Yellow passion fruits (Passiflora edulis spp) were obtained from Bishnupur district, Manipur, India. Uniform-sized yellow-colored ripe fruits were cut into halves and parts were separated. Fresh seeds were dried in an oven at 50 °C to constant weight.

The standards used for HPLC were purchased from Sigma. All other chemicals were of analytical grade.

The dried passion fruit seed was coarsely ground into powder to pass through a 40-mesh sieve. The powder was added to ethanol (extraction solvent) and placed in the ultrasonic bath (Riviera Glass Pvt. Ltd., India, Power 230 volts, 50 Hz). Solid-solvent ratio, extraction temperature, and extraction time were selected as extraction conditions. After the extraction, the ethanolic oil extract was recovered using a rotary evaporator followed by vacuum drying, until a constant weight (complete evaporation of ethanol) was obtained. This ethanolic extract was termed as UAEO.

The passion fruit seed oil was weighed and the extracted oil yield was calculated. A single factor analysis was performed to determine the solid-solvent ratio, extraction temperature, and extraction time. For the optimization of solid-solvent ratio, 5, 10, 15, 20, 25 and 30 g/100 ml solid-solvent ratios were selected and all extractions were performed for 30 min at 60 °C and the ratio at which maximum yield of oil occurred was selected as the optimized solid-solvent ratio. Similarly, for the selection of temperature, UAE was performed from 30 to 80 °C at each 10 °C interval using optimized solid-solvent ratio for 30 min. The extraction time periods used were 5, 10, 15, 20, 25, 30, 40, 50, 60 and 90 min at the optimized solid-solvent ratio and temperature conditions. For all experiments, oil yield was measured in triplicate.

Soxhlet extraction was performed using ethanol, as ethanol is the most widely used solvent because it is less harmful and has no impurities (Teng et al., 2016). The optimized solid-solvent ratio was chosen for each extraction process. The seed powder was extracted for 8 h, and yield was calculated at 10, 20, 30, 60, 90, 120, 180, 240, 300, 360, 420, and 480 min. The extracted oil was named SEO.

Optimized ultrasound extraction followed by Soxhlet extraction up to 4 h in total was performed and its kinetics studied in terms of ethanolic extract yield. and tThe ethanolic extract was coded as UAESO. All experiments were conducted in triplicate.

Moisture content, total lipids, crude fiber, crude protein, and ash content were measured using standard AOCS methods (Firestone, 2009). All experiments were performed in triplicate.

The extraction yield of ethanolic extracted oil was measured using Equation 1 in triplicate.

Where, Y_i and Y_f are the initial and extracted dried residual weights of passion fruit seed powder (g), respectively.

The refractive index was measured at 25 °C using a hand refractometer. The acid value was determined by titration with 0.1 mol/L NaOH using phenolphthalein as an indicator (Pereira et al., 2017). Iodine value (Firestone, 2009), peroxide value (Firestone, 2009), and conjugated diene (Farhoosh et al., 2012) were also determined. All experiments were carried out in triplicate.

TCC was determined as per the method described by da Silva and Jorge (2016) and expressed as µg of β-carotene/g of oil. The TPC of PFSO was determined by the Folin-Ciocalteu reagent method (Purohit et al., 2021) and reported in mg GAE/g of oil. All experiments were conducted in triplicate.

The phenolic acids were determined (Espin et al., 2016) using a HPLC instrument with a C18 column. Phenolic acids were identified against standards and quantified.

The COX value was calculated according to the formula in Equation 2, in triplicate.

The fatty acid profile of the PFSO samples was determined by preparation of methyl esters as described by Pereira et al. (2017) and separated by gas chromatography (GCMS-MS Model: GCMS-TQ8040, Shimadzu) and identified by comparison with retention times of the standard FAMEs. Each fatty acid was quantified as a percentage of the total methyl ester peak areas.

The mathematical model was based on the following assumptions (Milić et al., 2013; Marković et al., 2019): pseudo-single component of extracted oil; perfectly mixed solution; plant matrices were isotropic in nature; negligible resistance to the mass transfer; a fraction of the oils was located at the external surfaces, and rest was uniformly distributed in the matrices; extraction process occurred via two simultaneous mechanisms: (i) rapid penetration of solvent and dissolution of the oil located on or near surfaces of solid matrices known as washing, and (ii) mass transfer of oil soluble compounds from the plant matrices into the extractable solvent by diffusion and osmotic processes, well known as slow extraction; the fractions of oil extracted via washing, unhindered diffusion and hindered diffusion (f1, f2 and f3, respectively) assumed to be constant:

Where, w_p was the average content of ethanolic extract in the plant particles (g/100g) at time t, and process rate constant (k)

For, t = 0, w_p = w_∞, so the integration of Equation 3 gave the following expression for washing, unhindered diffusion and hindered diffusion, respectively:

Based on the assumption,

Where, f1 = w p , 1 w ∞ , f2 = w p , 2 w ∞ and f3 = w p , 3 w ∞ , and, f1 + f2 + f3 = 1 .

Hence, the amount of ethanolic extract extracted until time t, w = w ∞ − w p follows

Or

In general, for ideal conditions, it is assumed that hindered diffusion is negligible i.e. f ₃ = 0, the model can be represented by Equation 8:

if washing was faster than diffusion, (k ₁ >> k ₂>), then Equation 9 follows:

if f = 0, which implies that washing does not occur, means extraction yield increases exponentially due to diffusion:

Here Equation 8, Equation 9 and Equation 10 represent the Phenomenological model, model based on instantaneous washing followed by diffusion, and pseudo first order model, respectively.

The model parameters were estimated using MATLAB 7.14 (Release 2012a) and non-linear least squares regression using Microsoft Excel Solver (Microsoft office, USA). The performances of the developed model were statistically measured by coefficient of determination (R ²), mean squared error (MSE), and mean relative percent deviation (MRPD). The one-way analysis of variance (ANOVA) was performed to compare the means along with the Duncan’s multiple range test to evaluate the significant difference using IBM SPSS software at p < 0.05.

The PFS sample contained 8.64±0.64% moisture, 11.58±0.87% crude protein, 24.4 ±2.78% crude fat, 1.64±0.18% crude ash, and 51.59±3.81% crude fiber. The moisture content was 8.64%, whereas reported moisture content in yellow passion fruit seeds varied from 7.38 - 12.38% (Purohit et al., 2021; Reis et al., 2020; Malacrida and Jorge, 2012). The ethanolic extract content was 24.4%, in comparison, Purohit et al. (2021) found 24.6 and 28% oil in PFS cultivated in Manipur and Assam, respectively, and Malacrida and Jorge (2012) and Reis et al., (2020) obtained 30.39 and 23% oil, respectively, in Brazilian yellow PFS. PFS contained crude fiber (51.59 g/100g), crude protein (11.58 g/100g), and ash content (1.64 g/100 g), which agreed with previous reports (Malacrida and Jorge, 2012; Pereira et al., 2017; Purohit et al., 2021). Malacrida and Jorge, (2012) found 7.38% moisture, 30.39% lipids, 12.23% protein, 1.75% ash and 48.73% fiber in Brazilian yellow passion fruit. The species of passion fruit used, and the diverse geo-climatic conditions may be the cause of variations in proximate composition (Reis et al., 2020).

The best yield is achieved when the solution reaches saturation concentration. In our study, the solid-solvent ratio varied between 5 and 30 g/100 ml. The yield of ethanolic extract increased rapidly up to a solid-liquid ratio of 20 g/100 ml, after which a slight increase in yield was observed. At 20 g/100 ml solid-liquid ratio, 75.6% of total ethanolic extract was extracted by UAE; any further increase in the ratio increased the oil yield only marginally (Figure 1a). Therefore, a solid-liquid ratio of 20 g/100 ml was selected as the most suitable ratio for further tests. The most suitable liquid-solid ratio for UAE extraction of oil from tara seed using petroleum ether was 14:01 (Li et al., 2016).

UAE temperature affects the interaction between the solid and liquid. When the ultrasonic temperature was increased from 30 to 70 °C, the yield of ethanolic extract increased from 62.4 to 81.3% (Figure 1b), whereas extraction at 80 °C caused a slight decrease (0.1%) in yield. 70 °C was selected as the most suitable temperature for subsequent experiments. Solvent and ethanolic extract viscosity decrease as temperature rises, improving its fluidity and facilitating its flow through solid matrices as well as raising the diffusion coefficients of the compounds that can be extracted from lipophiles. But at temperatures > 70 °C, the dissolution of cell impurities and decomposition of some constituents may increase and too high temperatures would enhance the speed of bubble collapse in the solvent, which may hamper the strength of the bubbles and thus decrease the yield (Goula et al., 2017; Chutia and Mahanta, 2021).

Extraction times varying between 5 and 90 min were used to investigate the effect of ethanolic extract yield (Figure 1c), and the extraction yield was observed to increase with time. This might be because greater time allows the ultrasonic wave to penetrate cell walls and release the cell contents. The extraction yield was rapid up to around 30 min and thereafter the rate slowed down, which was also reported by Raj and Dash (2020). To extract approximately 75% of total oil, it required approximately 30 min and to extract the next 12% ethanolic extract, it required another 60 min. So, 30 min was chosen as the most suitable UAE time.

It is evident that when treatment time was extended at a steady temperature, oil extraction yield rose until it reached saturation (Figure 2a). Higher yield is a result of the solvent’s viscosity being reduced, with the extracted oil being more fluid and soluble, and the solvent’s diffusivity and mass transfer being improved. Similarly, higher treatment time means prolonged contact or exposure to the solvent (Goula et al., 2017; Chutia and Mahanta, 2021). To extract approximately 50% total extractable oil, it required approximately 120 min, to extract the next 30% oil approximately another 2 h was required. For a further 20% extract, it required another 4 h.

In this extraction process, optimized UAE extraction was performed for the first 30 min followed by Soxhlet extraction for 4 h (Figure 2a). The results revealed that combined UAE-Soxhlet extraction was able to extract approximately 95% of oil within a short period of time (4 h), and in comparison, the Soxhlet extraction process would require approximately 8 h. Using UAE only, it was very difficult to extract more than 85% of the total oil. The overall improvement may be due to the synergistic effect of ultrasound (due to breakdown of the cell and better penetration of the solvent) and Soxhlet extraction processes (due to high temperature and better solvent movement) (Reis et al., 2020).

Figure 2b illustrates the variation in oil yields from PFS using different extraction techniques with ethanol as solvent. For all extraction processes, oil yield increased with extraction time. Using UAES process, overall extraction rate was higher compared to SE and UAE alone. The positive effect of UAES process on the extraction yield may be because UAE disrupted the cell wall and enhanced the accessibility of oil, and the later use of higher temperature in the Soxhlet method decreased the viscosity of the solvent and enhanced the overall fluidity (Djenni et al., 2013) and solubility of the oil and gave synergetic effects. As a result, it enhanced the diffusivity and mass transfer (Hlaváč et al., 2019). This study suggested that combined ultrasound and Soxhlet with ethanol as solvent can be used as a novel technology for higher extraction efficiency within a short time. UAE followed by Soxhlet extraction can be easily scaled up for industrial use (Deenu et al., 2013).

The ethanolic extract extraction mechanism from PFS consisted of two stages. In the initial stage, rapid extraction of the extractable materials located at the surfaces occurred, and then the extractable materials from intact cells slowly diffused to the surface of the particles (Marković et al., 2019; Silou et al., 2021). This observation was very important for subsequent modelling of the extraction kinetics.

The kinetics of oil by UAE (UAEO), Soxhlet (SEO), and UAES (UAESO) was described by the phenomenological models represented in Table 1. The values for the model parameters of Equation 8 and the statistical criteria used for assessing the goodness of fit (R ², MRPD, and MSE) are presented in Table 1. The experimental and predicted values for oil yield agreed very well with each other (Figure 2c) as the high coefficient of determination (R ² > 0.997), low MSE (< 0.1118), and the acceptable MRPD (< ±1.79-1.55%) implied the good fitting that explained the extraction phenomena using the phenomenological model. This can be visually observed in Figure 2b. The ethanolic extract extraction yield at saturation point increased with extraction techniques, which may be due to the increase in the extraction temperature as Soxhlet extraction temperature was higher than UAE.

The fraction of washable extractable substances decreased with the increase in the extraction temperature, for both Soxhlet and UAE extraction techniques. The “f1” values were higher for the UAE (0.7320) followed by combined UAES (0.7259) and SE (0.0096) (Table 1), which demonstrated the influence of ultrasound on the washing extraction process (Milić et al., 2013). The k ₁ and k ₂ values for all extraction processes ranged from 0.0807 to 0.1139 min ^-1 and 0.0050 to 0.0054 min ^-1, respectively, and for all extraction processes k ₁ >> k ₂, which emphasized the predominance of the washing process. The k ₁ value was highest for UAES extraction followed by UAE and SE extraction processes, in that order, which indicated the synergistic effect of UAE and SE on the washing process as well as on the overall extraction process. Milić et al. (2013) observed a similar trend for UAE and conventional extraction processes. Marković et al. (2019) found similar values for phenomenological model parameters where k ₁ was 0.1753 min ^-1, k ₂ was 0.0173 min ^-1 and f1 was 0.619 during the extraction of essential oil from juniper.

Two simpler models were developed, namely the pseudo-first order model (Equation 9), where the washing process was neglected, and the model based on instantaneous washing followed by diffusion, where the diffusion process was neglected (Equation 10). These models were evaluated for the purpose of comparison with the phenomenological model (Equation 8). The kinetics parameters of these two simpler models are presented in Table 2. The k ₂ value for the pseudo first order model ranged between 0.0083 and 0.0754 min ^-1, whereas the same parameter value for the model based on instantaneous washing followed by diffusion ranged from 0.0081-0.0714 min ^-1 (Table 2). For each extraction process, the k ₂ value for the pseudo first order model was found higher than the model based on instantaneous washing followed by diffusion process. The pseudo-first order model and instantaneous washing followed by diffusion model had a small R ² value and relatively high MSE and MRPD (%) values compared to the phenomenological model, regardless of the adopted extraction techniques. Therefore, only the pseudo first order model and instantaneous washing followed by diffusion models were less efficient compared to dual extraction (phenomenological model) for describing the extraction kinetics of oil from PFS. Thus, the phenomenological model better explained the variation in oil yield than the simpler models and could be recommended for modelling extraction kinetics.

The physicochemical properties of the optimized ethanolic extract (UAESO) and conventional extraction (SEO) were calculated and are reported in Table 3. The acid values for the optimized UAESO and SEO were 2.4 and 2.8 mg KOH/ g oil, respectively (Table 3), which were within the permissible limit of 4.0 mg KOH/ g for crude oil as set by Codex Alimentarius (2015). Similarly, the peroxide value for the extracted oils was within the permissible range.

Between the two extraction methods, the UAESO demonstrated better oil quality. The low acid value and peroxide value for PFSO suggested the good quality and edibility of the oil. The scent and flavor of the oil are produced by free fatty acids. The percentage of unsaturated substance in oil is determined by the iodine value for the oil (Purohit et al., 2021). For combined extraction, the iodine value for UAESO was determined to be 130.2 g I₂/100 g oil and for SEO 139.4 g I₂/100 g oil (Table 3).

Purohit et al. (2021) discovered the value between 126.9 and 131.4 g I₂/100 g oil for yellow passion fruit seed oil. The PFS possessed high values for iodine (124.67±0.67-131.00±0.58 g I₂/100 g) and peroxides (1.43±0.12-3.23±0.12 meq O₂/kg), which were similar to other edible seed oils as reported by Ramaiya et al. (2019).

The refractive index of PFSO of UAESO and SEO was 1.469 and 1.54, respectively (Table 3), which was within the range of refractive index for most vegetable oils. This value indicated the presence of long-chain unsaturated fatty acids and also intermolecular interaction in PFSO (Purohit et al., 2021). The density of UAESO was found to be higher than SEO. The lower antioxidant activity observed in this study may be attributed to the high temperature used for extraction in the Soxhlet apparatus (Ramaiya et al., 2019).

The total carotenoid contents of UAESO and SEO were 5.8 and 5.1 µg β-carotene/g of oil, respectively, and β -carotene was the major carotenoid in the seed oils. Similarly, 6.70 µg β-carotene/g sample of oil was reported by da Silva and Jorge (2016). Total phenolic contents, total carotenoid contents and overall antioxidant activity of the oil extracted by the UAES technique was higher as compared to the Soxhlet process. Prolonged heating in the Soxhlet process (SEO) may degrade the total phenolic content and total carotenoids, which affected the overall antioxidant activity of the oil (Goula et al., 2017; Chutia and Mahanta, 2021).

The total phenolic content in PFSO was 33.57 mg GAE/g of oil, whereas 36.02-39.11 mg GAE/g of total phenolic content in yellow passion fruit seed oil from the northeast region was reported by Purohit et al. (2021).

Four major phenolic compounds were detected by the HPLC method, namely gallic acid (20%), salicylic acid (16.4%), piceatannol (54.6 %) and kaempferol (8%). da Silva and Jorge (2016) found three main phenolic acids: caffeic acid, coumaric acid and salicylic acid, with salicylic acid being the major phenolic acid. de Santana et al. (2017) identified stilbene piceatannol as the major compound in yellow passion fruit seed extract, accounting for 50% of the total phenolic content.

The fatty acid composition and its relative percentage in yellow PFSO are presented in Table 4. The fatty acid methyl ester (FAME) composition indicated the presence of unsaturated fatty acids, which was also observed from the high iodine value. The presence of long chain unsaturated fatty acids in the PFS indicated by the refractive index values was confirmed by the presence of high amounts of linoleic and oleic acids. The major fatty acids found in UAESO were linoleic acid (67.3%), oleic acid (16.1%), palmitic acid (13.4%), stearic acid (2.5%), and linolenic acid (0.6%). The total content of saturated fatty acids (SFA) was 15.9%, while the content of unsaturated ones (UFA) was 84%. Palmitic acid was predominant among the SFAs. The results agreed with the report of Purohit et al. (2021) and Pereira et al., (2017).

The COX value for UAESO was 7.2, which was calculated by considering the percentages of oleic , linoleic , and linolenic acids present in the PFSO. Higher unsaturated fats lead to an increase in the COX value in oils. UAESO was, therefore, stable and the process can be employed for the prevention of oxidative deterioration. Purohit et al. (2021) reported almost similar COX values (7.4) in PFSO. In comparison, the COX values for fresh palm oil, peanut oil, and camellia oil are 1.615, 4.631, and 1.772, respectively (Xu et al., 2015).