Separation and Purification Technology 169 (2016) 187195

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier.com/locate/seppur

Influence of the bed height on the kinetics of watermelon seed

oil extraction with pressurized ethanol
J. Colivet a,b, A.L. Oliveira a, R.A. Carvalho a,
a
Faculdade de Zootecnia e Engenharia de Alimentos, Universidade de So Paulo, FZEA/USP, Caixa Postal 23, CEP: 13635-900, Pirassununga, So Paulo, Brazil
b
Escuela de Zootecnia y Tecnologia de Alimentos, Universidad de Oriente, Campus Los Guaritos, cdigo postal 6201, Maturn, Monagas, Venezuela

a r t i c l e i n f o a b s t r a c t

Article history: The influence of the bed volume on the kinetics of watermelon seed oil extraction was evaluated in this
Received 15 April 2016 study. Watermelon seed oil was extracted using different extraction cells (S1 = 34 mL, S2 = 66 mL and
Received in revised form 3 June 2016 S3 = 100 mL) at different temperatures (40, 60 and 80 C) using the sample mass/solvent volume ratio
Accepted 6 June 2016
as the fixed sizing criteria (w/s = 0.30). The extraction kinetics were mathematically described using
Available online 7 June 2016
the Peleg, Fick and second-order models. Samples were extracted in batches using pressurized liquid
extraction (PLE) with ethanol as the solvent for different extraction times. Oil extraction yields ranged
Keywords:
from 24.69 to 37.21 g oil/100 g of seeds, and the concentration of total phenolic compounds ranged from
Oil
Extraction
2.44 to 3.84 mg of gallic acid equivalents (GAE)/g of seeds. All kinetic models showed a good fit to the
Citrullus lanatus experimental data, but the second-order model better predicted the behavior of data, with high coeffi-
Green process cient of determination (R2ajd ) and low root-mean-square deviation (RMSD) values. The different extraction
Polyphenols cells did not affect the total extraction yield, but affected the extraction parameters obtained in the
models. The effective diffusivities were dependent on temperature and ranged from 9.10
106 to
2.07
105 m2/s. The activation energy ranged from 11.43 to 18.54 J/mol.
2016 Elsevier B.V. All rights reserved.

1. Introduction The extraction and purification of active compounds from natu-

Watermelon is a fresh fruit consumed worldwide and is used in
the production of juices, jellies, marmalades, sauces and salads [1].
In addition, watermelon has been widely used as a medicinal plant
in African and Asian cultures due to the presence of several
compounds with phytochemical activities [2]. Watermelon seeds
have strong antioxidant activity [3,4], and due to their diuretic
and purgative activity, have also been used in the treatment of
gastrointestinal diseases [5], urinary infections [6], gonorrhea,
leukorrhea [7] and prostatic hyperplasia [8]. Watermelon seed oils
are classified as high quality oils due to the presence of x-3 and
x-6 fatty acids [9] and phenolic compounds such as gallic,
protocatechuic, p-hydroxybenzoic, vanillic, caffeic, syringic,
p-coumaric and ferulic acids [10]. The processing of the water-
melon fruit yields a large amount of seeds, which are usually trea-
ted as waste. The use of these seeds to produce oil and other
functional ingredients will add value to this waste produce that
currently has no specific use, so that it can become a raw material
used for the generation of a product with active compounds.
Corresponding author.
E-mail address: rosecarvalho@usp.br (R.A. Carvalho).
ral sources are important steps in the production of phytochemi-
cals for use in food supplements or nutraceuticals, functional
foods and pharmaceuticals [11]. In industry, oil extraction pro-
cesses are based on conventional methods that involve the use of
organic solvents which are heated at atmospheric pressure condi-
tions, and production can also occur via mechanical processing
such as pressing. High-pressure extraction processes, such as pres-
surized liquid extraction (PLE) and extraction with supercritical
CO2, are extraction techniques that are becoming more prominent
due to their ability to obtain specific target molecules and reduced
loss of solvents during the production process [12]. PLE is a tech-
nique that has emerged as an alternative to conventional extrac-
tion methods, such as maceration, percolation or reflux, as it
offers advantages for parameters including extraction time, solvent
consumption, extraction yield and reproducibility [13]. PLE uses
organic solvents at elevated pressure and temperature in order to
increase the extraction process efficiency, reduce viscosity and
improve solvent penetration and diffusivity, thereby reducing
extraction times and avoiding possible thermal degradation
[13,14]. PLE can be used for the extraction of polar compounds
such as polyphenols, however, this is a little studied technique
which still requires more research before its implementation on
a larger scale [12]. An advantage of PLE is the possibility of using

http://dx.doi.org/10.1016/j.seppur.2016.06.020
1383-5866/ 2016 Elsevier B.V. All rights reserved.
188 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195
Nomenclature
A average yield of oil extracted in the washing step
(g oil/100 gdb)
An, Bn model-fitting parameters (An, dimensionless; Bn, s1)
B0 extraction rate at the beginning of extraction (t = to)
C0 yield of oil extracted in the washing step (g oil/100 gdb)
Cs extraction capacity at equilibrium (g oil/100 gdb)
Csp extraction capacity of the Peleg model at equilibrium (g
oil/100 gdb)
Ct extraction yield at any time (g oil/100 gdb)
D0 pre-exponential diffusion constant (m2/s)
De effective diffusivity coefficient (m2/s)
IC50 antiradical activity (mg/L)
h initial extraction rate for the second-order model
(g/100 gdb minute)
k second-order extraction rate (100 gdb/g oil minute)
K1 constant of the Peleg model (min 100 gdb/g oil)
K2 constant capacity of the Peleg model (100 gdb/g oil)
any solvent for the fractionation of phytochemicals. An example of
this is the ability to perform successive extraction steps in the
same sample using the same or different solvent for each extrac-
tion step, which yields a variety of compounds with different prop-
erties. Another advantage of using PLE is that it significantly
modifies the properties of the extraction solvents, offering the pos-
sibility of using polar solvents for the extraction of compounds
with hydrophobic characteristics (e.g. lipids, pigments and vita-
mins). Pronyk and Mazza [15] reported that pressurized liquids
have high versatility due to the physicochemical properties of
the solvent, as the density, diffusivity and dielectric constant can
be controlled by varying the pressure and extraction temperature,
therefore, changing these properties can effectively control the sol-
vation power and selectivity of solvents.
Considering the advantages of using PLE as an extraction tech-
nique, it is important to carry out studies to establish equipment
design criteria, in addition to the industrial and pilot scale, in order
to obtain the information required for its implementation. The
design criteria takes into account the extraction processes and
the effect of the extraction bed geometry on the kinetic parameters
that define the process profile used, and this information can be
used in relation to the physicochemical properties of the products
obtained. Thus, the aim of this study was to investigate the influ-
ence of the bed height on the kinetics of watermelon seed oil
extraction with pressurized ethanol at different temperatures,
through the use of mathematical models that describe the extrac-
tion process. In addition, we also studied the effect of bed height
and temperature on the kinetic parameters, total content of pheno-
lic compounds and antiradical capacity of the oil obtained.
2. Materials and methods
2.1. Sample preparation
Watermelons of the Ruby variety were purchased from a local
market in the city of Pirassununga (Sao Paulo, Brazil). The seeds
were manually extracted by slicing the watermelons with a stain-
less steel knife, after which the juice was extracted from the pulp.
The seeds were separated from the residue using a sedimentation-
flotation system in plastic containers with potable water. The seeds
were then dried at 60 C for 24 h, after which they were ground
using a Wiley mill (Thomas Scientific, Philadelphia, PA, USA) with
a 1 mm sieve. Finally, the powder (1% moisture) was packaged
and stored under refrigeration at 5 C [16].
Rp average radius (m)
Rg gas constant (J/mol K)
T temperature (C or K)
t extraction time (min)
DEa activation energy (J/mol)
Abbreviations
Abs absorbance to 765 nm
DPPH 2,2-diphenyl-1-picrylhydrazyl radical
GAE gallic acid equivalents
GRAS generally recognized as safe
PLE pressurized liquid extraction
RMSD root-mean-square deviation
S1 34 mL extraction cell (5.2
2.8 cm)
S2 66 mL extraction cell (9.8
2.8 cm)
S3 100 mL extraction cell (15.8
2.8 cm)
2.2. Pressurized liquid extraction
Extractions were performed using an accelerated solvent
extraction system, ASE 150 (Dionex, Sunnyvale, CA, USA). Extrac-
tions were performed at three different temperatures (40, 60 and
80 C), using three cylindrical cells (S1, S2 and S3), keeping the
mass/solvent ratio constant (0.30) over 8 extraction cycles with
total solvent renovation at different times (3, 6, 12, 18, 24, 30, 36
and 42 min). The volume of solvent used in each extraction cycle,
was approximately 31, 52 and 78 mL for cells S1, S2 and S3, respec-
tively. In this study, ethanol was selected as it is a generally recog-
nized as safe (GRAS) solvent [17,18].
Initially, the fixed bed was packed with ground sample, then
solvent was placed into the cell and the system pressure was con-
trolled at 1500 psia (102.4 atm). Temperature conditions were
adjusted according to experimental conditions. N2 was used to dis-
charge the cell solvent, and the system was finally depressurized in
order to avoid the presence of the remaining extract in the cell.
After extraction, the cell was flushed thoroughly before the next
extraction cycle. A rotary evaporator (model IKA RV 05 IKA; IKA-
Werke, Staufen, Germany) was used to evaporate the solvent from
the extract at 40 C, and the yield was determined as the ratio
between the extract obtained from the seed mass present in the
fixed bed extractor.
2.3. Total phenolic content
The phenolic content was determined using the Folin-
Ciocalteau reagent [19]. An aliquot of extract sample was mixed
with distilled water (2 mL) and 1 mL of Folin-Ciocalteu reagent.
After 3 min, 2 mL of sodium carbonate (20%) was added and stirred
with a vortex. Solutions were stored at room temperature for 1 h in
the dark and the absorbance was determined at 760 nm. Gallic acid
(diluted in ethanol) was used as a standard solution for the prepa-
ration of the calibration curve at concentrations from 0 to 80 mg/L
(R2 = 0.997). Results were expressed as mg of gallic acid equiva-
lents (GAE) per gram of seed. The quantification of total phenolics
was performed in triplicate.
2.4. Antiradical activity
The antioxidant activity was determined using the 2,2-
diphenyl-l-picrylhydrazyl (DPPH
) reagent as the free radical [20].
For each extract obtained, different concentrations were tested
 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195 189

 
(in mg extract/L of solution). An aliquot (1 mL) of each of the C0
A 1 A1  eB1 t0 6
extracts was added to 3.0 mL of an ethanolic DPPH
solution Cs
(40 mg/L, diluted in ethanol). The decrease in absorbance was
determined at 517 nm after 90 min. The exact DPPH
concentration 6
A1 7
(C
DPPH) in the reaction medium was calculated using a calibration p2
curve Abs517nm 0:0227CDPPH
0:13 (R2 = 0.998), obtained by the
The B1 coefficient is associated with the effective diffusivity,
dilution of DPPH
in ethanol to concentrations ranging from 31 to
defined in Eq. (8)
8 ppm. The antiradical activity was defined as the amount of
extract required to reduce the DPPH
concentration by 50% (IC50, De p2
B1 8
mg/L). R2p

2.5. Kinetic models
In the modeling study of extraction kinetics, three mathemati-
cal models were used (Table 1). The second-order model (Table 1)
describes the diffusion rate of oil present in the ground water-
melon seeds. In this model, k is the second-order extraction rate
expressed as 100 g of seeds/g oil min, Cs is the extraction capacity
and Ct is the oil concentration at time t (minutes). Note that the
experimental data for oil concentrations are expressed in g of oil
per 100 g of seeds.
Considering the limiting conditions from t = 0 to t and Ct = 0 to
Ct, the second-order model may be transformed into Eq. (1), which
becomes Eq. (2) when linearized.
C 2s kt
Ct
1 C s kt
Rp is the average particle radius (m) and De is the effective diffusiv-
ity coefficient (m2/s) for the experimentally determined B1 values
after linearization of the Fick model (Table 1).
Finally, the Peleg model was also applied to predict the water-
melon seed oil extraction behavior. This model was used by
Bucic-Kojic et al. [23] in polyphenol extraction experiments with
a good fit to the experimental data. In this model, Ct is the oil con-
centration at time t, K1 is the Peleg constant ratio (min 100 g seeds/
g of oil) and K2 is the Peleg capacity constant (100 g seeds/g oil).
Constant K1 is related with extraction rate B0 at the beginning of
extraction, (t = 0), as shown in Eq. (9).
1
B0 g of oil=100 g of seed min 9
K1
The Peleg capacity constant K2 is related to the maximum
1
extraction yield, defined in Eq. (10).

t 1 t 1
2 C sp g of oil=100 g of seed 10
C t kC 2s C s K2

In the arrangement of Eq. (2), the extraction rate can be deter-

mined according to Eq. (3). 2.6. Statistical methods

Ct 1 To evaluate the effect of the extraction cell and temperature, a


 3
t 1
t randomized block design for two factors (extraction cells and tem-
kC 2s Cs
perature) was used. Analysis of variance (ANOVA) followed by
The initial extraction rate (h), with Ct/t at a time of approxi- Tukeys test were performed to determine the effect of these fac-
mately 0, is defined in Eq. (4). tors on the dependent variable at a 95% significance level using
2
the Statistica Software 10.0 (Stat Soft Inc., USA). All experiments
h kC s 4 were performed in triplicate.
Similarly, the oil concentration at any time can be calculated The mathematical models for the study parameters were deter-
according to Eq. (5). mined by fitting the experimental data using non-linear regression
by the Levenberg Marquardt method [24] using the Statgraphics
t Centurion XVI software (Statpoint Technology Inc., USA). The
Ct  
 5
1
h
t
Cs
agreement between the experimental data and calculated values
were determined by the determination coefficient (R2adj ) and the
All parameters of the second-order extraction model were root-mean-square deviation (RMSD), according to Eq. (11).
determined by the intersection of the line and the slope of the r
Ct/t vs. t graph. 1 Xn 2
Another model tested was the modification of Ficks law
RMSD experimental  calculated 11
n i1
(Table 1) [21]. In this model, Ct and Cs are the extraction yields at
time t and at equilibrium, and are both expressed as g oil/100 g
of solids (ground seeds). A and B1 are model coefficients where A 3. Results and discussion
is associated with the average concentration of oil extracted from
the surface of particles that are dragged during extraction (wash- 3.1. Total yield of oil and phenolic compounds from watermelon seeds
ing), described by Eq. (6). The value of A1 when a spherical surface
is adopted is described by Eq. (7) [22]. The aim of this study was to compare the size of the extraction
bed on oil extraction yield from watermelon seeds at different
temperatures, using the sample mass/solvent volume ratio as the
Table 1 fixed sizing criteria. The different lengths of the fixed bed extrac-
Models used to describe the kinetics of watermelon seed oil extraction.
tors did not significantly affect the watermelon seed oil yield
Model Equation Parameters Reference (p > 0.05) (Table 2). The extraction yield was mostly affected by
Second-order dC t
kC s  C t 2 k, h [44] the temperature at which the extraction was performed. Overall,
dt
Ficks law Ct B1 t A, B1 [21] the average extraction yields for all cells were 25.7, 30 and 37 g
C s 1  Ae
Peleg C t K 1 K
t K1, K2 [23,45]
oil per 100 g of seeds on a dry base for extraction temperatures
2t
of 40, 60 and 80 C, respectively. The extraction yield values are
190 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195
Table 2
Extraction yields of oils and phenolic compounds in watermelon seeds.
Temperature (C) Bed Yield (mean SD)*
Oil (g oil/100 gdb) Total phenolic (mg GAE/gdb)
40 S1 25.8 0.9 a 3.8 0.06 g
S2 24.7 0.5 a 3.1 0.06 c.d.e
S3 26.6 1.1 a.b 3.4 0.07 f
60 S1 28.7 1.4 b.c 3.0 0.04 c.d
S2 30.3 0.7 c 3.2 0.01 e.f
S3 31.1 0.6 c 3.0 0.07 b.c
80 S1 36.8 1.4 d 2.8 0.04 b
S2 37.0 0.8 d 2.4 0.04 a
S3 37.2 0.4 d 3.2 0.14 d.e.f
*
Experiments were performed in triplicate. The means within a column followed
by the same letter are not significantly different at the p = 0.05 level. S1 = 34 mL,
S2 = 66 mL, S3 = 100 mL.
comparable to the 32.16% yield previously reported by Conto et al.
[25] from seeds of the Top Gun watermelon variety, achieved
using Soxhlet extraction with hexane as the solvent. However,
the extraction yield values were lower than those obtained by Acar
et al. [3] (52.34%) and Rai et al. [1] (52.37%), who used the Soxhlet
extraction and supercritical fluid methods, respectively. This vari-
ation could be attributed to differences in the species of water-
melon and differences in growth and geographical conditions,
including soil properties and nutrients, water availability and sea-
sonal temperature, in addition to other relevant factors [26].
The concentration of phenolic compounds ranged from 2.4 to
3.8 mg GAE/g of the seeds, regardless of the different extraction
cells (Table 2), and this yield is higher than values which have pre-
viously been reported (62 mg GAE/g of seed) [3,27,28]. However,
when the effect of processing temperature on the concentration
of phenolic compounds was considered, we found that the concen-
tration of these compounds was higher for extraction temperatures
below 80 C, which could be related to thermal degradation of
phenols by the effects of high temperatures and long extraction
times [29].
 te et al. [30] suggested that the extraction tempera-
Kryzeviciu
ture can affect the content of phenolic compounds extracted from
pomace raspberry using PLE. However, this performance may
depend on the solvent used. The use of pressurized hexane resulted
in higher yields when temperatures were above 90 C, however,
the temperature did not significantly affect the results for pressur-
ized methanol, although the highest total yield of phenolics was
obtained at temperatures below 70 C with a longer extraction
time (15 min).
3.2. Antiradical activity
The effect of extraction time on the antioxidant activity of
watermelon seed oils is shown in Fig. 1(AC). The analysis demon-
strated that the different extraction cells did not affect the antiox-
idant activity of oils obtained. However, extraction temperature
and time had a significant effect on the antioxidant activity of com-
pounds in the oil extracts. Oil obtained at 60 C had a higher
antioxidant activity compared to that obtained at temperatures
of 40 or 80 C (Fig. 1AC). The oil obtained at 40 C showed higher
activity than that obtained at 80 C. Therefore, the antioxidant
behavior was found to be related to the extraction temperature,
as lower temperatures produced decreased amounts of bioactive
compounds. According to Mustafa and Turner [31], high tempera-
tures decrease the viscosity and increase diffusivity of the liquid by
the matrix, thus facilitating extraction, which was not observed at
40 C. For higher temperatures, although the diffusivity was higher,
extraction at 80 C may cause degradation of compounds with
antioxidant activity [29].
After the matrix was submitted to successive extractions with
pressurized ethanol, the antioxidant activity was found to increase
considerably for the 24 and 18 min extractions at temperatures of
40 and 60 C, respectively (equivalent to 4 and 5 extraction cycles,
respectively), and the activity was found to increase at a slower
rate up to 42 min for the extraction temperatures of 40 and
60 C. However, when the temperature used was 80 C, the maxi-
mum antioxidant activity was found to occur after 12 min of
extraction (3rd cycle), which was reduced due to the higher heat
exposure time. In PLE, static cycles are used to introduce the sol-
vent when the solubility of target compounds is limited, which
facilitates the penetration of the solvent into the matrix [32]. Thus,
this study showed that using different static cycles improved the
extraction capacity of the inhibitor compounds of the DPPH free
radical activity. Furthermore, according to Carabias-Martinez
et al. [33], the application of several extraction cycles and the
use of ethanol as solvent helps to reduce the presence of hydropho-
bic compounds (e.g. resins) in the extracts, which interfere with
the subsequent extraction of compounds with phytochemical
activity.
3.3. Extraction kinetics
Fig. 2 shows the oil extraction yields for the different cells (S1, S2
and S3) at the different extraction temperatures tested (40, 60 and
80 C). The experiments suggested that the extraction yields were
dependent on time for all three extraction cells. We observed that
the extraction rate was higher in the first 6 min, which was
reduced until equilibrium was reached (18 min). This behavior
is associated with the concentration gradient produced in the ini-
tial extraction stage, where the solvent extracts components to
the surface of the matrix and begins to penetrate into the solid
[34]. According to previous studies, the rapid increase in extraction
performance in the initial extraction stage is mainly attributed to
washing of extracted components located on the external surface
of particles and diffusion of the solute dissolved inside broken par-
ticles [3537]. However, the decrease in the extraction rate is
mainly controlled by diffusion of oil into the particles [3537].
Moreover, when the three extraction cells were compared there
were no differences in the behavior of the process. In this study,
the mass/solvent ratio was considered to be constant, indicating
that this can be used for scaling up in PLE. According to Zabot
et al. [38], assessment of the bed geometry in the extraction pro-
cess is important for scaling an extraction process. Similar extrac-
tion profiles should be obtained regardless of the geometric
characteristics of the bed. However, the appropriate scaling criteria
for similar kinetic parameters must be chosen, in addition to mod-
eling the process.
3.4. Mathematical models
3.4.1. Second-order model
The second-order model showed a good fit to the experimental
data (Fig. 2A), demonstrating that it is easily adjustable to extrac-
tion processes with pressurized liquids (R2adj > 0.964). The parame-
ters Cs (oil concentration at equilibrium) and k (extraction rate
constant) were determined by fitting the second-order model
(Table 3) to the experimental data. The parameter Cs was used to
measure the extraction efficiency compared to the total yield
obtained. Overall, these differences decreased with increasing tem-
perature, which indicates that a more severe heat treatment
improves the extraction efficiency, however, the extraction of com-
pounds other than oil cannot be excluded.
In relation to the constant k and extraction parameter h (initial
extraction rate) we found that the values of these parameters
increased with temperature, but there was not a linear relationship.
 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195 191

Fig. 1. IC50 index variations with changes in temperature and size of the extraction cell for (A) 40 C, (B) 60 C and (C) 80 C.

Although there is little information in the literature on the applica-
tion of second-order models to extraction systems with pressurized
fluids, the initial extraction rates and the extraction rates may be
higher than that observed for other extraction systems. Sayyar
et al. [39] used a second-order model to predict the extraction per-
formance of Jatropha curcas seed oils using a solid-liquid extraction
method, and they obtained initial extraction rates of 2.94, 3.60 and
4.21 g/L per minute for extraction temperatures of 45, 60 and 68 C,
respectively. This behavior indicates increasing extraction rates
with increasing temperature, which was also reported in the cur-
rent study.
The parameters Cs, k and h were minimally affected by the size
of the extraction cell, however, there was no linear relationship
between the length of the cells with these parameters at the dif-
ferent temperatures measured. At temperatures of 40 and 80 C,
the larger extraction cells resulted in a reduction in the extraction
rate (k) and initial extraction rate (h). This behavior has also been
shown by Berna et al. [40] and Kister et al. [41], who reported that
192 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195

Fig. 2. Extraction curves from watermelon seed oil, obtained with different mathematical models under different experimental conditions for the (A) second-order model, (B)
Peleg model and (C) Ficks model.

the mass transfer rate may decrease with increasing extraction
bed length. However, the same behavior was not observed at
60 C, which may be associated with the heterogeneity of the
extraction bed agglomeration, which makes mass transfer within
particles difficult [40].
3.4.2. Peleg model
The prediction of the watermelon seed oil extraction behavior
by PLE using the Peleg model showed a good fit to the experimen-
tal data (Fig. 2B). The parameters obtained from the model (K1, K2,
B0 and Csp), correlation coefficients and RMSD are shown in Table 4.
Correlation coefficients were high for all experiments (R2adj > 0.964),
suggesting a good fit of the experimental data to the model. K1 val-
ues represent the process speed and K2 values are related to the
maximum extraction yield [23]. The Csp and B0 values obtained by
the Peleg model have the same physical meaning of parameters Cs
and h found in the second-order model. These were temperature-
dependent, but could not be linearly related to temperature. The
extraction temperature affected both the parameters K1 and K2.
The values for the parameters K1 and K2 decreased with increasing
 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195 193

Table 3 C0
Thus, the ratio Cs
was calculated from coefficient A by Eq. (12).
Parameters of second-order models related to the extraction of watermelon seed oil.
Accordingly, the values obtained at 40 C were approximately zero,
Temperature Bed Cs (g oil/ k (100 gdb/g h (g/ R2adj RMSD and when calculated at the higher temperature the values ranged
100 gdb) oil min) 100 gdb
between 0.35 and 0.64. These values were higher than those
min)
reported by Fernndez et al. [42] for canola oil (0.24), but lower
40 C S1 29.857 0.0057 5.0007 0.964 1.5679
than those obtained by Perez et al. [21] (>0.68). Although these
S2 29.028 0.0047 3.9086 0.995 0.5267
S3 31.365 0.0040 3.9540 0.981 1.1270 previous studies did not find a linear relationship between this
parameter and the extraction temperature, these variations were
60 C S1 30.348 0.0142 12.8227 0.989 0.8995
S2 31.002 0.0278 26.7001 0.996 0.5861 related to the structure of the material [42].
S3 31.889 0.0282 28.6572 0.989 0.5884
C0 p2
80 C S1 37.567 0.0196 28.5180 0.992 0.7710 1A 12
Cs 6
S2 38.326 0.0167 24.5132 0.997 0.5792
S3 38.415 0.0162 23.8953 0.999 0.3380 From the coefficient B1, the effective diffusivity could be calcu-
lated using Eq. (9). Diffusivities increased with temperature, and
were 9.1
106, 9.70
106 and 2.10
105 m2/s for tempera-
temperature, indicating that the extraction rate and yield increased tures of 40, 60 and 80 C, respectively (see Table 5). These values
with increasing temperature (see Table 4). were much higher than those obtained by conventional extraction
There are limited previous studies which have used the Peleg techniques in different seeds. Perez et al. [21] and Fernndez et al.
model to predict seed oil extraction behavior, however, our results [42] found values between 1.3
1012 and 3
1012 and
show that PLE quickly increased the process efficiency compared to 1.4
1012 and 2.1
1012 m2/s for sunflower and canola oil
conventional methods. Amarante et al. [36] studied the process of extractions, respectively. These differences are probably related
castor oil extraction using a solid-liquid extraction method to stir to the fact that extraction under high pressure conditions increases
at atmospheric pressure, and they reported K1 values from the mass transfer rate during extraction.
0.8062 to 1.7921 min 100 gdb/g oil for extraction temperatures The effective diffusivity increased with increasing extraction
between 20 and 55 C, which is equivalent to the extraction speed temperature, which adjusted the Arrhenius equation (Eq. (13)).
of between 0.55 and 1.24 g of oil/100 gdb per minute obtained in
this experiment. D Ea
De D0 e R T
g 13
Regarding the size of the extraction cell, we observed that an
increase in the size of the extraction cell produced variations in where D0 (m /s) is the pre-exponential constant, DEa (J/mol) is the
2

K1 and K2. The Peleg (K1) constant slightly increased with the size activation energy, Rg (J/mol K) is the gas constant and T (K) is the
of the extraction cell at temperatures of 40 and 80 C, but the same absolute temperature.
effect was not observed at 60 C. The Peleg capacity constant (K2) After using the Arrhenius equation, the values of the coefficient
slightly decreased with increased size of the extraction cell, but were 1.014
102, 0.96
103 and 0.78
103 m2/s for the
these differences were not statistically significant for an effect on extraction cells S1, S2 and S3, respectively, with a decrease in this
the overall extraction yields. variable observed with decreasing size of the extraction cell. Sim-
ilarly, the activation energy also decreased with the size of the
extraction cell, yielding values of 18.55, 12.28 and 11.43 J/mol for
3.4.3. Ficks model
the extraction cells S1, S2 and S3, respectively (R2 > 0.80). These
Fig. 2C shows the kinetic curves constructed according to Ficks
values are 1000 times smaller than those reported by other authors
model. The predictions made by this model were well adjusted to
for different oils extracted by conventional techniques under
all experimental conditions (R2adj > 0.850), however, this model was
less accurate when compared to the second-order and Peleg models, Table 5
as shown by the overestimation of extraction yields under some Parameters of the Fick model related to the extraction of watermelon seed oil.
experimental conditions.
Temperature Bed B1 De (m2/s) DEa (J/mol) R2adj RMSD
The coefficient A values, related to the oil extraction in the ini-
6
tial phase (washing step) [21], decreased considerably with 40 C S1 0.0323 9.10
10 18.5479 0.871 0.0938
S2 0.0325 9.17
106 12.2821 0.908 0.0755
increasing temperature, but there were no significant differences S3 0.0337 9.50
106 11.4307 0.924 0.0728
among the extractions carried out from 60 to 80 C. The values 60 C S1 0.0350 9.85
106 18.5479 0.850 0.1037
within the period involving the initial washing step t0 were low, S2 0.0344 9.70
106 12.2821 0.922 0.0775
regardless of the temperature or bed size, confirming that the S3 0.0456 1.29
105 11.4307 0.955 0.1538
80 C S1 0.0736 2.07
105 18.5479 0.934 0.0728
washing step was very fast. Thus, to simplify the mathematical
S2 0.0560 1.58
105 12.2821 0.956 0.0595
model for this extraction step, it was considered that t0 ? 0, which S3 0.0554 1.56
105 11.4307 0.953 0.0617
changed the mathematical expression to eB1 t0 ! 1 (see Eq. (6)).

Table 4
Parameters of the Peleg model related to the extraction of watermelon seed oil.

Temperature Bed K1 (min 100 gdb/g oil) K2 (100 gdb/g oil) B0 Csp (g oil/100 gdb) R2adj RMSD
40 C S1 0.2231 0.0335 5.0007 29.851 0.964 1.5679
S2 0.2562 0.0342 3.9086 29.268 0.995 0.5267
S3 0.2569 0.0319 3.9540 31.348 0.981 1.1270
60 C S1 0.0788 0.0330 12.823 30.272 0.989 5.2941
S2 0.0375 0.0323 26.700 30.991 0.996 0.5861
S3 0.0351 0.0313 28.657 31.881 0.996 0.5884
80 C S1 0.0353 0.0263 28.518 38.095 0.992 0.7710
S2 0.0408 0.0261 24.513 38.314 0.997 0.5792
S3 0.0418 0.0260 23.895 38.412 0.999 0.3380
194 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195
atmospheric pressure. For example, values of 1334 kJ/mol have
been reported for sunflower seed oil extraction [21,43] and 14 kJ/-
mol for canola seed oil extraction [42].
4. Conclusions
The results showed that the temperature of extraction had a
significant effect on the extraction efficiency of oils and com-
pounds with antioxidant activity, with higher yields of oil obtained
at 80 C and better antioxidant activity at temperatures of 60 C.
However, increasing the length of the extraction cell and keeping
the constant mass/solvent volume ratio obtained similar yields.
In this study, the geometry of the extraction bed affected some
extraction parameters in the models studied, reducing the extrac-
tion rate and the initial extraction rate in the second-order model
and decreasing the constant capacity in the Peleg model. Based on
these models, the optimum conditions for the extraction of water-
melon seed oil occurred at higher temperatures (80 C). However,
extraction at lower temperatures yielded oils with a higher content
of phenolic compounds and antioxidant activity. Among the
extraction models, the second-order model best fit the experimen-
tal data (R2adj > 0.964). The Fick model confirmed that the PLE process
used to extract the watermelon seed oil was based on a superficial
extraction step followed by a diffusive step. Moreover, this model
allowed the estimation of the dependence of effective diffusivity
with temperature as an Arrhenius function, yielding values lower
than those previously reported in literature using conventional
extraction techniques.
Acknowledgments
The authors thank So Paulo Research Foundation (FAPESP,
2010/16665-3) for their financial support. Carvalho R.A. thanks
National Council for Scientific and Technological Development
(CNPq) for the productivity grant (307143/2013-9). J. Colivet
thanks the scholarship program of the Organization of American
States (OAS) and the Grupo Coimbra de Universidades Brasileiras
(GCUB) of 2012 for conducting the PhD program in Brazil.
The authors also thank Nilson Jos Ferreira for assistance with
the pressurized liquid extractions.
References
[1] A. Rai, B. Mohanty, R. Bhargava, Modeling and response surface analysis of
supercritical extraction of watermelon seed oil using carbon dioxide, Sep.
Purif. Technol. 141 (2015) 354365, http://dx.doi.org/10.1016/j.
seppur.2014.12.016.
[2] R. Simpson, G.A. Morris, The anti-diabetic potential of polysaccharides
extracted from members of the cucurbit family: a review, Bioact. Carbohydr.
Dietary Fibre 3 (2014) 106114, http://dx.doi.org/10.1016/j.bcdf.2014.03.003.
[3] R. Acar, M.M. zcan, G. Kanbur, N. Dursun, Some physico-chemical properties
of edible and forage watermelon seeds, Iran. J. Chem. Chem. Eng. 31 (2012) 41
47.
[4] H. Rahman, K. Manjula, T. Anoosha, K. Nagaveni, M.C. Eswaraiah, D. Bardalai,
In-vitro anti-oxidant activity of Citrullus lanatus seed extracts, Asian J. Pharm.
Clin. Res. 6 (2013) 152157.
[5] T.T.K. Ghosh, H. Rahman, D. Bardalai, F. Ali, In-vitro antibacterial study of
Aquilaria agallocha heart wood oil and Citrullus lanatus seed oil, Sch. J. Appl.
Med. Sci. 1 (2013) 1315.
[6] S. Varghese, R. Narmadha, D. Gomathi, M. Kalaiselvi, K. Devaki, Phytochemical
screening and HPTLC finger printing analysis of Citrullus lanatus (Thunb.) seed,
J. Acute Dis. 2 (2013) 122126, http://dx.doi.org/10.1016/S2221-6189(13)
60111-0.
[7] O.O. Lucky, U.O. John, I.E. Kate, O.O. Peter, O.E. Jude, Quantitative
determination, metal analysis and antiulcer evaluation of methanol seeds
extract of Citrullus lanatus Thunb (Cucurbitaceae) in rats, Asian Pac. J. Trop.
Biomed. 2 (2012) S804S808, http://dx.doi.org/10.1016/S2222-1808(12)
60269-2.
[8] A. Olamide, O. Olayemi, Effects of methanolic extract of Citrullus lanatus seed
on experimentally induced prostatic hyperplasia, Eur. J. Med. Plants 1 (2011)
171179.
[9] R.L. Jarret, I.J. Levy, Oil and fatty acid contents in seed of Citrullus lanatus
Schrad, J. Agric. Food Chem. 60 (2012) 51995204, http://dx.doi.org/10.1021/
jf300046f.
[10] K.L. Nyam, C.P. Tan, Y.B. Che Man, O.M. Lai, K. Long, Physicochemical properties
of Kalahari melon seed oil following extractions using solvent and aqueous
enzymatic methods, Int. J. Food Sci. Technol. 44 (2009) 694701, http://dx.doi.
org/10.1111/j.1365-2621.2008.01828.x.
[11] J.E. Cacace, G. Mazza, Mass transfer process during extraction of phenolic
compounds from milled berries, J. Food Eng. 59 (2003) 379389, http://dx.doi.
org/10.1016/S0260-8774(02)00497-1.
[12] H. Wijngaard, M.B. Hossain, D.K. Rai, N. Brunton, Techniques to extract
bioactive compounds from food by-products of plant origin, Food Res. Int. 46
(2012) 505513, http://dx.doi.org/10.1016/j.foodres.2011.09.027.
[13] B. Kaufmann, P. Christen, Recent extraction techniques for natural products:
microwave-assisted extraction and pressurised solvent extraction, Phytochem.
Anal. 13 (2002) 105113, http://dx.doi.org/10.1002/pca.631.
[14] A.L. Oliveira, E. Destandau, L. Fougre, M. Lafosse, Isolation by pressurised fluid
extraction (PFE) and identification using CPC and HPLC/ESI/MS of phenolic
compounds from Brazilian cherry seeds (Eugenia uniflora L.), Food Chem. 145
(2014) 522529, http://dx.doi.org/10.1016/j.foodchem.2013.08.065.
[15] C. Pronyk, G. Mazza, Design and scale-up of pressurized fluid extractors for
food and bioproducts, J. Food Eng. 95 (2009) 215226, http://dx.doi.org/
10.1016/j.jfoodeng.2009.06.002.
[16] M. Das, S.K. Das, S.H. Suthar, Composition of seed and characteristics of oil
from karingda [Citrullus lanatus (Thumb) Mansf], Int. J. Food Sci. Technol. 37
(2002) 893896, http://dx.doi.org/10.1046/j.1365-2621.2002.00638.x.
[17] F. Pouralinazar, M.A.C. Yunus, G. Zahedi, Pressurized liquid extraction of
Orthosiphon stamineus oil: experimental and modeling studies, J. Supercrit.
Fluids 62 (2012) 8895, http://dx.doi.org/10.1016/j.supflu.2011.12.009.
[18] M. Herrero, P.J. Martn-lvarez, F.J. Seorns, A. Cifuentes, E. Ibez,
Optimization of accelerated solvent extraction of antioxidants from Spirulina
platensis microalga, Food Chem. 93 (2005) 417423, http://dx.doi.org/
10.1016/j.foodchem.2004.09.037.
[19] V.L. Singleton, R. Orthofer, R.M. Lamuela-Ravents, Analysis of total phenols
and other oxidation substrates and antioxidants by means of Folin-Ciocalteu
reagent, Methods Enzymol. 299 (1998) 152178, http://dx.doi.org/10.1016/
S0076-6879(99)99017-1.
[20] W. Brand-Williams, M.E. Cuvelier, C. Berset, Use of a free radical method to
evaluate antioxidant activity, LWT Food Sci. Technol. 28 (1995) 2530,
http://dx.doi.org/10.1016/S0023-6438(95)80008-5.
[21] E.E. Perez, A.A. Carelli, G.H. Crapiste, Temperature-dependent diffusion
coefficient of oil from different sunflower seeds during extraction with
hexane, J. Food Eng. 105 (2011) 180185, http://dx.doi.org/10.1016/j.
jfoodeng.2011.02.025.
[22] J. Crank, The Mathematics of Diffusion, second ed., Clarendon Press, Oxford,
London, 1975.
[23] A. Bucic-Kojic, M. Planinic, S. Tomas, M. Bilic, D. Velic, Study of solidliquid
extraction kinetics of total polyphenols from grape seeds, J. Food Eng. 81
(2007) 236242, http://dx.doi.org/10.1016/j.jfoodeng.2006.10.027.
[24] D.W. Marquardt, An algorithm for least-squares estimation of nonlinear
parameters, J. Soc. Ind. Appl. Math. 11 (1963) 431441, http://dx.doi.org/
10.1137/0111030.
[25] L.C. Conto, M.A.L. Gragnani, D. Maus, H.C.I. Ambiel, M.C. Chiu, R. Grimaldi,
et al., Characterization of crude watermelon seed oil by two different
extractions methods, J. Am. Oil Chem. Soc. 88 (2011) 17091714, http://dx.
doi.org/10.1007/s11746-011-1850-8.
[26] T.G. Kemper, Oil extraction, in: F. Shahidi (Ed.), Baileys Ind. Oil Fat Prod, John
Wiley & Sons, Inc., Hoboken, NJ, USA, 2005, pp. 5798, http://dx.doi.org/
10.1002/047167849X.bio013.
[27] N. Jorge, A.C. da Silva, C.R. Malacrida, Physicochemical characterisation and
radical-scavenging activity of Cucurbitaceae seed oils, Nat. Prod. Res. (2015)
15, http://dx.doi.org/10.1080/14786419.2015.1007135.
[28] G.I.O. Badifu, Effect of processing on proximate composition, antinutritional and
toxic contents of kernels from Cucurbitaceae species grown in Nigeria, J. Food
Compos. Anal. 14 (2001) 153161, http://dx.doi.org/10.1006/jfca.2000.0964.
[29] C.F. Ross, C. Hoye, V.C. Fernndez-Plotka, Influence of heating on the
polyphenolic content and antioxidant activity of grape seed flour, J. Food Sci.
76 (2011) 884890, http://dx.doi.org/10.1111/j.1750-3841.2011.02280.x.
 te, P. Kraujalis, P.R. Venskutonis, Optimization of high pressure
[30] N. Kryzeviciu
extraction processes for the separation of raspberry pomace into lipophilic and
hydrophilic fractions, J. Supercrit. Fluids 108 (2016) 6168, http://dx.doi.org/
10.1016/j.supflu.2015.10.025.
[31] A. Mustafa, C. Turner, Pressurized liquid extraction as a green approach in food
and herbal plants extraction: a review, Anal. Chim. Acta 703 (2011) 818,
http://dx.doi.org/10.1016/j.aca.2011.07.018.
[32] B.E. Richter, D. Raynie, Accelerated solvent extraction (ASE) and high-
temperature water extraction, in: J. Pawliszyn (Ed.), Compr. Sampl. Sample
Prep., Elsevier, Waterloo, ON, Canada, 2012, pp. 105115, http://dx.doi.org/
10.1016/B978-0-12-381373-2.00047-8.
[33] R. Carabias-Martnez, E. Rodrguez-Gonzalo, P. Revilla-Ruiz, J. Hernndez-
Mndez, Pressurized liquid extraction in the analysis of food and biological
samples, J. Chromatogr. A 1089 (2005) 117, http://dx.doi.org/10.1016/j.
chroma.2005.06.072.
[34] M.M. Poojary, P. Passamonti, Extraction of lycopene from tomato processing
waste: kinetics and modelling, Food Chem. 173 (2015) 943950, http://dx.doi.
org/10.1016/j.foodchem.2014.10.127.
 J. Colivet et al. / Separation and Purification Technology 169 (2016) 187195
[35] J. Xi, L. Yan, L. He, Pressure-dependent kinetic modeling of solid-liquid
extraction of the major green tea constituents, Sep. Purif. Technol. 133 (2014)
155159, http://dx.doi.org/10.1016/j.seppur.2014.06.043.
[36] R.C.A. Amarante, P.M. Oliveira, F.K. Schwantes, J.A. Morn-Villarreyes, Oil
extraction from castor cake using ethanol: kinetics and thermodynamics, Ind.
Eng. Chem. Res. 53 (2014) 68246829, http://dx.doi.org/10.1021/ie500508n.
[37] G.C. So, D.G. Macdonald, Kinetics of oil extraction from canola (rapeseed), Can.
J. Chem. Eng. 64 (1986) 8086, http://dx.doi.org/10.1002/cjce.5450640112.
[38] G.L. Zabot, M.N. Moraes, A.J. Petenate, M.A.A. Meireles, Influence of the bed
geometry on the kinetics of the extraction of clove bud oil with supercritical
CO2, J. Supercrit. Fluids 93 (2014) 5666, http://dx.doi.org/10.1016/
j.supflu.2013.10.001.
[39] S. Sayyar, Z.Z. Abidin, R. Yunus, A. Muhammad, Extraction of oil from Jatropha
seeds-optimization and kinetics, Am. J. Appl. Sci. 6 (2009) 13901395, http://
dx.doi.org/10.3844/ajassp.2009.1390.1395.
[40] A. Berna, A. Trrega, M. Blasco, S. Subirats, Supercritical CO2 extraction of
essential oil from orange peel; effect of the height of the bed, J. Supercrit.
Fluids 18 (2000) 227237, http://dx.doi.org/10.1016/S0896-8446(00)00082-6.
195
[41] H.Z. Kister, P.M. Mathias, D.E. Steinmeyer, W.R. Penney, B.B. Crocker, J.R. Fair,
Equipment for distillation, gas absorption, phase dispersion, and phase
separation, in: R.H.P. Don, W. Green (Eds.), Perrys Chem. Eng. Handb, 8th
ed., McGraw-Hill, 2008, pp. 14.114.129, http://dx.doi.org/10.1036/
0071511377.
[42] M.B. Fernndez, E.E. Perez, G.H. Crapiste, S.M. Nolasco, Kinetic study of canola
oil and tocopherol extraction: parameter comparison of nonlinear models, J.
Food Eng. 111 (2012) 682689, http://dx.doi.org/10.1016/j.jfoodeng.2012.
01.036.
[43] E.R. Bamler, G.H. Crapiste, A.A. Carelli, Solvent extraction: kinetic study of
major and minor compounds, J. Am. Oil Chem. Soc. 87 (2010) 14891495,
http://dx.doi.org/10.1007/s11746-010-1637-3.
[44] Y.S. Ho, H.A. Harouna-Oumarou, H. Fauduet, C. Porte, Kinetics and model
building of leaching of water-soluble compounds of Tilia sapwood, Sep. Purif.
Technol. 45 (2005) 169173, http://dx.doi.org/10.1016/j.seppur.2005.03.007.
[45] M. Peleg, An empirical model for the description of moisture sorption curves, J.
Food Sci. 53 (1988) 12161217, http://dx.doi.org/10.1111/j.1365-2621.1988.
tb13565.x.