the elasticity of the layer decreases the interfacial mobility and the rate of film drainage between approaching droplets, leading to increasing emulsion stability by "stiffening" the interface (Lutz et al., 2009).Moreover, the typical triangular relation between the amount of surfactant, resulting particle size, and surface coverage, existing as a result of droplet break-up and the recoalescence mechanism of minidroplet formation was recognized to depend on the amount of added salt (Landfester, 2006).
Response surface methodology was applied in order to obtain relationships between the emulsion quality parameters (d 4,3 , span and SI90) and the composition of W/O emulsions (X 1 , X 2 , and X 3 ).The proportion of variance accounted for by the whole model, with 9 degrees of freedom, for d 4,3 , span and SI90 was 0.994 (adjusted r 2 =0.982), 0.958 (adjusted r 2 =0.881) and 0.979 (adjusted r 2 =0.942), respectively.However, all parameters of the whole model were by no means statistically significant, and the whole model was reduced to obtain regression functions with all the significant parameters.The developed regression models for d 4,3 , span and SI90 and the coded values of the independent variables X 1 , X 2 , and X 3 and their interdependence are shown in Eq. ( 6), ( 7) and ( 8), respectively.The values of the appropriate coefficients of the regression models are listed in Table 4.
Table 4: The values of the coefficients in Eq. ( 6), ( 7) and ( 8) and the proportion of variance accounted for by the models (r 2 ).mean volume diameter of water droplets, the span of the droplet size distribution and the sedimentation stability index after 90 days of storage were 155.8±12.3nm, 0.70 and 5.5%, respectively, and corresponded well to the values calculated by the objective functions.Droplet size distribution of the emulsion after six months of storage was recorded and confirmed good stability of the optimized formulation, which is intended for preparing double W/O/W emulsions.