The raw materials used were buriti oil (Amazon Oil Industry^TM, Ananindeua, Pará, Brazil), soy protein isolate (SPI) (Tovani Benzaquen Ingredientes^TM, São Paulo, São Paulo, Brazil) and high-methoxyl pectin (HMP) (CP Kelco^TM, Matão, São Paulo Brazil). Preparation of samples required deionized water, 1 M solutions of sodium azide, hydrochloric acid or sodium hydroxide (all Dinâmica^TM, Indaiatuba, São Paulo, Brazil) and McIlvaine buffer (pH=3.5) prepared using disodium phosphate and citric acid (Dinâmica^TM).

Oil-in-water emulsions were prepared with different mass fractions of oil (10-30%) and SPI in the continuous phase (50-80%), and different applied homogenization pressures (200·10⁵-400·10⁵ Pa). These were the three independent variables. The HMP content was not included as an independent variable in the experimental design since it was added in the amount necessary to attain a fixed total mass fractions of biopolymers (SPI+HMP) in the continuous phase of the emulsion.

All the emulsions were prepared using McIlvaine buffer at pH=3.5, which is below the isoelectric point of SPI and allows the biopolymers in the solution to acquire opposite charges in a sufficient amount to form electrostatic complexes that remain dispersed in the aqueous phase (13). Stock solutions of SPI (3 g/100 g) were prepared by dispersing the protein in half of the required water and adjusting pH=11.0±0.05 with NaOH for complete solubilization, then subjecting them to magnetic stirring for 3 h, leaving to rest overnight (25 °C) for complete hydration, and finally completing the water content with buffer pH=3.5. Stock solutions of HMP (2 g/100 g) were prepared by dispersion in buffer solution at pH=3.5, then 3 h under magnetic stirring, and overnight hydration at 25 °C. Sodium azide 0.04% was added to stock solutions to prevent microorganism growth (13).

Emulsions were produced by following the method described by Kaltsa et al. (27), with some modifications. Buriti oil was dispersed in the SPI stock solution using Ultra Turrax stirrer (T-25; IKA, Staufen, Baden-Württemberg, Germany) at 15 000 rpm for 4 min, followed by the addition of HMP stock solution, and dispersion at 15 000 rpm for 4 min using the same stirrer. After this procedure, the coarse emulsion was homogenized at high pressure in a two-stage homogeniser (Lab Series APV-2000; SPX Flow Technology, Charlotte, NC, USA), without sample recirculation.

The assays were carried out according to a 2³ rotatable central composite design (RCCD) to analyze the effects of three independent variables (28). Five assays were added at the central point of the experimental design, totalizing 19 assays (Table 1). The following polynomial model was fitted to the data:

where y is the response variable, x₁ is the oil mass fraction (m(oil)/m(emulsion))/%, x₂ is the SPI mass fraction (m(SPI)/m(SPI+HMP))/%, x₃ is the homogenization pressure (Pa), and β_i and β_ij are the fitted coefficients.

The results were subjected to analysis of variance (ANOVA) and the fitted model was tested regarding the lack of fit and coefficient of regression using the software STATISTICA v. 7.0 (29). For all fitted models, the coefficient of regression was significant, whereas the lack of fit was not significant, suggesting good fit and accuracy of the models.

The emulsions were stored at 25 °C for a 7-day period after production. Then they were analyzed in triplicate regarding stability, droplet size, electrical conductivity, electrical charge and morphology using optical microscopy and scanning electron microscopy (SEM) and rheological behaviour. Confocal laser scanning microscopy (CLSM) was performed on the same day of emulsion preparation with fluorescent dyes to avoid fluorescence losses. For samples that show phase separation after 7 days, the analyses were performed using the samples from the creamed phase. The responses for the experimental RCCD were stability, droplet size, electrical conductivity and electrical charge. The analytical procedures are described as follows.

Experimental validation of the models resulting from the RCCD (Eq. 1) was carried out based on the results of stability, droplet size, electrical conductivity and electrical charge. Selected values for buriti oil and SPI mass fractions, and homogenization pressure were applied to produce an emulsion and the corresponding experimental results were compared with those predicted by the fitted models. In addition, this emulsion was characterized by rheological and morphological analyses (as described above).

The creaming index was significantly affected (p≤0.10) by the mass fractions of oil (β₁) and SPI (β₂), and homogenization pressure (β₃). In addition, there were significant interactions between the oil and SPI mass fractions (β₁₂), as well as between SPI mass fraction and homogenization pressure (β₂₃) (Table 2). The oil content and homogenization pressure had negative effects, whereas SPI content had a positive effect on creaming index, being important to remember that the higher the CI, the higher the emulsion instability.

The increase of emulsion stability with increasing concentration of the dispersed phase is probably related to the higher viscosity associated with the increased concentration of emulsion droplets (19). In addition, the buriti oil used is a crude oil and has about 12% of diacylglycerols and monoacylglycerols, compounds that are recognized by their emulsifying properties (24). In such a way, it is possible to suggest that the increase in oil mass fraction leads to an increasing content of these compounds, which could contribute to lower CI values, or higher emulsion stability. On the other hand, lower SPI mass fractions in the continuous phase also imply higher HMP mass fractions (Table 1), which suggests that a minimum content of HMP is important to increase the continuous phase viscosity and prevent oil droplet diffusion, which could contribute to increase in the stability and lower CI values (38). In turn, higher homogenization pressures may contribute to the production of smaller oil droplets, resulting in a more homogeneous and, consequently, more stable system, with lower creaming index (27).

With respect to droplet size measured by OM, the variables with significant effects (p≤ 0.10) were the oil mass fraction (linear, β₁, and quadratic, β₁₁), homogenization pressure (β₃ and β₃₃), the interactions between the oil and SPI mass fractions (β₁₂) and between the SPI mass fraction and homogenization pressure (β₂₃). The oil mass fraction and the interaction between mass fractions of oil and SPI had positive effects on droplet size, while the homogenization pressure and the interaction between the SPI mass fractions and homogenization pressure had negative effects. This suggests that high HMP levels helped to stabilize the emulsions, although the smaller amount of SPI was not sufficient to produce emulsions with small droplets, since a larger amount of proteins would be needed to cover the larger interfacial area.

The droplet sizes determined by LD were represented by the distribution model and were significantly affected (p≤0.10) by the mass fractions of oil (β₁₁) and SPI (β₂), homogenization pressure (β₃ and β₃₃) and by the interaction between the oil and SPI mass fractions (β₁₂). The SPI content and homogenization pressure had positive effects, whereas the oil content, homogenization pressure, and the interaction between the oil and SPI mass fractions had negative effects. It is interesting to note the large difference between the results for droplet size determined by OM and LD. Differences could also be observed in the variables that exerted significant effects on these results. The laser diffraction technique applied requires sample dilution, which does not happen in the optical microscopy technique. This fact led to the hypothesis that the dilution was not enough to completely disperse the oil droplets, which remained arranged in clusters. This hypothesis was confirmed by observing the emulsions at the same dilution in the optical microscope. Despite the observed differences, the results from LD are relevant, since this is the standard technique used for droplet size analysis in emulsions (32, 39, 40).

The electrical conductivity was significantly affected (p≤0.10) by the mass fractions of oil (β₁) and SPI (β₂) and homogenization pressure (β₃). The oil and SPI mass fractions had negative effects and the homogenization pressure had positive effect on the electrical conductivity of emulsions or on cream phases. These results can be explained considering that the systems with higher SPI mass fractions had lower HMP mass fraction and, that pectin helps to maintain oil droplets apart from each other. Therefore, in these systems there was a great possibility that oil droplets were flocculated, or even merged (coalescence) in the creamed phase, leading to lower electrical conductivity values. On the other hand, in emulsions subjected to higher homogenization pressures, smaller oil droplets and greater homogeneity are expected, thus, resulting in higher electrical conductivity values.

According to Bruttel (41), in oil-in-water emulsions, water has a high dielectric constant and is very polar, whereas the oil has a low dielectric constant and thus low polarity. When the water phase is bound to stabilizers, that is, the emulsion has maximum viscosity, its stability will also be maximum and, due to this fact, the electrical conductivity is minimal.

The zeta potential, or electrokinetic potential, is related to the electrophoretic mobility, gives a measure of the net charge on the surface of a macromolecule, and is used to describe the surface charge of polyelectrolytes of dispersed droplets. Like the electrical conductivity, zeta potential is related to emulsion stability. The more charged the droplets, the greater is the repulsion among them, which contributes to a more stable suspension. In order to prevent irreversible oil droplet flocculation, an emulsion should have a zeta potential greater than 25 mV (in modulus), characterizing a metastable system (42). All the emulsions prepared according to the RCCD had negative zeta potential, varying in the range of (-22.33±0.70) to (-14.60±1.74) mV (Table 2). At pH=3.5, SPI is positively charged and HMP has a slightly larger negative charge than SPI (13), thus suggesting that the presence of pectin was determinant for the net charge of emulsions, reinforcing the hypothesis that protein acts at oil/water interface and pectin, in addition to interacting with protein, covers droplets as a second layer, also acting as a stabilizer in the continuous phase (4).

Stable emulsions (O26S56P240, O26S56P360 and O30S65P300) showed high chargemodulus (Table 2). However, these values were lower than 25 mV and this may be an indication that these emulsions would not remain stable for a long time. The homogenization pressure might have contributed to this lower negative charge, since the smaller droplets resulted in larger surface area covered by SPI and available to interact with HMP. This greater interaction might contribute to further neutralization of available negative charges. In agreement with the above discussion, the variables with significant effects (p≤0.10) on zeta potential were SPI mass fraction (β₂) and homogenization pressure (β₃₃), both having positive effects on the zeta potential, i.e. increasing SPI mass fraction and homogenization pressure caused lower charge modulus.

In order to produce a stable emulsion with high carotenoid and tocopherol contents, which corresponds to a higher buriti oil content, with reduced size droplets, low electrical conductivity, and high electrical charge, the optimal values for the independent variables were fixed at 28% buriti oil, 55% SPI, and homogenization pressure of 380·10⁵ Pa.

The observed results, as well as those predicted by the regression models (Table 2) for the validation assay are shown in Table 4, as well as the fitting and relative errors. The obtained results were consistent with those expected and, due to proximity of the experimental and predicted results, the relative errors were low, suggesting that the experimental design was adequate to obtain predictive models for the production of buriti oil emulsions.

The images obtained by optical microscopy, SEM and CLSM allowed to observe that the optimized buriti oil emulsion morphology is small and less flocculated.

The pseudoplastic behaviour of this system was confirmed by the good fitting of the Sisko viscosity model. The mechanical spectra revealed that the sample was located in the viscoelastic plateau region and its microstructure revealed weak gel characteristics.

The results obtained in the present work may be compared to similar investigations reported in literature. Goyal et al. (50) developed a stable flaxseed oil emulsion with 12.5% oil, 7.5% WPC-80 and 10% lactose (filler), homogenized at 207·10⁵ Pa, which was suggested to be potentially applied as ω-3-fatty acid delivery system in functional foods such as pastry, ice-creams, curd, milk, yogurt, cakes, etc. Djordjevic et al. (51) studied physical stability of whey protein‐stabilized O/W emulsions at pH=3 using corn oil, and concluded that the emulsions could be used to incorporate functional lipids that are sensitive to oxidation. In addition, Nikovska (52) studied the oxidative stability of walnut oil and O/W emulsions containing walnut oil stabilized by SPI and whey protein isolate, observing that the emulsions were more stable than pure walnut oil. Based on these observations, the next steps of the present study should focus on the investigation of chemical composition and chemical stability of carotenes and other relevant constituents of buriti oil emulsions produced applying the optimized formulation developed in this work.