International Journal of Heat and Mass Transfer 72 (2014) 274–287

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Evaluation of models for supercritical fluid extraction

Amit Rai a,⇑, Kumargaurao D. Punase b,1, Bikash Mohanty a,2, Ravindra Bhargava a,3
a
Department of Chemical Engineering, Indian Institute of Technology Roorkee, Roorkee 247667, Uttarakhand, India
b
Department of Chemical Engineering, University of Petroleum and Energy Studies, Deharadoon 248007, Uttarakhand, India

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

Article history: Various models available in the literature for the modeling of supercritical extraction process are studied
Received 22 April 2013 and validated using published experimental data. The first model considers internal mass transfer coef-
Received in revised form 1 December 2013 ficient as the controlling parameter for the extraction process. On the other hand, the second model ana-
Accepted 4 January 2014
lyzes the dynamic behavior of the extraction process by considering intra-particle diffusion and external
mass transfer. These models have also been studied to understand the effects of various model parame-
ters like intra-particle diffusion, mass transfer coefficients & operating parameters on cumulative extrac-
Keywords:
tion yield. The model proposed by Reverchon (1996) [13] predicts a cumulative yield within an error limit
Diffusion
Mass transfer
of +9% in the MATLAB simulation and +4% to 5% in the FEMLAB simulation. Also, the model proposed by
Numerical analysis Goto et al. (1993) [8] fits the experimental data of Kim et al. (2007) [19], Skerget and Knez (2001) [20],
Packed bed and Tonthubthimthong et al. (2004) [21] within an error band of +10% to 2%.
Simulation Ó 2014 Elsevier Ltd. All rights reserved.
Supercritical fluid

1. Introduction tions [3]. Single stage supercritical extraction and separation

In recent years, the use of supercritical fluid extraction (SFE)
process for the removal of organic compounds from different liquid
and solid matrices has received much attention. Because, super-
critical fluids have several distinctly advantageous properties, such
as liquid like density and gas like viscosity and diffusivity, they
have high mass-transfer characteristics and their effectiveness
can be controlled by small changes in temperature and pressure
leading to better fractional separation [1]. The mechanism of
supercritical fluid extraction process can be explain by the follow-
ing steps; (1) Transport of supercritical solvent to the particle sur-
face and then from particle surface to interior of particle by
diffusion. (2) Dissolution of the solute with the supercritical sol-
vent. (3) Transport of supercritical solvent with molecules from
interior of particle to particle surface. (4) Transport of supercritical
solvent and solute molecules from particle surface to bulk solvent.
Hence, the possibility of using supercritical solvents at the com-
mercial level has increased in the recent past [2]. To design an
extraction plant, it is necessary to have reliable mass-transfer mod-
els that will allow the determination of optimum operating condi-
⇑ Corresponding author. Mobile: +91 75790 75744; fax: +91 1332 276535.
E-mail addresses: mit.rai123456@gmail.com (A. Rai), kumargaurav04@gmail.-
com (K.D. Punase), bmohanty@iitr.ernet.in (B. Mohanty), ravibfch@iitr.ernet.in (R.
Bhargava).
1 Reverchon [13] took into account the shape of the particles
Mobile: +91 9917720952.
2
Tel.: +91 01332 5710.
3
Tel.: +91 01332 285382.
produces a quasi-solid extract, which consists of several compound
families. However, the yield data and the shape of the extraction
curve are influenced by the presence of undesired compounds [4].
Some authors have attempted to describe the evolution of the
extraction process by using empirical kinetic equations [5,6]. Heat
transfer analogy of a single sphere cooled in a fluid medium was
used by Reverchon et al. [7] to describe the extraction process.
However, this model describes a highly idealized situation and
the performance of the fixed bed of particles used is overestimated.
The extraction process was also modeled by integrating the dif-
ferential mass balances in the solid and fluid phase. Goto et al. [8]
described the extraction of peppermint essential oil as a desorption
process characterized by the attainment of an instantaneous equi-
librium by breaking the peppermint leaves into differential slab
elements.
Sovova [9,10] modeled the vegetable oil extraction process
based on the broken and intact cell model by considering the oil
contained as either accessible or inaccessible. The same model
was also proposed for pepper extraction [11], where the internal
and external mass transfer resistances were taken into account.
Goto et al. [12] proposed a shrinking core model and explained
the ginger rhizomes extraction considering effective diffusivity
and solubility as model parameters. However, the model was un-
able to describe the experimental results obtained for different
particle sizes.
(slabs) to obtain a good fit with the experimental data for large par-
ticles and found that internal mass transfer controlled the essential

0017-9310/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijheatmasstransfer.2014.01.011
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287
Nomenclature
Reverchon [13]
Ap total surface of particles (m2)
c extract concentration in the fluid phase (kg/m3)
cn fluid-phase concentration in the nth stage (kg/m3)
h spatial coordinate along the bed (m)
kp volumetric partition coefficient of the extract between
the solid and the fluid phase at equilibrium (–)
K internal mass-transfer coefficient (m/s)
n number of stages deriving from the bed subdivision (–)
q extract concentration in the solid phase (kg/m3)
qn solid phase concentration in the nth stage (kg/m3)
q ⁄
concentration at the solid–fluid interface (kg/m3),
t extraction time (s)
ti internal diffusion time (s)
u superficial velocity (m/s)
V extractor volume (m3)
W CO2 mass flow rate (kg/s)
e bed porosity (–)
q solvent density (kg/m3)
Goto et al. [8]
a1, a2 constants defined by Eq. (3.41) (–)
ap specific surface area (1/m)
A constant defined by Eq. (3.42) (–)
Ab bed cross section area (m2)
b, c constant defined by Eq. (3.43) (–)
C solute concentration in the solvent (kg/m3)
Cp solute concentration within the particle pore (kg/m3)
Cp0 solute concentration in the pore phase at t ¼ 0 (kg/m3)
Cps solute concentration in the pore space at the particle
surface (kg/m3)
Cs solute concentration in particle (kg/m3)
oil extraction from sage leaves. Goodarznia and Eikani [14] pro-
posed a model based on differential mass balance on a single parti-
cle as well as in the fluid phase and validated the experimental data
of Reverchon et al. [7,15] and Sovova [10]. The model also included
the effects of internal diffusivity and axial dispersion.
The phase equilibrium depends on solute composition, solvent
composition, extraction pressure and temperature. It controls the
initial extraction period when the fluid phase leaving the extractor
is either in equilibrium or is about to attain equilibrium with the sol-
ute in the solid phase. When the solute concentration in solid phase
is high, like that of vegetable oil in Canola seed, the fluid-phase equi-
librium concentration is independent of the matrix and equal to oil
solubility. When the initial solute concentration in the plant is low,
which is rare for vegetable oils, the equilibrium is usually controlled
by solute–solid interaction and the fluid-phase concentration is
much lower than the oil solubility. The equilibrium is expressed
as a linear relationship between the solid and fluid phase concentra-
tions and the proportionality constant is called the partition coeffi-
cient. Goto et al. [16] used the Brunauer–Emmett–Teller adsorption
isotherm to simulate a smooth transition between the equilibrium
of free solute at high concentrations and the equilibrium of sol-
ute–solid interaction at low concentrations.
Perrut et al. [17] considered a discontinuous equilibrium to
model the sunflower oil extraction process. The fluid-phase con-
centration is equal to the oil solubility above the discontinuity
and is determined by partition coefficient below it. The discontin-
uous equilibrium curve was successfully applied by Wu and Hou
[18] in the simulation of egg yolk oil extraction.
275
Cs0 solute concentration in the solid phase at t ¼ 0 (kg/m3)
C0 total solute concentration (kg/m3)
dp particle diameter (m)
DAB binary diffusion coefficient (m2/s)
Dax axial dispersion coefficient (m2/s)
De effective intraparticle diffusion coefficient (m2/s)
F cumulative fraction of solute extracted (–)
h height of the bed (m)
ka adsorption rate constant (1/s)
kp overall mass transfer coefficient (m/s)
kf external mass transfer coefficient (m/s)
K equilibrium adsorption coefficient (–)
r radial position in spherical particle (m)
R radius of spherical particle (m)
t time (s)
Us superficial velocity (m/s)
x dimensionless solute concentration in effluent (–)
x0 initial solute mass ratio in the solid phase (–)
xp dimensionless solute concentration in pore (–)
xs dimensionless solute concentration in solid particle (–)
y solute mass ratio in the fluid phase (kg/kg)
z bed height coordinate (m)
a bed void fraction (–)
b particle porosity (–)
/ dimensionless mass transfer coefficient (–)
h dimensionless time (–)
qs solid density without void volume of the solid matrix
(kg/m3)
s total bed volume/volumetric flow rate (s)
IC initial condition
BC boundary condition
It can be concluded that there are various models available in the
literature that differ not only from a mathematical point of view, but
also in terms of mass transfer mechanisms, which control the super-
critical extraction process of different matrices. Hence, a single
model cannot describe all the experimental results. In all published
models, the initial extraction process is governed by the solubility
equilibrium between the solute and the fluid phase, which, in most
cases, is assumed to be linear as detailed information is not available
for complex matrix systems. From a mathematical point of view, all
the proposed models are based on differential mass balance integra-
tion with some assumptions. Table 1 shows that the published
supercritical fluid extraction models differ in the description of
phase equilibrium, flow pattern, and solute diffusion in the solid
phase. As the experimental data considered in the present study
are related to seeds of sage, black pepper, nimbin and caffeine.
Amongst these sages, black pepper and nimbin belong to the cate-
gory of essential oil. The literature shows that for extraction of
essential oil as well as caffeine the model developed by Goto et al.
[7] and Reverchon [3,12] are suitable. Therefore, in the present pa-
per these models are considered for extraction of oil from seeds of
sage, black pepper, nimbin and caffeine.
2. Mathematical modeling
The initial distribution of the solute within the solid substrate
affects the selection of the possible models. The solute may be free
on the surface of the solid material, adsorbed on the outer surface,
located within pores or evenly distributed within plant cells. In the
276 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287

Table 1
The models based on differential mass balance equations and applied to the extraction of natural products with near-critical CO2.

References Solutea Equilibrium relationship Particleb Solvent flowc

[29] FO Constant No internal resistance PF
[8] EO Linear Porous slab M
[12] EO Linear Shrinking core AD
[30] OR, FO Linear Sphere, cylinder, slab PF
[13] EO Linear Sphere, cylinder, slab PF
[31] EO Constant Shrinking core AD
[32] EO Constant Shrinking core PF
[14] EO Linear Sphere AD
[33] EO, FO Linear Sphere PF
[34] EO Linear No internal resistance AD
[16] EO, W BET Porous particle M
[36] FO Combined Porous particle PF
[37] FO Combined No internal resistance PF
[10] FO Constant B+I PF
[43] FO Constant B+I PF
[10] EO Approx. linear B+I PF
[28] FO Combined B+I PF
[35] EO,FO Combined B+I PF
[38] FO Linear No internal resistance PF
[39] EO Linear B+I AD
[40] EO Polynomial Porous Spherical AD
[41] EO Freundlich isotherm Porous Spherical AD
[42] EO Linear Porous Spherical AD
a
C: caffeine; EO: essential oil; FO: fatty oil; OR: oleoresin; W: wax.
b
B + I: broken and intact cells.
c
PF: plug flow; M: mixer; AD: flow with axial dispersion.

present work, an in-depth evaluation of two models [8,4] based on

differential mass balance integration have been carried out using
Solute free solvent
the published experimental data of Kim et al. [19], Reverchon
et al. [4], Skerget and Knez [20], and Tonthubthimthong et al. [21].

2.1. The Reverchon model

A model based on the integration of differential mass balances

along the extraction bed is proposed with the following assump-
tions: (1) Plug flow exists in the bed. (2) The axial dispersion in
the bed is negligible. (3) The fluid flow rate, temperature, pressure δh
and bed properties are constant.
Based on the following assumptions, the model equations
(2.1.1) and (2.1.2) is obtained for the situation where the seeds
are stationary and oil free solvent is entered at the top of extraction
vessel, by taking solute mass balance on the solvent and solid
phase over an element of extractor of height dh respectively as de-
scribed in Fig. 1 [22].
@c @c
uV þ eV þ Ap Kðq  q Þ ¼ 0 ð2:1:1Þ Solute bearing solvent
@t @t
@q
ð1  eÞV ¼ Ap Kðq  q Þ ð2:1:2Þ
@t Fig. 1. Schematic diagram of the extraction vessel.

c ¼ 0; q ¼ q0 ; at t ¼ 0 ð2:1:3aÞ
cð0; tÞ ¼ 0 at h ¼ 0 ð2:1:3bÞ @q 1
¼  ðq  q Þ ð2:1:5Þ
⁄ @t ti
Assuming a linear relationship for SFE process between c and q due
to lack of experimental phase equilibrium data, Now, the fixed bed is divided into n stages and it is also assumed
 that the fluid and solid phase concentration is uniform in each
c ¼ kp  q ð2:1:4Þ
stage, which is approximated as a plug-flow extractor through a
In Eq. (2.1.2), the fraction ApK/(1  e)V depends on the geometry of series of mixed extractors. So model equations (2.1.1) and (2.1.5)
particles, though ApK and e are supposed to be constant within the are written as a set of 2n ODEs given by Eqs. (2.1.6) and (2.1.7),
bed. This fraction is dimensionally equal to l/s. Therefore, the inter- respectively.
nal diffusion time (ti) which, according to the hypothesis, is charac- W v dcn v dqn
teristic of the extraction process and is defined as ti = (1  e)V/ApK ðcn  cn1 Þ þ e þ ð1  eÞ ¼0 ð2:1:6Þ
q n dt n dt
and related with the internal diffusion coefficient (Di) by ti = r2/
dqn 1
15Di, where r is the mean particle radius. Therefore Eq. (2.1.2) be- ¼  ðqn  qn Þ ð2:1:7Þ
comes Eq. (2.1.5). dt ti
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287 277

Initial conditions are given as: C p ¼ C P0 at t ¼ 0; 0 6 r 6 R ð2:2:1:3aÞ

cn ¼ 0 and qn ¼ q0 at t ¼ 0 C s ¼ C s0 at t ¼ 0; 0 6 r 6 R ð2:2:1:3bÞ
 
@C p
Eqs. (2.1.6) and (2.1.7) are solved by using a fourth-order Runge–  De ¼ kf ðC ps  CÞ at r ¼ R for all t ð2:2:1:3cÞ
@r
Kutta method to obtain the solute concentration in solvent and so-  
lid phase, respectively with initial conditions. @C p
¼ 0 at r ¼ 0 for all t ð2:2:1:3dÞ
@r
2.2. The Goto model C 0 ¼ bC po þ ð1  bÞC s0

In this model, the solid particle is considered as a porous struc-

ture and the mechanism of the extraction of solute from porous 2.2.2. General fluid phase mass balance
structures is divided into five steps: (1) Transport of the supercrit- The concentration C of the solute in the fluid phase at height z in
ical solvent molecules from bulk solvent to the particle surface a bed of particulate material depends on the rate of mass transfer
through the boundary layer adjacent to the particle surface. (2) from particles at height z, the fluid flow rate and extent of mixing.
Penetration of the solvent molecules into the pores of the particle Eq. (2.2.2.1) is obtained by taking a mass balance around an ele-
(impregnation). (3) Diffusion of the extractable component i.e. sol- ment Dz of bed height as shown in Fig. 3 [23].
ute to the particle surface (internal diffusion). (4) Diffusion of the
@C @C @2C
solute from the particle surface through a stagnant (boundary) a ¼ U s þ Dax 2 þ ap ð1  aÞkf ðC ps  CÞ ð2:2:2:1Þ
@z @z @z
layer of the fluid (external diffusion), and (5) Convective transfer
of the solute in the intergraded space towards the extractor outlet. The specific surface area is defined as. ap = 6/dp. The initial and
The diffusion of the extractable substance through the boundary boundary conditions for Eq. (2.2.2.1) are:
layer is a relatively slow process. Diffusion in a solid, particularly
C ¼ 0 at t ¼ 0; 0 6 z 6 h ð2:2:2:2aÞ
in a nonporous solid, occurs even more slowly, and it is this step
that limits the extraction rate. C ¼ 0 at z ¼ 0 ð2:2:2:2bÞ
@C
¼ 0 at z ¼ h ð2:2:2:2cÞ
2.2.1. General solid phase mass balance @z
The transport of solute (at low concentration) in or out of the
particle is assumed to take place by diffusion through a network
2.2.3. Model simplification and analytical solution
of pores as described in steps 1–5. As the pore diameter consider-
In order to simplify the model equation and its initial and
ably varies, the diffusion is described in terms of an effective pore
boundary conditions, the following assumptions are made: (1) ax-
diffusion coefficient. The solute distribution is assumed to be radi-
ial dispersion is negligible, (2) radial dispersion is also neglected
ally symmetric. The fluid-solute mass balance can be represented
(small column diameter), (3) isothermal process, (4) the packed
for a porous spherical particle as shown in Fig. 2.
column is isobaric, (5) no interaction among solutes in the fluid
The differential mass balance equation for solute concentration
phase or solid phase, (6) local equilibrium adsorption between sol-
in the particle pores is written as:
  ute and solid in pore of material, (7) differential bed is gradient-
  2 @C p less in solid and fluid phase, (8) physical properties of the super-
@C p 1 @ r De @r @C s
b ¼ 2  ð1  bÞ ð2:2:1:1Þ critical fluid are constant. It is assumed that the combined internal
@t r @r @t
and external mass transfer processes are described by a linear driv-
The local extraction rate, which is equivalent to the desorption rate, ing force approximation, which is derived by assuming a parabolic
is assumed to be reversible and linear in terms of adsorption rate concentration profile within the particle.
constant ka and the adsorption equipment constant K given by Eq.
(2.2.1.2)
  Coutlet(t)
@C s Cs
¼ ka C p  ð2:2:1:2Þ z=h
@t K
Eq. (2.2.1.1) can be solved using the following initial and boundary
conditions.

r=R
Z+∆Z
C(t,r,z)
r+ r
Z
Particle surface

Cp(t,r,z) Solid
Fluid
Cps(t,R,z)
SCF Film
Cs(t,r,z)
z=0

Cinlet(t)

Fig. 2. Porous solid particle. Fig. 3. Schematic represent of packed bed element.
278 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287

100

90

80

70

Modelled
60
Experimental (Reverchon, 1996)
Yield, (%)

50

40

30

20

10

0
0 50 100 150 200 250 300 350 400 450

Time, (min)

Fig. 4. Comparison between experimental and model predicted data for MATLAB.

15De ven by C s ¼ KC p . To further simplify the model equations, some

kf ap ð1  aÞðC  C ps Þ ¼ ð1  aÞðC ps  CÞ ð2:2:3:1Þ
R2 dimensionless variables are defined as: x = C/C0, xp ¼ C p =C 0 ,
xp ¼ C p =C 0 , xs ¼ C s =C 0 , h = t/s and / = kpaps. In terms of dimension-
The overall mass transfer coefficient for a spherical particle is given less variables, Eqs. (2.2.2.1) and (2.2.1.1) can be written as Eqs.
by Eq. (2.2.3.2) (2.2.3.4) and (2.2.3.5):

kp ¼
kf
ð2:2:3:2Þ dx x /ð1  aÞ  xs 
1 þ Bi =5 þ ¼ x ð2:2:3:4Þ
dh a a K
dxs /  xs 
where Bi = kfh/De is Biot number. Average intra-particle and solid ¼ b  x ð2:2:3:5Þ
concentrations are evaluated using the parabolic profile. dh K
þ ð1  bÞ K
Z R Initial conditions are:
3
Cp ¼ 4pr 2 C p ðrÞdr ð2:2:3:3aÞ
4pRs 0 x ¼ 0 at h ¼ 0 ð2:2:3:6aÞ
Z R
3 K
Cs ¼ 4pr 2 C s ðrÞdr ð2:2:3:3bÞ xs ¼ at h ¼ 0 ð2:2:3:6bÞ
4pRs 0 ½b þ ð1  bÞK
It is assumed that the equilibrium in the pores is established instan- The analytical solution of Eqs. (2.2.3.4) and (2.2.3.5) with its initial
taneously for a relatively fast adsorption–desorption rate and is gi- conditions is given as:

100

90 +42% +9%

80

70
Predicted Yield, (%)

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Experimental Yield, (%)

Fig. 5. Error between predicted data from Reverchon model and experimental data.
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287 279

100
When the mass transfer resistance is negligible, i.e. when / = 1; Eq.
(2.2.3.9) reduces to Eq. (2.2.3.10).
90  
h
FðhÞ ¼ 1  exp  ð2:2:3:10Þ
80 ½b þ ð1  bÞKð1  aÞ þ a

70 The mass ratio of the solute in the fluid phase as a function of time
can be obtained as:
60
   

Yield, (%)

b x0 qs t t
yðtÞ ¼ þ ð1  bÞ A exp a1  exp a2 ð2:2:3:11Þ
50 K qCO2 s s
qCO y
40 where x ¼ c02 .
The mass of extract at the bed outlet can be calculated from Eq.
30
(2.2.3.12):
20 Z t
Modelled
mðtÞ ¼ yðtÞQ CO2 qCO2 dt ð2:2:3:12Þ
10 Experimental (Reverchon, 1996) 0

0 Using Eq. (2.2.3.8a) in Eq. (2.2.3.8b), we obtain

0 100 200 300 400
  

b 1 t
Time, (min) mðtÞ ¼ þ ð1  bÞ x0 qs Q CO2 As exp a1 1
K a1 s
 

Fig. 6. Comparison between experimental data and Reverchon model predictions 1 t
with FEMLAB. þ 1  exp a2 ð2:2:3:13Þ
a2 s
The ordinary differential equations, Eqs. (2.2.3.4) and (2.2.3.5), for
xðhÞ ¼ A½expða1 hÞ  expða2 hÞ ð2:2:3:7Þ the dimensionless solute concentration in the bulk fluid phase
where, and solid phase respectively are used as the model equations with
 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Eqs. (2.2.3.6a) and (2.3.3.6 b) as the initial conditions. The model
1 2 1 2 equations are simplified using Laplace transform to obtain the ana-
a1 ¼ b þ b  4c ; a2 ¼ b  b  4c ð2:2:3:8aÞ
2 2 lytical solution of the model in terms of the dimensionless solute
ð1  aÞ/ concentration in the bulk fluid phase, given by Eq. (2.2.3.7). Eq.
A¼ ð2:2:3:8bÞ (2.2.3.9), which gives the cumulative fraction of the solute extracted
½b þ ð1  bÞKaða1  a2 Þ
/ 1 /ð1  aÞ or extraction yield is used for model validation against experimen-
b¼ þ þ ð2:2:3:8cÞ tal results with the determination of the constants defined by Eqs.
b þ ð1  bÞK a a
2.2.3.8a, 2.2.3.8b, 2.2.3.8c, 2.2.3.8d.
/
c¼ ð2:2:3:8dÞ
½b þ ð1  bÞKa
2.2.4. Estimation of the physical properties and parameter
The cumulative fraction of solute extracted up to dimensionless Identification
time h is given by: 2.2.4.1. Estimation of the physical properties. The empirical correla-
Z h tion proposed by Jossi et al. can be used to calculate the viscosity
1 of CO2 as given below [14]:
FðhÞ ¼ xdh
1a 0

0:25
A expða1 hÞ  1 expða2 hÞ  1 ½ðlf  lf Þn þ 104  ¼ 0:10230 þ 0:023364qr
¼  ð2:2:3:9Þ
1a a1 a2 þ 0:058533q2r  0:040758qr3 þ 0:0093324q4r ð2:2:4:1:1Þ

100

90 +4

80 -5%

70
Predicted Yield, (%)

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90 100
Experimental Yield, (%)

Fig. 7. Error between predicted data from Reverchon model and experimental data.
280 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287

100

90

80

70
Model prediction by MATLAB
Model prediction by FEMLAB
60
Yield, (%)

50

40

30

20

10

0
0 50 100 150 200 250 300 350 400 450
Time, (min)

Fig. 8. Comparison between Reverchon model prediction with MATLAB and FEMLAB.

0.18
0.5
0.16
0.45

0.14 0.4
Solute Concntration, X (-)

Cumulative fraction, F (-)

0.12 0.35

0.3
0.1

0.25
0.08
0.2
0.06
0.15

0.04
0.1
ODE Solution
0.02 Analytical Solution Modeled
0.05
Experimental (Kim et al.,2007)

0 0
0 5 10 15 20 25 0 10 20 30 40 50
Time, θ (-)
Time (-)
Fig. 10. Comparison between experimental data and simulated data.
Fig. 9. ODE and analytical solutions for the overallsolute concentration in bulk fluid
phase.

The viscosity l⁄ at normal pressure can be calculated as follows: The axial dispersion coefficient in the fluid phase is calculated
using the correlation given by Tan and Liou [26]:
 
l ¼ 34:0 
fn T r 6 1:50 105 T 0:94
r ;
lf n ¼ 17:78  10 ð4:58T r  1:67Þ5=8 ; T r > 1:50
  5
Pe ¼ 1:634 Re0:265 Sc0:919 ð2:2:4:1:3Þ
1=6
 Tc
where n ¼ 1=2 2=s . where Peclet number Pe is obtained using the relation Dax(2UsR)/Pe.
M Pc
The particle porosity (b) and the bed void fraction (a) is com-
puted by using the relation b = 1  (qp/qs) and a = 1  (qb/qp),
respectively. The effective intra particle diffusion coefficient De is 2.2.4.2. Model parameters. The experimental data related to extrac-
estimated from De = DABb2. The binary diffusion coefficient DAB, tion of caffeine from green tea [19], sage oil [4], pepper [20], nim-
can be obtained using Riazi and Whitson correlation [24]. The bin from neem seed [21] is used for model validation. In the
external mass transfer coefficient, kf, is calculated with the Wakao present work, two models are selected and model equations are
and Kaquei correlation [8,25]. solved in MATLAB and FEMLAB environment. After solving the
model equations, the results are validated with the experimental
Sh ¼ 2 þ 1:1Sc1=3 Re0:6 ð2:2:4:1:2Þ
values using the tuning parameter ‘‘Di’’ and ‘‘K’’ of the model
where Sh = 2Rkf/De, Sc ¼ l=ðqCO2 De Þ, and Re ¼ ð2RU s qCO2 Þ=l. equations.
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287 281

0.9 +15%

P redicted cum ulative extraction yield, F (-)

-2%
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Experimental cumulative extraction yield, F (-)

Fig. 11. Error predictions from Goto model and experimental data of [19].

1.4 k1 ¼ hf ðxn ; yn Þ
 
Solute concntration in solid phase, Xs (-)

h k1
k2 ¼ hf xn þ ; yn þ
1.2 5 5
 
3h 3k1 9k2
k3 ¼ hf xn þ ; yn þ þ
1 10 40 40
 
4h 44k1 56k1 32k3
0.8 k4 ¼ hf xn þ ; yn þ  þ
5 45 15 9
 
8h 19372k1 25360k3 64448k3 212k1
0.6 k5 ¼ hf xn þ ; yn þ  þ 
9 6561 2187 6561 729
 
9017k1 355k1 46732k3 49k5 5103k5
0.4 k6 ¼ hf xn þ h; yn þ  þ þ 
3168 33 5247 176 18656
0.2
Solute Concentration
The 2n differential equations are solved by applying ode45 toolbox
0 of MATLAB 7.0 which works on the principle of Runge–Kutta (Dor-
0 5 10 15 20 25 30 mand–Prince) method. The advantages of using Dormand–Prince
Time, θ (-) method over other Runge–Kutta methods are that it matches more
terms in the Taylor series approximation by taking a weighted aver-
Fig. 12. Dimensionless average solute concentration profile in solid phase with
time. age of several derivative approximations.

3.2. The algorithm for solution technique employed for the Reverchon
model using MATLAB is as follows
3. Solution technique

Step 1: Define all the input parameters like bed void fraction, par-
The solution strategies used for the ordinary differential equa-
ticle diameter, length of the extractor, solvent flow rate
tions (ODE) and analytical equations are discussed in this section.
and density of SC-CO2.
Step 2: Define the ordinary differential equations, Eqs. (2.1.1) and
3.1. Solution technique for the solution of ODE (2.1.5), for solute concentration in bulk fluid and solid
phases depending upon the number of divisions in the bed.
The Dormand–Prince method of Runge–Kutta family [27] is Step 3: Give time interval step size for the time span up to which
used for the solution of the model ordinary differential equations. the solution is to be computed.
It uses a sampling of slopes through an interval and takes a Step 4: Define the initial conditions of the solute concentrations,
weighted average to determine the right end point. The method c0 and q0 in fluid and solid phases, respectively.
is given as: Step 5: Compute the solution for the solute concentration, c in
  bulk fluid and solute concentration q in solid phase with
35 500 125 2187 11
ynþ1 ¼ yn þ k1 þ k3 þ k4 þ k5 þ k6 respect to time using Runge–Kutta (Dormand–Prince)
384 1113 192 6784 84
method.
where k1–k6 is given as: Step 6: Print the calculated values.
282 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287

1 10 MPa
[2.89, 1.84; 2.64]
0.9 18 MPa
0.9 [4.16, 2.66; 3.83] 20 MPa
[9.39, 6.01; 8.71] 26 MPa
0.8 [2.89, 1.84; 2.64] (Kim et al., 2007) 0.8
10 MPa (Tonthubthimthong et al., 2004)
Cumulative fraction, F (-)

[4.16, 2.66; 3.83] (Kim et al., 2007) 18 MPa (Tonthubthimthong et al., 2004)
0.7 20 MPa (Tonthubthimthong et al., 2004)
[9.39, 6.01; 8.71] (Kim et al., 2007) 0.7
26 MPa (Tonthubthimthong et al., 2004)

Cumulative fraction, F (-)

0.6
0.6
0.5

0.4 0.5

0.3
0.4
0.2
0.3
0.1

0 0.2
0 10 20 30 40 50
Time, θ (−) 0.1

Fig. 13. Combined effects of mass transfer coefficients and effective diffusivity at
0
P = 40 MPa and T = 313–353 K. Parameters in brackets are. [kf  105 (m/s);
kp  105 (m/s); De  109 (m2/s)]. 0 2 4 6 8 10 12 14
Time, θ (-)
1
[29.66, 19.00; 27.50]
Fig. 16. Effect of pressure on extraction yield at constant temperature 328 K and
[6.31, 4.03; 5.80]
0.9
[4.16, 2.66; 3.83]
flow rate 0.62 cm3/min.
[4.16, 2.66; 3.83] (Kim et al., 2007)
0.8 [6.31, 4.03; 5.80] (Kim et al., 2007)
[29.66, 19.00; 27.50] (Kim et al., 2007) 3.3. The solution technique and algorithm employed for the Reverchon
Cumulative fraction, F (-)

0.7 model using FEMLAB is as follows

0.6 The partial differential equations, Eqs. (2.1.1) and (2.1.5), are
solved by FEMLAB using Eqs. (2.1.3a) and (2.1.3b) as initial and
0.5
boundary conditions, respectively. Firstly, Eqs. (2.1.1) and (2.1.5)
0.4 are converted into a dimensionless form by introducing some var-
iable such as T ¼ Ks t, H ¼ Ku h.
0.3 Then Eq. (2.1.1) is written as:
  
0.2 @c @c Ap c
þ þ q ¼0 ð3:3:1Þ
@T @H V kp
0.1    
@q Ap e  c
¼ q ð3:3:2Þ
0 @T V 1e kp
0 10 20 30 40 50
Time, θ (−) Eq. (2.1.5) was solved in FEMLAB after adding a diffusion term to it,
which is a function of the Peclet number, defined in terms of an
Fig. 14. Effect of mass transfer coefficients and effective diffusivity at T = 323 K and effective diffusion coefficient. Eq. (3.3.2) becomes:
P = 10–40 MPa. Parameters in brackets are [kf  105 (m/s); kp  105 (m/s);
De  109 (m2/s)].

1 3
1.24 cm /min
3
0.62 cm /min
1 0.9 3
0.24 cm /min
3
0.24 cm /min (Tonthubthimthong et al., 2004)
0.9 0.8 3
0.62 cm /min (Tonthubthimthong et al., 2004)
3
0.8 0.7 1.24 cm /min (Tonthubthimthong et al., 2004)
Cumulative fraction, F (-)
Cumulative fraction, F (-)

0.7
0.6
0.6
0.5
0.5
0.4
0.4
333 K 0.3
0.3
323 K
0.2 308 K 0.2
308 K (Tonthubthimthong et al., 2004)
0.1 323 K (Tonthubthimthong et al., 2004) 0.1
333 K (Tonthubthimthong et al., 2004)
0
0
0 2 4 6 8 10 12
0 2 4 6 8 10 12
Time, θ (-)
Time, θ (-)
Fig. 15. Effect of temperature on extraction yield (F) at pressure = 20 MPa, solvent
flow rate = 0.62 m3/min. Fig. 17. Effect of solvent flow rate on extraction yield at P = 20 MPa and T = 328 K.
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287 283

0.9

0.8

Cumulative fraction, F (-)

0.7

0.6

0.5

0.4 1.85 mm
1.445 mm
0.3 1.015 mm
0.575 mm
0.2 1.850 mm (Tonthubthimthong et al., 2004)
1.445 mm (Tonthubthimthong et al., 2004)
0.1 1.015 mm (Tonthubthimthong et al., 2004)
0.575 mm (Tonthubthimthong et al., 2004)
0
0 2 4 6 8 10 12 14 16 18
θ (-)
Time,θ

Fig. 18. Effect of particle size on extraction yield at P = 20 MPa and T = 308 K.

Table A1 Eq. (3.3.3) can be rearranged to Eq. (3.3.4) so that it resembles the
Experimental conditions and process parameters for sage oil extraction [13]. standard form of convection diffusion equation in FEMLAB with
S. No. Parameter Magnitude transient analysis.
!   
1 P 9 MPa
@c 1 @2c Ap c @c
2 T 323 K  ¼ q  ð3:3:4Þ
3 Q 8.83 g/min @T Pe @H2 V kp @H
4 e 0.4
5 q 285 kg/m3 Eqs. (3.3.2) and (3.3.4) are solved using FEMLAB. The algorithm used
6 l 2.31  105 Pa s for the solution of the above partial differential equations is given
7 dp 0.75  103 m
as:
8 DL 0.24  105 m2/s
9 De 8.48  1012 m2/s Step 1: Open FEMLAB and choose Chemical Engineering Module,
10 Di 6  1013 Mass Balance, Convection and Diffusion, and Transient
11 u 0.455  103 m/s Analysis. The default variable name c is used for Eq. (3.3.4).
12 kf 1.91  105 m/s
Step 2: For the solute concentration in the solid phase, use the
13 dE 0.06 m
14 He 0.17 m Multiphysics menu, and choose the same equation. This
15 W 0.160 kg time, however, change the name of the variable to q. Add
16 kp 0.2 this equation to the problem.
Step 3: Draw a line from x = 0 to x = 1 and set the mesh to have 60
elements, or 121 nodes. For the two problems, then, there
are 242 degrees of freedom.
Table A2
Step 4: Ensure that the concentration equation for c is selected
Experimental conditions and process parameters for pepper extraction [20].
(under the Multiphysics menu). Under the Physics/Subdo-
S. No. Parameter Magnitude main settings, the following equation is displayed:
1 a 0.26
2 b 0.3 @c
3 dp 0.25  103 m dts þ r  ðDrcÞ ¼ R  u  rc
4 ap 24,000 m1
@t
5 Co 60 mg/g of pepper
Rearrange this equation in such a way that Eq. (3.3.4) is obtained.
6 P 475 bar
7 T 80 °C
This is done by setting the following parameters:
8 Qv 32.6 1/h
9 s 110.43 s
1 Ap
dts ¼ 1; D ¼ ; R¼  rate; u¼u
10 Re 2.652 Pe V
11 kf 0.6213  105 m/s
12 Bi 1.987 Also, select the artificial diffusion option and choose the Petrov–
13 kp 0.4446  105 m/s Galerkin method to consider the effect of additional diffusion in
14 De 39.08  109 m2/s the problem. Choose the Init tab and set the initial concentration
15 / 11.78 to zero.
16 K 5.70
Step 5: Under Options/Expressions/Subdomain Expressions,
define the following term for the equilibrium equation.
   rate ¼ ðq  c=kp Þ
@c @c Ap c 1 @2c
þ þ q ¼ ð3:3:3Þ Step 6: Under the Multiphysics menu, select the second equation
@T @H V kp Pe @H2
for q. The equation is of a similar nature and uses the fol-
lowing parameters:
284 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287

Table A3
Experimental conditions and process parameters in the model at various conditions for caffeine extraction [19].

P (MPa) T (K) q (kg/m3) Q  106 (kg/h) s (s) Re kf  105 (m/s) Bi kp  105 (m/s) De  109 (m2/s) / K
Effect of pressure
10 323 390 28.08 227.64 0.32 29.66 2.81 1.84 27.50 4.32 3181
20 323 790 28.08 462.95 0.23 6.31 2.83 2.66 5.80 1.87 784
40 323 920 28.08 542.56 0.19 4.16 2.83 6.01 3.83 1.44 229
Effect of temperature
40 313 960 28.08 566.15 0.18 2.89 2.80 1.84 2.64 1.04 564
40 323 920 28.08 542.56 0.19 4.16 2.83 2.66 3.83 1.44 229
40 353 820 28.08 843.59 0.22 9.39 2.85 6.01 8.71 2.91 210

b = 0.61, a = 0.84, ap = 11,538 m1, dp = 0.520 mm and C0 = 34.832 mg/g of green tea.

Table A4
Experimental conditions and process parameters in the model at various conditions for nimbin extraction [21].

P (MPa) T (K) q (kg/m3) Q  106 (m3/min) s (s) Re kf  105 (m/s) Bi kp  105 (m/s) De  109 (m2/s) / K
Effect of flow rate
20 328 759 1.24 213.8 0.84 2.40 6.63 1.03 1.09 34.9 59.56
20 328 759 0.62 427.7 0.42 1.91 5.28 0.93 1.09 15.3 59.56
20 328 759 0.24 1105 0.16 1.49 4.14 0.82 1.09 8.5 59.56
Effect of pressure
10 328 330 0.62 427.7 0.43 4.35 4.57 2.27 2.86 97.4 1321
18 328 727 0.62 427.7 0.42 2.09 5.20 1.02 1.21 43.8 79.06
20 328 794 0.62 427.7 0.42 1.91 5.28 0.93 1.09 39.7 59.56
26 328 825 0.62 427.7 0.39 1.53 5.47 0.73 0.83 31.1 43.67
Effect of temperature
20 308 875 0.62 427.7 0.38 1.41 5.53 0.67 0.77 11.1 21.51
20 323 791 0.62 427.7 0.41 1.86 5.29 0.90 1.06 14.9 36.76
20 333 729 0.62 427.7 0.42 1.96 5.26 0.95 1.12 15.7 56.67

b = 0.6143, a = 0.6142, ap = 3857 m1, dp = 0.6 mm and C0 = 0.2646 mg/g of neem kernel powder.

Table A5
Experimental conditions and process parameters for particle size in the model nimbin extraction [21].

P (MPa) T (K) dp (mm) qb (kg/m3) s (s) Re kf  105 (m/s) Bi kp  105 (m/s) / K ap a b

20 308 1.840 435 444.8 1.18 0.69 7.93 0.26 1.43 21.51 1203 0.629 0.629
20 308 1.445 452 427.7 0.92 0.80 7.52 0.32 2.20 21.51 1607 0.614 0.614
20 308 1.015 458 422.6 0.65 0.99 6.70 0.43 4.16 21.51 2308 0.610 0.610
20 308 0.575 458 422.6 0.37 1.45 5.54 0.69 11.9 21.51 4074 0.610 0.610

   
Ap v oidage
dts ¼ 1; D ¼ 0; R¼   rate
V 1  v oidage
Step 7: Set the boundary conditions for the first equation to have
concentration equal to 1.0 at the left-hand side and equal
to the convective flux at the right-hand side. Since the
solid material cannot flow in or out when it is in the
extractor, the boundary conditions for the second equation
are Insulation/Symmetry (i.e. no flux) at both ends.
Step 8: Define the parameters used in both Eqs. (3.3.2) and (3.3.4)
under the Options/Constants.
Step 9: The time dependent analysis is obtained using Solve/Sol-
ver parameters. Define the time interval for which the
solution is to be obtained and then press ‘=’ to solve the
problem.
Step
10: The
extract concentration in the fluid phase and the extract concen-
tration in the solid phase can be plotted by choosing Plot
Parameters. In the Line tab, select either c or q as required.
The variation of extract concentration with respect to time
can be obtained under Post-processing/Domain Plot Parameters
in Line tab.
3.4. The algorithm for solution technique of ODE employed for the Goto
model using MATLAB is as follows
Step 1: Define all the input parameters like bed void fraction, par-
ticle diameter and porosity, overall mass transfer coeffi-
cient, solvent flow rate and volume of the bed.
Step 2: Define the ordinary differential equations, Eqs. (2.2.3.4)
and (2.2.3.5), for solute concentration in bulk fluid and
solid phases respectively.
Step 3: Give time interval step size for the time span up to which
solution is to be computed.
Step 4: Define the initial conditions, Eqs. (2.2.3.6a) and (2.2.3.6b),
of the solute concentrations in both the phases.
Step 5: Compute the solution for the solute concentration in bulk
fluid and solid phases with respect to time using Runge–
Kutta (Dormand–Prince) method.
Step 6: Print the calculated values.
3.5. The algorithm for analytical solution technique employed for the
Goto model using MATLAB is as follows
The analytical solution of the model is obtained from Eq.
(2.2.3.9) with equilibrium adsorption constant, K, as a fitting
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287
parameter. The minimum average absolute relative deviation
(AARD) between the experimental extraction yield and predicted
extraction yield is chosen to determine the best fit value of K. Aver-
age absolute relative deviation (AARD) is given as:
Kim et al.
N
X
100
yieldcalc:  yieldexp:
To examine the reliability of the model of Goto et al. [8], the re-
AARDð%Þ ¼
N i¼1
yieldexp:
sults obtained are validated with the experimental data of Kim
The algorithm for the solution technique employed is as follows:
Step 1: Input the parameters like bed void fraction, particle diam-
eter, porosity, overall mass transfer coefficient, solvent
flow rate and volume of the bed.
Step 2: Define the equilibrium adsorption constant, K.
Step 3: Compute the constants used for the solution of Eq.
(2.2.3.9)
Step 4: Solve the model equation (2.2.3.9) to compute the extrac-
tion yield.
Step 5: Compare the predicted extraction yield with experimental
yield using AARD.
Step 6: If AARD is minimum, stop computing otherwise go to step
2.
4. Results and discussion
The properties of SFE and experimental data related to the
extraction of sage oil, pepper, caffeine from green tea, nimbin from
neem seed, used for model validation are given in Appendix A. The
results of the available models along with their inherent weak-
nesses are discussed in this section.
4.1. Results obtained from model proposed by Reverchon
Fig. 4 shows the variation of percentage normalized yield with
time for operating conditions of T = 323 K, P = 9 MPa, solvent flow
rate of 8.83 g/min and the particle size of 0.75 mm. The experimen-
tal data points are superimposed on the simulated results and it is
observed that the predicted results bear a close semblance to the
experimental results.
Fig. 5 shows a parity plot to compute the extent of error. From
the parity plot, it can be seen that the model represents 86% exper-
imental data within +9% errors. It can be concluded that the model
over predicts the experimental values slightly.
Fig. 6 shows the predicted percentage yield and experimental
data with respect to time for the model of Reverchon [4] on FEM-
LAB. From Fig. 6, it can be seen that the predicted results match
with the experimental data within an acceptable error limit.
Fig. 7 shows that the deviation of the calculated results from the
experimental data for the Reverchon [4] model on FEMLAB and it
can be seen that the results match with the experimental values
within an error band of +4% to 5%.
The results obtained by solving ODE and PDE are compared with
the experimental data of Reverchon [4] as shown in Fig. 8. The max-
imum deviation between these two methods of solution is found to
be about 28%. It should be noted that the solution obtained by solv-
ing linear and non-linear coupled partial differential equations
using FEMLAB which uses finite element method are expected to
be more accurate. The results clearly show that the computer pro-
gramme developed in the present work under MATLAB environ-
ment for ODE is moderately good as far as accuracy is concerned.
4.2. Results obtained from the model proposed by Goto
The model proposed by Goto et al. [8] is solved numerically and
analytically. The analytical solution proposed by Goto et al. [8] for
the above model is represented by Eq. (2.2.3.7). Fig. 9 is drawn to
show the comparison between dimensionless overall solute
285
concentration within the bulk fluid phase (x) and dimensionless
time (h) for numerical and analytical ODE solutions.
4.2.1. Validation of model proposed by Goto with experimental data of
et al. [19]. To validate the model, presented by Eqs. (2.2.3.4) and
(2.2.3.5) along with initial and boundary conditions given by Eqs.
(2.2.3.6a) and (2.2.3.6b), respectively, The cumulative extraction
yield computed using Eq. (2.2.3.9) and Fig. 10, is graphically repre-
sented by plotting the cumulative fraction of the solute extracted
(F) on the vertical axis and the dimensionless time (h) on the hor-
izontal axis along with experimental data of [19]. From Fig. 10, it is
evident that model predictions match excellently with that of
experimental data.
To compute the extent of error, a parity plot is drawn in Fig. 11
and it is clear that the model predictions are within +15% to 2% of
the experimental values. It can be concluded that the model
slightly over predicts the experimental values.
The model proposed by Goto et al. [8] considers the transfer of
solute from solid matrix to bulk liquid phases through different
seamless steps. The solute in the solid phase is transferred to the
supercritical fluid present in the pores of the solid matrix, which
is then transported to the liquid bulk. Various curves are plotted,
as given below, to effectively understand the depletion of the sol-
ute from the solid phase and depict how it is transferred to the
supercritical fluid in the pores.
4.2.2. Variation of dimensionless average solute concentration profile
in solid phase with time
The solute concentration profile (xs) in the solid phase of the
particle with respect to the dimensionless time (h) is shown in
Fig. 12. It can be seen that the solute concentration is decreasing
with respect to time due to the diffusion of solute from solid phase
to the pores and then to bulk fluid by convective mass transfer.
4.2.3. Effect of model parameters on cumulative extraction yield
Once the model has been validated using the experimental data
of Kim et al. [19] for caffeine extraction, it can now be used to study
the effect of the operating parameters on the extraction yield, as
shown below. The effect of various system parameters such as
effective intra-particle diffusion coefficient (De), external mass
transfer coefficients (kf), and overall mass transfer coefficients
(kp) on the cumulative extraction yield were studied using the
model of Goto et al. [8] and then validated using the experimental
data for caffeine extraction by Kim et al. [19].
4.2.3.1. Effect of temperature on mass transfer coefficient and effective
diffusivity. Fig. 13 shows the combined effects of mass transfer
coefficients coupled with effective intra-particle diffusion coeffi-
cient on cumulative extraction yield (defined in bulk liquid) for a
pressure of 40 MPa and temperature ranging from 313 to 353 K.
It is noted that for a given dimensionless time, the cumulative
extraction yield (F) increases with increase in temperature. The
model predictions match excellently with experimental data of
Kim et al. [19]. The observation can be explained by the fact that
the mass transfer coefficients increases with rise in temperature
as a result of high diffusivity of caffeine in SC-CO2 at higher tem-
peratures. Thus the rate of mass transfer of solute to bulk liquid
phase increases with increase in temperature.
4.2.3.2. Effect of pressure on mass transfer coefficient and effective
diffusivity. Fig. 14 shows that for a given dimensionless time, the
cumulative extraction yield (F) increases with increase in pressure.
The observation can be explained by the fact that the solubility of
286 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287
caffeine in SC-CO2 increases at higher pressures despite the fact
that mass transfer coefficients decreases with rise in pressure. Con-
sequently, the mass transfer resistance increases resulting in the
decrease of the extraction rate.
4.2.4. Effect of operating parameters on cumulative extraction yield
In this section, the effects of temperature, pressure, solvent flow
rate and particle size on the cumulative extraction yield are investi-
gated using the data obtained for nimbin extraction from neem seed.
4.2.4.1. Effect of temperature on cumulative extraction yield. Fig. 15
shows the effect of temperature on cumulative extraction yield
at pressure of 20 MPa and a solvent flow rate of 0.62 cm3/min.
Fig. 15 shows that for a given dimensionless time, the cumulative
extraction yield (F) increases with increase in temperature. How-
ever, in this respect a reversal trend is observed from experimental
investigation due to Tonthubthimthong et al. [21].
4.2.4.2. Effect of pressure on cumulative extraction yield. Fig. 16
shows the influence of pressure on the extraction yield at constant
temperature of 328 K and solvent flow rate of 0.62 cm3/min. For a
given dimensionless time (h), the cumulative extraction yield (F)
increases with increase in pressure. The observation can be ex-
plained by the fact that with the increase in pressure from 10 to
26 MPa, the solubility increases rapidly. It has a positive effect on
extraction process. However, the effective diffusivity and mass
transfer coefficients decreased with higher pressure. The mass
transfer resistance increased resulting in a decrease in the extrac-
tion rate showing negative effect of pressure on the extraction pro-
cess. At low pressures, the positive effects on the extraction
process overcome the negative effects and hence the cumulative
extraction yield (F) increases with increasing pressure.
4.2.4.3. Effect of solvent flow rate on cumulative extraction yield.
Fig. 17 shows the effect of flow rate on extraction yield at
P = 20 MPa and T = 328 K. For a given dimensionless time, the
cumulative extraction yield, F increases with increase in solvent
flow rate. This observation can be explained by the fact that with
the increase in SC-CO2 flow rate from 0.24 to 1.24 cm3/min, the
mass transfer coefficient increases and thus the solute transported
per unit time to bulk liquid increases leading to higher ‘F’ values on
increasing the flow rate for a given time.
4.2.4.4. Effect of particle size on cumulative extraction yield. Fig. 18
shows the effect of particle size on nimbin extraction at 308 K,
20 MPa and a flow rate of 0.62 cm3/min. For a given dimensionless
time (h), the cumulative extraction yield (F) increases with de-
crease in particle size. The observation can be explained by the fact
that smaller particles offer higher interfacial area for mass transfer.
When the particle size decreases, the interface area increases lead-
ing to higher mass transfer. This, in turn, the ‘F’ value increases
with time.
5. Conclusions
5.1. Supercritical fluid extraction is a green and effective extrac-
tion process which produces an extract with little or no
residual solvent, superior purity and high yield under lower
operating cost as compared to solvent-based methods. A
large number of useful species can be extracted using SFE
technique.
5.2. The models proposed by Goodarznia et al. [14], Goto et al.
[8,12], Marrone et al. [28], Reverchon [13], and Sovova
[9,10] are studied in the present investigation to compute
the extraction yield. Out of these models, the model based
on differential mass balances proposed by Reverchon [13]
and the local adsorption equilibrium model proposed by
Goto et al. [8] are found to be best as these address the inter-
nal diffusion and the dynamic behavior of the extraction
process incorporating intra-particle diffusion as well as
external mass transfer of SFE respectively.
5.3. The model of Reverchon [13] fits 86% experimental data
within +9% error when MATLAB environment was used for
solving the model equations and +4% to 5% error when it
was modeled using in FEMLAB environment.
5.4. The numerical solution provided by FEMLAB is found to be
better than the solution obtained by MATLAB environment
showing a maximum deviation of 28%.
5.5. The model of Goto et al. [8] fits the experimental data given
by Kim et al. [19], Skerget and Knez [20], and Tonthubthim-
thong et al. [21] within an error band of +10% to 2%.
5.6. At high temperature, pressure and solvent flow rate for
smaller particle size, the highest yield of the product is
obtained.
Appendix A
The data used for the model validation and discussion of para-
metric effects is given in this section (see Tables A.1–A.5).
References
[1] M. Mukhopadhyay, Natural Extracts using Supercritical Carbon Dioxide, CRC
Press, Boca Raton, FL, 2000. pp. 1–9.
[2] P. Munshi, S. Bhaduri, Supercritical CO2: a twenty-first century solvent for the
chemical industry, Curr. Sci. 97 (2009) 63–72.
[3] D. Ghonasgi, S. Gupta, K.M. Dooley, F.C. Knop, Measurement and modeling of
supercritical carbon dioxide extraction of phenol from water, J. Supercrit.
Fluids 4 (1991) 53–59.
[4] E. Reverchon, M. Poletto, Mathematical modelling of supercritical CO2
fractionation of flower concretes, Chem. Eng. Sci. 51 (1996) 3741–3753.
[5] S.N. Naik, H. Lentz, R.C. Maheshawari, Extraction of perfumes and flavours
from plant materials with liquid carbon dioxide under liquid–vapor
equilibrium conditions, Fluid Phase Equilib. 49 (1989) 115–126.
[6] M. Spiro, M. Kandiah, Extraction of ginger rhizome: partition constants and
other equilibrium properties in organic solvents and in supercritical carbon
dioxide, Int. J. Food Sci. Technol. 25 (1990) 566–575.
[7] E. Reverchon, G. Donsi, L.S. Osseo, Modeling of supercritical fluid extraction
from herbaceous matrices, Ind. Eng. Chem. Res. 32 (1993) 2721–2726.
[8] M. Goto, M. Sato, T. Hirose, Extraction of peppermint oil by supercritical carbon
dioxide, J. Chem. Eng. Jpn. 26 (1993) 401–407.
[9] H. Sovova, Rate of the vegetable oil extraction with supercritical CO2—I.
Modelling of extraction curves, Chem. Eng. Sci. 49 (1994) 409–414.
[10] H. Sovova, J. Kucera, J. Jez, Rate of the vegetable oil extraction with
supercritical CO2—II. Extraction of grape oil, Chem. Eng. Sci. 49 (1994) 415–
420.
[11] H. Sovova, J. Jez, M. Bartlova, J. Stastova, Supercritical carbon dioxide
extraction of black pepper, J. Supercrit. Fluid 8 (1995) 295–301.
[12] M. Goto, B.C. Roy, T. Hirose, Shrinking-core leaching model for supercritical-
fluid extraction, J. Supercrit. Fluids 9 (1996) 128–133.
[13] E. Reverchon, Mathematical modelling of supercritical extraction of sage oil,
AIChE J. 42 (1996) 1765–1771.
[14] I. Goodarznia, M.H. Eikani, Supercritical carbon dioxide extraction of essential
oils: modeling and simulation, Chem. Eng. Sci. 53 (1998) 1387–1395.
[15] E. Reverchon, S.L. Osseo, Supercritical CO2 extraction of basil oil:
characterization of products and process modeling, J. Supercrit. Fluids 7
(1994) 185–190.
[16] M. Goto, B.C. Roy, A. Kodama, T. Hirose, Modeling of supercritical fluid
extraction process involving solute–solid interaction, J. Chem. Eng. Jpn. 31
(1998) 171–177.
[17] M. Perrut, J.Y. Clavier, M. Poletto, E. Reverchon, Mathematical modelling of
sunflower seed extraction by supercritical CO2, Ind. Eng. Chem. Res. 36 (1997)
430–435.
[18] W. Wu, Y. Hou, Mathematical modeling of extraction of egg yolk oil with
supercritical CO2, J. Supercrit. Fluid 19 (2001) 149–159.
[19] W. Kim, J. Kim, S. Oh, Supercritical carbon dioxide extraction of caffeine from
Korean green tea, Seper. Sci. Technol. 42 (2007) 3229–3242.
[20] M. Skerget, Z. Knez, Modelling high pressure extraction processes, Comput.
Chem. Eng. 25 (2001) 879–886.
[21] P. Tonthubthimthong, P.L. Douglas, S. Douglas, W. Luewisutthichat, W.
Teppaitoon, L. Pengsopa, Extraction of nimbin from neem seeds using
 A. Rai et al. / International Journal of Heat and Mass Transfer 72 (2014) 274–287
supercritical CO2 and a supercritical CO2–methanol mixture, J. Supercrit. Fluid
30 (2004) 287–301.
[22] N.R. Bulley, M. Fattori, Supercritical fluid extraction of vegetable oil seeds, J.
Am. Oil Chem. Soc. 61 (1984) 1362–1365.
[23] M.A.A. Melreles, G. Zahedi, T. Hatami, Mathematical modeling of supercritical
fluid extraction for obtaining extracts from Vetiver root, J. Supercrit. Fluids 49
(2009) 23–31.
[24] D. Mongkholkhajornsilp, S. Douglas, P.L. Douglas, A. Elkamel, W. Teppaitoon, S.
Pongamphai, Supercritical CO2 extraction of nimbin from neem seeds – a
modelling study, J. Food Eng. 71 (2005) 331–340.
[25] H. Peker, M.P. Srinivasan, J.M. Smith, B.J. McCoy, Caffeine extraction rates from
supercritical carbon coffee beans with dioxide, AIChE J. 38 (1992) 761–770.
[26] C. Tan, D. Liou, Modeling of desorption at supercritical conditions, AIChE J. 35
(1989) 1029–1031.
[27] J.R. Dormand, P.J. Prince, A family of embedded Runge–Kutta formulae, J.
Comput. Appl. Math. 6 (1980) 19–26.
[28] C. Marrone, M. Poletto, E. Reverchon, A. Stassi, Almond oil extraction by
supercritical CO2: experiments and modeling, Chem. Eng. Sci. 53 (1998) 3711–
3718.
[29] A.K.K. Lee, N.R. Bulley, M. Fattori, A. Meisen, Modelling of supercritical carbon
dioxide extraction of canola oilseed in fixed beds, Journal of American Oil
Chemist Society 63 (1986) 921–925.
[30] O.J. Catchpole, J.B. Gray, B.M. Smallfield, Near-critical extraction of sage, celery,
and coriander seed, Journal of Supercritical Fluids 9 (1996) 273–279.
[31] B.C. Roy, M. Goto, T. Hirose, Extraction of ginger oil with supercritical carbon
dioxide: experiments and modeling, Industrial and Engineering Chemistry
Research 35 (1996) 607–612.
[32] M. Akgun, N.A. Akgun, S. Dincer, Extraction and Modeling of Lavender Flower
Essential Oil Using Supercritical Carbon Dioxide, Industrial and Engineering
Chemistry Research 39 (2000) 473–477.
[33] J.M. del Valle, J.M. Aguilera, An Improved Equation for predicting the solubility
of vegetable oils in supercritical CO2 Industrial and Engineering Chemistry
Research 27 (1988) 1551–1553.
287
[34] E. Reverchon, C. Marrone, Supercritical extraction of clove bud essential oil:
isolation and mathematical modeling, Chemical Engineering Science 52 (1997)
3421.
[35] E. Reverchon, G. Della Porta, G. Lamberti, Modelling of orange flower concrete
fractionation by supercritical CO2, Journal of Supercritical Fluids 14 (1999)
115–121.
[36] M. Perrut, J.Y. Clavier, M. Poletto, E. Reverchon, Mathematical modelling of
sunflower seed extraction by supercritical CO2, Industrial and Engineering
Chemistry Research 36 (1997) 430–436.
[37] W. Wu, Y. Hou, Mathematical modeling of extraction of egg yolk oil with
supercritical CO2, Journal of Supercritical Fluids 19 (2001) 149–159.
[38] M.J. Cocero, J. Garcia, Mathematical model of supercritical extraction applied
to oil seed extraction by CO2 + saturated alcohol- I, Desoption Model. Journal of
Supercritical Fluid 20 (2001) 229–243.
[39] E.M.C. Reis-Vasco, J.A.P. Coelho, A.M.F. Palavra, C. Marrone, E. Reverchon,
Mathematical modelling and simulation of pennyroyal essential oil
supercritical extraction, Chemical Engineering Science 55 (2000) 2917–
2922.
[40] I. Zizovic, M. Stamenic, A. Orlovic, D. Skala, Supercritical carbon dioxide
extraction of essential oils from plants with secretory ducts: Mathematical
modelling on the micro-scale, Journal of Supercritical Fluids 39 (2007) 338–
346.
[41] S.M. Ghoreishi, S. Sharifi, Modeling of supercritical extraction of mannitol from
plane tree leaf, Journal of Pharmaceutical and Biomedical Analysis 24 (2001)
1037–1048.
[42] M.A.A. Melreles, G. Zahedi, T. Hatami, Mathematical modeling of supercritical
fluid extraction for obtaining extracts from vetiver root, Journal of
Supercritical Fluids 49 (2009) 23–31.
[43] J. Stastova, J. Jez, M. Bartlova, H. Sovova, Rate of the vegetable oil extraction
with supercritical CO2— III. Extraction from sea buckthorn, Chemical
Engineering Science. 51 (1996) 4347–4352.