Date seeds were collected randomly from the Nefzaoua area, located in the South West of Tunisia. The selected seeds were soaked in water, and then washed to remove any adhered date flesh and air dried. They were further oven-dried at 60 ºC for 24 h. The dried seeds were milled and sieved using a 1-mm mesh sieve and stored at -20 ºC until use.

Steam treatments were carried out using a 100-L capacity reactor that can work at temperatures between 50 and 190 ºC by direct heating with saturated steam, and at a maximum pressure of 9 kg/cm². Date seed samples (10 Kg) were treated at temperatures between 140 and 190 ºC, as preliminary experiments. After treatment, the samples were collected, vacuum filtered through filter paper, dried, milled and sieved using a 1-mm mesh sieve and stored at -20 ºC.

Date seed oil extraction was carried out on a solvent extractor equipped with six Soxhlet posts. 10 g of date seed powder were added to a cellulose thimble and placed into a Soxhlet apparatus that was connected to a 250-mL round-bottom flask containing 200 mL hexane. The extraction process lasted for 8 h and until the solvent in the reflux pipe became colorless. The extract was evaporated in a rotary vacuum evaporator (model R-210, Buchi, Flawil, Switzerland) at 40 °C, after which the collected oil was stored in vials at -20 °C until further analysis.

The date seed powder was put into a screw-cap flask (250 mL), made up to required volume with hexane, and sonicated in an ultrasonic bath (Thermo-10D; 40 kHz frequency; 240 W ultrasonic output power; 500×300×150 mm internal dimensions). The seed and solvent mixture was irradiated at different levels of sonication time, temperature, and hexane-to-seed (H/S) ratio. The bath water was placed in an external thermostatic device to avoid auto-heating effects during sonication. Upon extraction completion, the extract was passed through filter paper in a funnel and the solvent was removed using a rotary vacuum evaporator (model R-210, Buchi, Flawil, Switzerland) at 40 °C. The extracted oils were stored at -20 °C until further analysis.

Response surface methodology with a three-factor, three-level Box-Behnken design was employed to study the effect of UAE parameters on the recovery of date seed oil. Three independent factors, namely sonication time (X₁, min), sonication temperature (X₂, °C), and H/S ratio (X₃, mL/g) were used. Three levels were applied to each factor (Table 1). The ranges of the variables studied were chosen according to the preliminary experiments. The response was oil recovery (%) referred to Soxhlet extraction yield. The experimental design and statistical analysis were created and performed using the Statgraphics Plus Version 2.1 software (Statgraphics Technologies Inc., The Plains, Virginia). The experimental design was composed of 15 experimental runs. The three replicates of the center points evaluated the repeatability of the designed method and were used to examine experimental error. A second-order polynomial model was used to fit the experimental data.

The experiments were randomly conducted in order to avoid the effects of unexpected variability in the observed responses which resulted from extraneous factors. Analysis of variance (ANOVA) was used to analyze the individual linear, quadratic and interaction regression coefficients. The values of R² (coefficient of determination) and F-test were used to examine the adequacy of the designed model. Regression analysis and response surface plot were used to examine the optimal conditions and the effects of the interactions among independent variables. The confidence level was set at 95%. The validity of the designed model was confirmed by performing triplicate analyses on the generated optimal parameters and one-sample t-test was used to verify the differences between experimental and predicted values.

The oxidative stability of the oil (OOS) was evaluated by applying a modification of the thiobarbituric acid (TBA) reactive species method (Rodríguez et al., 2007). The degradation compounds (aldehydes, ketones, etc.) synthesized during oil oxidation reacted with TBA to give a pink-colored complex whose concentration was measured at 540 nm. Sixty microliters of oil and 5 μL of ABAP were added to an Eppendorf tube (1.5 mL capacity) and made up to 0.1 mL with distilled water (in quadruplicate). Afterwards, 150 μL 20% acetic acid (pH 3.5) and 150 μL 0.8% w/v TBA in 1.1% sodium dodecylsulphate w/v were measured into each tube. This mixture was stirred in a Vortex and heated at 80 ºC for 1 h. After cooling at room temperature, 0.5 mL of 1-butanol were added, stirred and centrifuged at 12,100 g for 3 min. The absorbance of the butanol layer was measured at 540 nm. High absorbance values indicated high concentrations of TBA-reactive species quantified in the reaction medium and therefore low oxidative stability in the assayed oil. As the lower the absorbance the higher the stability, for reasons of clarity the final oil oxidative stability (OOS) was expressed as:

To assess the differences among samples, a multiple sample comparison was performed using the Statgraphics® Plus program Version 2.1. A multifactorial analysis of variance (ANOVA), followed by Duncan’s multiple comparison test was performed to differentiate the samples. The level of significance was p < 0.05.

The control and hydrothermally-treated samples were subjected to Soxhlet extraction and the obtained yields were considered as reference values for the ultrasound-assisted experiments. After preliminary experiences, the highest oil yield by the Soxhlet method was obtained at 180 ºC. This fact suggests a very intense disorganization of seed tissues which enabled the extraction of higher amounts of oil, together with the enrichment in oil of the solid recovered after pre-treatment (Fernández-Bolaños et al., 2004) due to the partial solubilization of the solid material into the treatment liquid. In a previous work (Mrabet et al., 2015), cellulose, lignin and mannans coming from date seeds were quantified in the solid residue recovered after hydrothermal treatments of complete date fruits, pointing to a partial degradation of seed structure. This disorganization of lignocellulosic material (the solubilization of lignin and hemicelluloses and the enrichment with cellulose and oil of the resultant solid) have been described when working with other agricultural by-products (Rubio-Senent et al., 2015; Fuentes-Alventosa et al., 2013). For the native date seed, a lower oil yield was obtained (7.67 ± 0.06%) compared to pre-treated ones (Besbes et al., 2004) or 11.8-15.6% (Mrabet et al., 2015) for some Tunisian varieties, and the lowest for some Algerian ones (5.4-5.6%) (Boukouada et al., 2014). Surface response methodology was applied to samples, both the native seeds and 180 ºC-treated ones, for oil recovery optimization.

The different parameters and levels summarized in Table 1 were applied to both samples (native date seeds and with hydrothermal pretreatments at 180 ºC), and two different Box-Behnken designs were developed (Tables 2 and (3). Each model was composed of fifteen experiments plus three at optimal conditions to validate the designed model. The oil yield was the response variable studied because it is the main parameter considered to study the economic viability of the extraction process. The obtained oil recovery by UAE was between 42-73% and 57-81% of the oil extracted by Soxhlet for native seeds and those treated at 180 ºC, respectively. In spite of these yields, the savings in energy, solvent, and working time could make this system valuable. In addition, the chemical composition of UAE oils should be less modified when compared to the corresponding Soxhlet ones, probably due to a gentle exposure to temperature; but it is also important to keep in mind that ultrasound application could decrease the oxidative stability of oil (Samaram et al., 2015; Böger et al., 2018). The extra energy applied to plant materials by ultrasound is one of its advantages, because interesting biomolecules are extracted in higher concentrations, e.g. phenols and other antioxidants (Bimakr et al., 2012). However, this energy is also the activation energy for the appearance of free radicals, molecules that trigger oxidation chain reactions, which are responsible for the characteristic rancid color and flavor. These defects negatively affect the shelf-life and consumer acceptance of food products and are a major cause of food waste (Böger et al., 2018).

The high energy of ultrasonic sound waves provokes the disruption of the cell wall, thus making the oil more accessible and more easily extracted (Vinatoru, 2001). During sonication, cycles of compression and decompression are originated in the liquid. During compression, little bubbles are generated in the solvent that lead to localized pressure increases. But during decompression, the bubbles implode, collapse and originate rarefactions. These cycles are also known as cavitation phenomenon, which is responsible for cell wall disruption, oil release into the solvent and destruction of oil emulsion (Lou et al., 2010). The date seed is a very hard by-product from date fruit processing with a chemical composition that suggests a very complex and closed cell wall structure: 20-25% lignin, 17-20% cellulose and 11-20% hemicelluloses, 60% of which are mannans (Mrabet et al., 2015). Mannans establish many closed complexes with cellulose microfibers, which tighten the cell wall network (Whitney et al., 1998). These facts could hinder oil extraction. Hydrothermal pre-treatments change the structure of the plant material, dissolving most of the lignin and loosening the lignocellulosic materials (Fernández-Bolaños et al., 2004; Jaramillo-Carmona et al., 2019). The added effects of both processes, ultrasound and hydrothermal treatments, could be the reason for the higher extraction yields and oil recoveries in the sample treated at 180 ºC.

A multifactorial ANOVA was applied to all these assays, and the application of a pre-treatment was included as a factor. The results are presented in Table 4. The sonication temperature was the only factor that did not show significance, although it effects were variable when working with other vegetal products, e.g. winter melon seeds (Bimakr et al., 2012) and caper seed (Ara et al., 2014). In the cited works, the yield increased with temperature until it reached its maximum. This fact could be related to an increased diffusivity of the solvent into cells and to an enhanced solubility of oil from the cells into solvent. At higher temperatures, the yield decreases were probably due to a decrease in the number of acoustic bubbles generated by sonication. H/S ratio and the pre-treatment had the greatest influence (Table 4). At a high H/S ratio, there was sufficient solvent capacity for oil transfer and enhanced diffusion through a viscosity reduction. Both circumstances led to increases in the concentration gradient and to a more efficient extraction (Samaran et al., 2015). All these phenomena were eased by the disruption of the matrix structure caused by hydrothermal pre-treatments. The effect of the different significant factors is graphically displayed in Fig 1 as the recovery average values and LSD intervals.

The best-fitting models were generated by including only the significant terms (p < 0.05) in the generated quadratic models. Table 5 shows the results of the ANOVA analysis for the reduced quadratic model of the samples analyzed. These models can explain most of the variations in oil recovery, as shown by the high F values in the models (816.22 and 25.40 for native and 180 ºC-treated date seeds, respectively). This value decreased with pretreatment, and the same happened with the coefficient of determinations (R² and adjusted R²). This suggested a lower fit for treated seeds than for the native ones. Nevertheless, the models displayed high adjusted R², with values > 0.80.

For native date seeds, the H/S ratio (X₃) was the most influential factor (Table 5), followed by the second-order term of sonication temperature (X₁ ²) and interaction term between temperature and H/S ratio (X₁X₃). In addition, sonication time (X₂), interaction term between time and H/S ratio (X₂X₃), second-order term of H/S ratio (X₃ ²), and sonication temperature (X₁) were significant. The significant interactions among variables are presented in Figure 2 as response surfaces. The interaction between temperature and H/S ratio (X₁X₃) (Sub-Figure 2.1) showed that at low and high temperatures the recovery increased, especially at a high H/S ratio. The interaction between time and H/S ratio (X₂X₃) (Sub-Figure 2.2) was less significant, and the recovery decreased at a low H/S ratio and high sonication time, showing the opposite effect at a high H/S ratio. The reduced quadratic model that shows the relationship between the different factors and oil recovery is:

For pre-treated seeds, the results were much simpler. For 180 ºC-treated seeds, only sonication time (X₂) and H/S ratio (X₃) showed significance. The reduced quadratic model that shows the relationship between the different factors and the oil recovery is:

In Figure 3, the main effects are presented graphically. It is clear that for native seeds, sonication temperature and H/S ratio led to variations in oil recovery. By examining the H/S ratio, it seems that recovery did not reach its maximum and that, at higher ratios, recovery would increase. This aspect needs further confirmation. This factor was also the most influential for treated seeds, reaching maximum recovery at close to 7. For 180 ºC-treated seeds, sonication time also produced significant increases in recovery, as addressed in Table 5.

In summary, the H/S ratio was the most influential factor, although in some cases it was the time and/or temperature during sonication treatments as well. This ratio is also discussed in most published works, and is identified as the most significant factor in the ultrasound-assisted extraction of oil from capers (Ara et al., 2014), black seeds (Abdullah and Koc, 2013), and rapeseed flakes (Perrier et al., 2017). Sonication time and temperature, and their interactions with the H/S ratio had different effects, depending on materials and other experimental conditions, as occurred in our study.

A statistical study of optimization was carried out to find the optimal extraction conditions for each sample. To validate the model, three experiments were developed at optimal conditions and a t-test was used to verify significant differences (confidence level 95%) between the predicted values and the experimental ones.

For native seeds, the highest predicted recovery (74.32%, i.e. 5.42% oil yield) was found at sonication temperature and time of 45 ºC and 15 min, and at the H/S ratio of 7. The experimental average was 71.03±2.33% (5.18±0.17% as oil yield). For treated seeds at 180 ºC, the highest recovery (80.93%, i.e. 6.71 oil yield) was found with 45 min sonication, at 35 ºC and H/S ratio of 7. A mean value of 79.28±0.83% recovery (6.58±0.07% oil yield) was obtained. In any case, there were no significant differences between predicted and experimental values for recovery percent, so the validity of the designed models was confirmed. The recovery increased with the hydrothermal pre-treatment, increasing from 71% in untreated date seeds to 80% in those treated at 180 ºC. This higher oil yield could be related to the disorganization of the seed matrix due to hydrothermal pre-treatment, which facilitated oil extraction.

Similar recoveries (70-80%) were found in the bibliography when working with black seeds, 74.77% (Abdullah and Koc, 2013) or grape seeds, 74.77-82.9% (Böger et al., 2018; Malicanin et al., 2014), the highest recoveries reported. Oils from winter melon (Bimakr et al., 2012) or pomegranate seeds (Goula, 2013) were recovered in lower amounts, 43.2 and 59.8%, respectively. For H/S ratio, most authors found that the bests conditions were at high ratios, e.g. 20 for kiwi and pomegranate seeds (Cravotto et al., 2011; Goula, 2013), or even higher (40 for rapeseed flakes (Perrier et al., 2017)). It is clear that the volume of hexane used for extraction has a positive influence on oil recovery, but some authors pointed to its negative effect on the oxidative stability of the oil (Samaram et al., 2015). Sonication temperature and time at optimum conditions were very similar to ours, varying between 20 and 50 ºC and 20 and 45 min.

As pointed out throughout this work, ultrasound could affect the oil’s oxidative stability under certain conditions. To confirm the influence of ultrasounds on the stability of date seed oil, an assay was developed in which the absorbance at 540 nm was inversely proportional to oxidative stability. The compounds resulting from oil oxidation (aldehydes, ketones, etc.) are TBA-reactive species, therefore unstable oils are easily oxidized at TBA assay conditions and give high 540 nm absorbance values. The results for OOS are shown in Figure 4. For native seeds, no differences were found between Soxhlet and UAE oils. The extractions at relatively low H/S ratio, as presented in this work, could maintain oil stability better than when extracted at higher ratios (>10) (Samaram et al., 2015).

Significant decreases in oil stability were quantified in the case of 180 ºC-treated seeds. These results suggested that, in any case, ultrasounds did not lead to a loss in stability when compared to Soxhlet, but hydrothermal pretreatment did. With 180 ºC pre-treatment, the oil recovered by both sonication and Soxhlet was less stable. The higher resistance to oxidation in UAE oils from pre-treated seeds (180-US) when compared to its corresponding Soxhlet oil (180-SX) could be due to a more effective extraction of antioxidant molecules which are originated during the hydrothermal pre-treatment of date by-products (Mrabet et al., 2015). Pre-treatments could also destroy the naturally present antioxidants in date seed oil, but this loss could be partially counteracted by the appearance of new antioxidant compounds during hydrothermal treatments.