Brown linseed (Linum usitatissimum) samples (5 kg) were obtained from the 2014–2015 harvest. The commercial linseed samples came from a local market (Pazze Foods Company), cultivated in the Panambi municipality (28°17´33´´ latitude and 53°36´47´´ longitude), Rio Grande do Sul State, Brazil. Mechanically pressed brown linseed oil (150 mL) came from Pazze Foods Company (Panambi, RS, Brazil).

The linseed samples were oven-dried (MA035/1; Marconi Equipamentos para Laboratórios Ltda., Piracicaba, SP, Brazil) at 50 °C for 48 h, vacuum packed using a Selovac 200B (Selovac, São Paulo, SP, Brazil) and stored in a freezer (Brastemp, São Paulo, SP, Brazil) at –18 °C. In order to perform oil extractions, samples were crushed in a mixer Cuisinart Brasil (São Paulo, SP, Brazil), and sieved through 16 and 32 mesh sizes, using the fraction retained in the 32 mesh sieve after passing the 16 mesh one, obtaining particle sizes of 500 to 1000 µm.

A 2² experimental design with 3 central points using pressure of 8·10⁶, 10⁷ and 1.2·10⁷ Pa and temperature of 20, 40 and 60 °C for 90 min evaluated the effects of temperature and pressure on the extraction yield using subcritical propane and pressurized ethanol.

Acidity, refraction index, and density of the oil obtained with both extraction methods and the commercial oil were analyzed according to the methods described by the Adolfo Lutz Institute, São Paulo, Brazil (15). All physical and chemical analyses were performed in triplicate. Titratable acidity was determined by titrating a sample with 0.1 M sodium hydroxide solution (NaOH; Synth, Diadema, SP, Brazil) and phenolphthalein as indicator. The refractive index was measured with an Abbé refractometer (Delta, São Paulo, SP, Brazil). Density was determined using a density bottle (4500 density meter; DMA, São Paulo, SP, Brazil) at 25 °C.

The samples were esterified following the method set by the Adolfo Lutz Institute (15). A mass of 0.1 g oil was mixed with 0.2 mL (2 M) methanolic potassium hydroxide solution (KOH; Synth, and methanol; Merck, São Paulo, Brazil) and 2 mL n-hexane, and homogenized in a vortex mixer (Q220M; Quimis) for 30 s. After that, 3 mL saturated sodium chloride solution (NaCl; Cinética, Londrina, PR, Brazil) were added and the supernatant was then removed.

Qualitative analysis of essential fatty acids followed the instructions of Khattab and Zeitoun (14), in a gas chromatograph (CG-2010 Plus model; Shimadzu do Brazil, Barueri, SP, Brazil) equipped with a polar column (Rtx-Wax Restek, 30 m×0.25 mm i.d., 0.25 mm thickness) and a flame ionization detector, at the following temperatures: 80 °C (5 min), 80–200 °C (20 °C/min), 200 °C (5 min), 200–230 °C (5 °C/min), 230 °C (10 min), injector temperature at 250 °C, detector at 275 °C, with a 50:1 split ratio. Nitrogen and synthetic air (White Martins Gases Industriales Ltda., 99.99% purity) were the carrier gases, with a 1.5 mL/min flow rate. A diluted sample (0.4 mL) in dichloromethane (1:10; Merck) was injected. The sample components were identified by comparing their mass spectra with the mass spectra and retention indices of linolenate methyl ester (ω-3), linoleatemethyl ester (ω-6) and oleatemethyl ester (ω-9) standards (Sigma-Aldrich, Merck KGaA, St. Louis, MO, USA).

It is important to determine the relationship between the high extraction yield using propane and production costs. Based on the information generated by CONTAGRI software (16) the economic analysis generally depends on the total fixed and variable costs. In this way, the production costs included fixed and variable costs such as the raw material (seed), chemicals, reagents and gas (propane), equipment investment, and labour cost. Cost analysis of 16 working hours a day and 240 days a year of extraction unit operation took into consideration 1% for repair and maintenance, 10% for depreciation and 2% interest over a 10-year period.

The data from the experimental design were subjected to analysis of variance (ANOVA), and their average values were compared using the Tukey’s test at a 5% probability level for oil physicochemical analysis with Statistica v. 5.0 (17).

Linseed oil extraction using organic solvent (n-hexane) gave a 36.1% yield in 14-hour extraction. Khattab and Zeitoun (14) obtained a 42.4% yield of linseed oil using the same extraction process; however, they did not mention the process extraction time. Table 1 shows the 2² experimental design matrix for linseed oil yield using subcritical extraction.

High extraction yields were obtained in the central point (assays 5–7) and in assay 4 (Table 1). These results were similar to the ones obtained by Zanqui et al. (2), who verified a 28.8% yield of linseed oil using subcritical extractions with propane at 1.2·10⁷ Pa and 60 °C. Pradhan et al. (3) and Khattab and Zeitoun (14) also reported similar extraction yields using supercritical CO₂ respectively at 3·10⁷ Pa and 50 °C for 4 h (35.3%) and 4·10⁷ Pa and 50 °C for 2 h (36.9%). Such high yields might be due to high operational pressure and extraction time conditions, and solubility of the different compounds in the extraction solvent. In addition, the amount of solvent for complete extraction (12, 18, 19) was lower when using subcritical propane than when using supercritical CO₂, making propane more attractive for industrial application.

Analysis of variance (ANOVA) served to see the effects of variables on the extraction yields of linseed oil using subcritical propane, observable in the Pareto chart (Fig. 1). Both temperature and pressure had significantly positive effects (p<0.1) on the extraction yield during 90-min extraction (Fig. 1).

The following equation shows the first order-coded model for extraction yield as a function of temperature and pressure within the studied ranges:Yield/%=27.8+0.48t+0.22p /2/where t is extraction temperature (°C) and p is extraction pressure (Pa).

The independent variables showed a positive effect (p<0.1), whereas the interactions were not significant. Analysis of variance (ANOVA) validated the model and the non-significant parameters were added to the lack of fit. The correlation coefficient was 0.9, and the F calculated value was 0.25 times higher than the F tabulated one (Table 2), allowing the contour curve construction (Fig. 2).

Fig. 2 shows the increase of the extraction yield of linseed oil using subcritical propane with the increase of temperature and pressure.

Extraction using pressurized ethanol was performed under similar experimental conditions as the one using subcritical propane. The experimental design results (Table 1) showed that the highest yield (approx. 8%) was obtained at the central point (assays 5, 6 and 7). According to Table 1, the extraction using pressurized ethanol showed a lower yield than those using subcritical propane and organic solvent (n-hexane), which could be due to a short-chain alcohol that did not present any good solvency for triglycerides under the applied conditions. Besides that, this process requires, after extraction, an additional step for ethanol removal via evaporation, which is seen as a disadvantage, since it may cause loss or damage to certain oil components.

Jesus et al. (5) conducted a comparative study to evaluate palm oil extraction capacity with ethanol and propane under similar conditions (10⁷ Pa and 20 °C) yielding 53.3% with ethanol and 72.8% with propane. The authors observed that the solubility obtained with propane was higher than with ethanol, allowing total oil miscibility, and improving extraction yield.

Statistically treated results of linseed oil extraction with pressurized ethanol showed that both pressure and its interaction with temperature had a significantly positive effect (p<0.1), observable in the Pareto chart (Fig. 3). The correlation coefficient was 0.92 and the F calculated was 1.44 times lower than the F tabulated value, hence, neither validating the model nor allowing the contour curve construction. Based on that, in order to increase the yield, the extraction time was prolonged to 3 h under the experimental design central point conditions (40 °C and 10⁷ Pa). The improvement in the yield was 12.3%; however, even lower than in the extraction with subcritical propane (28.4%).

Table 3 shows the percentage of essential fatty acid esters obtained with different extraction methods from linseed oil, with methyl esters ranging from 49.8 to 55.0%.

Using propane as a solvent, Zanqui et al. (2) reported the yields higher than 48.7% ω-3 and 12.3% ω-6 from linseed oil. Results closer to those reported in this study were found by Ivanov et al. (20), who obtained 53.7% ω-3, 16.17% ω-6 and 19.4% ω-9 from linseed oil extracted using CO₂, however, the used pressure ranged from 4.5 to 6.2·10⁹ Pa, which may be explained by the extraction processing conditions and seed origin.

The statistical analysis of extraction using pressurized ethanol did not show any significant differences among acid esters, whereas the extraction with subcritical propane resulted in significant differences in linolenate (ω-3) and linoleate (ω-6) concentration. Extraction yields demonstrated that the concentrations of ω-3, -6 and -9 from linseed oil extracted using pressurized ethanol were similar to the ones found using subcritical propane (Table 3).

There were no statistical differences in the amounts of essential fatty acid esters obtained with the organic solvent and mechanical pressing, whose major component was linolenate (approx. 55%), corresponding to the ω-3 fatty acid, followed by oleate (19%) (matching the ω-9 fatty acid). Such results corroborated those found in the literature. Pradhan et al. (3) obtained 53.8% ω-3 and 15.6% ω-6 from linseed oil extracted using mechanical pressing and 50% ω-3 and 14.4% ω-6 in the oil extracted using organic solvent. Linseed oil extracted with n-hexane contained 56.28 and 15.6% ω-3 and ω-6, respectively (14).

The area sum (%) of fatty acid esters (linolenate (ω-3), linoleate (ω-6) and oleate (ω-9)) was around 86% (Table 3). The remaining 14% may be related to the amounts of both palmitic and stearic acids. Studies in the literature of fatty acid composition from linseed oil showed 45.1–73.4% linolenic, 11.9–20% linoleic, 17.1–24.3% oleic, 4.6–6.9% palmitic and 0.25–4.6% stearic acids (21-23).

In this work, the ω-6/ω-3 family polyunsaturated fatty acid ratio was around 0.24. An ω-6 to ω-3 essential fatty acid ratio of approx. 1 is considered nutritionally adequate (24). A lower ω-6/ω-3 fatty acid ratio is more desirable in reducing the risk of chronic diseases since polyunsaturated fatty acids are considered essential to the human body with nutritional advantages as long chain fatty acid (linoleic and linolenic acids) precursors. They can improve the brain, heart and immune system performance in humans, as well as have suppressive effects on the onset of cardiovascular, inflammatory and cancer diseases (24, 25).

The main advantage of extraction with subcritical propane are milder operating conditions than with supercritical CO₂. The use of propane reduced drastically operation pressure, and enabled its industrial application based on a cost analysis for implementation feasibility.

Physicochemical analyses are crucial for oil quality evaluation. Table 4 shows brown linseed oil acidity, refraction indices and density values after different extraction methods. The samples with high yield (central point, assays 5, 6 and 7) were used for extractions with subcritical propane and pressurized ethanol.

As observable in Table 4, the density and refractive indices were not statistically significant (p<0.05) for all linseed oil samples using the different extraction methods, as confirmed by Tukey’s test, with 95% confidence level. According to the Adolfo Lutz Institute, Brazil (15), the refraction index is a good oil quality indicator and is affected by oxidation, heat treatment and free fatty acid content from oil. Gunstone et al. (26) and Firestone (27) recommend values from 0.924 to 0.930 g/cm³.

Linseed oil acidity was statistically different among all extraction methods studied. The highest (1.76%) and lowest (0.95%) values were obtained using pressurized ethanol and subcritical propane, respectively. The highest acidity values were found after extractions with organic solvent (1.46%) and pressurized ethanol (1.75%). Such results were obtained since heating was used in both methods. Extractions using n-hexane (68 °C) and pressurized ethanol have a solvent evaporation step after the extraction. Heating can cause a triglyceride chain break during extraction, hence increasing acidity through fatty acid release (28). Thus, the physicochemical parameters indicate that oil extraction using subcritical propane was preferable due to fatty acid quality with low acidity.

The oil yield obtained with extraction using subcritical propane was around 28.3%. Table 5 (29) shows fixed and variable costs, as well as oil total cost per litre. Fixed oil unit production cost was around US$ 56 540.00 and plant age was approx. 10 years. The total oil cost per litre is high when produced in a laboratory scale unit, as it yields small quantities of linseed oil (0.48 L/day).

Galvão et al. (30) reported a US$ 13.21/kg specific cost when evaluating the supercritical technology (CO₂ solvent) costs for flaxseed oil extraction, whereas linseed oil obtained by cold mechanical pressing is marketed at US$ 54.00/kg in Brazil (31). These costs when compared to those for propane extraction make it a non-market-competitive technology. However, industrial-scale propane extraction (increased capacity) could reduce subcritical extraction costs, making it an attractive method.