Olive fruits, Olea europaea L., were hand-picked from a traditional grove in Sierra Mágina (Jaén, Spain). The cultivated variety is Picual with a 4.9 maturity index, determined according to Uceda and Frias and described by Espínola et al. (2009); 486 g/kg moisture content, determined by drying milled paste at 105 ºC, and a 284 g/kg oil content, determined by the Soxhlet method. The olives were collected from unirrigated land.

Oils were obtained under laboratory-scale conditions using an Abencor centrifugal system (Abencor analyzer, MC2, Ingeniería y Sistemas S.L., Seville, Spain) (Espínola et al., 2011). The oils obtained were decanted into a graduated test tube for at least three hours, paper filtered and stored in amber glass bottles, under N₂ atmosphere, at ‑18 ºC until they were analyzed.

The method proposed by the International Olive Council (COI/T.20/Doc No 29) was used to determine the phenolic compounds present in virgin olive oils via High Performance Liquid Chromatography (HPLC). The equipment used was a liquid chromatograph (Shimadzu Corp., Kyoto, Japan) with the essential components: an elution pump (model LC-20AD), solvent degasser (model DGU-20A5), a refrigerated automatic injector (model SIL-20ACHT), a column oven (model CTO-10AsvpC), a diode array detector (model SPD-M20) and LC LabSolutions V.5.42.SP3 software (Shimadzu Corp., Kyoto, Japan). The column used was BDS Hypersil C18 (Thermo Scientific, USA), the particle size was 5 μm, the column size was 25 cm and the internal diameter was 4.6 mm. The mobile phase was a ternary gradient made up by orthophosphoric acid-water to 0.2% (A), Methanol (B) and Acetonitrile (C). The proportions of the phases are changed over time by means of a flow ramp. The initial proportion of phase A was 96%, B and C were 2%. At minute 40, the proportion of phase A was 50%, B and C were 25%. At minute 45, the proportion of phase A was 40%, B and C were 30%. At minute 60, the proportion of phase A was 0%, B and C were 50%. From minute 72 to 80 the proportions of the phases were identical to the initial ones; minute 80 was the end of the chromatogram. The elution flow was 1 mL/min. The oven temperature was set at 30 ºC and the injected volume of sample was 20 μL. The detector UV provided a signal at 280 nm. The phenolic compounds were quantified through the addition of syringic acid as internal standard and tyrosol as external standard. The results obtained were expressed as mg of tyrosol per kg of oil. The phenolic compounds were identified through comparison with the following analytical standards: Syringic acid, vanillin, luteolin, vanillic acid, oleuropein, trans-ferulic acid and trans-cinamic acid, all purchased from Fluka (Milan, Italy). Tyrosol, pinoresinol, caffeic acid, apigenin and p-coumaric acid were purchased from Sigma-Aldrich (St. Louis, MO, USA). o-coumaric acid and hydroxytyrosol were purchased from Extrasynthese (Genay Cedex, France). 3,4‑DHPEA‑EDA (oleacein) and p‑HPEA-EDA (oleocanthal) were identified by analytical standards supplied by the Department of Organic Chemistry from the University of Jaén. p-HPEA-EA and 3,4‑DHPEA-EA were identified using the method proposed by COI, Determination of bio-phenols in olive oils by HPLC, (COI/T.20/Doc No 29).

The Folin-Ciocalteau reagent was used to determine the total quantity of phenolic compounds in the olive oil, with slight adjustments (Vázquez-Roncero et al., 1973). The weight of the oil samples was 1 g. The sample was placed in a tube and dissolved in 5 mL of hexane. The liquid-liquid extraction was performed by repeating contact with a 2 mL methanol-water solution (60:40). This extraction was repeated three times. Each contact was shaken for 2 min with a Vortex type agitator. The methanolic extracts were collected in a test tube of 10 mL. To conclude, the methanol-water solution was added to complete 10 mL. The methanolic extracts were reacted with the Folin-Ciacolteau reagent. The absorbance of the solutions was measured at a wavelength of 725 nm. The equipment used was a UV spectrophotometer, model Shimadzu UV-Spectrophotometer 1800 (Kyoto, Japan). A standard curve was used to determine the quantity of phenolic compounds, using caffeic acid as external standard. The results were expressed as mg of caffeic acid per kg of olive oil.

The antioxidant potential was determined as DPPH free radical scavenging activity. A DPPH stock solution (0.1 mmol/L) was prepared in methanol, and further diluted to 1.0 AU at 515 nm prior to use. Aliquots of extract (20-2000 mL) were adequately diluted to a volume of 200 mL and added to 1.5 mL of DPPH solution cuvettes, shaken and kept in the dark for 60 min, and then the absorbance was measured at 515 nm using methanol as a blank. The percentage of scavenged DPPH radicals was calculated according to Equation 1:

where A₀ and A_sample stand for the absorbance of the control and sample, respectively. The percentage of inhibition was converted into antioxidant activity by using Trolox as standard antioxidant.

Headspace solid-phase micro extraction (HS-SPME) and the gas chromatography-flame ionization detector (GC-FID) technique were used for the analysis of volatile compounds.

Two sample grams were placed in a 20 mL amber glass vial tightly capped with polytetrafluoroethylene (PTFE)/silicone septum and a magnetic cap. The vial was heated up to 40 ºC for 10 minutes to reach the equilibration of volatile compounds in the headspace. Afterwards, the SPME needle was inserted through the septum and the fiber was exposed for 40 min. The SPME fiber (2 cm length and 50/30 μm film thickness), purchased from Supelco (Bellefonte, PA, USA), was composed of Carboxen/DVB/polydimethylsiloxane (PDMS). Previously, the fiber had been conditioned following the instructions of the supplier.

The GC-FID analysis was performed using a gas chromatograph model 7890B (Agilent Technologies, Santa Clara, CA, USA). The gas chromatograph was equipped with a split/splitless injector and a flame ionization detector. The volatile compounds adsorbed in the fiber were desorbed into the injector port for 1 min in splitless mode. The DB-WAXetr polyethylene glycol capillary column (30 m length, 0.25 mm internal diameter, 0.25 μm coating) (Agilent Technologies, USA) was used for the chromatographic separation. The carrier gas was helium at a flow rate of 1 mL/min. The injector temperature was 260 ºC and the detector temperature was 280 ºC. The oven temperature was initially 40 ºC for 10 min. Afterwards, the temperature was increased with a ramp of 3 ºC/min up to 160 ºC and immediately increased with a ramp of 15 ºC/min up to 200 ºC and held for 5 min to the end. The integrations were performed with Agilent OpenLAB ChemStation C.01.06 Software (Agilent Technologies, CA, USA).

The chromatographic peaks were quantified by the Internal Standard Method. 4-Methyl-2-pentanol was the internal standard and each and every one of the compounds was used as external standard. The results obtained were expressed as mg of each standard compound per kg of oil.

For the analysis of volatile compounds, 39 analytical standards were used: acetic acid, trans-2-pentenal, 1-penten-3-one, pentanal, 1-penten-3-ol, pentan-1-ol, 2-methyl-1‑butanol, hexanal, cis-3-hexen-1-ol, hexan-1-ol, heptanal, octane, octanal, 1-octen-3-ol, cis-3-hexenyl acetate, hexyl acetate, trans,trans-2,4-decadienal, propionic acid, 2‑methylpropan-1-ol and butanoic acid, all supplied by Fluka (Buchs, Switzerland); 4‑methylpentan-2-ol, trans,trans-2,4-heptadienal, trans-2-heptenal, heptan-2-ol, trans-2-penten-1-ol, 2-methylbutanal, 3‑methylbutanal, 3-methyl butanoic acid, ethyl acetate, cis-3-hexenal, 2‑methylpropanoic acid, ethyl butanoate, trans,trans-2,6-nonadienal, and trans-2-hexenyl acetate were supplied by Sigma-Aldrich (St. Louis, MO, USA); pentan-3-one and nonanal were supplied by Supelco (Bellefonte, PN, USA); cis-2-penten-1-ol and trans-2-hexen-1-ol were supplied by SAFC; trans-2-hexenal was supplied by Acrös Organics (Geel, Belgium).

The Statistical Design of Experiments (SDE) and Response Surface Methodology (RSM) were used to plan and analyze the experiments. Both constitute a planning methodology and analysis based on statistical tools, where SDE selects the optimal experimental strategy to obtain the desired information with the minimum cost of analysis and RSM evaluates the experimental results ensuring maximum reliability in the conclusions (Box et al., 2005).

In this process, a Box-Behnken design, with five repetitions of central points, for three factors was used: diameter of the holes of the hammer-crusher, temperature and malaxation time. The range of variation of these factors is 4.5 to 6.5 mm, 20 to 40 ºC and 30 to 90 minutes, respectively. Table 1 shows the different trials proposed by the design and their order of execution. Likewise, Table 1 also shows the actual and coded values of the factors used in each trial. The design tests allow to determine the influence of these technological factors on the phenolic and volatile compounds obtained in the olive oils. The factors were coded according to the following transformation, Equation 2.

Where x_i is the dimensionless coded value of the factor X_i, X₀ is the value of X_i at the center point or average level of the factor and ∆X the half-step change. The factors are usually coded because they provide a uniform framework to investigate the effects of factors and the coefficients of the models can be easily compared.

The experimental results were analyzed by Design-Expert v. 8.0.7.1 software (Stat-Ease, Inc., Minneapolis, MN, USA). The adequacy of the proposed model was determined by evaluating the lack of fit, coefficient of determination (R²) and the Fisher value (F-value) obtained from the analysis of variance (ANOVA). A quadratic model for each response studied was used according to Eq. 3:

Where: D is the hole diameter of the crusher (mm), T is the malaxation temperature (ºC) and t is the malaxation time (min). The predicted response (Y) was correlated with the set of coefficients (β): the intercept (β₀), linear (β₁, β₂, β₃), interaction (β₁₂, β₁₃, β₂₃) and quadratic (β₁₁, β₂₂, β₃₃). SD is the standard deviation of the model.

The statistical significance of the model and model coefficients were determined at 5% probability level (p-value = 0.05). The models for each response were expressed in terms of actual factors and without taking into account terms which were not statistically significant.

Phenolic compounds play a very important role in VOO due to their high antioxidant activity, which contributes to the shelf life of the oils and, in addition, gives them their typical bitter taste (Zribi et al., 2013).

Table 1 shows the total phenols as determined by HPLC (mg/kg tyrosol), which are the sum of the individual phenolic compounds, total phenols determined by Folin-Ciocalteau reagent (mg/kg caffeic acid) and the antioxidant potential determined by DPPH free radical scavenging activity (µmol/kg). Table 2 shows the content of individual phenolic compounds identified by HPLC (mg/kg tyrosol). Table 3 shows the models obtained for all the responses, which can be used later to predict the answers once the factors have been fixed. All the responses in Tables 1, 2 and 4 were determined in duplicate; the average values are shown in these tables as well. The software generates the regression equations after discarding the terms which are not statistically significant (p-value > 0.05); however, in order to support hierarchy, some linear terms were not eliminated from the models, although they were not statistically significant. As can be seen in Tables 1 and 2, all oils meet the condition of being healthy because all of them contain more than 250 mg/kg of Hydroxytyrosol and its derivatives, according to the Commission Regulation (EU) No 432/2012.

Figure 1 shows the model surface obtained for the total HPLC phenol response; for total Folin phenols (total phenols determined with the Folin-Ciocalteau method), the surface is similar, as can be deduced from the model equation in Table 3. According to the models, the temperature during malaxation has a major influence on the presence of phenolic compounds, which increase as temperature increases. On the contrary, the phenolic compounds decrease as the diameter of the crusher hole increases, but this has less influence. Similar results were reported by other authors studying malaxation and milling separately (Vekiari and Koutsaftakis, 2002; Ranalli et al., 2003). However, Ben Brahim et al. (2015), using Response Surface Methodology (RSM), indicated that the malaxation time does not significantly affect the phenol content.

Figure 2 shows the relationship between phenolic compound content and the antioxidant potential DPPH, both for total HPLC phenols and total Folin phenols. A very good correlation was observed in it, as is also apparent from the models in Table 3. On the other hand, from Figure 2 and Table 3 a very good correlation between both methods of quantifying total phenolic compounds in the oils is deduced, although the Folin-Ciocalteau method overestimated the content regarding HPLC.

Table 3 shows the proposed models in terms of actual factors for individual phenols; each phenol had a different model, but after examining the response surfaces, small differences were observed among them. The secoiridoid derivative class (3,4-DHPEA-EDA, 3,4-DHPEA-EA, p-HPEA-EDA and p-HPEA-EA) was clearly predominant and all of them increased with temperature. The results are similar to those obtained by other authors (Fregapane and Salvador, 2013; Gómez-Rico et al., 2009; Kalua et al., 2006). In agreement with other papers (Rodis et al., 2002), this increase can be attributed to the increased partition coefficient between the oil and water phases, which causes an increase in the solubility of these compounds in the oil phase.

The malaxation time had a positive influence on secoiridoid derivatives from ligstroside (p‑HPEA-EDA and p-HPEA-EA), but did not show a significant influence on secoiridoid derivatives from oleuropein (3,4-DHPEA-EDA and 3,4-DHPEA-EA). This is in contradiction to Ranalli et al. (2003), who observed a reduction in these secoiridoid derivatives with the time of malaxation, and attributed it to an increase in the oxidative reactions catalyzed by the activity of oxidoreductase enzymes present in the olive fruit such as peroxidase and polyphenoloxidase, due to the fact that olive paste was exposed to air longer when the malaxation time increased. In our case, the positive influence of malaxation time may be due to the increased activity of the β-glucosidase enzyme that hydrolyzed ligstroside and oleuropein, and the greater phenol content of the Picual variety. Figure 3 shows the response surface for the influence of the temperature and time on oleocanthal (p-HPEA-EDA) content, and of temperature and diameter on oleacein (3,4-DHPEA-EDA) content. For these models, factors that are not considered are not significant.

Different behaviors are also observed in other compounds; for example, pinoresinol decreased when temperature increased with a malaxation time longer than 60 minutes; and luteolin was unaltered. In addition, as seen in Table 3, some models present interaction among the factors; for example, diameter-time in alcohols (hydroxytyrosol and tyrosol) and ligstroside aglycone (p-HPEA-EA), or diameter-time and temperature-time in pinoresinol. Finally, an interaction was observed between temperature and time in oleocanthal (p-HPEA-EDA), in agreement with De Torres et al. (2016).

Table 4 shows the volatile compound contents grouped according to the most probable precursor molecule and Table 3 shows the models. Only 16, of the 39 analytical standards used were identified in the samples because some analytical standards corresponded to compounds present in olive oils with defects, which are unwanted in extra virgin olive oils. Trans-2-hexenal was clearly predominant. The volatiles arising from the lipoxygenase (LOX) cascade (Angerosa et al., 2004; Kalua et al., 2007) decreased with temperature, but increased with the crusher hole diameter. The malaxation time was not significant. Figure 4 shows the response surface for the influence of diameter and malaxation temperature on total LOX volatile contents. The influence of temperature was observed to be much greater than that of diameter. When temperature was increased from 20 ºC to 40 ºC, the content of trans-2-hexenal was reduced by 62%, while the content of 1‑penten-3-ol only fell by 16%. The total volatile compounds decreased by 36%, in agreement with many research papers (Fregapane and Salvador, 2013; Gómez-Rico et al., 2009; Angerosa and Basti, 2001; Ranalli et al., 2001), as a result of the inactivation of hydroxidelyase enzymes (Salas and Sánchez, 1999).

For most of the studied volatiles, a positive dependence was observed with malaxation time, except for 1-penten-3-one, for which one slight decrease was observed. Overall, the total volatile compounds were not significantly affected by the duration of the malaxation.

Finally, some of them increased with the diameter crusher holes: trans-2-hexenal, cis-3-hexenol, hexanol and trans-2-hexenol; while others decreased: 1-penten-3-ol, 1-penten-3-one and cis-2-pentenol. The end result would focus on an increase in total volatile compounds due to the importance of the trans-2-hexenal and cis-3-hexenol. In both cases, the maximum concentration corresponded to a diameter of 6.5 mm and a temperature of 20 ºC.

In order to achieve the optimal conditions of a balanced EVOO, we aimed to maximize the content in total HPLC phenols and total LOX volatiles and, by extension, the antioxidant activity given its direct dependence with the content of total phenolic compounds (Figure 2). Table 5 shows the optimal values obtained from the derived mathematical models using the Design-Expert software. It can be observed that the optimal conditions are at the contour limits. As can be determined from the models in Table 3, and is seen in Figure 2 and Table 5, the antioxidant activity is closely linked to the total phenol content in such a way that as these were increased the antioxidant activity also increased. In contrast, the variation in total LOX volatiles was completely opposite to that of total phenols, according to the findings of Gómez-Rico et al. (2009) and Inarejos-García et al. (2011).

Therefore, obtaining balanced olive oils only depends on what is considered a balanced oil and on the operating conditions that are set to obtain the desired content in its different components. In Table 5, the total phenol content, total LOX volatiles and antioxidant activity for different operating conditions were calculated using the models in Table 3. In order to obtain an extra virgin olive oil which is balanced and of high quality from the Picual variety, it should be elaborated with a malaxation temperature between 20 and 25 ºC and preferably with a hammer-crusher hole diameter of 6.5 mm. Thus, for a hammer-crusher hole diameter of 6.5 mm 337 and 356 mg/kg total HPLC phenols were obtained for malaxation temperatures of 20 and 25 °C, respectively and, likewise, 12.7 and 11.5 mg/kg total LOX volatiles. These oils will be fragrant, healthy and, if they do not have defects, of high quality.

When the malaxation temperature is increased healthier but less fragrant oils were obtained and, on the contrary, less healthy but more fragrant oils were obtained at lower temperatures.