The ANOVA (not presented) carried out on the color parameters L*, a*, b* evidenced significant effects for ecotype and for ecotype x temperature interaction, but not for oil extraction temperature per se. Because the interaction was relevant, Table 2 reports the results separately for 60 and 80 °C. At 60 °C the luminosity was similar except for La Carbonera, which was slightly darker; while at 80 °C Huarmey grande was more luminous than the others, and La Carbonera was still the darkest. The other two parameters showed that a* was slightly higher (i.e. redder) for La Carbonera oil, while b* was greater (i.e. yellower) for Huarmey chico, Casma and Carhuaz (at 60 °C) and for Huarmey grande (at 80 °C). The low luminosity of the samples is partially a consequence of the extraction method utilized. Mechanical pressing tends to drag different components present in the seeds, such as gums: these impurities hinder the penetration of the light, thus reducing the luminosity. In fact, solvent-extracted oil is markedly clearer, as noticed by Conceição et al. (2007) and by Palconite et al. (2018).

Figure 2 presents the physicochemical characteristics of the castor oil extracted at two different temperatures (60 and 80 °C). The ANOVAs (not presented) showed significant effects due to the ecotype as well as the extraction temperature; the ecotype always had the largest effect, except for oil density when the temperature was preeminent. The interaction between the two factors was also significant but generally of minor relevance except for the peroxide and iodine indices.

Compared to 60 °C, the 80 °C treatment led to better oil yield, superior oil density and kinematic viscosity, slightly higher refraction index, and inferior acidity index and free fatty acid content. For the peroxide and iodine indices, the interaction between temperature and ecotype was more relevant than the temperature per se, as each genotype behaved differently. The p-anisidine was always undetectable.

Heating influences the process of rupture of lipid bodies, thus favoring oil extraction; the better yields at the superior temperature were probably linked to more drastic shattering of lipid bodies and superior oil fluidity. Interestingly, the yield was significantly correlated to the weight of the seeds (r = 0.78 and r = 0.83 at 60 and 80 °C, respectively). At 60 °C the oil yield varied from 33.4 (Huarmey chico) to 39.0% (Huarmey grande), and at 80 °C from 33.5 to 42.0% (same ecotypes). These values match those reported by Perdomo et al. (2013), who obtained a yield of 36.6% using a mechanical press at 60 °C on a Mexican variety. Similarly, our results are within the variation (31.99-48.39%) summarized in their review by Ahmad et al. (2020). In general, yield seems to be influenced by the size of the seed (Huarmey grande seeds were exceptionally large).

At 60 °C, the oil density of Carhuaz and La Carbonera was the highest (0.949 and 0.949 g/cm³), while Casma had the lowest (0.947 g/cm³); at 80 °C Carhuaz and Huarmey grande showed superior results (0.949 and 0.949 g/cm³). The differences between the results of the two extraction temperatures may be related to the fact that higher temperatures favor the migration of compounds such as gums, proteins, fiber, carbohydrates, etc. into the oil, thus increasing its density (Vasco Leal et al., 2017). Our results are within the variation (0.946-0.958 g/cm³) described by Conceição et al. (2007), Panhwar et al. (2016) and Vasco Leal et al. (2017) for Brazilian, Pakistani and Mexican accessions.

At 60 °C the kinematic viscosity of Huarmey grande was the highest (227.1 mm²/s), followed by those of Huarmey chico and La Carbonera, while Casma presented a relatively low value (213.6 mm²/s, respectively). At 80 °C La Carbonera showed the maximum value (228.4 mm²/s) and Casma, again, the minimum (215.8 mm²/s). These results are marginally lower than those (239.39 mm²/s) reported by Conceição et al. (2007) but are far inferior to those (250.04 to 265.84 mm²/s) by Vasco Leal et al. (2017), although these last researchers used different analytical equipment.

The refraction index at 60 °C was greatest in La Carbonera (1.477), and at 80 °C in Huarmey grande and Huarmey chico (1.477 in both samples) but the differences among ecotypes were generally minimal. A similar value (1.479) was reported by Canoira et al. (2010); while an inferior result (1.431) was observed by Panhwar et al. (2016).

Concerning the acidity index and the free fatty acid content, only La Carbonera evidenced high values, while all the other ecotypes showed 0.420-0.727 mg NaOH/g and 0.211-0.366 g/100 g, respectively; the minimum values were those of Carhuaz at 60 °C and of Huarmey grande at 80 °C. In general, lower values were observed at 80 °C extraction. Perdomo et al. (2013) stated that the temperature affects the free fatty acid content and thus the acidity index. In fact, all seed oils have lipolytic enzymes which release free fatty acids. However, if the enzymes are inactivated by high temperatures, hydrolysis is limited, free fatty acids are scarce and acidity is low. The results of this research are comparable to those (0.542-1.218 mg KOH/g) reported by Vasco Leal et al. (2017) and to those (0.823 to 1.390 mg KOH/g) determined by Perdomo et al. (2013). Therefore, the acidity indices of the five Peruvian ecotypes are ideal and in general the oil has a low content in free fatty acids. Poor post-harvest handling of the seeds and poor storage of the oil often affect this quality indicator, hence low acidity values are achieved when the raw material are damaged or exposed to extreme conditions (high temperatures, humid environments, etc.).

The lowest peroxide index scores were reached by La Carbonera and Casma at 60 °C, and by Carhuaz and La Carbonera at 80 °C. A similar situation was evident also for the iodine index: Casma and La Carbonera had the lowest values at 80 °C while the two Huarmey and Carhuaz had the smallest at 60 °C. The peroxide values were lower than those (4.63 to 4.90 meq O₂/kg oil) reported by López-Ordaz et al. (2019) but similar to the one (2.25 meq O₂/kg) described by Panhwar et al. (2016), while the iodine values (Figure 2) were higher than those scored by López-Ordaz et al. (2019), Torrentes-Espinoza et al. (2017), and Panhwar et al. (2016), i.e. 84.9-85.9, 91.0 and 83.6 g I₂/100 g lipid, respectively, probably because of different ecotypes and origins of the seeds as well as different extraction methods

The castor oil contained mainly ricinoleic acid (on average, 86.8%), followed by linoleic acid (4.2%), oleic acid (3.1%) and other minor fatty acids (Table 3). The one-way ANOVA (not presented) did not show significant differences for ricinoleic or linoleic acids among samples; the variation of the other fatty acids, even if significant, was very limited. Overall, the oil of our ecotypes was constituted by about 91% mono-unsaturated fatty acids (MUFA), 7% poly-unsaturated fatty acids and 2% saturated fatty acids. Similar percentages for ricinoleic acid (from 75.0 to 90.0%), linoleic acid (4.1-9.7%) and oleic acid (3.0%-7.7%) are reported in the literature (Ahmad et al., 2020; da Silva Ramos et al., 1984; Harhar et al., 2016; Torrentes-Espinoza et al., 2017; Velasco et al., 2015; Wang et al., 2011).

The ANOVA (not shown) demonstrated the existence of significant differences merely among ecotypes. No differences were observed between the two extraction temperatures, suggesting good thermal stability of the compounds. Only tocopherols were detected (Table 4); the most abundant homologue was δ-tocopherol (on average, 520.2 mg/kg), followed by γ-tocopherol (366.8 mg/kg), α-tocopherol (9.5 mg/kg) and β-tocopherol (7.5 mg/kg), for an average total tocopherol content of 904.1 mg/kg. Interestingly, and unlike the other ecotypes, Huarmey grande had more γ-tocopherol than δ-tocopherol (506.8 vs. 448.7 mg/kg). The Huarmey chico samples showed the highest total tocopherol content (1038.4 mg/kg) and Carhuaz the lowest (780.1 mg/kg). Very little information is available in the literature on the presence of tocopherols in castor oil. Total tocopherol content of all ecotypes was higher than the values reported by Ananth et al. (2019), i.e. 461.3 mg/kg, and by Harhar et al. (2016), i.e. 183 mg/kg but lower than the levels recorded by Velasco et al. (2005) in a natural high-oleic acid mutant (2617 mg/kg) and in a standard castor oil line (1345 mg/kg). Velasco et al. (2015) observed a very large variation (99.6-282.2 mg/kg) for total tocopherol content in the seeds of 121 accessions from southern Spain.

On average, α- and β-tocopherol showed a minimal percentage (1.1 and 0.8%, respectively) of total tocopherols, while γ-tocopherol represented 40.2% and δ-tocopherol 59.7%. Similarly, Velasco et al. (2005) and Velasco et al. (2015) identified γ-tocopherol and δ-tocopherol as the two most relevant homologues. Nevertheless, Ananth et al. (2019) and Harhar et al. (2016) identified γ-tocopherol as the most abundant homologue.

The antioxidant capacities of the hydrophilic and the lipophilic extracts are presented in Table 4. The ANOVA (not presented) highlighted a relevant ecotype influence and a significant temperature effect only for the DPPH results of hexane extracts. The interaction between ecotype and extraction temperature was predominant in the 80% methanol extracts for the ABTS test, although it was less important for DPPH. At first glance, it is easy to see that most of the antioxidant capacity was due to the hydrophilic compounds, extracted by the 80% methanol solution. We tested the total polyphenol content by the Folin-Ciocalteu method as described by Varas Condori et al. (2020), but we did not observe detectable values (not reported). Therefore, the high antioxidant capacity of this fraction is probably attributable to the ricinoleic acid. The antioxidant capacity in lipophilic extract is likely due to the tocopherols.

In general, Huarmey grande showed the highest antioxidant capacities in the methanolic phase at both extraction temperatures. For the ABTS test, Huarmey chico and Carhuaz also had top results at 60 °C and La Carbonera at 80 °C. For DPPH, instead, the other best performers (along with Huarmey Grande) were Huarmey chico and Casma, at 80 °C. For the lipophilic extracts, Huarmey grande, Huarmey chico and Casma had the best ABTS antioxidant capacity, in accordance with their tocopherol contents. The situation was somewhat different for DPPH because of La Carbonera’s outstanding values.

It is difficult to compare our results with the scarce data in literature because of the different variables involved. Ananth et al. (2019) observed that the ABTS radical scavenging capacity of hydrophilic castor oil extract was superior to that of the lipophilic extract (0.63±0.02 vs. 0.13±0.04 μM TEAC/g oil DM), but the opposite was true for the DPPH assay (0.18±0.02 vs. 0.61 ± 0.03 μM TEAC/g oil DM). Santos et al. (2018), working with methanolic extracts, observed a free radical inhibitory capacity of 15.31±0.13 mg ascorbic acid equivalent/100 mL for ABTS, while for DPPH a 1.87±0.18% inhibition was found.

The OSI values of the oils analyzed by Rancimat at 150, 160 and 170 °C, with a constant 20 L/h airflow, are presented in Table 5 while the extrapolated shelf life at 20, 25 and 30 °C is reported in Table 6. The ANOVA (not presented) recorded a significant effect of the ecotype and, to a lesser degree, of the extraction temperature; the interaction was significant only for OSI and of minor relevance. The effect of the extraction temperature was higher for shelf-life than for OSI.

Huarmey grande generally showed the highest oxidative stability, and La Carbonera the lowest (Table 5). The huge difference (with a ratio around 2.0) between these two ecotypes at 150 °C decreased with increasing temperatures, reaching the lowest ratio (1.38) at 170 °C. To allow comparison with the values reported in the literature, the extrapolations at 110 and at 120 °C were performed. The OSI of the oils extracted at 80 °C ranged from 66.2 h to 145.6 h at 110 °C, and from 32.4 h to 69.2 h at 120 °C. Overall, the oxidative stability of the Peruvian accessions was high, in line with or often better than those reported by other authors. For example, Harhar et al. (2016) found an OSI of 35.5±4.0 h at 110 °C in Moroccan castor oil, while Adam Ali et al. (2016) obtained a value of 48.0 h at 120 °C in solvent-extracted castor oil. The oxidative stability of castor oil is therefore superior to that of other oils, such as chia (1.49 h at 110 °C; Villanueva et al., 2017), sacha inchi (0.493 h at 110 °C; Rodríguez et al., 2015), argan, olive, sesame and nigella (31.0, 27.0, 28.5 and 17.0 h at 110 °C, respectively (Harhar et al., 2016). Furthermore, all the samples exceeded the minimum 8 h oxidation stability at 110 °C requirement prescribed in the EN 14214 specification for biodiesel.

The shelf-life of all castor oils was enhanced when the seeds were extracted at 80 °C, indicating that the temperature played an important role in stability, probably through enzyme inactivation. Carhuaz and Huarmey grande consistently had the longest shelf-life, whereas La Carbonera had the shortest, particularly when extracted at 60 °C. The oxidation of polyunsaturated fatty acids generates volatile compounds that impart undesirable flavors and aromas, endangering the nutritional quality of the oil and limiting its shelf-life (Ixtaina et al., 2012). However, the hydroxyl behavior of the unsaturated ricinoleic acid makes castor oil a natural polyol, boosting oxidative stability and shelf-life compared to other vegetable oils. For example, the shelf-life of sacha inchi oil is 3.29, 1.79 and 0.79 years at 20, 25 and 30 °C, respectively (Rodríguez et al., 2015), and that of cottonseed oil is 46 days at 25 °C (Kurtulbaş et al., 2018). At 25 °C the shelf-life as determined by the Rancimat extrapolation of the biodiesel oil obtained from the transesterification of cooking oil waste was lower (3.1 months), and even after the addition of different strong antioxidants (BHA, BHT, PY, PG, and TBHQ), it only reached a maximum of 37.9 months (3.2 years; Zhou et al., 2016). Therefore, the excellent shelf-life of castor oil guarantees a protracted storage without appreciable deterioration.