Sie sind auf Seite 1von 14

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Contents lists available at ScienceDirect

Chemical Engineering Research and Design


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

Review of kinetic models for supercritical uid extraction


Eduardo L.G. Oliveira, Armando J.D. Silvestre, Carlos M. Silva
Department of Chemistry, CICECO, University of Aveiro, Campus Universitrio de Santiago, 3810-193 Aveiro, Portugal

a b s t r a c t
The supercritical uid extraction (SFE) of liquids and solid materials is gaining increasing interest and commercial
application in last decades, most particularly under the recent concept of green chemistry and biorenery. Hence,
it is fundamental to provide adequate modeling of the SFE, in order to optimize operating conditions and simulate
the global process. This work intends to review the most signicant and physically sound models published in the
literature for countercurrent liquid-supercritical uid extraction and SFE of solid matrices, such as the linear driving
force, shrinking core, broken and intact cells, and the combination of BIC and shrinking core models. The main
assumptions and mass transfer expressions are presented and discussed.
2010 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Keywords: Supercritical uid; Extraction; Mathematical modeling; Mass transfer; Convective transport; Diffusion

1.

Introduction

In recent years there has been an increasing interest to replace


chemically synthesized compounds by natural equivalents
that can be found mostly in plant material (Herrero et al.,
2010). Conventional separation techniques such as solvent
extraction and distillation usually have the drawback of leaving trace amounts of organic solvents or to cause thermal
degradation (Bertucco and Vetter, 2001). Supercritical uid
extraction using carbon dioxide is a promising alternative for
the extraction of high value added products since mild temperatures are used (<343 K), CO2 is inert in the conditions
employed, is a by-product of many processes and thus widely
available, and it is considered benign since it is already present
in the atmosphere.
There are two main applications of the supercritical
extraction technology: countercurrent extraction from liquid
streams and extraction from beds of solid particles.
Countercurrent supercritical extraction has been studied
and applied both to the purication of edible oils, and the
extraction of high value-added components from vegetable or
sh oils (Brunner, 2009). During the production of olive oil,
extracts can present a wide range of acidity (0.54.0%) due to
the presence of free fatty acids. Oils with acidities above 2.0%
cannot be used for human consumption, and high quality oil

(extra virgin) must have free fatty acids contents lower than
0.8% (European Council Regulation, 2001; Vzquez et al., 2009).
Vzquez et al. (2009) have shown that acidity could be reduced
to less than 0.7% by countercurrent extraction with supercritical CO2 at 313 K. The main drawback was that the yield was a
function of the acidity of the feed stream.
An example of the extraction of high value-added
compounds from low value vegetable oil residues by countercurrent supercritical extraction is the recovery of tocopherols
and sterols from methyl esteried soybean oil deodorizer distillate studied by Fang et al. (2007) in a 2.4 m long column
lled with Dixon packing. It was reported that the majority of
fatty acids methyl esters were removed in an initial extraction
step at 16 MPa by imposing a temperature gradient inside the
extractor (313348 K). After this initial step, the pressure was
increased to 20 MPa and 80% of the high value tocopherols
were recovered. Another example of countercurrent extraction is the recovery of squalene from olive oil residue. Ruivo
et al. (2004) reported a 96% squalene recovery in the ranate at
313 K and 11.5 MPa. In the case of the extraction of polyunsaturated fatty acids from sh oil ethyl esters, Riha and Brunner
(2000) obtained (at 333 K and 14.5 MPa) a stream rich in the
high value fatty acids (100%).
Supercritical uid extraction is also employed to solid
materials, either to remove unwanted compounds or to obtain

Abbreviations: SFE, supercritical uid extraction; CO2 , carbon dioxide; SC, supercritical; AOC, anti-oxidant capacity; BIC, broken plus
intact cells model; SEM, scanning electron microscopy.

Corresponding author. Tel.: +351 234 401 549; fax: +351 234 370 084.
E-mail address: carlos.manuel@ua.pt (C.M. Silva).
Received 26 August 2010; Received in revised form 25 October 2010; Accepted 29 October 2010
0263-8762/$ see front matter 2010 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.cherd.2010.10.025

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Nomenclature
A
a
ae
af
Ap
ap
a5
Ci
Ci

Ci 

broken


C 

i intact

Cbroken,i
Cintact,i
Cp
Cp,s
Cs,i
C s,i
Cs,i
D12
Dax,i
dc
De
dp
F12
Fr
g
hSC
hL
hwall
I1
j
Jf
Ji
Js
Ki,SC
kB
KCO2 ,L
Ki
kf

cross-sectional area of the extractor in m2


packing specic surface area in m1
packing effective surface area (area/volume) in
m1
specic surface area related to the broken cells
in m1
external area of the particle in m2
particle specic surface area (area/volume) in
m1
specic surface area related to the intact cells
in m1
concentration of i in the SC phase in mol/m3
uid phase concentration in equilibrium with
the solid phase in mol/m3
uid phase concentration in equilibrium with
the concentration in the broken cells in mol/m3
uid phase concentration in equilibrium with
the concentration in the intact cells in mol/m3
solute concentration in the broken cells in
mol/m3
solute concentration in the intact cells in
mol/m3
specic heat capacity at constant pressure in
J/kg
specic heat capacity at constant pressure of
the packing in J/kg
solute concentration in the pore network of the
particle in mol/m3
average solute concentration in the pores in
mol/m3
pore phase concentration in equilibrium with
the concentration in the core in mol/m3
solute diffusion coefcient in m2 /s
axial dispersion coefcient in m2 /s
cell diameter measured by SEM
effective diffusion coefcient in m2 /s
particle diameter in m
binary correction factor of hard sphere system
Froude number
gravitational acceleration in m/s
convective heat transfer coefcient to the
supercritical phase in W/(m2 K)
liquid hold-up
convective heat transfer coefcient to the
extractor wall in W/(m2 K)
constant parameter of Eq. (7)
layer number (BIC plus shrinking core model)
solute mass ux between solid and uid phase
in mol/(m3 s)
mass ux in kg/(m2 s)
solute mass ux from intact cells in mol/(m3 s)
overall solute mass transfer coefcient in m/s
Boltzmann constant (1.380658 106 g cm2 /
(s2 K))
overall CO2 mass transfer coefcient in m/s
solute partition coefcient in m3 /kg
convective lm mass transfer coefcient in m/s

kf

kf,m
KLDF

kLDF
ks
ks

Lc
m

m
Mw
Na
P
Pc,i
Pe
q i
qi
qi,0
r
rc
Rcol
Re
Rp
Sc
Sh
t
T
Tc,i
u
Vbp,solute
Vc,i
Vi
Vp
We
xi

xCO

yi
yi
z

1105

pseudo mass transfer coefcient combining the


convective mass transfer resistance in the lm
side and the combined mass transfer resistances of the depleted layers in m/s
pseudo mass transfer coefcient of the internal
layers in m/s (BIC plus shrinking core model)
global linear driving force coefcient combining internal and lm mass transfer resistances
in s1
linear driving force mass transfer coefcient
coefcient in s1
pseudo mass transfer coefcient of the intact
cells side in m/s
pseudo mass transfer coefcient combining the
convective mass transfer resistance in the lm
side and the mass transfer resistance of the
wall of intact cells in m/s
extractor length in m
total number of layers (BIC plus shrinking core
model)
mass owrate in kg/s
molecular weight in g/mol
Avogadro constant, 6.0221367 1023 mol1
pressure in MPa
critical pressure of i in MPa
Peclet number, Pe = uLc /Dax,i
average solute concentration in the solid phase
in mol/kg
solid phase concentration in equilibrium with
SC phase in mol/kg
initial solute concentration in the particle in
mol/kg
radial coordinate of the particle
radius of the boundary between the core and
the external shell in m
radius of the extractor column in m
Reynolds number
particle radius in m
Schmidt number
Sherwood number
extraction time in s
temperature in K
critical temperature of i in K
interstitial uid velocity in m/s
solute molar volume at its normal boiling point
in cm3 /mol
critical molar volume of i in cm3 /mol
molar volume of i in cm3 /mol
particle volume in m3
Webber number
solute mass fraction in the liquid phase
carbon dioxide mass fraction in the liquid in
equilibrium with the SC phase
solute mass fraction in the supercritical
phase
solute mass fraction in the SC phase in equilibrium with the liquid
extractor axial coordinate

1106

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Greek letters
P
extractor pressure drop in bar (0.1 MPa)
Pdry
extractor pressure drop for dry column in bar
(0.1 MPa)
i /kB
LennardJones energy parameter in K
column porosity
c
p
particle porosity
thermal conductivity of the SC phase in
SC
W/(m K)
L
combined thermal conductivity of the liquid
and the packing in W/(m K)

viscosity in Pa s
c,i
viscosity of i at the critical point in Pa s

packing ow angle in

density in kg/m3
c,i
density of i at the critical point in kg/m3

surface tension of the liquid in kg m/s2
c
critical surface tension of the packing in
kg m/s2
i
molecular diameter in cm
binary hard sphere molecular diameter in cm
 12
packing packing tortuosity
friction factor

volume fraction of broken cells in the particle



dimensionless association factor of the solvent,
( CO2 = 1)
Subscripts
L
liquid phase
S
solid phase
SC
supercritical phase
wall
extractor wall

high value added compounds. The decaffeination of coffee


beans or tea leaves (Bertucco and Vetter, 2001) is already
carried out industrially by SFE-CO2 . Several patents propose
ranges of temperature and pressure for caffeine extraction from coffee beans using SC-CO2 : 12.218.2 MPa/313353 K
(Zosel, 1971); 7.540 MPa/304368 K (Margolis and Chiovini,
1980); PT < 40 MPa/343413 K (Katz, 1989); 7.530 MPa/293353 K
(Coenen and Ben-Nasr, 1993). The CO2 is regenerated by
washing with water or by adsorption in activated carbon.
The removal of cholesterol from egg yolk powder (Wu and
Hou, 2001) is another example of SFE processing to remove
unwanted components. Wu and Hou (2001) studied this
extraction at 318 K, 32 MPa and different feed owrates and
particle sizes. Results showed that the rate controlling step
of the process was the external mass transfer. The authors
assume that there are two types of oil: weakly adsorbed and
strongly adsorbed. When the weakly adsorbed oil is being
extracted, the supercritical phase equilibrium composition is
the oil solubility. However, when the weakly adsorbed oil is
exhausted, the equilibrium composition in the uid phase is
a linear function of the solid phase concentration, by y* = Kx.
The equilibrium constant, K, increases with increasing pressure but decreases with increasing temperature in the range
of the operating conditions studied.
Different types of matrixes have been studied such as
seeds, leaf, bark or roots to obtain oil, sterols, terpenes, free
fatty acids, vitamins, colorants such as -carotene, etc. For
example, Sovov et al. (1994b) studied the effect of milling,

CO2 owrate and direction of ow (upow or downow) on


the extraction of grape seed oil at 28 MPa and 313 K. Results
conrm that lower particle sizes have higher mass transfer
kinetics, and show that upow extraction is less efcient due
to the effect of natural convection. Passos et al. (2009b) studied the effect of an enzymatic pre-treatment on the extraction
of grape seed oil, and obtained yield increments up till 163%
for the most favorable conditions. The pre-treatment of triturated seeds has been performed with a cell wall degrading
enzyme cocktail containing cellulase, protease, xylanase, and
pectinase, in order to enlarge the broken/intact cells ratio
to increase oil availability. For the pre-treatment conditions
adopted, the SFE yield increased 43% at 313 K, independently
of the CO2 pressure in the range 1620 MPa (Passos et al.,
2009a). In another work, Passos et al. (2010) studied the inuence of pressure and temperature on the anti-oxidant capacity
of grape seed oil extracted with SC-CO2 . They have shown
that the AOC (anti-oxidant capacity) of extracts increased
with both pressure and temperature though the temperature effect was more important. Experimental measurements
of the AOC along extraction curves at 20 MPa/313.15 K and
20 MPa/323.15 K showed that the following strategy may be
adopted with advantage: employ higher temperatures at the
start of the extraction (e.g., 323.15 K), to obtain higher antioxidants concentration in the nal extract, and then decrease
temperature (e.g., to 313.15 K) to reduce CO2 consumption by
increasing oil solubility.
del Valle et al. (2004) studied the extraction of rose hip
seeds at laboratory and pilot scales. Bench-scale experiments
(30 MPa/313 K and 40 MPa/323 K) were employed to t the solid
mass transfer coefcient. This was then used to simulate
the operation of a 2.6 dm3 pilot-scale extractor. However, the
experimental extraction kinetics was slower than expected,
most likely due to heterogeneity in the uid ow that was not
accounted for in the model. Other examples of applications of
supercritical extraction of high value-added compounds can
be found in various review works (Herrero et al., 2010; Pronyk
and Mazza, 2009; Sahena et al., 2009).
From an engineering point of view, the scale-up of SFE units
requires both the knowledge of the relevant process parameters such as equilibrium and mass transfer kinetics, and also
the optimum operating conditions. These parameters may be
obtained by using an accurate mathematical description of the
extraction process and experimental laboratory data.
The mathematical modeling of supercritical uid extraction of high value-added products from plant matrixes such
as herbs, seeds, leafs or bark is difcult due to the existence of
different structures in the material. Prior to the extraction, the
material is usually grinded to increase the surface area in contact with the SC solvent and also increase the accessibility of
the solute inside cell structures, thus increasing mass transfer
kinetics. This enables the description of the resulting particles using the basic geometries of slab, cylinder or sphere. To
reduce the complexity of the model it is also necessary to make
simplifying assumptions about how the solute is distributed
inside the solid adsorbed in pore network, tied inside cell
structures or homogeneously distributed inside particles and
about the mechanisms involved in the mass transfer kinetics internal mass transfer resistance, external mass transfer
resistance or a combination of various resistances.
In 2006, del Valle and De La Fuente (2006) reported several
kinetic supercritical extraction models. That work was limited to supercritical uid extraction from plant matrixes and
did not include the most recent revision of the successful bro-

1107

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

the solute diffusion coefcients (D12 ). Examples of the best


and/or well known correlations for D12 are also reported in
Table 1. It should be noted that the correlations of Wilke and
Chang (1955), Catchpole and King (1994) and He et al. (1998)
do not require any adjustable parameters but have lower
accuracy when compared to the correlations of Liu et al. (Liu
et al., 1997; Magalhaes et al., in press-b) and Magalhaes et al.
(in press-a, in press-b) that require one adjustable parameter
(k12,d , k12 or ED ) but have average absolute relative deviations
around 4%; nonetheless an extensive list of these parameters
may be found in the original publications. In Eq. (1), A is the
cross-sectional area of the extractor, c is the column porosity,
hL is the liquid hold-up, SC is the density of the SC phase,
yi is the solute fraction in the supercritical phase, t is the
SC is the SC uid mass ow rate, z is the
extraction time, m
extractor axial coordinate, Ji is the mass ux, ae is the packing
effective surface area, and Dax,i,SC is the axial dispersion
coefcient in the SC phase.
Similarly, the material balance to the liquid phase is:

Fig. 1 Countercurrent supercritical extraction diagram.


ken and intact cell model by Sovov (2005) and the new model
by Fiori et al. (2009). The objective of this work is to review the
most important mathematical models applied to the supercritical uid extraction of both liquids and solids, and their
main assumptions, with particular emphasis on the kinetics.
The procedure employed to calculate the main kinetic parameters in supercritical conditions is also briey described.

2.
Mathematical modeling of
countercurrent supercritical uidliquid
extraction
The objective in countercurrent extraction is to promote the
highest possible contact between the two phases (liquid and
supercritical uids) which may be achieved by employing
structured packings (Bertucco and Vetter, 2001; Brunner, 2009).
A schematic diagram of a countercurrent extractor is shown
in Fig. 1, where mass ow rates of the inlet and outlet liquid
(L) and supercritical uid streams are reported.
In the development of the material and heat balances
inside the extractor it is assumed that the liquid is dispersed
as a thin layer at the surface of the packing (Fernandes et al.,
2007; Ruivo et al., 2004) and thus it is considered to be in thermal equilibrium with the packing. At the interface of the two
uids, thermodynamic equilibrium is assumed. Also, in the
energy balance, heat is supplied to the extractor through the
wall. In some cases, a temperature gradient is imposed from
the bottom to the top of the extractor to modify the solubility
of the different solutes in the SC uid (Fang et al., 2007; Kondo
et al., 2002; Sato et al., 1998).
The general mass balance of component i in the supercritical phase is:
Ac

SC yi )
((1 hL ) SC yi )
(m
+
= Ji Aae
t
z

(1)

When axial dispersion is not negligible, the term


Dax,i,SC SC ( 2 yi / z2 ) should appear in the right hand side of
Eq. (1). Ruivo et al. (2004) lumped the axial dispersion term
in the mass transfer one (Ji ) due to the difculty of estimating accurate Dax,i,SC values. A few examples of correlations
employed to calculate the axial dispersion coefcient are
reported in Table 1 . In the determination of the axial dispersion coefcient it is also necessary to accurately estimate

Ac

L xi )
(m
(hL L xi )

= Ji Aae
t
z

(2)

Once more, the axial dispersion term in the liquid phase,


Dax,L L ( 2 xi / z2 ), can be explicitly included in the right hand
side of the mass balance. L is the liquid phase density, xi is
L is the liquid mass
the solute fraction in the liquid phase, m
ow rate, Ji is the mass ux, and Dax,L is the axial dispersion
coefcient of the liquid.
The solute mass ux is a function of the packing effective surface area (ae ), the velocity of both phases, and the
thermodynamic equilibrium of the system, which may be calculated adopting a approach using an accurate equation
of state such as the SoaveRedlichKwong (Ruivo et al., 2004).
The mass ux of solutes from the liquid to the SC phase, Ji , is
expressed as:


Ji = Ki,SC SC


SC uSC 
yi yi
L SC uL

(3)

and the CO2 mass ux to the liquid phase is:


JCO2 = KCO2 ,L L


SC uSC 
xCO2 xCO2
L SC uL

(4)

where Ki,SC and KCO2 ,L are the overall mass transfer coef
is the CO2 mass fraction in the liquid in
cients, xCO
2
equilibrium with the SC phase, yi is the solute mass fraction
in the SC phase in equilibrium with the liquid, and uSC and
uL are the supercritical and liquid phase interstitial velocities
corrected for the presence of the packing (Ruivo et al., 2002).
Regarding the energy balance, Fernandes et al. (2007) proposed a simplied differential balance that considered only
axial temperature variation. The liquid and the packing are
lumped in a pseudo-homogeneous phase since it is assumed
the temperature is the same. The energy balances to both
phases are shown below.

Ac

(1 hL ) Cp,SC SC TSC

Ac
packing z

(1 hL ) SC

SC TSC
Cp,SC m


TSC

= Aae hSC (TL TSC ) + 2 Rcol hwall (Twall TSC )

(5)

1108

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Table 1 Correlations for axial dispersion coefcient (Dax,i ) and binary diffusion coefcient (D12 ).
Ref.
Axial dispersion correlations
c Dax,i

0.3c dp
Re

D12
c Dax,i

0.5c dp Re Sc2
Re Sc+3.8

Re = 0.008 400, Sc = 0.28 2.2

Wen and Fan (1975) and Yang (1999)

= 20 + 0.5 Re Sc

D12
c Dax,i

D12
c Dax,i

D12
c Dax,i

Wakao and Funazkri (1978) and Yang (1999)

0.085c 0.811 0.590


Re0.914 Sc,
d0.526
c0.725 0.676
p
c
2
Re+0.199
c 157.7 Re +0.284
, dp
Re+0.7

= 1.317(c Re Sc)

D12
c Dax,i

= 0.58c +

D12

1.392

dp (mm) = 0.52, Re = 0.1830

Tan and Liou (1989)

< 1 mm, Re = 280, Sc = 15.77

Catchpole et al. (1996a)

c Re Sc > 0.3, Sc = 3.9665

(Re Sc)2
1.65c 1+Re Sc ,

Funazukuri et al. (1998)

Re Sc = 0.0230

Yu et al. (1999)

Correlations for binary diffusion coefcient


WilkeChang Equation
D12 = 7.4 109

Liong et al. (1991), Reid et al. (2000) and


Wilke and Chang (1955)

SC Mw,SC

0.6
SC Vbp,solute

CO2 = 1
Catchpole and King correlation

D12 = 5.152Dc

Tc,SC

Dc = 4.300 107
v,CO2 = 26.9,

X=

R=

2/3

v,SC c,SC

0.4510

1/3
2 

/ 1+

SC
< 2.5
1<
c,SC

Mw,SC
Mw,solute

1/2

1.0 0.1, X < 2


0.664X0.17 0.1, 2 < X < 10

 

k
B T/
D12 = A 1011 VSC

SC /c,SC 1.2


1/6

1+

12,eff 12,LJ 2

0.75832

Mw,SC Vc,SC /10Pc,SC

1/6

 1000R T 
g

12,LJ = (1 k12,d )

1+

Mw,12


exp

1.3229Ti

1/6

1.3229T12

0.75SC

1.2588 SC

Tc,i
0.049029
10Pc,i

Tc,i
10Pc,i


SC
= SC
SC,eff

T
i,LJ /kB
T
=
12,LJ /kB

Ti =

i,LJ /kB = 0.774Tc,i


12,LJ /kB =
M12 =

Liu et al. (1997) and Magalhaes et al. (2010a)

3
3
SC,LJ
(SC,LJ /kB ) solute,LJ
(solute,LJ /kB )

MSC Msolute
MSC + Msolute

Rg = 8.3144 J/ (mol K)

(SC,LJ + solute,LJ )

2


108

Tc,i
< 1000
Pc,i
Tc,i
1000
Pc,i


SC
= SC Na / 10 Mw,SC

0.27862

T12

1/6

0.809 108 Vc1/3


 3


SC /c,SC 0.21

SC,LJ + solute,LJ

2


3 0.17791 + 11779

T12

SC /c,SC 0.21

TLSMd model
21.16 104
D12 =

2
SC
12,eff
i,eff i,LJ 2

SC /c,SC < 1.2

Mw,SC ,

A = 0.29263 + 1.6736 exp


B = 0.077Tc,SC ,

He et al. (1998)
Mw,solute

k = 1 + (SC /c,SC 1.2) /

i,LJ =

Catchpole and King (1994, 1997)


R
,
X

Dc,CO2 = 4.9368 108 m2 /s

(Reid et al., 2000)

HeYuSu correlation
k = 1,

2/3

SC
C,SC

0.5
0.75
Mw,SC
Tc,SC

Vc,solute
Vc,SC

1+

1109

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Table 1 (Continued )
Ref.
LJ1 model
D12 =

Magalhaes et al. (in press-a)


kB T


2
(8/3)SC
12,eff
(2 MW,12 kB T)

i,eff 1.1532i,LJ 1 + (1.8975Ti )

104

(g(12,eff )/F12 ) + (0.4/Ti1.5 )

1/2 1/6

12,eff 1.153212,LJ 1 + (1.8975Ti )


12,LJ = (1 k12 )

1/2

1/2 1/6

SC,LJ + solute,LJ
2

i,LJ = 0.7889 108 Vc,i

1/3

SC = SC Na /(103 Mw,SC )


Ti kB T/i,LJ

T12
kB T/12,LJ

i,LJ /kB = Tc,i /1.2593

12,LJ /kB =

(SC,LJ /kB ) (solute,LJ /kB )

Mw,SC Mw,solute
Na (Mw,SC + Mw,solute )

Mw,12 =

g(12,eff ) =

SC

1
(1 SC )

2SC
1 + SC,eff /solute,eff

1 SC +


1 SC +

SC
1 + SC,eff /solute,eff

3
SC Na SC,eff
=
6 103 Mw,SC

F12 =

1.7
a ln(solute,eff /SC,eff ) + b ln (solute,eff /SC,eff ) + c ln(Mw,solute /Mw,SC )
F11 + SC
2

3.0
1 + SC
d ln(solute,eff /SC,eff )

a = 1.676382SC
+ 1.638561

b = 8.516830SC
+ 8.631536

c = 1.320347SC
+ 1.351067

d = 5.062546SC
+ 5.409662

Ruckenstein and Liu (1997)

1.5
3
5
7
F11 = 1 + 0.94605SC
+ 1.4022SC
5.6898SC
+ 2.6626SC
,

SC

3
SC,eff
SC Na /(103 Mw,SC )

Na = 6.0221367 1023 mol1


LJ1 Model
D12 =

3.0

2
8SC
12,eff

kB T
2 Mw,12

1/2

i,eff = i,LJ 21/6 1 + 1.3229Ti

i,LJ =

0.809 108 V 1/3


c,i

ED
Rg T

Magalhaes et al. (in press-b)


104

1/2 1/6

SC,LJ + solute,LJ
2



3 0.17791 + 11.779

1/2 1/6

12,eff = 12,LJ 21/6 1 + 1.3229T12

12,LJ =

 F12  exp

g 12,eff

Tc,i
0.049029
10Pc,i

Tc,i
10Pc,i

2


108 ,

Tc,i
1000
Pc,i


SC
= SC Na /(103 Mw,SC )

Ti kB T/i,LJ

T12
kB T/12,LJ

i,LJ /kB = 0.774Tc,i


12,LJ /kB =
Mw,12 =

(SC,LJ /kB ) (solute,LJ /kB )


Mw,SC Mw,solute

6.0221367 1023 (Mw,SC + Mw,solute )

Rg = 8.31541 107 erg/ (mol K)

g 12,eff

, F12 and Mw,12 are the same as in LJ1 model

Tc,i
< 1000
Pc,i

1110

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

((1 hL ) c Cp,L L + (1 c ) Cp,s s ) TL

Ac
z

 T 
L
L

L TL )
(Cp,L m
z

z = L c , t

= Aae hSC (TSC TL ) + 2 Rcol hwall (Twall TL )

TG

z = 0
xi = xi,feed

(12)

TL = TL,feed

(6)
where Cp,SC , Cp,L , and Cp,s are the specic heat capacity at constant pressure of the supercritical, liquid and solid phases, s
is the solid phase density, TSC , TL , and Twall are the temperatures of the SC uid, liquid plus packing and extractor wall,
packing is the packing tortuosity, Rcol is the column radius, SC
is the thermal conductivity of the SC phase, L is the combined
thermal conductivity of liquid and packing, hSC is the convective heat transfer coefcient to the supercritical phase, and
hwall is the convective heat transfer coefcient to the extractor
wall.
The pressure drop (P) and liquid holdup (hL ) inside the
extractor can be estimated using the models proposed by
Stichlmair et al. (1989), Stocketh and Brunner (2000) and
Ruivo et al. (2004) shown next.
P
=
Pdry
I1
=
=



hL
c

1 c 1



1
1 c

(2+)/3  h 4.65
L
hL c
(7)

3.
Mathematical modeling of supercritical
uid extraction of plant material

dp
Pdry 4.65
4
c
3 Lc (1 c ) SC (uSC c )2


hL = 0.7

3L a2 c uL
L g sin 

1/3 

 
ae 2/3
a


1 + 0.474 Re0.05
SC

L
L SC

(8)
Pdry is the pressure drop for dry column, is a friction factor, I1 is a constant parameter, dp is the particle diameter, Lc is
the column length, L is the liquid viscosity, a is the specic
surface area of the packing, g is the gravitational acceleration,
 is the packing ow angle, and ReSC = uSC Lcorrugation SC /SC is a
Reynolds number of the supercritical phase. The effective surface area (ae ) can be calculated using the correlation of Onda
et al. (1968) for random packings, and were applied by Ruivo
et al. (2002, 2004) for Sulzer EX structured gauze packing.
ae
= 1 exp
a

  0.75

1.45

c


0.05
Re0.1
We0.2
L FrL
L

G,0

z = 0, t

Ci
C
2 C
= uG i + c Dax,i 2i + (1 c ) Jf ap
t
z
z

(10)

where Ci is the concentration of component i in the supercritical phase, Dax,i is the axial dispersion coefcient of component
i, Jf is the solute mass ux from solid to uid phase, and ap
is the particle specic surface area, which for spheres equals
3/Rp . The following initial and boundary conditions (Dankwertz (Rice and Do, 1995)) are required to solve Eq. (10):
t = 0, z

Ci = Ci,0

z = 0, t

c Dax,i

= TL,0

TL

z = 0
yCO = 1

TG = TG,feed

(13)

(9)

yi = yi,0
xi = xi,0

There are various models employed to describe the supercritical uid extraction of oils or other compounds from plant
material. All of them consider that the particles are packed
inside an extractor column. The simplifying assumptions
employed by most authors are: isothermal operation, negligible pressure drop across the extractor, and constant bed
porosity and solid density along extraction. Furthermore, it
is usually assumed that the solute loading in the supercritical
uid is low and, therefore, uid density, axial dispersion, and
uid velocity remain approximately constant. Such assumptions reduce the number of equations necessary to describe
the extraction process to mass balances, equilibrium relations,
and kinetics laws. The general mass balance to the uid phase
of a SFE column is:
c

where  is the surface tension of the liquid,  c is the critical surface tension of the packing, ReL = uL Lcorrugation L /L , FrL =
u2L /Lcorrugation g, and WeL = u2L L Lcorrugation / are the Reynolds,
Froude, and Weber numbers for the liquid.
Finally, the initial and boundary conditions necessary to
solve the model are shown below.

t = 0, z

Ruivo et al. (2004) employed the model to the extraction of


squalene from olive oil residue (modeled as a mixture of squalene and methyl oleate). In their work it was considered that
the extractor was operating isothermally. The experiments
were performed at temperatures between 313 and 333 K and
pressures in the range 11.518.5 MPa. Results show that the
model was able to provide good description of the squalene
ranate compositions. There was, however, a higher deviation
between experimental and simulated extract compositions
that the authors claim was due to a poor correlation of the
squalene composition in the extract. Fernandes et al. (2007)
employed the complete model to the extraction of the same
mixture (squalene and methyl oleate) as Ruivo et al. (2004). The
assumption of non-isothermal operation resulted in a more
accurate prediction of the liquid and supercritical phase compositions than when isothermal operation was considered.
Optimization of the operating conditions showed that squalene purity and recovery higher than 90% can be achieved at
SC /m
L = 75 and Lc = 3m.
13.5 MPa, 330 K, m

z = Lc , t
(11)



Ci
= u Ci,feed Ci
z

Ci
=0
z

(14)

(15)

(16)

Furthermore, various authors consider the axial dispersion


negligible (Wu and Hou, 2001). This assumption is usually valid
for long columns or when large interstitial velocities are

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

employed:
Pe =

uSC Lc
100
Dax,i

(17)

where Pe is the Peclet number. From this equation, an equivalent relation can be obtained.
Lc
50
dp

(18)

The advantage of this inequality is that it does not depend


upon compositions, pressure or temperature, and thus it provides a very simple and fast criterion to evaluate the existence
of axial dispersion. If Pe < 100, the axial dispersion coefcient
for supercritical phases can be estimated by the correlations
of Funazukuri et al. (1998):
c Dax,i
= 1.317(c Re Sc )1.392 ,
D12

c Re Sc > 0.3

(19)

or Catchpole et al. (1996a) for small particles (dp < 1 mm)


Dax,i
0.018
10Re
,
=
+
uG Lc
Re
Re + 0.7

Re = 280

(20)

where D12 is the solute diffusion coefcient, Re = uSC SC dp /SC


is the Reynolds number and Sc = SC /SC D12 is the Schmidt
number.

3.1.

Mass balance to the solid phase

The mass balance to the solid phase inside extractor is conditioned by the simplifying assumptions taken by each author
regarding the solid matrix composition and the extraction
kinetics. The following four different models are described:
linear driving force, shrinking core, broken plus intact cells,
and a combination of the broken plus intact cells with

1111

shrinking core. The mass balances described in the following


sections assume spherical particle. A diagram of the different
models is shown in Fig. 2.

3.1.1.

Linear driving force model (LDF)

The linear driving force model was originally developed


by Glueckauf (Glueckauf and Coates, 1947; Glueckauf, 1955;
Ruthven, 1984) in his Theory of Chromatography publications. This model is a simplication of the general mass
balance to a solid particle. It is assumed that the mass transfer ux is proportional to the difference between an average
solute concentration in the particle, q i , and the solute concentration in equilibrium with the uid phase, qi = qi (Ci ),
as shown in Fig. 2a. It can be demonstrated that when the
solute concentration in the particle follows a parabolic prole,
the linear driving force model is exact (Liaw et al., 1979). Do
and Rice (1986) have shown that these assumptions are valid
except at the beginning of the extraction of each particle in
the column, when the concentration prole is still very sharp.
The minimum extraction time to validate the LDF assumption
(tmin ) is (Yang, 1999):
tmin = 0.4

d2p
De

(21)

where De is the effective diffusion coefcient. In the simplest


case, when the external resistance in the lm is negligible, the
mass balance becomes:



q i
= kLDF q i qi
t

(22)

which embodies the LDF mass transfer coefcient calculated


by:
kLDF = 15

De
Rp2

Fig. 2 Different models for supercritical extraction of solutes from solid matrixes.

(23)

1112

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Rp is the particle radius. The equilibrium is approximately linear for low concentrations, being described by the partition
coefcient Ki :

(Ghoreishi et al., 2009) and with subcritical water (Ghoreishi


and Shahrestani, 2009).

3.1.2.
qi = Ki Ci

(24)

Taking into account the previous relations, the solute mass


ux, Jf , corresponds to:

Jf ap = s kLDF q i qi

(25)

In many cases, the solid particles are porous and thus the
solute is present both in the solid phase and in the uid inside
the pores (Peker et al., 1992; Recasens et al., 1989). Therefore,
two linear driving force approximations are employed (see
Fig. 2b): one between the uid phase and the uid in the pores
of the solid, and the other between the uid phase in the pores
and the solid phase. The mass balances to the particle and to
the solid phase of the particle are:



C s,i
q
+ (1 p ) s i = KLDF C s,i Ci
p
t
t

(26)



q i
= kLDF q i qi
t

(27)

1
KLDF

Rp
1
5
=
+
KLDF =
kf ap
kf ap
5De ap
5 + (kf Rp /De )

(28)

where KLDF is the global linear driving force coefcient combining internal and lm mass transfer resistances and p is
the porosity of the particle, C s,i is the average solute concentration inside pores, and kf is the convective lm mass transfer
coefcient. In this case, the equilibrium between solid and
pore phase concentration can be expressed as:
qi

= Ki C s,i

(29)

The solute mass transfer from the particle to the uid


phase is

Jf ap = KLDF C s,i Ci

(30)

At the start of the extraction, the solute concentrations


inside the particle are qi,0 and C s,i,0 :


t=0

C s,i = C s,i,0
q i = qi,0

(31)

Recasens et al. (1989) employed the linear driving force


approximation to model supercritical CO2 extraction of ethyl
acetate from activated carbon at temperatures in the range
300338 K and 8.8313.1 MPa (experimental data from Tan and
Liou (1988)). Since the experimental results indicated the presence of lm resistance and the solid was considered porous,
using two linear driving force approximations yielded a good
description of the experimental results.
Peker et al. (1992) adopted the linear driving force model
to describe experimental supercritical extraction of caffeine
from coffee beans. The model was able to accurately describe
the extraction of caffeine beans soaked with water using CO2
or CO2 with 1% steam. The extraction rate was higher when
the steam modied CO2 was employed. More recently, the LDF
model was also employed in the mathematical modeling of
extraction of mannitol from olive leaves with supercritical CO2

The shrinking core model

The shrinking core model assumes that there is a sharp


boundary between the extracted and non-extracted parts of
the particle. As extraction proceeds, the boundary recedes
until it reaches the center of the particle and all solute is
exhausted. The position of that boundary is given by rc in the
spherical particle shown in Fig. 2c.
In the development of the mass balances to the core and
extracted phases, it is assumed that there is no solute accumulation in the extracted region and the solute diffuses from
the boundary to the surface through the pore network. Therefore, all solute leaving the particle by convection at surface
comes from the core. The mass balances to the core and to
the extracted phases of a spherical particle are given by (Goto
et al., 1996):
q i
= kf ap
t
De
r2 r

 C 
s,i
2
r

Cs,i 

r=Rp

Ci

=0

(32)

(33)

where Cs,i is the solute concentration in the pore network and


r is the radial coordinate of the particle. The average solute
concentration is a function of rc :


q i = qi,0

rc
Rp

3
(34)

The solute mass ux from the particle to the supercritical


uid is given by
Jf = kf

Cs,i 

r=Rp

Ci


(35)

The initial and boundary conditions in the solid phase are:


r , t = 0 qi = qi,0
t , r = rc
t , r = Rp

Cs,i = Cs,i
De

Cs,i
= kf (Cs,i Ci )
r

(36)
(37)
(38)

The pore phase concentration at the boundary in equilibrium with the core phase concentration, i.e. Cs,i , is usually
assumed to be the solute solubility in the supercritical uid.
Goto et al. (1996) applied the shrinking core model to
rape seed oil extraction data reported by Brunner (1984) at
20.5 MPa/324 K and 35 MPa/316 K. The model offered a good
description of the extract mass owrate and extraction rate
at both conditions by taking the effective diffusivity as tting
parameter. The effective diffusivity dependence with temperature and pressure was similar to the predictions of the Wilke
and Lees correlation (Reid et al., 2000; Wilke and Chang, 1955).
Stber et al. (1997) employed the shrinking core model to
obtain extraction proles of benzene, toluene, ethylbenzene
and 1,2-dichlorobenzene from cylindrical particles with open
or sealed ends. Since cylindrical particles were employed,
ap = 2/Rp . The shrinking core model was also employed by
Abaroudi et al. (1999) to study the extraction of -naphthol
from cylindrical particles. Since the ends of the cylinders were
not sealed, these were included as new transfer sources.

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

3.1.3.

Broken plus intact cells model (BIC)

The cell walls of plant materials act as a barrier to solute


extraction. Hence, they are previously grinded to reduce particle size and thus promote higher extraction rates and yields.
Sovov (1994) argue that this pre-treatment breaks part of the
cell walls, resulting in particles with intact and broken cells.
The assumption is that the extraction kinetics from broken
cells is faster than that from intact cells, since cell walls introduce an additional mass transfer resistance. Accordingly, from
the broken cell the mass transfer mechanism is convection,
whereas through the intact inner core is molecular diffusion.
In the mathematical model, Sovov (1994) proposed a mass
transfer term considering a lower resistance until the broken cells are depleted. Then, the mass transfer resistance is
increased to account for the extraction from the intact cells.
The mass balance to the solid phase is shown below.
Cs,i
= Jf ap
t

(39)



kf Ci Ci Cs,i solubility
Jf =


ks Ci Ci Cs,i < solubility
Ki

where ks is a pseudo mass transfer coefcient of the intact


cells side, Ci is the uid phase concentration in equilibrium
with the solid phase taken as the solute solubility in the
work of Sovov (1994).
This initial model was rstly revised by Sovov et al. (1994a)
and then by Sovov (2005) to include additional assumptions.
It is claimed that the solute from broken cells diffuses directly
to the uid phase while the solute from intact cells diffuses
only to the broken cells, which congures the series resistance
model schematically shown in Fig. 2d. Based on these hypothesis, it is necessary to develop two material balances, one to
the broken cells and another to the intact cells:

Cbroken,i
= Jf af + Js as
t

Cintact,i
= Js as
(1
)
t

(41)

Js = ks (Cintact,i Cbroken,i )

Cbroken,i
= Jbroken ap
t

(1
)

(45)

Cintact,i
= Jintact ap
t

Jbroken = kf

Ci 

broken

Jintact = ks Ci 

intact

Ci

Ci

(42)

(43)
(44)

Since it is difcult to experimentally measure the values of the


specic surface areas, both kf af and ks as are usually tted as
lumped parameters.
Reverchon and Marrone (2001) proposed a parallel resistances mechanism where both broken and intact cells transfer
solute to the uid with different kinetics. A representation of
this model is shown in Fig. 2e.
The difference between the series and the parallel models
is that the solute from intact cells is extracted directly to the

(46)
(47)
(48)

and total ux to the uid phase reduces to


Jf = Jintact + Jbroken

(49)

where ks is the pseudo mass transfer coefcient combining


the convective mass transfer resistance in the lm side and
 the
mass transfer resistance of the wall of intact cells, and Ci 
broken

and Ci 
are the uid phase concentrations in equilibrium
intact
with the solute concentration in the broken and the intact
cells, respectively. When comparing the values of the mass
transfer coefcients, ks is smaller than kf due to the higher
resistance to mass transfer imposed by the cell walls. In the
parallel model, the specic surface area (ap ) is the same for
the broken and intact cells (Ap /Vp ).
The uid phase concentration in equilibrium with the
solid phase (Ci ) is a function of the solute-matrix interaction.
Following Perrut et al. (1997), at low concentrations the equilibrium relation is frequently linear, i.e. Ci = Ki Csolid , while above
a critical solid loading a maximum concentration is attained
whose value coincides with solubility.
At the start of the extraction process, the solid and SC
phases are in equilibrium, which means the concentration
inside the intact and the broken cells are different due to the
initial solute distribution. Accordingly, the initial conditions
should be written as:

where
is the volume fraction of broken cells in the particle,
Cbroken,i and Cintact,i are the solute concentrations in broken and
intact cells, and af and as are the specic surface areas related
to the broken and the intact cells, respectively. In the uid
mass balance, the particle specic surface area (ap ) has to be
replaced by af when this model is employed. Jf and Js are the
solute mass uxes from broken cells to the SC solvent, and
between intact and broken cells, respectively:
Jf = kf Ci Ci

uid phase in the second one. The mass balance and mass
ux equations to the broken and to the intact cells are:

(40)

1113

t=0

Cbroken,i = Cbroken,i,0
Cintact,i = Cintact,i,0

(50)

The solution of the BIC series model requires the knowledge of several parameters: the external volumetric mass
transfer coefcient (kf af ), the internal volumetric mass transfer coefcient (ks as ), the fraction of broken cells (
) and,
in some cases, the partition coefcient (Ki ). Sovov (2005)
developed a simplied approach with approximate models to
obtain initial guesses of ks as and
. Furthermore, kf af may be
also estimated using well know correlations from the literature. A few examples are shown in Table 2. With these initial
estimates, the complete model is then used to optimize all
parameters. Sometimes, the external mass transfer coefcient
is not optimized, being xed to the initial value given by correlations (Silva et al., 2009).
Reverchon and Marrone (2001) determined the value of the
fraction of broken cells by scanning electron microscopy. Using
this technique, the number of parameters to be optimized is
reduced. In the SEM images it is clearly visible that the oil
is trapped inside the structures that we refer as cells and
that some of them are broken, validating the assumptions
of the BIC model. The BIC parallel model coupled with SEM
was employed to represent the experimental data of SFE of
oils from different seeds (Reverchon and Marrone, 2001): sun-

1114

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Table 2 Correlations for convective mass transfer coefcient (kf ).


Sh =
kf
u
kf
u

kf 2Rp
D12

= 2.0 + 0.6 Re1/2 Sc1/3

= 0.4548 Re0.4069 Sc2/3

Ranz and Marshall (1952) and Yang (1999)


Dwivedi and Upadhyay (1977)

= 0.765 Re0.82 + 0.365 Re0.386 Sc2/3

Dwivedi and Upadhyay (1977)

Sh = 2.0 + 1.1 Re0.6 Sc1/3 , Re = 310000


Sh = 0.38 Re0.83 Sc1/3 , Re = 240, Sc = 220
Sh = 0.82 Re0.6 Sc1/3 , Re = 170, Sc = 311

Wakao and Funazkri (1978) and Yang (1999)


del Valle and De La Fuente (2006) and Tan et al. (1988)
del Valle and De La Fuente (2006) and King and Catchpole (1993)

kf
uc

= 1.17 Re0.42 Sc2/3


Sh = 0.206 Re0.8 Sc1/3 , Re = 10100, Sc < 10

Cussler (1997)
Puiggen et al. (1997)

ower (Perrut et al., 1997), coriander (Catchpole et al., 1996b),


grape (Sovov et al., 1994b), tomato (Roy et al., 1994), peanut
(Goodrum et al., 1996), almond (Marrone et al., 1998) and fennel
(Reverchon et al., 1999). The internal mass transfer coefcient
was taken as tting parameter, assuming that the lm resistance was small, i.e. very big values of kf were employed. The
values of the internal mass transfer coefcient were in the
range (9.224) 108 m/s. When the fraction of broken cells
was also used as a tting parameter, the values of the internal
mass transfer coefcient of the different matrices are scattered over a wider range.
Silva et al. (2009) compared the performance of the BIC
series and parallel models to t the experimental data of grape
seed oil supercritical extraction at 313 K and 1820 MPa (Passos
et al., 2009a). In that work, kf af was estimated using the correlation of Tan et al. (1988). It was also assumed that the solutematrix interactions were negligible, reducing the number of
adjustable parameters to two (ks as and
). The parameters of
the series model were optimized rst and then employed as
initial guess in the optimization of the parallel model.
In Fig. 3 it can be observed that the parallel model gives
slightly lower values of the extraction yield during the fast
extraction period until the broken cells are depleted. In the
slow extraction step, both models are in good agreement with
experimental results.

3.1.4. Combined model: broken plus intact


cells + shrinking core
Fiori et al. (2009) proposed a new model combining the concepts of the BIC and the shrinking core models. It is assumed
0.7

E (kg solute/kg seed)

0.6
0.5
0.4
0.3

that, in a particle obtained from milled grape seed, there are


N concentric layers (where N is the ratio of the particle radius
to the cell diameter). The cells can either be broken or intact
by the milling process, as reported by Reverchon and Marrone
(2001) and Fiori et al. (2009) by SEM. It is assumed that the broken cells are located in the outer layer of the particle, as shown
in Fig. 2f.
During the extraction process, each layer is depleted in
sequence, from the outer layer to the center of the particle,
as assumed by the shrinking core model. The retraction of the
core increases the mass transfer resistance to the uid phase,
since the solute has to diffuse across each depleted layer. Fiori
et al. (2009) argued that the concentration inside the core is
higher than the solute solubility in the SC uid. Therefore,
according to Perrut et al. (1997), the uid phase concentration in equilibrium with the boundary concentration (Ci ) is
the solute solubility in the SC uid. The mass balance to the
core of the particle is:
q s,i
= Jf ap
t

Jf = kf Ci Ci

(51)

(52)

where kf is the pseudo mass transfer coefcient combining


the convective mass transfer resistance in the lm side and
the combined mass transfer resistances of the depleted layers.
According to Fiori et al. (2009), the resistance in the rst layer
(from the surface) is lower due to the presence of broken cells
containing free oil, as observed by SEM. After the broken cells
in the rst layer have been depleted of free oil, the extraction
of the intact cells in the inner layers is initiated. These layers
are composed of intact cells that have higher resistance to
mass transfer due to the presence of the cell walls. This is
modeled by decreasing the pseudo mass transfer coefcient
proportionally to the distance to the particle surface.
The initial solute concentration in the particle is required
to solve the mass balance:
t = 0 qi = qi,0

0.2

Exp 200 bar


Exp 180 bar

0.1

Series model
Parallel model

0
0

200

400

600

800

t (min)
Fig. 3 Comparison between the series and parallel BIC
models used to t experimental data of grape seed oil
extraction at 18 and 20 MPa, and 313.15 K. Data from Passos
et al. (2009a).

Fiori et al. (2009) proposed three approaches to calculate the


value of the pseudo mass transfer coefcient (kf ): (a) employ a
continuous function in all particle to calculate the dependence
of the coefcient with core radius continuous model; (b) use a
constant kf until the rst layer is exhausted (coincident with
kf ) and then use a continuous function of the core radius
semi-continuous model; (c) use a constant value of the coefcient for each layer discrete model. The equations employed
to calculate the pseudo mass transfer coefcient and average
solute concentration are listed in Table 3.

1115

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Table 3 Mathematical models developed to calculate the pseudo mass transfer coefcient, kf , and the average solute
concentration for spherical particles (Fiori et al., 2009).
Approach
Continuous

Model

 
r R

1
1
1 Rp
=
+ 
kf
kf,m dc
kf (r)

q i  = qi,0

 3
rc
Rp

k (r)

Semi-continuous

Conclusions

In this work, a review of the relevant mathematical models


employed for supercritical uid extraction of solutes from liquids or solids was presented. In the case of countercurrent
liquid-supercritical uid extraction, a typical model combining material and energy transfer was shown. The expressions
of the mass and heat uxes necessary for the calculations
were also presented. Regarding SC extraction of solid matrices, the most relevant mass transfer models linear driving
force, shrinking core, broken and intact cells and the combination of BIC and shrinking core models were described as
well as their assumptions and a few cases of their application.

r
Rp

1
1
1
=
+ 
kf,j
kf
kf,m
q i  = qi,0

4.

Rp dc r Rp

 3
= qi,0

R

(53)

(54)

Rp

In these equations j is the layer number, m is the total


number of layers, qi,0 is the initial solute concentration in the
particle, dc is the cell diameter measured by SEM, and kf,m
is the pseudo mass transfer coefcient of the internal layers.
The rst term on the right hand side of Eqs. (53), (55) and (57)
is the mass transfer resistance in the rst shell (convective
mass transfer coefcient) while the second term represents
the added resistance as each shell is depleted.
Fiori et al. (2009) compared the three approaches proposed, and shown that the continuous one underestimated
the coefcient initially, especially for small particle diameters. Simulation results showed that, despite relatively small
differences in the total extraction time (20%), there are big
differences in the predictions of the rst part of the extraction
curves (70%) between the continuous and discrete models.
The combined model was employed to describe the extraction of grape seed oil at 55 MPa and 313 K (Fiori, 2007) and the
extraction of almond oil at 35 MPa and 313 K (Marrone et al.,
1998). A discrepancy between the experimental and modeling results was observed. Fiori et al. (2009) concluded that
such disagreement was due to the existence of broken cells in
internal layers of the particle. To overcome this limitation, the
rst layer (where extraction kinetics is faster) was assumed to
have twice the thickness of the other layers. With this assumption the accuracy of the tting of the experimental results was
greatly improved.

0 r Rp

0 r Rp

1
,
kf

 
R r Rp

 1 = 1 + 1 p
, 0 r Rp dc
r
kf
kf,m dc
kf (r)

 R d 3

q i  = qi,0 p c , Rp dc r Rp


q i 

Discrete

(55)

(56)
,

m1 


0 r Rp dc
Rp
Rp jdc

2
(57)

j=1

jdc
Rp

3
(58)

Acknowledgements
The research leading to these results has received funding from the European Communitys Seventh Framework
Programme FP7/2007-2013 under grant agreement no CP-IP
228589-2 AFORE

References
Abaroudi, K., Trabelsi, F., Calloud-Gabriel, B., Recasens, F., 1999.
Mass transport enhancement in modied supercritical uid.
Ind. Eng. Chem. Res. 38, 35053518.
Bertucco, A., Vetter, G., 2001. High Pressure Process Technology:
Fundamentals and Applications, rst ed. Elsevier Science,
Amsterdam.
Brunner, G., 1984. Mass transfer from solid material in gas
extraction. Ber. Bunsenges. Phys. Chem. 88, 887891.
Brunner, G., 2009. Counter-current separations. J. Supercrit. Fluids
47, 574582.
Catchpole, O.J., Bernig, R., King, M.B., 1996a. Measurement and
correlation of packed-bed axial dispersion coefcients in
supercritical carbon dioxide. Ind. Eng. Chem. Res. 35, 824828.
Catchpole, O.J., Grey, J.B., Smalleld, B.M., 1996b. Near-critical
extraction of sage, celery, and coriander seed. J. Supercrit.
Fluids 9, 273279.
Catchpole, O.J., King, M.B., 1994. Measurement and correlation of
binary diffusion coefcients in near critical uids. Ind. Eng.
Chem. Res. 33, 18281837.
Catchpole, O.J., King, M.B., 1997. Measurement and correlation of
binary diffusion coefcients in near critical uids. Ind. Eng.
Chem. Res. 36, 401314013.
Coenen, H., Ben-Nasr, H., 1993. Decaffeination of raw coffee
beans. German Pattent Ofce, DE 3744988 (C2).
del Valle, J.M., De La Fuente, J.C., 2006. Supercritical CO2
extraction of oilseeds: review of kinetic and equilibrium
models. Crit. Rev. Food Sci. Nutr. 46, 131160.
del Valle, J.M., Rivera, O., Mattea, M., Ruetsch, L., Daghero, J.,
Flores, A., 2004. Supercritical CO2 processing of pretreated
rosehip seeds: effect of process scale on oil extraction
kinetics. J. Supercrit. Fluids 31, 159174.
Dwivedi, P.N., Upadhyay, S.N., 1977. Particle-Fluid Mass Transfer
in Fixed and Fluidized Beds. Ind. Eng. Chem. Proc. Des. Dev.
16, 157165.

1116

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Do, D.D., Rice, R.G., 1986. Validity of the parabolic prole


assumption in adsorption studies. AIChE J. 32, 149154.
Fang, T., Goto, M., Wang, X., Ding, X., Geng, J., Sasaki, M., Hirose,
T., 2007. Separation of natural tocopherols from soybean oil
byproduct with supercritical carbon dioxide. J. Supercrit.
Fluids 40, 5058.
Fernandes, J., Ruivo, R., Mota, J.P.B., Simes, P., 2007.
Non-isothermal dynamic model of a supercritical uid
extraction packed column. J. Supercrit. Fluids 41, 2030.
Fiori, L., 2007. Grape seed oil supercritical extraction kinetic and
solubility data: critical approach and modeling. J. Supercrit.
Fluids 43, 4354.
Fiori, L., Basso, D., Costa, P., 2009. Supercritical extraction kinetics
of seed oil: a new model bridging the broken and intact cells
and the shrinking-core models. J. Supercrit. Fluids 48,
131138.
Funazukuri, T., Kong, C., Kagei, S., 1998. Effective axial dispersion
coefcients in packed beds under supercritical conditions. J.
Supercrit. Fluids 13, 169175.
Ghoreishi, S., Shahrestani, R.G., Ghaziaskar, H.S., 2009.
Experimental and modeling investigation of supercritical
extraction of mannitol from olive leaves. Chem. Eng. Technol.
32, 4554.
Ghoreishi, S.M., Shahrestani, R.G., 2009. Subcritical water
extraction of mannitol from olive leaves. J. Food Eng. 93,
474481.
Glueckauf, E., 1955. Theory of chromatography. Part 10. Formulae
for diffusion into spheres and their application to
chromatography. Trans. Faraday Soc. 51, 15401551.
Glueckauf, E., Coates, J.I., 1947. Theory of chromatography. Part
IV. The inuence of incomplete equilibrium on the front
boundary of chromatograms and on the effectiveness of
separation. J. Chem. Soc., 13151321.
Goodrum, J.W., Kilgo, M.K., Santerre, C.R., 1996. Oilseed solubility
and extraction modelling. In: King, J.W., List, G.R. (Eds.),
Supercritical Fluid Technology in Oil and Lipid Chemistry.
AOCS Press, New York, pp. 101131.
Goto, M., Roy, B.C., Hirose, T., 1996. Shrinking-core leaching model
for supercritical-uid extraction. J. Supercrit. Fluids 9, 128133.
He, C.-H., Yu, Y.-S., Su, W.-K., 1998. Tracer diffusion coefcients of
solutes in supercritical solvents. Fluid Phase Equilib. 142,
281286.

Herrero, M., Mendiola, J.A., Cifuentes, A., Ibnez,


E., 2010.
Supercritical uid extraction: Recent advances and
applications. J. Chromatogr. A 1217, 24952511.
Katz, S.N., 1989. Method for decaffeinating coffee with a
supercritical uid. European Pattent Ofce, EP 0331852 (A2).
King, M.B., Catchpole, O.J., 1993. Physico-chemical data required
for the design of near-critical uid extraction process. In:
King, M.B., Bott, T.R. (Eds.), Extraction of Natural products
using near-critical solvents. Blackie Academic & Professional,
London, pp. 184231.
Kondo, M., Akgun, N., Goto, M., Kodama, A., Hirose, T., 2002.
Semi-batch operation and countercurrent extraction by
supercritical CO2 for the fractionation of lemon oil. J.
Supercrit. Fluids 23, 2127.
Liaw, C.H., Wang, J.S.P., Greenkorn, R.A., Chao, K.C., 1979. Kinetics
of xed-bed adsorption: a new solution. AIChE J. 25, 376381.
Liong, K.K., Wells, P.A., Foster, N.R., 1991. Diffusion in
supercritical uids. J. Supercrit. Fluids 4, 91108.
Liu, H., Silva, C.M., Macedo, E.A., 1997. New equations for tracer
diffusion coefcients of solutes in supercritical and liquid
solvents based on the LennardJones uid model. Ind. Eng.
Chem. Res. 36, 246252 (corrected in Ind. Eng. Chem. Res.
1998, 37(1), 308).
Magalhaes, A.L., Cardoso, S.P., Figueiredo, B.R., Da Silva, F.A.,
Silva, C.M., 2010a. Revisiting the LiuSilvaMacedo model for
tracer diffusion coefcients of supercritical, liquid, and
gaseous systems. Ind. Eng. Chem. Res. 49, 76977700.
Magalhaes, A.L., Da Silva, F.A., Silva, C.M., in press-a. New models
for tracer diffusion coefcients of hard sphere and real
systems-application to gases, liquids and supercritical uids.
J. Supercrit. Fluids, doi:10.1016/j.supu.2010.09.031.

Magalhaes, A.L., Da Silva, F.A., Silva, C.M., in press-b. New tracer


diffusion correlation for real systems over wide ranges of
temperature and density. Chem. Eng. J.,
doi:10.1016/j.cej.2010.09.069.
Margolis, G., Chiovini, J., 1982. Decaffeination process. European
Pattent Ofce, EP 0010637 (A1).
Marrone, C., Poletto, M., Reverchon, E., Stassi, A., 1998. Almond
oil extraction by supercritical CO2 : experiments and
modelling. Chem. Eng. Sci. 53, 37113718.
Onda, K., Takeuchi, H., Okumoto, Y., 1968. Mass transfer
coefcients between gas and liquid phases in packed
columns. J. Chem. Eng. Jpn. 1, 5662.
Passos, C.P., Silva, R.M., Da Silva, F.A., Coimbra, M.A., Silva, C.M.,
2009a. Enhancement of the supercritical uid extraction of
grape seed oil by using enzymatically pre-treated seed. J.
Supercrit. Fluids 48, 225229.
Passos, C.P., Silva, R.M., Da Silva, F.A., Coimbra, M.A., Silva, C.M.,
2010. Supercritical uid extraction of grape seed
(Vitis vinifera L.) oil. Effect of the operating conditions upon
oil composition and antioxidant capacity. Chem. Eng. J. 160,
634640.
Passos, C.P., Yilmaz, S., Silva, C.M., Coimbra, M.A., 2009b.
Enhancement of grape seed oil extraction using a cell wall
degrading enzyme cocktail. Food Chem. 115, 4853.
Peker, H., Srinivasan, M.P., Smith, J.M., McCoy, B.J., 1992. Caffeine
extraction rates from coffee beans with supercritical carbon
dioxide. AIChE J. 38, 761770.
Perrut, M., Clavier, J.Y., Poletto, M., Reverchon, E., 1997.
Mathematical modeling of sunower seed extraction by
supercritical CO2 . Ind. Eng. Chem. Res. 36, 430435.
Pronyk, C., Mazza, G., 2009. Design and scale-up of pressurized
uid extractors for food and bioproducts. J. Food Eng. 95,
215226.
Puiggen, J., Larrayoz, M.A., Recasens, F., 1997. Free
liquid-to-supercritical uid mass transfer in packed beds.
Chem. Eng. Sci. 52, 195212.
Ranz, W.E., Marshall, W.R., 1952. Evaporation from drops. Chem.
Eng. Prog. 48, 173180.
Recasens, F., McCoy, B.J., Smith, J.M., 1989. Desorption processes:
supercritical uid regeneration of activated carbon. AIChE J.
35, 951958.
Regulation, E.C., 2001. Descriptions and Denitions of Olive Oils
and Olive-Pomace Oils.
Reid, R.C., Prausnitz, J.M., Poling, B.E., 2000. The Properties of
Gases and Liquids, fth ed. McGraw-Hill Professional, New
York.
Reverchon, E., Daghero, J., Marrone, C., Mattea, M., Poletto, M.,
1999. Supercritical fractional extraction of fennel seed oil and
essential oil: experiments and mathematical modeling. Ind.
Eng. Chem. Res. 38, 30693075.
Reverchon, E., Marrone, C., 2001. Modeling and simulation of the
supercritical CO2 extraction of vegetable oils. J. Supercrit.
Fluids 19, 161175.
Rice, R.G., Do, D.D., 1995. Applied Mathematics and Modeling for
Chemical Engineers, rst ed. John Wiley & Sons, Inc, New
York.
Riha, V., Brunner, G., 2000. Separation of sh oil ethyl esters
with supercritical carbon dioxide. J. Supercrit. Fluids 17,
5564.
Roy, B.C., Goto, M., Hirose, T., Navaro, O., Hortacsu, O., 1994.
Extraction rates of oil from tomato seeds with supercritical
carbon dioxide. J. Chem. Eng. Jpn. 27, 768772.
Ruckenstein, E., Liu, H., 1997. Self-diffusion in gases and liquids.
Ind. Eng. Chem. Res. 36, 39273936.
Ruivo, R., Cebola, M.J., Simoes, P.C., Nunes da Ponte, M., 2002.
Fractionation of edible oil model mixtures by supercritical
carbon dioxide in a packed column. 2. A mass-transfer study.
Ind. Eng. Chem. Res. 41, 23052315.
Ruivo, R., Paiva, A., Mota, J.P.B., Simes, P., 2004. Dynamic model
of a countercurrent packed column operating at high pressure
conditions. J. Supercrit. Fluids 32, 183192.
Ruthven, D.M., 1984. Principles of Adsorption and Adsorption
Processes, rst ed. John Wiley & Sons, New York.

chemical engineering research and design 8 9 ( 2 0 1 1 ) 11041117

Sahena, F., Zaidul, I.S.M., Jinap, S., Karim, A.A., Abbas, K.A.,
Norulaini, N.A.N., Omar, A.K.M., 2009. Application of
supercritical CO2 in lipid extraction a review. J. Food Eng. 95,
240253.
Sato, M., Kondo, M., Goto, M., Kodama, A., Hirose, T., 1998.
Fractionation of citrus oil by supercritical countercurrent
extractor with side-stream withdrawal. J. Supercrit. Fluids 13,
311317.
Silva, C.M., Passos, C.P., Coimbra, M.A., Da Silva, F.F.A., 2009.
Numerical simulation of supercritical extraction processes.
Chem. Prod. Process Model. 4, Article 9.
Sovov, H., 1994. Rate of the vegetable oil extraction with
supercritical CO2 . I. Modelling of extraction curves. Chem.
Eng. Sci. 49, 409414.
Sovov, H., 2005. Mathematical model for supercritical uid
extraction of natural products and extraction curve
evaluation. J. Supercrit. Fluids 33, 3552.
Sovov, H., Komers, R., Kucera, J., Jez, J., 1994a. Supercritical
carbon dioxide extraction of caraway essential oil. Chem. Eng.
Sci. 49, 24992505.
Sovov, H., Kucera, J., Jez, J., 1994b. Rate of the vegetable oil
extraction with supercritical CO2 . II. Extraction of grape oil.
Chem. Eng. Sci. 49, 415420.
Stichlmair, J., Bravo, J.L., Fair, J.R., 1989. General model for
prediction of pressure drop and capacity of countercurrent
gas/liquid packed columns. Gas Sep. Purif. 3, 1928.
Stocketh, R., Brunner, G., 2000. Holdup, pressure drop, and
ooding in packed countercurrent columns for the gas
extraction. Ind. Eng. Chem. Res. 40, 347356.
Stber, F., Julien, S., Recasens, F., 1997. Internal mass transfer in
sintered metallic pellets lled with supercritical uid. Chem.
Eng. Sci. 52, 35273542.

1117

Tan, C.-S., Liang, S.-K., Liou, D.-C., 1988. Fluid-solid mass transfer
in a supercritical uid extractor. Chem. Eng. J. 38, 1722.
Tan, C.S., Liou, D.C., 1988. Desorption of ethyl acetate from
activated carbon by supercritical carbon dioxide. Ind. Eng.
Chem. Res. 27, 988991.
Tan, C.S., Liou, D.C., 1989. Axial dispersion of supercritical carbon
dioxide in packed beds. Ind. Eng. Chem. Res. 28, 1246
1250.
Vzquez, L., Hurtado-Benavides, A.M., Reglero, G., Fornari, T.,

Ibnez,
E., Senorns,
F.J., 2009. Deacidication of olive oil by
countercurrent supercritical carbon dioxide extraction:
experimental and thermodynamic modeling. J. Food Eng. 90,
463470.
Wakao, N., Funazkri, T., 1978. Effect of uid dispersion
coefcients on particle-to-uid mass transfer coefcients in
packed beds: correlation of sherwood numbers. Chem. Eng.
Sci. 33, 13751384.
Wen, C.Y., Fan, L.T., 1975. Models for Flow Systems and Chemical
Reactors, rst ed. Dekker, New York.
Wilke, C.R., Chang, P., 1955. Correlation of diffusion coefcients
in dilute solutions. AIChE J. 1, 264270.
Wu, W., Hou, Y., 2001. Mathematical modeling of extraction of
egg yolk oil with supercritical CO2 . J. Supercrit. Fluids 19,
149159.
Yang, R.T., 1999. Gas Separation by Adsorption Processes, second
ed. Imperial College Press, London.
Yu, D., Jackson, K., Harmon, T.C., 1999. Dispersion and diffusion
in porous media under supercritical conditions. Chem. Eng.
Sci. 54, 357367.
Zosel, K., 1971. Decaffeinating green coffee with moist carbon
dioxide in the supercritical state. German Pattent Ofce, DE
2005293 (A1).

Das könnte Ihnen auch gefallen