Professional Documents
Culture Documents
Theoretical Models For Supercritical Fluid Extraction
Theoretical Models For Supercritical Fluid Extraction
Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma
Review
a r t i c l e i n f o a b s t r a c t
Article history: For the proper design of supercritical fluid extraction processes, it is essential to have a sound knowledge
Available online 17 April 2012 of the mass transfer mechanism of the extraction process and the appropriate mathematical represen-
tation. In this paper, the advances and applications of kinetic models for describing supercritical fluid
Keywords: extraction from various solid matrices have been presented. The theoretical models overviewed here
Supercritical fluid include the hot ball diffusion, broken and intact cell, shrinking core and some relatively simple models.
Extraction kinetics
Mathematical representations of these models have been in detail interpreted as well as their assump-
Carbon dioxide
tions, parameter identifications and application examples. Extraction process of the analyte solute from
Mathematical model
Mass transfer
the solid matrix by means of supercritical fluid includes the dissolution of the analyte from the solid, the
Diffusion analyte diffusion in the matrix and its transport to the bulk supercritical fluid. Mechanisms involved in
a mass transfer model are discussed in terms of external mass transfer resistance, internal mass transfer
resistance, solute–solid interactions and axial dispersion. The correlations of the external mass trans-
fer coefficient and axial dispersion coefficient with certain dimensionless numbers are also discussed.
Among these models, the broken and intact cell model seems to be the most relevant mathematical model
as it is able to provide realistic description of the plant material structure for better understanding the
mass-transfer kinetics and thus it has been widely employed for modeling supercritical fluid extraction
of natural matters.
© 2012 Elsevier B.V. All rights reserved.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. General aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1. Solid pre-treatment and particle characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. Solute–matrix interactions and phase equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3. Mass transfer and axial dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3. Relatively simple theoretical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1. Diffusion layer theory (DLT) model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.2. Desorption model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3.3. Partitioning coefficient model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4. Hot sphere diffusion model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.1. Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
4.2. Mathematical representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
4.3. Parameter identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.4. Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
5. BIC model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.1. Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
5.2. Mathematical representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.3. Parameter identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4. Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
∗ Corresponding author at: Tianjin Key Laboratory of Refrigeration Technology, Tianjin University of Commerce, Tianjin 300134, China. Tel.: +86 13752549698;
fax: +86 22 26686529.
E-mail address: huangzhen50@yahoo.com (Z. Huang).
0021-9673/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.chroma.2012.04.032
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 3
6. Shrinking model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.1. Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.2. Mathematical representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.3. Parameter identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.4. Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
parameters along with thermodynamics restrictions could be used Reverchon et al. [43] developed a hot ball model for SFE process and
to economically evaluate the extraction process from laboratory treated the solute extraction from a solid matrix as a hot ball cooling
to pilot and industrial scale. In the meantime, the understanding down in a uniform medium. Veress [44] assumed a diffusion-layer
of various process variables and how they can be connected to a theory (DLT) model to quantify the extraction process of cannabi-
theoretical model to describe the extraction kinetics may be also noids from marihuana and hashish by SCCO2 [44]. Tan and Liou [45]
obtainable. For these reasons many mathematical models for SFE proposed a single parameter desorption model for describing SFE
have been proposed and developed, as summarized in two recent process as desorption of the analyte from the parent matrix.
review papers [36,37]. In what follows we will review mathematical models for SFE
process classified as: (1) relatively simple models, (2) hot sphere
model, (3) BIC model, and (4) SC model. Here we would like to
2. General aspects highlight those models with analytical solution although they are
valid at strictly limited conditions because they can largely simplify
Mathematical modeling of experimental data is a common task the estimation of model parameters from the shape of experimental
in the field of engineering and applied sciences. A number of kinetic extraction curve.
extraction models can be found in specialized scientific literatures In dealing with mathematical modeling of SFE, several general
that have been proposed for the SFE of packed beds of solid matri- aspects have to be taken into account [11,46]:
ces [36,37]. Through fitting experimental data, these models are
attempted to not only provide a quantitative interpretation of the 2.1. Solid pre-treatment and particle characteristics
experimental data, and discover the true values of the parameters
that are physically meaningful, but also complete data at points Solid substrates are commonly subjected to mechanical pre-
where measurements are not carried out. The most important goal treatments such as grinding, milling, cutting, chopping, crushing or
of modeling is the prediction of large scale extraction curves as other size-reduction methods. These pretreatments have multiple
functions of the experimental parameters as described above, since purposes, including releasing of solutes from cells, facilitating sol-
real experiments at large scale are extremely expensive. vent flow through the packed bed, increasing substrate loads into
Of these models, some are empirically simple in nature and thus extraction units, and improving the rate and yield of the subsequent
of limited viability as they have no physical meanings correspon- extraction process. As a consequence, superficial vegetable cells
dent to the materials and the process studied, and sometimes they may be considerably broken, rendering the analyte to be extracted
may deal with SFE extraction curves very well. However, these free on the surface of the solid material. Reverchon et al. [47–49]
models are not adequate for scaling-up predictions because they used scanning electron microscope (SEM) and found broken cells
do not provide a description of underlying mass transfer phenom- on the particle surface. If the broken cells were assumed to form a
ena that occur also in industrial-size extraction units. For examples, single layer, the authors were able to calculate volumetric broken-
Naik et al. in 1989 [38] proposed a 2-parameter Langmuir-type to-intact cell ratio and thus reduce the number of model parameters
model for representing perfume extraction from several vegetable evaluated from extraction curves. Generally, the content of the
materials by subcritical CO2 . Barton et al. in 1991 [39] suggested a extractable substances oil due to cell breaking is inversely propor-
2-parameter kinetic model for describing the SFE process of vanilla tional to the final particle diameter and then very small particles
oleoresin as a first-order chemical reaction, bearing a form [40] are preferred to improve the extraction rate and yield. But the use
similar to one-parameter desorption model discussed later. These of small particles is not always encouraged since substrate agglom-
empirical models will not be given further consideration in this erating or channeling may occur in tightly packed beds especially
review. when particle size is very small or when the extractor diameter
Therefore, extraction models should not be a mere mathemat- is large, leading to the poor solvent flow and poor performance.
ical tool, but should reflect the physical insight arising from the In addition, particles may be spherical, cylinders, slabs, and other
knowledge of the solid structure and from experimental observa- shapes as a result of the grinding process and their shapes may
tions. From this regards, the most valuable models to represent the influence the diffusion of the supercritical solvent in different ways
overall fixed bed behavior are those rigorous differential mass bal- [36,50–52].
ance models, developed from differential mass balance equations
for the packed bed of solid substrates. By integration of those dif- 2.2. Solute–matrix interactions and phase equilibrium
ferential equations time-dependent concentration profiles in both
phases can be obtained and the extraction curve could then be cal- Depending on interactions between the solute and the solid
culated from fluid-phase concentration at the extraction unit outlet structure, different phase equilibria may be involved. Usually, if
[36]. Note that these models contain two differential solute mass the solute has no interactions with matrix, equilibrium solubility is
balances for the solute in supercritical phase and in solid phase readily taken into account. This may be true in the case of vegetable
respectively, in addition to a solute equilibrium desorption that oil in seeds where the oil content in solid phase is relatively high and
describes the interactions between solute and solid matrix. the fluid-phase equilibrium concentration is then independent of
Arising from different mass transfer mechanisms and equilib- matrix and equal to its solubility in SCCO2 . Consequently, it appar-
rium relationships, a varsity of mass balance models have thus ently controls the first period of the SFE process as the fluid phase
been proposed to characterize the mass transfer kinetics [41–44]. leaving the extractor could be reasonably be assumed to be in equi-
The most important idea involved for developing these models is librium with solid phase at the extraction unit outlet. Therefore, the
the concept of broken and intact cells (BIC), firstly introduced by initial part of extraction curve contains information on the type
Sovová [41]. In the BIC model, the solid phase is considered present of equilibrium between the solid and fluid phases. On the other
between broken and intact cells and thus the total oil available hand, if the solute–matrix interactions are not negligible, then the
within the plant cells are divided by easily accessible oil on the cell extraction rate should depend on how the analyte is bound to the
surface and difficultly accessible oil in the intact cell. Shrinking core solid matrix such as either on the outer surface or inside the solid
(SC) model is the model which describes the situation of irreversible structure. Thus the solute desorption from the matrix and its par-
desorption followed by diffusion in the porous solid through the tition equilibrium between the SCCO2 and the solid matrix should
pores. This model assumes that the solute inside the particle is be considered. This case may be possible for essential oils in the
located within a core that shrinks as the solute is extracted [42]. plant where its initial content is only a few percent or less, and the
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 5
Table 1
Applications of the Tan–Liou model for different SCCO2 extraction systems.
Shiitake mushroom Shiitake oil 303–323 15–30 0.190–0.487 Kitzberger et al. [3]
Activated carbon Toluene 308–338 6.08–17.23 2.12–11.1 Tan and Liou, [45]
Turmeric rhizomes Curcumins 343–378 25.0–30.0 3.47–12.2 Chassagnez-Méndez et al. [63]
Pupunha Oil 318–323 25.0–30.0 5.33–6.67 Araújo et al. [64]
Carqueja Oil 303–313 10.0–30.0 3.82–9.28 × 10−2 Silva et al. [66]
Marigold flowers Oleoresin 299–313 12.0–20.0 0.17–0.33 Campos et al. [65]
Croton Oil 288 6.7 0.95–1.01 Sousa et al. [116]
Vetiver roots Volatile oils 313 20.0 1.15–2.33 Talansiera et al. [117]
Fennel seeds Oil 303–313 10.0–25.0 0.93–3.68 Moura et al. [135]
methanol flow rate. More interestingly, the thickness (h) of the dif- Alternatively, some researchers have attempted a so-called one-
fusion layer can be calculated from the ˇ value, along with the De site desorption model to quantitatively evaluate the dissolution and
obtained from the hot ball model. Then the thickness of the diffu- desorption behavior of chemicals under supercritical conditions
sion layer obtained is in the order of 10 nanometers, spanning from [67,68]. If rapid equilibrium between the analyte and the matri-
9.8 to 12.8 nm. ces can be assumed during the extraction process, this model may
be given below to define the extraction curve:
3.2. Desorption model mt
= 1 − Ad · exp(−kd t) (9)
m0
For this kind of desorption models, the SFE process is assumed
where Ad is a dimensionless model coefficient. This model could be
to be desorption of the analyte from the parent matrix. In 1989,
valid for SFE of chemicals freshly added to soils, chemicals with low
Tan and Liou proposed a single parameter desorption model for
hydrophobicity, and soils with low organic carbon content [24].
describing desorption of activated carbon loaded with toluene by
However, for many other systems, nonequilibrium disso-
SCCO2 [45]. This model may be simply derived as follows.
lution/desorption may have been applied. In this case, the
Mass balance for the bulky fluid phase:
nonequilibrium behavior can be represented using a two-site
∂y ∂y 1 − ε s ∂x model in which one fraction of the solute in the matrices is quickly
+u =− · · (5)
∂t ∂z ε f ∂t released and another fraction is slowly released [67–70]. Generally
speaking, the fraction of the total solute that is released quickly
with boundary conditions: t = 0, y = 0 and z = 0. Note that y (g
(Fr ) from the matrices is commonly assumed to be representative
solute/g solvent) and x (g solute/g solute free solid bed) are the
of equilibrium release conditions whereas the remaining fraction
analyte concentration in the SCCO2 fluid phase and the solid phase,
of the total mass that is released slowly (1 − Fr ) is considered to be
respectively; f and s are the density of the SCCO2 fluid and the
kinetically rate-limited. With this model, dissolution/desorption is
solid (g/cm3 ), respectively; ε is the porosity of the extraction bed,
assumed to occur in two different sites in the parent matrices. These
u is the superficial velocity of the SCCO2 solvent (m/s), and z is the
are conceptual “sites” since it is recognized that analyte desorption
axial position (m).
undoubtedly occurs from many actual locations in the matrices.
Mass balance for the solid phase:
The two-site kinetic desorption model can be written, consisting
∂x of two first-order expressions, as follows:
(1 − ε)s = −kd x (6)
∂t mt
= 1 − Fr · exp(−kd1 t) − (1 − Fr ) · exp(−kd2 t) (10)
with t = 0, x = x0 . x0 is the initial solute concentration in the solid m0
phase (g solute/g solute free solid bed), and kd is the desorption where Fr is the fraction of the analyte released quickly, (1 − Fr ) is
rate constant (s−1 ). the fraction of the analyte released slowly; kd1 is the first-order
The time-dependent solute concentration at the exit of the
rate constant describing the quickly released fraction (min−1 ); and
packed bed can then be solved, resulting in
kd2 is the first-order rate constant describing the slowly released
εk H
1 − ε s d fraction (min−1 ).
y(z = H, t) = x0 exp − 1 exp(−kd t) (7)
ε f u This model has thus been applied in the extraction SCCO2 extrac-
tion of ginger rhizome [67], Morinda citrifolia roots [68], savory
where H is the height of extraction bed (m). Integration of Eq. (7)
[69], seeds of rosehip, loquat and physic nut [70]. Table 2 presents
from the beginning to time t gives the overall extraction curve:
the regressed results using the two-site kinetic desorption model.
t
1 − ε s 1 Obviously, the fast desorption rate constant (kd1 ) is at least one
mt (h = H) = QCO2 y (z = H, t)dt = QCO2 · x0 order of magnitude higher than the slow rate constant (kd2 ).
0
ε f kd
Note that the kinetic model includes no factor describing extrac-
εHAf tion flow-rate, but relies solely on time. Therefore, doubling the
× 1 − exp [exp(−kd t) − 1] (8) volumetric SCCO2 flow-rate should have little effect on the extrac-
QCO2
tion efficiency per unit time if the extraction efficiency is controlled
where A is the cross-sectional area of the extraction vessel (m2 ), by the kinetics of the initial desorption step and the other extraction
QCO2 is the mass flow rate of the SCCO2 solvent (g/s). parameters remain constant. Thus, in terms of the solvent flow-rate
This model was initially proposed by Tan and Liou in 1989 for effect it would be better to compare this model with the partition-
the removal of toluene from activated carbon with SCCO2 [45], ing coefficient model that will be discussed in the next section.
and has been applied to oil from Shiitake mushroom [3], curcum-
ins from turmeric rhizomes [63], oil from pupunha [64], oleoresin 3.3. Partitioning coefficient model
from marigold [65], and oil from carqueja [66]. Table 1 shows the
regressed results for the Tan–Liou model, and more details on the Partitioning coefficient model, proposed by Kubátová et al. [69],
model can be found in these publications. has been developed on a thermodynamic partitioning coefficient
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 7
Table 2
Values of Fr , kd1 and kd2 obtained with the two-site kinetic model for different SCCO2 extraction systems.
Matrix Solute T (◦ C) P (MPa) CO2 flow rate Fr kd1 (min−1 ) kd2 (min−1 ) Reference
(ml/min)
Ginger rhisome Oleoresin 50–65 12.7–19.6 – 0.52–0.89 1.75–26.4 × 10−3 1.1–2.2 × 10−4 Kandiah and
Spiro, [67]
Morinda Damnacanthal 150–220 4.0 1.6–4.0 0.76 0.0357–0.090 0.0078–0.029 Anekpankul
citrifolia et al. [68]
roots
Savory Essential oil 50 20.0 0.25–4.0 Kubátová et al.
[69]
Thymol 0.37–0.48 0.09–0.12 0.005–0.006
Carvacrol 0.36–0.48 0.09–0.12 0.005–0.005
Borneol 0.25–0.37 0.07–0.12 0.006–0.007
Thymoquinone 0.20–0.53 0.05–0.15 0.011–0.013
p-Cymene 0.26–0.50 0.06–0.11 0.006–0.007
Sosehip seed Oil 40–80 15–49 3.0 0.75 0.0014–0.1803 1.4 × 10−6 –0.0193 Machmudah
et al. [70]
Loquat seed Oil 3.0 0.75 0.0063–0.0263 0.0007–0.0108
Physic nut seed Oil 3.0 0.75 0.0054–0.0517 1.9 × 10−5 –0.040
Kp , which is defined as the analyte concentration in the solid matrix flow-rate should double the extraction rate for the same period
phase over that in the SCCO2 fluid phase at equilibrium: if the extraction is described by thermodynamic partitioning and
x all other extraction parameters hold the same. Therefore, if the
Kp = (11) cumulative amount of solute extracted is plotted against extrac-
y
tion time, the curve for the higher flow rate will lie above that for
This model assumes that the initial desorption step and the the lower flow rate. On the other hand, if the same data are plotted
subsequent fluid–matrix partitioning are rapid, and thus do not in terms of volume, the curves from all flow rates will overlap com-
significantly affect the extraction rate. For SFE process, this type of pletely. Obviously, this case is different from that for the two-site
behavior may occur if the initial solute concentration in the plant desorption model as discussed previously.
matrix is small. Thus, if the size of the solid particles is uniform,
the mass of the solute in each unit mass of extraction fluid and that 4. Hot sphere diffusion model
remaining in the matrix for particular extraction period can be cal-
culated based on the Kp value determined for the solute. Then, the 4.1. Assumptions
extraction curve over the entire extraction process can be obtained
by: The SFE process may be treated analogously as a heat transfer
mb 1 − ma /m0 ma phenomenon by considering the solid particles as hot balls cooling
= + (12) down in a uniform surrounding. Then the extracted particles are
m0 Kp m/[(Vb − Va )f ] + 1 m0
assumed to be spherical with initially uniformly distributed solute
where ma and mb are the cumulative mass (g) of the analyte diffusing through the matrix analogously to the heat diffusion and
extracted after the extraction fluid volume Va and Vb (m3 ), respec- the solute concentration in the solvent is assumed to be close to
tively. Thus mb /m0 and ma /m0 are the cumulative fractions of the zero. Thus, Bartle et al. [71] in 1990 proposed a hot ball diffusion
analyte extracted by the extraction fluid of the volume Vb and Va , (HBD) model for describing supercritical fluid extraction processes.
respectively. The data points Vb and mb is the next sequential data They have insightfully discussed the effect of matrix shape, size
point after Va and ma , and m are the mass of the loaded sample variation, solute distribution and solubility limitations on the hot
(g). ball model for real systems [71,72]. Fig. 2 graphically shows the
For this model, the parameter to be regressed against the extrac- particle scheme for the hot ball diffusion (HBD) model. Unlike the
tion data is Kp . Table 3 shows calculated results reported in the BIC situations where some of the solute is readily accessible to the
literature [68–70] for using the partitioning coefficient model to solvent on particle surface [41], the hot ball model is suggested
fit with experimental data. It should be noted from the model for a matrix which contains small quantities of extractable materi-
that the cumulative amount of solute extracted is described as als such that the extraction is not limited by solubility. When the
a function of SCCO2 volume passed the extractor, and does not intraparticles or internal diffusion controls the mass transfer pro-
depend on the extraction time. Thus, doubling the extraction fluid cess, it may be appropriate to start studying mass transfer between
Table 3
Applications of the partitioning coefficient model for different SCCO2 extraction systems.
Morinda citrifolia roots Damnacanthal 150–220 4.0 1.6–4.0 43.27–66.46 Anekpankul et al. [68]
Rosehip seed Oil 40–80 15–49 3.0 18.5–337.1 Machmudah et al. [70]
Loquat seed Oil 29.8–233.6
Physic nut seed Oil 29.5–485.9
8 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
M = ke (c|R − c) (13b)
∂ c|R 1 ∂ ∂c|R
= De r2 (14)
∂t r 2 ∂r ∂r
a single particle and the supercritical solvent, and then extend the
results to the whole bed. The extension of the results to the whole This expression is mathematically analogous to the represen-
bed of particles is immediate since all the particles are assumed to tation of the unsteady cooling of a hot sphere suddenly immersed
be at the same extraction conditions. For these reasons, the HDB into a cold fluid. Applying the Fourier transforms and the heat–mass
model has been further elaborated by Reverchon et al. [43] and transfer analogy, it is possible to obtain a general solution for the
Esquível et al. [73], and attempted by a number or researchers to concentration based on the extraction time and the intraparticle
simulate various extraction processes of natural materials [74–81]. position as given below [43,73,79].
4R sin ˇk − ˇk cos ˇk
More recently, Huang et al. [62] have further extended this model ∞
c|R − c ˇk r
for describing amine extraction from freshly synthesized hexago- = sin
cs0 − c r 2ˇ2 − ˇk sin 2ˇk R
nal mesoporous silica materials by means of methanol-enhanced k k
supercritical CO2 . 2
In this model, the extraction process from a single particle may ˇk
× exp − De t (15)
be described by the following stages in sequence: (1) CO2 diffusion R
in the SCCO2 solution film around the particle; (2) CO2 adsorption
on the particle surface and diffusion into the particle; (3) analyte where the terms ˇk allow the particularization for a given system
dissolving; (4) analyte diffusion out of the solid matrix; (5) analyte and each term of the series is defined by the periodic Eq. (16):
diffusion through the SCCO2 solution film into the bulky solvent.
R
Besides, some assumptions should be included (a) each single ˇk cot ˇk = 1 − kf (16)
De
particle has the same shape and same diameter, (b) extraction of
different analyte components are parallel and non-interacting pro- The mass flux density M may be obtained from the variation
cesses, which is particularly well suited in the SFE of essential oils of the solute concentration at the particle surface with extraction
due to the fact that the essential oil concentration is very low if time, giving [73]
compared with the equilibrium solubility, (c) diffusivities of differ- t
ent analytes in the parent matrix are independent from the particle ∂c|R
M = −De dt (13c)
radius, and (d) different analytes are treated as a single chemical 0
∂r
compound.
From Eqs. (6) and (8), the general solution can be given as follows
[43,73]:
4.2. Mathematical representation
∞
(sin ˇk − ˇk cos ˇk )
2
M = 4R(cs0 − c)
The material balance across an internal particle surface of radius 2ˇk4 − ˇk3 sin 2ˇk
k
r (m), according to the first Fick’s law, for a constant density and 2
diffusivity in the solid and spherical coordinates (r, , ), can be ˇk
· 1 − exp − De t (13d)
given below [43]: R
Similarly, the total weight of the solute in the fixed bed, m0 , may is the rate controlling mode of the process. However, their two-part
be expressed as models have not led to satisfactory performance as the deviation
4 3 can be higher than 50%.
m0 = Np · R cs0 (19)
3
4.3. Parameter identification
Then, it is possible to obtain the fraction of the solute that is
extracted from the bed against time if the concentration of solute
The extraction calculations obtained either from Eq. (20a) or
in the bulk fluid phase is very small, given by
2
from Eq. (20) along with Eq. (16) can be thus compared to the exper-
mt (sin ˇ − ˇ cos ˇ )
∞ 2
ˇk imental data by applying the least-squares method in which the
k k k
= 12 · 1 − exp − De t solute solid diffusivity De used as the model parameter to obtain
m0 ˇ3 (2ˇk − sin 2ˇk ) R
k k the best fit of the experimental data. In order to model the extrac-
(20) tion curve with Eq. (20), ˇk must be calculated from Eq. (16), thus
the external mass transfer coefficient kf has to be evaluated. Usu-
Further, if the solute film resistance is neglected, ke /Dm → ∞, ally, kf can be estimated by using a number of correlations that
then ˇk = n, thus Eq. (20) can be further condensed to the familiar relate this coefficient with dimensionless numbers. These correla-
expression: tions have extensively been used for the solid–fluid transfer case in
a SFE process, as summarized in recent review papers [36,37]. For
2
6 1
∞
mt n examples, the correlation proposed by Tan et al. [36,37,84] can be
=1− 2 · exp − De t (20a) given as:
m0 n2 R
n=1
1/3
which may also be written in the equivalent form Sh = 0.3Re0.83 Sc (valid for 2 < Re < 40, and 2 < Sc < 20)
mt
De t
1/2 1 ∞
nR 3De t
The other expressions commonly used are presented as below.
=6 √ + i erfc − (20b) King and Catchpole:
m0 R2 De t R2
n=1
1/3
Sh = 0.82Re0.6 Sc (valid for 1 < Re < 70, and 3 < Sc < 11)
For short time, a simplified form for the initial region of the
extraction curve may be obtained by neglecting the higher order Puiggené et al.:
term in Eq. (20b).
1/3
D t 1/2 Sh = 0.206Re0.8 Sc (valid for 10 < Re < 100, and Sc < 10)
mt 6 e
= √ (20c)
m0 R2 Wakao and Funazkri:
In the long time region, all terms except the first in the series of Sh = 2.0 + 1.1Re0.6 Sc
1/3
(valid for 3 < Re < 10, 000)
exponential terms in Eq. (20a) become negligible and the extraction
yield curve can be recast as: where Sh = dp ke /D12 , Re = f udp /
, Sc =
/f D12 . Sh , Re , and Sc are
the dimensionless numbers of Reynolds, Sherwood, and Schmidt,
mt 6 2 De t
= 1 − 2 exp − (20d) respectively.
is the viscosity of the supercritical solvent phase
m0 R2 (g/(m s)), and D12 is the external diffusion coefficient (m2 /s) of the
From which it is apparent that a plot of ln(1 − mt /m0 ) versus solute to the supercritical fluid or called the solute–solvent binary
t will approach a straight line with a slope of −2 De /R2 . Ohira diffusivity coefficient as well. More recently, Mongkholkhajornsilp
et al. [80] have treated Eq. (20d) as a one-site kinetic desorption et al. have proposed a new expression of mass transfer coefficient
model with a different form of mt /m0 = 1 − exp(− kt), similar to that in order to improve the correlation performance [85]. In conjunc-
reported by Mehr et al. [74] and Subra et al. [75]. However, some tion with experimental results for nimbin extraction neem seeds,
studies [75,81–83] show that a single line may not properly be the correlation obtained by using non-linear regression analysis is
resulted when ln(1 − mt /m0 ) is plotted against time. In stead, the Sh = 0.135Re0.5 Sc0.33 with Reynolds number from 0.169 to 1.292 and
extraction data points may fall onto two intersecting straight lines, Schmidt number from 6 to 25.
the first with a relatively steep slope and the second with a rela- The values of parameter D12 may be similarly evaluated using
tively shallow one. Therefore, these studies suggest that it may be existing correlations and available data. For example, Reverchon
necessary to consider the presence of two parallel diffusion pro- et al. [43] have assumed a D12 value of 1 × l0−6 m2 /s for describing
cesses inside the solid, with one faster and one slower. It follows SCCO2 extraction of essential oils and waxes from basil, marjo-
that: ram and rosemary leaves. Otherwise, certain correlations should
be attempted and these D12 involved correlations have been sum-
mt 6 2 De1 t 2 De2 t marized in recent review [37]. Two of them are exampled here;
=1− 2 f1 exp − + f2 exp (20e)
m0 R2 R2 one is the well-known Stokes–Einstein equation D12 = kB T/(6
)
where is the solute molecule radius and kB is Boltzman constant.
where f1 and f2 are the fractions of the solute, which are extracted
The other correction is given below as it is suitable for estimating
with different diffusion coefficients De1 and De2 , respectively.
binary diffusivity of solute in a wide variety of supercritical solvents
Using Eq. (20e) Hojnik et al. [81] have modeled extraction of
[37,52]:
lutein from Marigold flower petals and found that higher De and
lower De values are in the range of 1.45–9.09 × 10−12 m2 /s and c r r −2/3 r
0.9–5.8 × 10−14 m2 /s, respectively. Interestingly, Čretnik et al. [76] D12 = 5.152Dscf Tscf [(scf ) − 0.4510] · , 1 < scf < 2.5
have mathematically modeled separation of parthenolide from
feverfew by dividing the extraction into two continuous parts. The
c )0.75
Mscf · (Tscf
first part represents the period of constant extraction rate, where c −7
Dscf = 4.30 × 10 · 2/3 ,
steady state mass transfer prevails and the film resistance controls vscf c
· scf
the rates. The second part represents the period of falling extraction
rate with unsteady state mass transfer where the internal diffusion c
for SCCO2 , Dscf = 4.937 × 10−8 m2 /s
10 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
Table 4
Calculated De values with the HBD models for different SCCO2 extraction systems.
Shiitake mushroom Shiitake oil 303–323 15–30 7.35–38.5 × 10−14 Kitzberger et al. [3]
Baizhu herb Extracted 313–333 15–45 1.47–8.13 × 10−13 Huang et al. [40]
Basil leaves Essential oil 313 10 1.5 × 10−13 Reverchon et al. [43]
Cuticular waxes 313 10 1.6 × 10−17
Porous silica material Dodecylamine 323–378 15–25 1.19–6.72 × 10−12 Huang et al. [62]
Oleoresin Marigold flowers 299–313 12.0–20.0 1.30–2.55 × 10−13 Campos et al. [65]
Olive husk Olive husk oil 313 12–18 0.1–25.9 × 10−13 Esquivel et al. [73]
Guaraná seed Caffeine 308–328 13.8–27.6 4.4–4.6 × 10−12 Mehr et al. [74]
Feverfew flower heads Extracted 313–353 20–30 2.74–3.59 × 10−14 Čretnik et al. [76]
Parthenolide 1.28–13.36 × 10−14
Cow brain Cholesterol 323–343 23.0–27.0 1.90–2.80 × 10−13 Vedaraman et al. [77]
Soybeans Oil 313 28 2.2–36 × 10−14 Rochová et al. [78]
Nannochloropsis gaditana Carotenoid 313–333 10–50 0.08–114 × 10−19 Macías-Sánchez et al. [79]
Synechococcus sp. Carotenoid 313–333 10–50 0.02–64 × 10−19
Dunaliella salina Carotenoid 313–333 10–50 0.22–830 × 10−19
Peach almond oil Peach almond 313 15.0–25.0 0.041–2.28 × 10−10 Mezzomo et al. [94]
Guava leaves Extract 308–328 30.0 0.75–0.77 × 10−12 Moura et al. [114]
Vetiver roots Volatile oils 313 20.0 0.7–2.2 × 10−13 Talansier et al. [117]
5.2. Mathematical representation overall extraction curve, e (mass of the extract, g/g solute-free feed),
corresponding to each extraction period, is then determined as
The BIC model describes a situation when a SCF solvent flows H
axially with superficial velocity u through a bed of pretreated matri- 1
e(t) = x0 − x(z, t) dz (26)
ces in a cylindrical extraction unit. Assuming that the concentration H 0
of the readily accessible solute and that of the less accessible solute
are xb and xi at the beginning of the extraction, respectively, the For evaluation of experimental data, extraction curve should be
total initial concentration of the extractable solute x0 has x0 = xb + xi . expressed in terms of real time or amount of the SCF solvent con-
The fraction, r, of the less accessible solute can be given as r = xi /x0 sumed. Eq. (26) can be analytically solved for each stage, expressing
and sometimes, a term of the grinding efficiency, 1 − r, is used the amount of extract as a function of the specific amount of solvent
instead for expressing the fraction of the readily accessible solute in q, concentrations x0 , xi and y* and parameters Z and W.
the bed. Fig. 3 shows the particle scheme for the broken and intact For stage one q < qCER or t < tCER ,
cell model. It is seen that the easily accessible solute from broken
e = q × y∗ × [1 − exp(−Z)] (27a)
cells is transferred directly to the fluid-phase while the solute from
intact cells diffuses first to broken cells and then to the fluid-phase. For stage two: qCER ≤ q < qFER or tCER ≤ t < tFER
Besides, some researchers suggest that the solute from intact cells
may also diffuse directly to the fluid phase as well [37]. e = y∗ × [q − qm × exp(Zw − Z)] (27b)
Based on differential mass balances, taken both within the solid
For stage three: q ≥ qFER or t ≥ tFER ,
matrix and in the surrounding solvent phase, the BIC model may
be derived as follows [41]. y∗
x0
Solid phase: e = x0 − ln 1 + exp W + ∗ −1
W y
∂x
xi
−s (1 − ε) = f (x, y) (22) × exp W × (qm − q) × for t ≥ tFER (27c)
∂t x0
Fluid phase: where
∂y ∂y mCO2 q x0 − xi
f u + f ε = f (x, y) (23) q= , q̇ = and qCER = (28a)
∂z ∂t mbed t y∗ · Z
where f(x, y) is the interfacial mass transfer rate. Usually for SFE
1 xi + (x0 − xi ) · exp(W · x0 /y∗ )
the fluid phase can be treated as a diluted solution, therefore the qFER = qCER + · ln (28b)
W x0
solvent density CO2 can replace the fluid phase density.
The second term in the fluid phase balance Eq. (23) is often
kf a0 f
neglected as plug flow is assumed, giving Z= (28c)
[q̇(1 − ε)s ]
f u
∂y
= f (x, y) (23a) Z · y∗
x · exp[W · (q − q )] − x
0 CER i
∂z Zw = · ln (28d)
W · x0 x0 − xi
The boundary conditions are
W=
ks a0
(28e)
x(z, t = 0) = x0 [q̇(1 − ε)]
y(z = 0, t) = 0
where mCO2 is the mass (g) of SCCO2 solvent consumed and mbed is
To solve Eqs. (22) and (23a), Sovová [41] has defined f(x, y) as a the mass (g) of solute-free feed loaded in the bed. q and q̇ are the
function of the concentration difference using a local mass transfer specific mass (g solvent/g solute-free feed) and specific mass flow
coefficient for both phases. The easily accessible solute which sur- rate (g solvent/g solute-free feed in unit time) of solvent passed
mounts only the diffusion resistance in the solvent is extracted first through the extraction, y* is the equilibrium solubility of the solute
and this part of solute crosses the interfacial boundary fast enough in the SCCO2 solvent (g solute/g solvent), qm is the q value when
to kept the solvent at the boundary saturated, the mass transfer extraction begins inside the particles, qn is the q value when the
rate is easily accessible part of solute is all extracted, Zw is the dimen-
sionless axial co-ordinate between fast and slow extractions, Z is
f (x > xi , y) = kf a0 f (y∗ − y) (24) the dimensionless mass transfer parameter in the fluid phase, W is
But when the solid-phase concentration decreases to xi , the the dimensionless mass transfer parameter in the solid phase, ks
mass transfer is then retarded by the diffusion in the solid phase, is the solid-phase mass transfer coefficient (m/s), kf a0 is the volu-
subsequently resulting in slow extraction process. The mass trans- metric solvent phase mass transfer coefficient (s−1 ), and ks a0 is the
fer rate could thus be expressed as: volumetric solid phase mass transfer coefficient (s−1 ).
If extraction curve were expressed in terms of real time used,
f (x ≤ xi , y) = ks a0 s (x − xf ) (25) two terms of tCER and tFER , i.e., constant extraction rate (CER) and
falling extraction rate (FER) periods, may be used. In other words,
Considering that the solid-phase diffusion resistance is much
tCER is the time at which the extraction of the oil from the inside of
large than the resistance in the supercritical phase, the interfacial
particles starts (min) and tFER is the time at which the extraction of
concentration xf (g solute/g solute-free feed) could be neglected in
easily accessible solute (min) ends. Their quantitative expression
comparison with the concentration inside particles x, and Eq. (25)
may be given as:
becomes
x0 − xi
f (x ≤ xi , y) = ks a0 s x (25a) tCER = (28f)
y∗ · Z · q̇
The concentration profiles in the sold and solvent phases can
1 xi + (x0 − xi ) · exp(W · x0 /y∗ )
then be calculated by integrating from Eqs. (22) and (23a) after tFER = tCER + · ln (28g)
W · q̇ x0
substituting Eqs. (24) and (25a) for the rate of mass transfer. The
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 13
Reference
achieved through from Eqs. (27a)–(27j) with a number of adjustable
parameters regressed against the experimental data. As we can see,
once adjustable parameters include x0 , xi , y*, Z and W are known,
the q-dependent or t-dependent extraction curve can be readily
obtained.
As we know, the time-dependent extraction curve usually
ks (×10−9 m/s)
2
1.8–5.0 × 10
0.038–2.88
exhibits an initial linear extraction period followed a smooth
0.88–2.87
2.36–6.81
0.90–20.7
60.0–90.0
asymptotic plateau for SCCO2 extraction of various plant substrates.
3.49
0.51
Thus there may be two ways to estimate x0 . One common way is to
11
estimate x0 from the exhaustive extraction yield, taking it to be the
maximum yield over all the conditions investigated [40,78,90]. The
7.73–35.0 × 102
other way is to experimentally determine it with organic solvent
3.88–9.28 × 10
0.016–0.059
kf (×10−8 m/s)
[91,92]. The solute solubility y* can be directly measured in SCCO2
4.02–6.14
under low flow rates to ascertain the solute saturation [92]. It can
344–547
291–847
also be evaluated it from the experimental plot of the extraction
yield as a function of the mass of solvent flowed [90]. Since the
466
541
102
initial part of the extraction curves with their linear shape strongly
suggests that equilibrium conditions may have been attained in the
fluid coming out of the extractor and then the initial oil concentra-
tion in the fluid x0 can be approximately by x0 = e/q.
0.598–0.823
0.187–0.918
0.511–0.988
0.587–0.673
0.35–0.62
As for the parameter xi , it is commonly fitted from the exper-
imental data. Instead, it may be replaced by r (r = xi /x0 ) to be
0.123
0.321
0.555
0.335
regressed [92–95]. Note that x0 , xi or r in the modeling should be
r
taken as a constant at different conditions for its resolved phys-
ical meanings, but sometimes they may be adjustably regressed
and then strongly dependent on the extraction conditions (see
Tables 5–7).
y* (×10−3 g/g
24.2–144.4
0.70–11.7
7.35–14.0
The solvent-phase mass transfer coefficient kf and the solid-
2.1–4.69
1.5–25.0
solvent)
phase mass transfer coefficient ks can be regressed from the data
14.47
20.74
17.36
5.05
fitting [51,93,96,97]. In the calculations, parameter a0 must be
solved before using the BIC model and it may be estimated using:
6(1 − ε) 3(1 − ε)
a0 = = (29)
dp R
x0 (×10−2 g/g
10.26–49.01
7.27–8.56
2.20–13.1
solute-free
7.24
8.18
4.25
feed)
11.9
12.8
bed may be obtained using the actual mass of feed and the volume
of the extraction unit.
As discussed previously, kf can also be obtained from certain
correlation [36,37]. Table 5 shows calculated kf and ks values
30.0–40.0
for the BIC model. As seen in Table 5, the kf and ks values are
P (MPa)
15–45
15–45
10–20
15–30
45.0
308–338
313–333
303–323
313–323
313
333
313
323
Rosehip seeds
nelumbinis
Corn germ
Calendula
italicum
flowers
flowers
Baizhu
Matrix
seed oil [98], parsley seed oil [99], orange peel oil [88], safflower
Oil
Oil
Oil
Oil
Oil
Oil
Oil
seed oil [100] and Olive husk oil [73]. Similarly, the Z q̇ is much
14 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
0.25–6.21
0.42–1.46
2.88–4.33
0.44–2.11
0.15–0.84
0.16–0.66
0.23–1.06
coefficient ks a0 are the most preferably used parameters
instead, to avoid the effect of the different ways for spe-
cific interfacial area calculations, and thus widely reported
in the literatures. Table 7 summarizes calculated results by
using the BIC model for oil extraction of various vegetable
matters [3,29,41,51,53,54,65,66,78,84,87,89–92,94–96,102–119].
Z q̇ (×10−2 s−1 )
0.011–0.068
0.12–5.12
0.64–5.70
0.10–517
1.5–25.7
1.9–14.5
higher than those ks a0 ones, due to the fact the former is strongly
related to the diffusion of the released oil close to the particle sur-
14.5
0.421–0.479
0.511–0.588
0.046–0.479
0.319–0.603
0.072–0.45
0.533–0.60
2.12–9.55
2.12–9.55
2.30–4.93
content (see Fig. 4b) and extraction temperature (see Fig. 4c), due
2.8–33.0
solvent)
12–110
to the increase in driving force and convection. But they are both
13.2
19.3
decreased with the increase in the particle dimension (see Fig. 4e),
95
due to the decrease in the specific surface area. The increase in the
diffusion path length may also contribute the decrease observed in
ks a0 . However, extraction pressure is seen to influence kf a0 and ks a0
in two opposite trends, as shown in Fig. 4d. The ks a0 increases with
x0 (×10−2 g/g
pressure but the kf a0 decreases with it. The reason for it is due to the
solute-free
fact that the increased oil solubility (i.e., increase in driving force)
24–32
10–15
feed)
The Z q̇ and W q̇ values obtained with the BIC models for different SCCO2 extraction systems.
40.8
5.8
12.6
50
10
22–31
22–28
10–20
20.0
20.0
P (MPa)
5.4. Applications
318–328
308–318
308–333
T (K)
323
313
313
cept of broken and intact cells has been developed into the BIC
model, which has been verified by Reverchon and Marrone [48].
These researchers have employed SEM analyzer to successfully
identify broken cells on the particle surfaces, and then the volumet-
Naveline cultivar
Satsuma cultivar
orange peel
orange peel
on the particle surfaces. The BIC model has taken into account the
Matrix
solute solubility y* in the solvent phase and the mass transfer coef-
ficient both in the fluid kf and in the solid phases ks , resulting in
an analytical solution for plug flow SFE processes if neglecting the
accumulation of the solute in the fluid phase. Physically speaking,
Safflower seed oil
the BIC model [41,53] has reasonably described the sudden reduc-
Parsley seed oil
Orange peel oil
Celery seed oil
tion in extraction rate after the initial fast extraction period through
Olive husk oil
Essential oil
Essential oil
proposing the broken and intact cell concept. This is actually sup-
ported by many oil-rich examples where the solute concentration
Solute
Table 6
at the outlet from the extraction unit is near constant in the early
stages of the extraction and then falls progressively as extraction
Table 7
The ks a0 and kf a0 values obtained with the BIC models for different SCCO2 extraction systems.
Solute Matrix T (K) P (MPa) x0 (×10−2 g/g y* (×10−3 g/g r kf a (×10−2 s−1 ) ks a (×10−4 s−1 ) Reference
solute-free solvent)
feed)
Shiitake oil Shiitake mushroom 303–323 15.0–30.0 0.55–1.00 2.31–2.52 0.367–0.968 0.010–0.044 0.38–0.87 Kitzberger et al. [3]
Oil Wormwood 313–323 9.0–18.0 3.56 0.25–0.79 0.13 52.0 0.0065–0.108 Martín et al. [29]
Grape oil Grape seeds 313 28 14.4 6.85 0.139–0.681 4.06–9.50 0.065–0.15 Sovová et al. [41,53]
Oil Corn germ 313 45.0 10.26–49.01 14.47 0.123 1.64–2.60 0.124–0.317 Nagy et al. [51]
Oil Pumpkin seeds 333 45.0 47.8 20.74 0.321 1.72 0.406
Oil Calendula flowers 313 45.0 8.18 5.05 0.555 2.3 0.148
Oil Paprika fruit 323 45.0 11.9 17.36 0.335 24.78 0.124
Oleoresin Black pepper 297–333 28 1.3–4.5 1.3–40.0 0.18–0.50 0.27–1.50 0.5–2.0 Sovová et al. [54]
Oleoresin Marigold flowers 299–313 12.0–20.0 3.23–9.98 2.804 0.186–0.510 5.17–76.8 × 10−3 0.83–19.0 × 10−2 Campos et al. [65]
Carqueja oil Carqueja 303–313 10.0–30.0 1.21–2.36 2.83 0.289–0.873 1.46–1.78 17.6–17.8 Silva et al. [66]
Oil Soybeans 20.5 7.00 0.33–0.992 0.03–2.8 0.0058–0.26 Rochová et al. [78]
Essential oil and Chamomile flowers 303–313 10.0–20.0 0.91–4.76 1.15–6.32 0.398–0.522 0.024–0.099 0.126–0.249 Povh et al. [84]
oleoresin
Sea buckthorn oil Sea buckthorn 298–333 17.4–27 0.8–1.34 3.0–8.2 0.33–0.62 0.53–1.20 0.05–0.30 Štastová et al. [87]
Sunflower oil Sunflower seeds 303–333 8.0–25.0 41 2.26–6.0 0.31–0.44 0.41–0.64 0.42–1.12 Nimet et al. [95]
*In subcritical 303–333 8.0–25.0 41 880–1500 0.38–0.45 0.18–0.38 2.33–6.50
propane
Oil Helichrysum 313–333 10–20 4.25 1.5–25.0 0.511–0.988 4.84–6.09 2.01–4.28 Ivanovic et al. [96]
italicum flowers
Fig leaf gourd seeds Fig leaf gourd seeds 308–318 18–20 45.3 2.83–6.86 0.386 2.34–4.57 1.73–4.34 Bernardo-Gil et al. [102]
oil
Hazelnut oil Hazelnut 308–321 18.0–23.4 1960 4.68–9.40 0.332 1.29–3.65 0.897–2.81 Bernardo-Gil and Casquilho [103]
Walnut oil Walnut 308–321 18.0–23.4 2450 7.0 – 15.0 0.419–0.661 1.04–3.07 0.517–1.60
Acorn oil Acorn 308–333 15–21 12.1 6.2–48.2 0.33–0.61 0.134–0.602 0.292–5.22 Bernardo-Gil et al. [104]
Rosemarry oil Rosemarry leaves 313–333 10.0–18.0 3.2 0.238–0.457 0.281–0.625 0.64–2.75 0.837–1.17 Bensebia et al. [105]
Sesame oil Sesame seeds 323–343 25.0–35.0 50.6 6.5–12.6 0.195–0.459 0.11–1.15 2.60–10.33 Döker et al. [106]
Oil Buriti chips 313–328 20–30 8.11 1.74–6.82 0.15 0.195–0.449 0.65–6.5 de França et al. [107]
Essential oil Black pepper 303–323 15–30 2.20–13.1 24.2–144.4 0.35–0.62 0.016–0.059 0.038–2.88 Ferreira & Meireles [108]
Seed oil Grape seeds 313 16.0–20.0 32.9–42.3 1.99–3.41 0.27–0.43 3.52–4.06 0.41–6.27 × 10−2 Passos et al. [109]
Oleoresin Ginger 293–313 15.0–25.0 2.0–3.0 4.12–7.20 – 3.33–4.50 × 10−2 0.98–1.69 Martínez et al. [110]
Carqueja essential Carqueja 313–343 9.0 2.34 6.66–19.1 0.034–0.363 0.82–2.31 2.8–6.9 Vargas et al. [111]
oil
Pigments Bixa orellana seeds 313 20.0 1.2 0.032 – 0.82–1.05 0.5 Nobre et al. [112]
Stevia glycosides Stevia ginger 303 20.0 3.04 2.27 0.23 3.02 × 10−2 0.189 Pasquel et al. [113]
Extract Guava leaves 308–328 30.0 2.8–3.06 3.00 0.51–0.54 0.099–0.0126 0.42–0.44 Moura et al. [114]
Canola seed oil Canola seed 313–333 20.0–25.0 58.7 2.40–5.53 0.587–0.710 0.58–2.78 0.011–0.059 Pederssetti et al. [115]
*In subcritical 303–343 8.0–312.0 58.7 212–360 0.631–0.743 0.037–0.14 0.31–0.72
propane
Oil Croton 288 6.7 3.3 28.3 – 0.013–0.030 0.83–1.11 Sousa et al. [116]
Volatile oils Vetiver roots 313 20.0 2.9–4.7 1.76–10.01 0.172–0.313 0.328–0.515 0.276–0.633 Talansiera et al. [117]
Oleoresin Ginger 298–308 15.0–25.0 2.48–3.46 4.20–8.23 0.355–0.522 0.020–0.034 0.024–0.089 Zancan et al. [118]
Oil Rosehip seeds 313–323 30.0–40.0 7.27–8.56 7.35–14.0 0.587–0.673 2.28–10.1 1.77–2.31 del Valle et al. [119]
15
16 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
Fig. 4. Influences of various variables on the ks a0 and kf a0 obtained using the BIC model for apricot kernel oil extraction with SCCO2 .
Data from [92].
processes. The analytical model resulted has greatly simplified the The BIC model has many different versions as different assump-
calculations of model parameters from the overall extraction curve. tions about the mass transfer mechanism within the solid particles
Thus it is the most widely used models up to now (see Tables 5–7) have been employed [75]. Reverchon and Marrone [48] have mod-
and now it seems to be a very convenient tool for scaling-up design ified the BIC model with the SEM analysis of the microstructure
purposes [90,94]. to acquire cell structure data and greatly reduced the number of
However, the equilibrium relationship, the equilibrium solubil- parameters that should be regressed from experimental data. They
ity y*, for free solute has been incorporated into the BIC model have retained only the internal mass transfer coefficient as one
without considering the solute–sold interactions. As discussed pre- adjustable parameter while the film mass transfer coefficient can
viously, the interactions between the solute and solid matrices be estimated from dimensionless correlations. A fair good fitting of
will affect the solute desorption equilibrium in both the fluid and all the available experimental results has been obtained for SCCO2
solid phases. To account for these interactions, several researchers extraction process of sunflower, coriander, grape, tomato peanut,
[41,48,53,57,121–123] have proposed a linear partition relation- and almond fennel seeds. However, this modification requires SEM
ship in terms of partitioning coefficient between the solid and fluid expertise to avoid any confusion between the oil bearing cells and
phase for modeling the SCCO2 process. Goto et al. [56] have applied biological cells and it could not give the accurate results for the
a BET-type adsorption isotherm to simulate a smooth transition seeds having very low initial oil contents. Besides, in this work
between the equilibrium of free solute at high concentrations and the authors had applied a size reduction pretreatment, which may
the equilibrium of solute–matrix interaction at low concentrations. be of limited practical significance for oilseed extraction. Later
Besides, Perrut et al. [58] has proposed a discontinuous equilibrium on, Fiori [125] has applied the Reverchon and Marrone version
relationship to cover the solute–sold interactions. Above the point of the BIC model and a new version for grape seed oil extrac-
of discontinuity the fluid-phase concentration is equal to solubility tion, resulting in a value of the effective diffusivity in the range
and below the point it is determined by partition coefficient. The of 10−12 –10−10 m2 /s and that of the internal transport coefficient
discontinuous equilibrium curve has been later applied by Wu and in the range of 10−8 –10−7 m/s, respectively. The modeled results
Hou [59] in the simulation of egg yolk oil extraction and by Sajfr- have also shown that the choice of the value of the maximum
tová et al. for near-critical extraction of -sitosterol and scopoletin extractable oil yield seems to be of great importance for determin-
from stinging nettle roots [124]. Concerning the solute interactions ing model parameter and using the maximum experimental value
with the solid matrix, Sovová [46] has later proposed a general is preferable. Sovová has recast the original BIC model by fully con-
new approach for seed oil and essential oil extraction. The new sidering the solute–solid interactions. The modified version seems
feature of the approach is the description of the first extraction to be versatile for the SFE simulation as it includes flexible rela-
period considering different types of phase equilibria: indepen- tionships both for phase equilibrium and flow pattern. Thus it can
dent on matrix in terms of solubility equilibrium, adsorbed on the be advantageously used to simulate different types of extraction
matrix in terms of partition between the two phases, and different curves involving mass transfer coefficients and equilibrium con-
flow patterns such as dispersion. The model has also been veri- stants. However, the new model has no more analytical solution but
fied on data sets selected from literature for oil extraction from numerically solved. Furthermore, the number of model parameters
almond seeds and extraction of essential oils from orange peels and becomes more and varies in accordance with varying complex-
pennyroyal. ity of simulated extraction processes in the range 1–3 for phase
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 17
The authors [129] then have applied the proposed model with
two adjustable parameters to supercritical extraction of different
Fig. 5. Influences of CO2 flow rates on the kf and Dl obtained with the BIC model for
SCCO2 processing of pretreated rosehip seed.
biomass. Results for cashew nut shells and black pepper oil show
Data from [119]. that the model agrees well with the experimental results for the
seeds having both high as well as low initial oil contents. The com-
parison of the proposed model results with the BIC model [41,53]
results indicate that the accuracy of proposed model is of same
equilibrium, 2–3 for mass transfer, and 0–1 for flow pattern.
order although the proposed model is simplified by considering
Recently, Sovová et al. [126] have attempted this new model involv-
two-section kinetics as against three section kinetics in the origi-
ing solute–solid interactions for describing multi-component
nal one. Likewise, rather similar two-stage modifications have also
extraction of vegetable substrates. But this model still presents a
been reported in the literatures [29,98,130–134].
limit in the case of essential oils extraction, when the extractable
Similarly, Martínez et al. have developed another simple modi-
material is located only inside the matrix: the concept of broken
fication of the BIC model, which is taken as the logistic model (LM)
and intact cells is no more applicable and the first part of extraction
here for better description. The LM model has neglected the accu-
controlled by equilibrium does not apply [11].
mulation and dispersion in the fluid phase as these phenomena
If the solute accumulation in the fluid phase is taken into account
have no significant influence on the process while compared to the
along with the other same assumptions, Eq. (25a) may be given [99]
effect of convection. Following this model, the mass balance Eq.
y
(23a) for the fluid phase may be rearranged as [110]
f (x ≤ xi , y) = ks a0 s x 1 − (25b)
y∗ ∂y
uε = F(x, y) (23b)
Therefore the system of differential Eqs. (22) and (23a), coupled ∂z
with the expressions (Eqs. (24) and (25b)) for f(x, y) proposed by The interfacial mass-transfer term, represented as F(x,y), can be
Sovová [41], can be solved numerically by employing a fourth order interpreted as a function Si (t) that depends on the composition of
Runge–Kutta method and assuming that the solid bed is divided the extract along the process. The extract composition may be pos-
into a number of 10 stages where the solute concentration can be sibly divided into a few groups of chemically similar compounds,
considered uniform in each stages. Following this treatment, Louli whose composition could be measured along the SFE process. The
et al. [99] have attempted the modeling of SCCO2 extraction of pars- logistic equation, usually applied to model time-dependent popu-
ley seed oil but found that the fitted parameters are consequently lation growth, is then chosen to describe the variation of the extract
same with those of the analytical BIC model, which means that composition, giving:
the accumulation term has not affected the description of process
significantly.
n
n
Ai bi exp(bi tmi − bi t)
F(x, y) = Si (t) = (30)
del Valle et al. [119] have modified the BIC model by considering [1 − exp(bi tmi − bi t)]
2
the axial dispersion and then Eq. (23) may be given as follows. i=1 i=1
imental CER period values to get model parameters. Ferreira and QCO2 H 1 1
= − (32)
Meireles [89] obtained the mass transfer coefficient from the CER uε 1+ exp(bi tmi − bi t) 1+ exp(bi tmi )
period using a logarithmical solute mass ratio difference. Martínez
18 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
As it can be further derived that when time tends to infinity, the solute concentration in pores at the particle surface. ε is noted here
mass of the extracted component i tends to a fixed value, asymp- to be the porosity of bed including the intrinsic particle porosity.
totically. This value can be considered as the total extractable mass Time dependence of the solid-phase concentration, i.e., average
of this compound at the given process conditions. Taking this into solute concentration in a particle, is assumed to be equal to the rate
account and with rearrangement of Eq. (32), a new equation for the of mass transfer of the solute within external film surrounding in
extraction curve of a single group of compounds was defined: the particle, giving
Ci = 1 at = c (36a)
6.2. Mathematical representation
∂Ci
= Bi[C − Ci ( = 1)] (37a)
Considering axial dispersion, the mass balance of the solute in ∂
=1
the bulk fluid phase is described by [42]
¯ = 3
(38a)
c
∂c ∂c ∂2 c 1 − ε 3kf
+u = Dl 2 − [c − ci (R)] (33)
∂t ∂z ∂z ε R at = 0, c = 1,
= 1 and C = 0 (39a)
Table 8
Applications of the LM model for different SCCO2 extraction systems.
Shiitake mushroom Shiitake oil 303–323 15.0–30.0 −5037.9 to 63.4 1.35–3.18 Kitzberger et al. [3]
Baizhu herb Extract 313–333 15.0–45.0 −114.2 to −31.7 0.80–2.88 Huang et al. [40]
Marigold flowers Oleoresin 299–313 12.0–20.0 −500.2 to 508.2 0.45–0.82 Campos et al. [65]
Peach almond Oil 313 15.0–25.0 −2522.0 to 56.0 1.23–6.23 Mezzomo et al. [94]
Croton Oil 288 6.7 −1042.6 to −25.6 1.55–2.75 Sousa et al. [116]
Guava leaves Extract 308–328 30.0 −1600 1.37 Moura et al. [135]
If axial dispersion is negligible, a quasi-steady state solution may When the solute–solid interaction may not be negligible, the
be obtained by assuming that changes in c during the residence solute equilibrium in the solid phase and within the pores should be
time ( = 1/ = De H/uR2 ) of solvent in the column are negligible and taken into account in the SC model. The initial solute in the porous
that changes in the axial direction of the column at any time are solid may be either adsorbed on solid sites, ca0 (g/m3 ), or being free
very small. Then integration of Eq. (43) gives within the pores, ci0 (g/m3 ). The total initial solute concentration
1−ε in a particle, cs0 (g/m3 ), is composed of two parts:
3Bi Z
C = 1 − exp − · (46)
ε 1 − Bi (1 − 1/c ) cs0 = (1 − εp )ca0 + εp ci0 (49)
20
Table 9
The De , kf and Dl values obtained with the SC model for different SCCO2 extraction systems.
Solute Matrix T (K) P (MPa) Csat (×10−3 g/g solvent) De (m2 /s) kf ( m/s) Dl (m2 /s) Reference
−9 −4
Nutmeg oil Nutmeg 313–323 10.0–20.0 9.0–82.0 1.90–76.9 × 10 2.65–21.4 × 10 5.49–69.9 × 10−5 Machmudah
et al. [70]
Sesame oil Sesame seeds 323–343 25.0–35.0 6.5–12.6 4.43–5.57 × 10−10 5.11–6.05 × 10−6 2.69–3.06 × 10−11 Döker et al.
[106]
ˇ-Carotene Apricot bagasse 313–333 30.4–50.7 4.99–32.13 × 10−3 4.66–8.66 × 10−11 0.97–1.70 × 10−11 3.87–13.01 × 10−11 Döker et al.,
where εp is the porosity of the solid particle. Then, the mass balance binary diffusion coefficient (D12 ), and the effective solute diffu-
of the solute in the solid phase may be written as [55,56,106] sivity in the particles (De ). These parameters are evaluated using
existing correlations and available data. The methods for obtain-
∂ci ∂cs De ∂ ∂ci
εp + (1 − εp ) = 2 r2 (34a) ing the values of kf and D12 have been outlined in previous section.
∂t ∂t r ∂r ∂r As for De , it may be directly taken as an adjustable parameter and
Likewise, Eq. (44) may be expressed as regressed from the experimental extraction data. Alternatively, it
can be readily estimated as a function of D12 for vegetable oils in
∂c Bi (1 − C) SCCO2 and the value of the particle porosity that is determined by
= (44a)
∂ εp Bi (Ci − 1) + (1 − εp )ωc [c + Bi (1 − c )] using mercury porosimetry assay:
If the interaction of solute with the fluid and solid phase is De = ε2p D12 (56)
assumed to be fast adsorption/desorption equilibrium, a linear
expression can be given by [55]: Similarly, the solute axial dispersion coefficient in the super-
critical phase in packed beds, Dl , can be estimated with several
cs = Kci (50) correlations that are summarized in this review [36,37]. The corre-
lation proposed by Funazukuri et al. [136] may be written in terms
where K is the equilibrium coefficient between the solute in the
of dimensionless Reynolds number and dimensionless Schmidt
fluid and solid phases.
number, giving
The linear driving force approximation may be used to combine
internal and external processed and defined by [55] εDL
= 1.317(ε Re Sc)1.392 for ε Re Sc > 0.3 (57)
kf 1 − ε 15De D12
[c − ci (R)] = (1 − ε)[ci (R) − ci ] (51)
R ε R2 These researchers have measured the Dl values of several
Then, Eqs. (33) and (34a) can be further reduced without con- species in SCCO2 under pressures from 11 to 35 MPa, using the
sidering axial dispersion as follows chromatographic method in a packed bed (ε = 0.38) with particles
c of 0.05–0.15 mm diameter. The proposed correlation is found to fit
∂c c 3kp s the experimental data quite well and the reported average abso-
ε + = (1 − ε) −c (33a)
∂t R K lute deviation is 25% over 280 data points with Schmidt numbers
ε ∂c 3kp
cs
ranging from 3.9 to 665. del Valle and de la Fuente [36] have also
p i
+ (1 − εp ) = c− (34b) addressed that this correlation may provide consistent values for
K ∂t R K
the dispersion of solutes in SC–CO2 in a packed bed for a wide range
where kp is the overall mass transfer coefficient for a spherical of values for Reynolds number.
particle, given by To evaluate the importance of axial dispersion, a dimension-
kf less Peclet number can be introduced and evaluated through the
kp = (52) following expression:
1 + Bi /5
The exact solution of Eqs. (33a) and (34b) is Hu
Pe = (58)
Dl
(1 − ε)cs0
c= (em1 t − em2 t ) (53) As discussed previously, several studies [42,48,60,61] have
ε[εp (1 − εp )K](m1 − m2 )
shown that the axial dispersion may have little effect on SCCO2
where extraction as if Pe or Pe /ε is generally larger than 100, due to the
fact that the size of solid particles typically loaded in the bed is
− + 2 − 4˚
m1 = (54a) very small as compared to the diameter of the extraction unit
2 with a ratio of the former over the latter less than 0.1 and then
− − 2 − 4˚ the axial dispersion problem may have been avoided during SFE
m2 = (54b) process.
2
= kp ap (54c)
6.4. Applications
1 (1 − ε)
= + + (54d)
ε ε εp + (1 − εp )K The SC model has been extensively used for simulating SFE
processes of herbaceous matrices such as sesame seeds [106],
˚= (54e) sunflower seeds [137], rapeseed [138], neem seeds [35,85,139],
ε[εp + (1 − εp )K] nutmeg seeds [140], grape seed [141], caraway seed [141], canola
The cumulative fraction of solute extracted is analytically given seed [52], rosehip seeds [52], corn germ [52], olive husk [52], ginger
by [142], soybean meal [143], apricot bagasse [120], and peppermint
t
[55]. Tezel et al. [144] have further developed a multi-component
cs0 − cs 1 (1 − ε) SC model for seed and essential oil extractions and Machmudah
= c dt =
cs0 (1 − ε)cs0 0
ε[εp + (1 − εp )K](m1 − m2 ) et al. [140] have used it for modeling extraction of two compo-
nents from nutmeg seeds. In their work [140], the SC model has
em1 t −1 em2 t − 1 been used twice for the lighter components and heavier compo-
× − (55)
m1 m2 nents, resulting in the De values in the range of from 4.33 × 10−9 to
7.69 × 10−8 m2 /s and from 1.90 × 10−9 to 3.20 × 10−8 m2 /s, respec-
tively.
6.3. Parameter identification Tables 9 and 10 summarized the correlated results reported
in the literatures [3,35,52,70,85,106,116,117,120,135,137–143].
The mass transfer parameters involved in the shrinking core Among these works, several researchers have systematically inves-
model are the mass transfer coefficient to the bulk phase (kf ), axial tigated the influences of various process parameters such as
dispersion coefficient in the fluid phase (Dl ), the solute–solvent extraction pressure, temperature, particle size and CO2 flow rate on
22 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
Table 10
The calculated and K values with the SC model for different SCCO2 extraction systems.
Shiitake oil Shiitake mushroom 303–323 15–30 0.15–0.39 1.15–1.77 Kitzberger et al. [3]
Nimbin Neem seeds 308–333 10.0–26.0 7.7–97.4 0.132–0.596 Tonthubthimthong et al. [35]
Volatile oils Croton 288 6.7 0.42–6.96 0.42–0.50 Sousa et al. [116]
Volatile oils Vetiver roots 313 20.0 0.196–1.964 0.735–0.854 Talansier et al. [117]
Oil Fennel seeds 303–313 10.0–25.0 0.088–0.202 0.400–0.476 Moura et al. [135]
Fig. 7. Influences of various variables on the kf and Dl obtained using the SC model for extraction of sunflower oil with SCCO2 .
Data from [137].
Fig. 8. Influences of various variables on the kf and Dl obtained using the SC model for nimbin from neem seeds with SCCO2 .
Data from [139].
the mass transfer rates [35,137,139]. Fig. 7 shows modeled results different trends have been obtained and their calculations are
for extraction of sunflower oil with SCCO2 [137]. As we can see shown in Fig. 8. Their results show that (1) the increase in
from Fig. 7a–c, the kf , De and Dl are all increasing with the extrac- particle size has led to the reduction of the kf values but the
tion pressure and increase in temperature has also promote the increase in the Dl , (2) the increase in CO2 flow rate has made
increase of kf and De . Fig. 7d shows the increases in CO2 flow rate kf and the Dl both being increased, (3) the increase in pres-
can result in increased mass transfer rate but render the axial dis- sure has resulted in the decrease in the kf and De as well.
persion reduce. Similar observations have applied to the case of The reason for it may be due to the different assumptions
increase in particle size, as shown in Figs. 7e and f. However, in involved and further exploiting is needed. Furthermore, in the
the studies of Tonthubthimthong et al. [35] and Ajchariyapagorn study of Machmudah et al. [140], the calculations show that the
et al. [139] for SCCO2 extraction of nimbin from neem seeds, Bi value thus obtained varies from 14.9 to 107 and is larger
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 23
than 10, seemingly indicating that the internal mass transfer seed oil [125]. But these results show the model performance is
resistance is much greater than the external mass transfer resis- dependent on the real granulometric size distribution other than
tance. In addition, the Pe estimated spans from 7.5 to 26.3 and far the weight-averaged particle dimension.
smaller than 100, indicative pronounced axial dispersion on the
mass transfer rate.
Apart from the above-mentioned process parameters, the other 7. Conclusions
factors, based on the SC model calculations, may also influence the
extraction process as well, like the microstructure, drying and flow This paper has summarized representative developments of
type. del Valle et al. [52] have researched microstructural effects on theoretical models for supercritical fluid extraction of solutes from
internal mass transfer of vegetable oil from seeds subjected to vari- various solid matrices. These models thoroughly discussed here
ous high-shear pretreatments using the SC model and explored the include the hot ball diffusion, broken and intact cell, shrinking core
application of microstructure–extractability relationships for mod- and some relatively simple models. These models are physically
eling SCCO2 extractions. The authors have quantitatively related meaningful and valuable to represent the overall extraction pro-
the effective diffusivity (De ) inside the porous particles with the files as they are developed from differential mass balance equations
oil diffusivity (D12 ) in SCCO2 through a microstructural correction for the packed solid bed. Due to different assumptions adopted for
factor, which is estimated as the ratio between the porosity and mass transfer kinetics, these models thus bear different mathe-
tortuosity of the particle itself. Besides, Balachandran et al. [142] matical expressions. Special emphasis is given to the models with
have investigated supercritical extraction of oleoresin from fresh, analytical solutions as they can readily be used for simulating
oven-dried and freeze-dried ginger samples. They have found that experimental data. The interactions between the solute and solid
water-rich fresh ginger rhizomes lead to higher extraction yields matrices and the possible axial dispersion are also discussed. Of
and a more pungent and naturally flavored extract relative to the these models, the broken and intact cell model is probably the
other two dried samples. The modeled results with the SC model most appropriate one as it seems to capture the solid structure and
confirm that this yield enhancement may have resulted principally the extraction process kinetics and may enable us to develop the
from an enhancement in the effective diffusivity within the gin- scale-up of laboratory data for industrial design purposes.
ger particle and that the mass transfer rate would decline once Despite great achievements made so far, further efforts and
dried ginger samples was used. Germain et al. [141] have investi- developments in modeling and scaling-up of SCCO2 extraction pro-
gated SCCO2 extraction of essential oils and lipids from grape seeds, cesses for various solid matrices are still required. These works
caraway seeds and rapeseeds under solvent flow conditions (down- may include elaborating the relationship among effective solute
flow and up-flow) and low superficial velocities. Their works show diffusivity, microstructural features of vegetable substrates and
that there are no differences in mass-transfer rates between solvent sample treatment, developing convenient and reliable optimiza-
down-flow and up-flow for at a high Reynolds number (Re = 14.1). tion methods for theoretical mass transfer models involved with
But kf at lower Re of 1.57 was 3.6 times smaller when using solvent axial dispersion and solute–solid matrix interactions.
up-flow than solvent down-flow condition. That is to say that, the
extraction is slower for solvent up-flow rather than solvent down- Nomenclature
flow condition when using a superficial solvent velocity below a
threshold value between 0.17 and 1.53 mm/s. The reduction in
the extraction rates is probably related to the negative influence
of natural convection phenomena as the loaded SCCO2 phase when a0 specific surface area of a solid substrate particle (m−1 )
flowing slowly from the bottom of the vertically positioned extrac- A cross sectional area of the extraction vessel (m2 )
tion unit can be affected negatively by gravity. Thus the interstitial Ad a dimensionless model coefficient (Eq. (9))
solvent velocities should be set to values high enough to make the Ai characteristic parameters of the logistic model (Eq. (30))
effect of natural convection on the mass-transfer rates negligible. b characteristic parameter of the logistic model (Eq. (32b))
However, at higher velocities, the unsteady extraction may be pro- (s−1 )
duced by solvent flow asymmetries or channeling phenomena due bi characteristic parameter of the logistic model (Eq. (30))
to agglomeration and deformation of particles, increase bed poros- (s−1 )
ity or fluidized or float particles under up-flow conditions.Recently, Bi dimensionless Biot number defined as Bi = 2kf R/De
Machmudah et al. [140] have compared the performances of the c concentration of solute in the bulk SC CO2 phase (g/m3 )
BIC and SC models for SCCO2 extraction of nutmeg oil at pressures C (=c/csat ) dimensionless solute concentration in the bulk SC CO2
of 15–20 MPa and temperatures of 313–323 K. Their work shows phase
that the BIC model with discontinuous phase equilibrium between c0 the solute concentration c at t = 0, usually assumed to be
fluid and solid phases [58] could only describe the data at lower 0
extraction yields well whereas the SC model could describe the ca0 initial solute concentration adsorbed on solid sites (g/m3 )
experimental data well for all extraction conditions. Then, Fiori c|R concentration of the solute at the particle surface (Eqs.
et al. [145] have attempted to bridge the BIC model and SC model (13)–(15)) (g/m3 )
for a better evaluation and proposed a new model. This model ci solute concentration in pores within particles (g/m3 )
has plausibly taken into account the inner structure of the seeds ci (R) the solute concentration in pores at the particle surface
consisting of the vegetable cells bearing the oil, and schematized (g/m3 )
ground seed particles as spherical particles composed of concen- Ci (=ci /csat ) dimensionless solute concentration in pores within
tric shells each one consisting of an oil-bearing cell layer. As the particles
extraction process advances, each shell is irreversibly depleted of ci0 initial solute concentration within the pores of the parti-
its oil, from the outer shell consisting of broken cells to the inner cles (g/m3 )
shells consisting of intact cells. These considerations have incorpo- cs solute concentration in the solid phase (g/m3 )
rated the ‘shrinking-core’ concept with the ‘broken and intact cell’ c̄s averaged solute concentration in the solid phase (Eq. (34))
one for describing the supercritical extraction kinetics of seed oils. (g/m3 )
This model has been verified by modeling extraction of grape seeds cs0 initial concentration of solute in the solid or adsorbed
[145] and later evaluated for supercritical extraction of sunflower onto the solid (g/m3 )
24 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
csat concentration of solute that is in equilibrium with a solid Si (t) population growth function defined in Eq. (30) (s−1 )
surface (g/m3 ) t extraction time (s)
D12 binary diffusion coefficient of solute and SCF (m2 /s) T extraction temperature (K)
De effective diffusion coefficient in a solid substrate (m2 /s) tCER time for constant extraction rate (Eq. (28f)) (s)
De1 , De2 different effective diffusion coefficient (Eq. (20e)) (m2 /s) te time required for complete equilibrium extraction (Eq.
Dl solute axial dispersion coefficient in the SCCO2 solvent (48a)) (s)
(m2 /s) tFER time used till the extraction of easily accessible solute
dp particle diameter (m) ends (Eq. (28g)) (s)
dpi the diameter of the sieve i (Eq. (21)) (m) tm characteristic parameter of the logistic model (Eq. (32b))
e time-dependent extraction yield, (g solute/g solute-free (s)
feed) tmi characteristic parameter of the logistic model (Eq. (32))
F volumetric flow rate of the SCCO2 solvent (m3 /s) (s)
Fr the fraction of the analyte released quickly (Eq. (10)) u the superficial velocity of the SCCO2 solvent (m/s)
H height of the packed bed (m) V void volume of the extraction chamber (m3 )
h thickness of the diffusion layer assumed in the DLT model Va , Vb volume of the SCCO2 solvent consumed (Eq. (12)) (m3 )
K equilibrium coefficient in Eq. (50) W dimensionless mass transfer parameter in the solid phase
kB Boltzman constant (J/K) (Eq. (28e))
kd desorption rate constant (s−1 ) x concentration of solute in the solid phase (g solute/g
kd1 the first-order rate constant describing the quickly solute free solid bed)
released fraction (Eq. (10)) (min−1 ) x0 the initial analyte concentration in the solid phase (g
kd2 the first-order rate constant describing the slowly solute/g solute free solid bed)
released fraction (Eq. (10)) (min−1 ) xb initial concentration of the readily accessible solute (g
kf film mass transfer coefficient (m/s) solute/g solute free solid bed)
kf a0 volumetric mass transfer coefficient in the solvent phase xf interfacial solute concentration (g solute/g solute-free
(s−1 ) feed) (Eq. (25))
ks solute mass transfer coefficient in the solid phase (m/s) xi initial concentration of the less accessible solute (g
ks a0 volumetric mass transfer coefficient in the solid phase solute/g solute free solid bed)
(s−1 ) xr , x , xϕ the dimensionless concentrations along coordinates r,
kp overall mass transfer coefficient for a spherical particle and ϕ (Eq. (13))
(Eq. (52)) (m/s) y (=c/f ) concentration of solute in the SCCO2 fluid phase (g
Kp partition coefficient between tied solute in the solid and solute/g solvent)
dissolved solute in the SCCO2 solvent (Eq. (11)) y (z = H, t) time-dependent solute concentration at the exit of the
m0 mass of solute to be extracted and present in the parent packed bed (g solute/g solvent)
matrix (g) y* (=csat /f ) solubility of solute in the SCCO2 fluid phase (g solute/g
m1 , m2 dimensionless term defined in Eqs. (54a) and (54b) solvent)
ma , mb the cumulative mass (g) of the analyte extracted after z axial position along the bed (m)
consuming Va , Vb of the SCCO2 solvent (Eq. (12)) Z (=z/H) dimensionless axial position used in the SC model
mbed mass of solute-free feed loaded in the bed (Eq. (28a)) (g) Z dimensionless mass transfer parameter in the fluid phase
mCO2 mass of SCCO2 solvent consumed (Eq. (28a)) (g) (Eq. (28c))
mi (z = H, t) extracted mass of the group i of compounds (Eq. (32)) Zw dimensionless axial co-ordinate between fast and slow
(g) extractions (Eq. (28d))
mi0 total extractable mass of the group i of compounds (Eq.
(32a)) (g) Greek letters
mt time dependence of the mass of the extracted solute (g) ˇ characteristic parameter of the DLT model ˇ = 4R2 De /Vh
mt (z = H) time dependence of the mass of the extracted solute (g) (s−1 )
Np number of particles loaded in the bed defined by Eq. (17) ˇk kth term of the series solution defined by the periodic
P extraction pressure (MPa) equation (16)
Pe dimensionless Peclet number defined as Pe = Hu/Dl m the mass of the loaded sample (Eq. (12)) (g)
q specific mass of solvent consumed (g solvent/g solute-free mi the mass fraction of the solid retained by the sieve i (Eq.
feed) (21))
q̇ (=q/t) specific mass flow rate of solvent used (g solvent/g solute- ε porosity of the packed bed
free feed/s) εp porosity of the solid particle
QCO2 (=F × f ) the mass flow rate of the SCCO2 solvent (g/s) ϕ one of the coordinates
qm q value when extraction begins inside the particles (g sol-
viscosity of the SCCO2 solvent (Pa s)
vent/g solute-free feed) one of the coordinates
qn q value when the easily accessible part of solute is all (=De t/R2 ) dimensionless time used in the SC model
extracted (g solvent/g solute-free feed) e dimensionless time required for complete equilibrium
r radial position within the particle (m) extraction (Eq. (48b))
r (=xi /x0 ) fraction of the less accessible solute a apparent density of the solid loaded in the extraction bed
R radius of spherical solid particles (m) (g/m3 )
solute molecule radius (m) s density of solid substrate particle (g/m3 )
rc radius of a shrinking core (m) f density of the SCCO2 fluid (g/m3 )
Re (=udp /
) dimensionless Reynolds number M the solute mass flux density per unit of area (Eq. (13)) (g
Sc (=
/f D12 ) dimensionless Schmidt number solute/m2 s)
Sh (=dp kf /D12 ) dimensionless Sherwood number (=r/R) dimensionless axial coordinate within a particle
(=rc /R) dimensionless axial position within a particle
Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26 25
(=uR2 /De H) dimensionless term used in the SC model [45] C.S. Tan, D.C. Liou, AIChE J. 35 (1989) 1029.
(=De H/uR2 ) dimensionless residence time in the bed [46] H. Sovová, J. Supercrit. Fluids 33 (2005) 35.
[47] E. Reverchon, A. Kaziunas, C. Marrone, Chem. Eng. Sci. 55 (2000) 2195.
ω (= csat /cs0 ) dimensionless term used in the SC model [48] E. Reverchon, C. Marrone, J. Supercrit. Fluids 19 (2001) 161.
dimensionless term defined in Eq. (54c) [49] E. Reverchon, J. Daghero, C. Marrone, M. Mattea, M. Poletto, Ind. Eng. Chem.
(= cs /cs0 ) dimensionless solute concentration in the solid phase Res. 38 (1999) 3069.
[50] F. Gaspar, T. Lu, R. Santos, B. Al-Duri, J. Supercrit. Fluids 25 (2003) 247.
¯ dimensionless averaged solute concentration in the solid [51] B. Nagy, B. Simándi, C.D. András, J. Food Eng. 88 (2008) 104.
phase (Eq. (38a)) [52] J.M. del Valle, J.C. Germain, E. Uquiche, C. Zetzl, G. Brunner, J. Supercrit. Fluids
37 (2006) 178.
dimensionless term defined in Eq. (54d)
[53] H. Sovová, J. Kučera, J. Jež, Chem. Eng. Sci. 49 (1994) 415.
˚ dimensionless term defined in Eq. (54e) [54] H. Sovová, J. Jež, M. Bártlová, J. St’astová, J. Supercrit. Fluids 8 (1995) 295.
[55] M. Goto, M. Sato, T. Hirose, J. Chem. Eng. Jpn. 26 (1993) 401.
[56] M. Goto, B.C. Roy, A. Kodama, T. Hirose, J. Chem. Eng. Jpn. 31 (1998) 171.
Acknowledgements
[57] O.J. Catchpole, J.B. Grey, B.M. Smallfield, J. Supercrit. Fluids 9 (1996) 273.
[58] M. Perrut, J.Y. Clavier, M. Poletto, E. Reverchon, Ind. Eng. Chem. Res. 36 (1997)
The authors would like to thank the financial support of National 430.
Natural Science Foundation of China through the research grant [59] W. Wu, Y. Hou, J. Supercrit. Fluids 19 (2001) 149.
[60] F. Recasens, B.J. McCoy, J.M. Smith, AIChE J. 35 (1989) 951.
NSFC-20676107 and State Education Ministry of China through the [61] V.M. Rodrigues, E. Sousa, A.R. Monteiro, O. Chiavone-Filho, M.O.M. Marques,
project sponsored by SRF for ROCS, SEM and Tianjin University of M.A.A. Meireles, J. Supercrit. Fluids 22 (2002) 21.
Commerce. [62] Z. Huang, L. Xu, J. Li, Chem. Eng. J. 166 (2011).
[63] A.L. Chassagnez-Mendez, N.C.F. Corrêa, L.F. França, N.T. Machado, M.E. Araújo,
Braz. J. Chem. Eng. 17 (2000) 315.
References [64] M.E. Araújo, N.T. Machado, L.F. França, M.A.A. Meireles, Braz. J. Chem. Eng. 17
(2000) 297.
[1] L.T. Taylor, Supercritical Fluid Extraction, John Wiley & Sons, Toronto, 1996. [65] L.M.A.S. Campos, E.M.Z. Michielin, L. Danielski, S.R.S. Ferreira, J. Supercrit.
[2] J.A. Wood, M.A. Bernards, W. Wan, P.A. Charpentier, J. Supercrit. Fluids 39 Fluids 34 (2005) 163.
(2006) 40. [66] D.C.M.N. Silva, L.F.V. Bresciani, R.L. Dalagnol, L. Danielski, R.A. Yunes, S.R.S.
[3] C.S.G. Kitzberger, R.H. Lomonaco, E.M.Z. Michielin, L. Danielski, J. Correia, S.R.S. Ferreira, Food Bioprod. Process. 87 (2009) 317.
Ferreira, J. Food Eng. 90 (2009) 35. [67] M. Kandiah, M. Spiro, Int. J. Food Sci. Technol. 25 (1990) 328.
[4] W. Chen, C. Chen, C.J. Chang, Y. Chiu, D. Hsiang, J. Supercrit. Fluids 51 (2009) [68] T. Anekpankul, M. Goto, M. Sasaki, P. Pavasant, A. Shotipruk, Sep. Purif. Tech-
174. nol. 55 (2007) 343.
[5] F. Temelli, J. Supercrit. Fluids 47 (2009) 583. [69] A. Kubátová, B. Jansen, J.F. Vaudoisot, S.B. Hawthorne, J. Chromatogr. A 975
[6] M. Herrero, J.A. Mendiola, A. Cifuentes, E. Ibanez, J. Chromatogr. A 1217 (2010) (2002) 175.
2495. [70] S. Machmudah, M. Kondo, M. Sasaki, M. Goto, J. Munemasa, M. Yamagata, J.
[7] F. Sahena, I.S.M. Zaidul, S. Jinap, A.A. Karim, K.A. Abbas, N.A.N. Norulaini, A.K.M. Supercrit. Fluids 44 (2008) 301.
Omar, J. Food Eng. 95 (2009) 240. [71] K.D. Bartle, A.A. Clifford, S.B. Hawthorne, J.J. Langenfeld, D.J. Miller, R. Robin-
[8] F. Mattea, A. Martin, M. Jose Cocero, J. Food Eng. 93 (2009) 255. son, J. Supercrit. Fluids 3 (1990) 143.
[9] S.M. Pourmortazavi, S.S. Hajimirsadeghi, J. Chromatogr. A 1163 (2007) 2. [72] K.D. Bartle, T. Boddington, A.A. Clifford, S.B. Hawthorne, J. Supercrit. Fluids 5
[10] M. Herrero, A. Cifuentes, E. Ibanez, Food Chem. 98 (2006) 136. (1992) 207.
[11] E. Reverchon, I. De Marco, J. Supercrit. Fluids 38 (2006) 146. [73] M.M. Esquível, M.G. Bernardo-Gil, M.B. King, J. Supercrit. Fluids 16 (1999)
[12] G. Brunner, J. Food Eng. 67 (2005) 21. 43.
[13] S. Sarrade, C. Guizard, G.M. Rios, Sep. Purif. Technol. 32 (2003) 57. [74] C.B. Mehr, R.N. Biswal, J.L. Collins, H.D. Cochran, J. Supercrit. Fluids 9 (1996)
[14] C. Turner, C.S. Eskilsson, E. Bjorklund, J. Chromatogr. A 947 (2002) 1. 185.
[15] Q.Y. Lang, C.M. Wai, Talanta 53 (2001) 771. [75] P. Subra, S. Castellani, P. Jestin, A. Aoufi, J. Supercrit. Fluids 12 (1998) 261.
[16] R.M. Smith, J. Chromatogr. A 856 (1999) 83. [76] L. Čretnik, M. Škerget, Ž. Knez, Sep. Purif. Technol. 41 (2005) 13.
[17] E. Reverchon, J. Supercrit. Fluids 10 (1997) 1. [77] N. Vedaraman, C. Srinivasakannan, G. Brunner, B.V. Ramabrahmam, P.G. Rao,
[18] J.Q. Albarelli, R.B. Rabelo, D.T. Santos, M.M. Beppu, M.A.A. Meireles, J. Supercrit. J. Supercrit. Fluids 34 (2005) 27.
Fluids 58 (2011) 343. [78] K. Rochová, H. Sovová, V. Sobolík, K. Allaf, J. Supercrit. Fluids 44 (2008) 211.
[19] Z. Huang, L. Xu, J. Li, S. Kawi, A.H. Goh, Sep. Purif. Technol. 77 (2011) 112. [79] M.D. Macías-Sánchez, C.M. Serrano, M.R. Rodríguez, E. Martínez de la Ossa,
[20] R. van Grieken, G. Calleja, G.D. Stucky, J.A. Melero, R.A. García, J. Iglesias, Chem. Eng. J. 150 (2009) 104.
Langmuir 19 (2003) 3966. [80] H. Ohira, N. Torii, T.M. Aida, M. Watanabe, R.L. Smith, Sep. Purif. Technol. 69
[21] S.M. Talebi, M. Abedi, M.M. Gheisari, M. Saber-Tehrani, Asian J. Chem. 22 (2009) 102.
(2010) 971. [81] M. Hojnik, M. Škerget, Ž. Knez, LWT – Food Sci. Technol. 41 (2008) 2008.
[22] J. Sunarso, S. Ismadji, J. Hazard. Mater. 161 (2009) 1. [82] C.S. Tan, D.C. Liou, Ind. Eng. Chem. Res. 28 (1989) 1246.
[23] D. Garcia-Rodriguez, A.M. Carro-Diaz, R.A. Lorenzo-Ferreira, J. Sep. Sci. 31 [83] O.J. Catchpole, M.B. King, Ind. Eng. Chem. Res. 33 (1994) 1828.
(2008) 1333. [84] N.P. Povh, M.O.M. Marques, M.A.A. Meireles, J. Supercrit. Fluids 21 (2001)
[24] G. Anitescu, L.L. Tavlarides, J. Supercrit. Fluids 38 (2006) 167. 245.
[25] M.D.A. Saldana, V. Nagpal, S.E. Guigard, Environ. Technol. 26 (2005) 1013. [85] D. Mongkholkhajornsilp, S. Douglas, P.L. Douglas, A. Elkamel, W. Teppaitoon,
[26] Z. Huang, L. Huang, S.C. Shen, C.C. Poh, K. Hidajat, S. Kawi, S.C. Ng, Microporous S. Pongamphai, J. Food Eng. 71 (2005) 331.
Mesoporous Mater. 80 (2005) 157. [86] E. Uquiche, J.M. del Valle, J. Ortiz, J. Food Eng. 65 (2004) 55.
[27] Z. Huang, D.Y. Luan, S.C. Shen, K. Hidajat, S. Kawi, J. Supercrit. Fluids 35 (2005) [87] J. Št’astová, J. JeŽ, M. Bártlová, H. Sovová, Chem. Eng. Sci. 51 (1996) 4347.
40. [88] B. Mira, M. Blasco, S. Subirats, A. Berna, J. Supercrit. Fluids 9 (1996) 238.
[28] C. Erkey, J. Supercrit. Fluids 17 (2000) 259. [89] V.M. Rodrigues, P.T.V. Rosa, M.O.M. Marques, A.J. Petenate, M.A.A. Meireles, J.
[29] L. Martín, A.M. Mainar, A. González-Coloma, J. Burillo, J.S. Urieta, J. Supercrit. Agric. Food Chem. 51 (2003) 1518.
Fluids 56 (2011) 64. [90] M.R. Garcia-Risco, E.J. Hernandez, G. Vicente, T. Fornari, F.J. Senorans, G.
[30] Y.H. Lee, A.L. Charles, H.F. Kung, C.T. Ho, T.C. Huang, Ind. Crops Prod. 31 (2010) Reglero, J. Supercrit. Fluids 55 (2011) 971.
59. [91] H.K. Kiriamiti, E. Rascol, A. Marty, J.S. Condoret, Chem. Eng. Process. 41 (2002)
[31] Y.B. Zuo, A.W. Zeng, X.G. Yuan, K.T. Yu, J. Food Eng. 89 (2008) 384. 711.
[32] H.S. Park, H.K. Choi, S.J. Lee, K.W. Park, S.G. Choi, K.H. Kim, J. Supercrit. Fluids [92] S.G. Özkal, M.E. Yener, L. Bayındırlı, J. Supercrit. Fluids 35 (2005) 119.
42 (2007) 205. [93] D. Jia, S. Li, L. Xiao, J. Supercrit. Fluids 50 (2009) 229.
[33] U. Topal, M. Sasaki, M. Goto, K. Hayakawa, J. Agric. Food Chem. 54 (2006) 5604. [94] N. Mezzomo, J. Martínez, S.R.S. Ferreira, J. Supercrit. Fluids 51 (2009) 10.
[34] M. Sun, F. Temelli, J. Supercrit. Fluids 37 (2006) 397. [95] G. Nimet, E.A. Da Silva, F. Palu, C. Dariva, L.d.S. Freitas, A. Medina Neto, L.
[35] P. Tonthubthimthong, P.L. Douglas, S. Douglas, W. Luewisutthichat, W. Tep- Cardozo Filho, Chem. Eng. J. 168 (2011) 262.
paitoon, L. Pengsopa, J. Supercrit. Fluids 30 (2004) 287. [96] J. Ivanovic, M. Ristic, D. Skala, J. Supercrit. Fluids 57 (2011) 129.
[36] J.M. del Valle, J.C. De La Fuente, Crit. Rev. Food Sci. Nutr. 46 (2006) 131. [97] S.R.S. Ferreira, Z.L. Nikolov, L.K. Doraiswamy, M.A.A. Meireles, A.J. Petenate, J.
[37] E.L.G. Oliveira, A.J.D. Silvestre, C.M. Silva, Chem. Eng. Res. Des. 89 (2011) 1104. Supercrit. Fluids 14 (1999) 235.
[38] S.N. Naik, H. Lentz, R.C. Maheshwari, Fluid Phase Equilib. 49 (1989) 115. [98] I. Papamichail, V. Louli, K. Magoulas, J. Supercrit. Fluids 18 (2000) 213.
[39] K. Nguyen, P. Barton, J.S. Spencer, J. Supercrit. Fluids 4 (1991) 40. [99] V. Louli, G. Folas, E. Voutsas, K. Magoulas, J. Supercrit. Fluids 30 (2004) 163.
[40] Z. Huang, M. Yang, S. Liu, Q. Ma, J. Supercrit. Fluids 58 (2011) 31. [100] X. Han, L. Cheng, R. Zhang, J. Bi, J. Food Eng. 92 (2009) 370.
[41] H. Sovová, Chem. Eng. Sci. 49 (1994) 409. [101] A. Berna, A. Tárrega, M. Blasco, S. Subirats, J. Supercrit. Fluids 18 (2000)
[42] M. Goto, B.C. Roy, T. Hirose, J. Supercrit. Fluids 9 (1996) 128. 227.
[43] E. Reverchon, G. Donsi, L.S. Osséo, Industrial and Engineering Chemistry [102] M.G. Bernardo-Gil, M. Casquilho, M.M. Esquível, M.A. Ribeiro, J. Supercrit.
Research, 1993, Medium: X; 2721 pp. Fluids 49 (2009) 32.
[44] T. Veress, J. Chromatogr. A 668 (1994) 285. [103] M.G. Bernardo-Gil, C. Miguel, AIChE J. 53 (2007) 6.
26 Z. Huang et al. / J. Chromatogr. A 1250 (2012) 2–26
[104] M.G. Bernardo-Gil, I.M.G. Lopes, M. Casquilho, M.A. Ribeiro, M.M. Esquível, J. [124] M. Sajfrtová, H. Sovová, L. Opletal, M. Bártlová, J. Supercrit. Fluids 35 (2005)
Empis, J. Supercrit. Fluids 40 (2007) 344. 111.
[105] O. Bensebia, D. Barth, B. Bensebia, A. Dahmani, J. Supercrit. Fluids 49 (2009) [125] L. Fiori, J. Supercrit. Fluids 50 (2009) 218.
161. [126] H. Sovová, A.A. Galushko, R.P. Stateva, K. Rochová, M. Sajfrtová, M. Bártlová,
[106] O. Döker, U. Salgın, N. Yildiz, M. Aydoğmuş, A. Çalımlı, J. Food Eng. 97 (2010) J. Food Eng. 101 (2010) 201.
360. [127] L.F. de Franca, M.A.A. Meireles, J. Supercrit. Fluids 18 (2000) 35.
[107] L.F. de Franca, G. Reber, M.A.A. Meireles, N.T. Machado, G. Brunner, J. Supercrit. [128] J. Martínez, J.M. Martínez, Comp. Chem. Eng. 32 (2008) 1735.
Fluids 14 (1999) 247. [129] R.N. Patel, S. Bandyopadhyay, A. Ganesh, Energy Convers. Manage. 52 (2011)
[108] S.R.S. Ferreira, M.A.A. Meireles, J. Food Eng. 54 (2002) 263. 652.
[109] C.P. Passos, M.A. Coimbra, F.A. Da Silva, C.M. Silva, Chem. Eng. Res. Des. 89 [130] E. Langa, G. Della Porta, A.M.F. Palavra, J.S. Urieta, A.M. Mainar, J. Supercrit.
(2011) 1118. Fluids 49 (2009) 174.
[110] J. Martínez, A.R. Monteiro, P.T.V. Rosa, M.O.M. Marques, M.A.A. Meireles, Ind. [131] C. Perakis, V. Louli, E. Voutsas, K. Magoulas, J. Supercrit. Fluids 55 (2010)
Eng. Chem. Res. 42 (2003) 1057. 573.
[111] R.M.F. Vargas, E. Cassel, G.M.F. Gomes, L.G.S. Longhi, L. Atti-Serafini, A.C. Atti- [132] C. Grosso, J.P. Coelho, F.L.P. Pessoa, J.M.N.A. Fareleira, J.G. Barroso, J.S. Urieta,
Santos, Braz. J. Chem. Eng. 23 (2006) 375. A.F. Palavra, H. Sovová, Chem. Eng. Sci. 65 (2010) 3579.
[112] B.P. Nobre, R.L. Mendes, E.M. Queiroz, F.L.P. Pessoa, J.P. Coelho, A.F. Palavra, [133] M.P. Corso, M.R. Fagundes-Klen, E.A. Silva, L. Cardozo Filho, J.N. Santos, L.S.
Braz. J. Chem. Eng. 23 (2006) 251. Freitas, C. Dariva, J. Supercrit. Fluids 52 (2010) 56.
[113] A. Pasquel, M.A.A. Meireles, M.O.M. Marques, A.J. Petenate, Braz. J. Chem. Eng. [134] O. Boutin, A. De Nadai, A. Garcia Perez, J.H. Ferrasse, M. Beltran, E. Badens,
17 (2000) 271. Chem. Eng. Res. Design 89 (2011) 2477.
[114] P.M. Moura, G.H.C. Prado, M.A.A. Meireles, C.G. Pereira, J. Supercrit. Fluids 62 [135] L.S. Moura, R.N. Carvalho, M.B. Stefanini, L.C. Ming, M.A.A. Meireles, J. Super-
(2012) 116. crit. Fluids 35 (2005) 212.
[115] M.M. Pederssetti, F. Palú, E.A. da Silva, J.H. Rohling, L. Cardozo-Filho, C. Dariva, [136] T. Funazukuri, C.Y. Kong, S. Kagei, J. Supercrit. Fluids 13 (1998) 169.
J. Food Eng. 102 (2011) 189. [137] U. Salgın, O. Döker, A. Çalımlı, J. Supercrit. Fluids 38 (2006) 326.
[116] E.M.B.D. Sousa, J. Martínez, O. Chiavone-Filho, P.T.V. Rosa, T. Domingos, M.A.A. [138] G.A. Núñez, C.A. Gelmi, J.M. del Valle, Comp. Chem. Eng. 35 (2011) 2687.
Meireles, J. Food Eng. 69 (2005) 325. [139] A. Ajchariyapagorn, T. Kumhom, S. Pongamphai, S. Douglas, P.L. Douglas, W.
[117] E. Talansier, M.E.M. Braga, P.T.V. Rosa, D. Paolucci-Jeanjean, M.A.A. Meireles, Teppaitoon, J. Supercrit. Fluids 51 (2009) 36.
J. Supercrit. Fluids 47 (2008) 200. [140] S. Machmudah, A. Sulaswatty, M. Sasaki, M. Goto, T. Hirose, J. Supercrit. Fluids
[118] K.C. Zancan, M.O.M. Marques, A.J. Petenate, M.A.M. Meireles, J. Supercrit. Flu- 39 (2006) 30.
ids 24 (2002) 57. [141] J.C. Germain, J.M. del Valle, J.C. de la Fuente, Ind. Eng. Chem. Res. 44 (2005)
[119] J.M. del Valle, O. Rivera, M. Mattea, L. Ruetsch, J. Daghero, A. Flores, J. Supercrit. 2879.
Fluids 31 (2004) 159. [142] S. Balachandran, S.E. Kentish, R. Mawson, Sep. Purif. Technol. 48 (2006)
[120] O. Döker, U. Salgın, İ. Şanal, Ü. Mehmetoğlu, A. Çalımlı, J. Supercrit. Fluids 28 94.
(2004) 11. [143] T. Kumhom, A. Elkamel, P.L. Douglas, S. Douglas, S. Pongamphai, W. Tep-
[121] C. Marrone, M. Poletto, E. Reverchon, A. Stassi, Chem. Eng. Sci. 53 (1998) 3711. paitoon, Chem. Eng. J. 172 (2011) 1023.
[122] J.M. del Valle, P. Napolitano, N. Fuentes, Ind. Eng. Chem. Res. 39 (2000) 4720. [144] A. Tezel, A. Hortacsu, O. Hortacsu, J. Supercrit. Fluids 19 (2000) 3.
[123] E.M.C. Reis-Vasco, J.A.P. Coelho, A.M.F. Palavra, C. Marrone, E. Reverchon, [145] L. Fiori, D. Basso, P. Costa, J. Supercrit. Fluids 48 (2009) 131.
Chem. Eng. Sci. 55 (2000) 2917.