You are on page 1of 13

J. Sep. Sci.

2002, 25, 477–489 Al-Jabari 477

Maher Al-Jabari Kinetic models of supercritical fluid extraction


Department of Chemical The scope of this paper is to review the theoretical basis for modeling various applica-
Engineering, An-Najah National
tions of supercritical fluid extractions (SFE), including industrial processes and analy-
University, Nablus,
tical tests. Most of the models for SFE are concerned mainly with one stage SFE in
P.O.Box 7-707, Palestine
packed beds. However, in SFE tests, an initial static stage (with no flow) is usually
implemented for the purpose of achieving higher yields with pre-extractions. The
importance of accounting for the initial static stage is illustrated and a basis for mode-
ling such a stage is presented.
Theoretically, SFE models are similar to those for conventional transport processes
in that they are based on differential mass balance for solute in both phases and
invoke a rate equation (based on desorption kinetics or mass transfer) as well as a
thermodynamic equation to describe the interactions at the interface. Local equili-
brium theory can be implemented using linear or non-linear isotherms. A conceptual
scheme for developing SFE models from these general equations is suggested.
The similarity between modeling SFE processes and reversible adsorption/desorption
processes is demonstrated and shown to provide a useful basis for developing new
models for the SFE process. The concept of analogy is presented and used to deve-
lop a new model for SFE in packed beds from a previous deposition model.
Recent models for SFE tests with CSTR flow field are presented, including a mass
transfer model with local linear equilibrium and a model based on the Langmuir iso-
therm. A new equilibrium CSTR model is also presented and compared to these two
models.

Key Words: Modeling; Extraction; Supercritical fluid; Industrial processes; Analytical


tests;
Received: November 28, 2001; revised: January 29, 2002; accepted: February 2, 2002

1 Introduction and HPLC. Then, the interest developed rapidly into appli-
cations in SF separations, and particularly into supercriti-
A fluid is considered to be supercritical when its tempera- cal fluid extraction (SFE) [3].
ture is raised above its critical temperature and its pres-
Conventional extraction methods using liquid solvents are
sure raised above its critical pressure. Under these condi-

Review
slow and usually require several hours or sometimes even
tions, the fluid characteristics combine those for liquids
days to effect complete extractions. They also result in
and for gases. It is considered to be similar to liquids in
dilute extracts, thus necessitating concentration of the
having high density and thus high solvent strength for var-
extracted solute in the liquid solvent, and they may not
ious solutes, which largely increases with increasing pres-
result in quantitative recovery of target solutes. In certain
sure [1]. However, since supercritical fluids (SF) are not
cases, solvents are associated with hazardous effects.
liquefied under pressure, they retain many attractive char-
Moreover, an ideal extraction method should be rapid,
acteristics of the gases such as low viscosity resulting in
yield quantitative recovery of target solute without degra-
high penetration capabilities and in high associated mass
dation, and the solvent should be easily separated from
transfer coefficients. All these factors have given SF great
the solute. All these facts supported the development of
potential for a wide range of processes [2]. Over the last
alternative extraction methods. Supercritical fluid extrac-
20 years, there has been an intense interest in the use of
tion (SFE) technology offered scientists attractive fea-
SF in separation science [3]. SF were first used in chro-
tures overcoming many of the limitations of conventional
matography as alternative media to gases or liquids in GC
extractions. Nowadays, SFE is used in many applications
Correspondence: Prof. Maher Al-Jabari, Department of [5 – 21], with the extraction of natural products [4] being of
Chemical Engineering, An-Najah National University, Na- major economic importance. This includes recovery of
blus, P.O.Box 7-707, Palestine. caffeine from coffee beans, extraction of nicotine from
E-mail: aljabari@mail.com tobacco wastes [6], isolation of essential oils from plants
Fax: +972 9-2387982 [7], as well as the extraction of carotene and lipids from

i WILEY-VCH Verlag GmbH, 69469 Weinheim 2002 1615-9306/2002/0806–0477$17.50+.50/0


478 Al-Jabari J. Sep. Sci. 2002, 25, 477–489

pressed palm oil fibers [7], extractions of chemicals from the basis of analogies between particle deposition and
forest products and wood pulp [8 – 11]. In addition, SFE molecular adsorption.
has been used as analytical tool to quantify metal content In this paper, the general equations that govern all kinetic
in liquid and solid samples using in-situ complexation with models are first introduced, followed by presentation of a
a suitable chelating agent [15 – 21]. New developments conceptual scheme for developing particular SFE models.
include fractionation of products, dyeing of fibers, treat- Then, various kinetic SFE models are presented and dis-
ment of contaminated soils [13, 14], production of pow- cussed.
ders in micron and sub-micron range, and reactions in or
with supercritical fluid [2]. SFE is established on the indus-
trial scale and as an analytical tool [15 – 17, 22, 23],
although its industrial scale application is still limited due 2 General equations
to the need for high-pressure equipment.
The general equations that govern the process of SFE are
similar to those that govern many mass transport opera-
This wide range of applications have motivated research-
tions involving solids and fluids such as leaching and
ers to develop models that would describe the various pro-
adsorption processes. Those equations include two differ-
cesses and provide a tool for evaluating the performance
ential solute mass balances for the solvating phase and
of SFE and optimizing operating conditions. There are
for the treated phase, in addition to a kinetic equation that
various models available in the literature for SFE [22 – 34].
describes the rate of solute transfer between the two
In general, modeling SFE involves two different theoreti-
phases. At the interface, equilibrium may be assumed to
cal approaches, viz. kinetic and thermodynamic modeling.
exist, and thus a thermodynamic equation is needed. In
Thermodynamic modeling is concerned with estimating
this section, equations are developed for SFE processes
equilibrium states [13, 16, 31, 32], i. e. determining the
in general without being applied to a certain system. The
solubility of solutes in SF which is beyond the scope of this
subsequent section illustrates how they can be used for
paper. Kinetic models are concerned with following the
developing certain models for certain systems. Equations
progress of extraction with time, or in other words, the
presented here are restricted to cases where the solid
dynamics of SFE process. In this paper, SFE modeling
matrix is inert and non-swelling.
refers to kinetic models.

The theoretical basis underlying all kinetic SFE models is 2.1 Solute mass balance for supercritical phase
established in this paper. The paper illustrates how pre-
For an SFE in a continuous contact apparatus, a differen-
vious models can be derived from such a basis and how
tial shell mass balance for solute in the supercritical phase
new models can be developed. Most of the existing SFE
yields the following partial differential equation (PDE):
models are based on the mass transfer approach [23 –
30], while other models are basically empirical [7, 16, 33,
34]. A model based on Langmuir kinetics has recently qc qc q2 c ð1 ÿ eÞ
been developed by the author [22]. Many of the models þU ÿ Da ¼ÿ aNA ð1Þ
qt qx qx 2 e
were concerned with operation in packed beds [24 – 30];
the author had developed two models for SFE as an ana- Where c is the concentration of solute in the bulk of the
lytical tool [22, 23]. supercritical phase, U is the superficial velocity of SF
through the process vessel, Da is the axial dispersion coef-
As a transport process, SFE is similar to conventional ficient, x is the linear position in the vessel measured from
extraction and reversible adsorption/desorption pro- SF inlet, e is the volume fraction of SF in the vessel, t is the
cesses. This similarity justifies drawing an analogy time, a is the specific surface area of the solid, and NA is
between SFE and conventional reversible adsorption pro- the flux of solute from solid towards SF expressed per unit
cesses, which would enable researchers to establish new surface area of the solid. Equation (1) is referred to as the
models for SFE. This approach is demonstrated in this axial dispersion equation and usually used to describe the
paper. Previous adsorption models [35 – 41] were based transport of a solute in a fluid flowing through continuous
mainly on mass transfer approaches similar to those of contact apparatus (like packed beds). This equation is
SFE models. Thus, the analogy between these two pro- valid for SFE in industrial processes or their simulations
cesses can lead to new SFE models. Such analogy is not on the lab scale.
limited to molecular adsorption; the paper also highlights In analytical tests, the extraction vessel is usually small,
the analogy with models for particle deposition developed having a small tube inlet and a small tube outlet. Such a
by the author and co-workers [42 – 44] and by other design creates flow expansion upon entrance of SF into
researchers [45, 46]. In fact, those previous particle the test vessel and contraction upon leaving. Both of these
deposition models were developed, to certain extent, on effects result in considerable flow mixing and hence the
J. Sep. Sci. 2002, 25, 477–489 Kintic models of supercritical fluid extraction 479

flow field in the vessel is considered to be similar to that (mass transfer). In such cases, diffusion and/or internal
for a continuous stirred tank reactor (CSTR). The differen- flow resistance for the fluid in the pores may be encoun-
tial shell mass balance on solute in the supercritical phase tered. These effects could be lumped together in one term
for a CSTR system yields the following simplified ordering accounting for internal solute transport where a modified
differential equation (ODE): internal mass transfer coefficient is used.
Consequently, the type of expression used for such a
Q dc ð1 ÿ eÞ sink-source term would depend on what step is the rate
cþ ¼ÿ aNA ð2Þ determining one. For certain systems, the rate-limiting
eV dt e
step may be switched from one mechanism into another
Where Q is the volumetric flow rate of SF through an as the SFE proceeds with time (as in drying processes).
extraction vessel of volume (V ). For example, de Franc et al. [7] demonstrated that SFE
extraction curves of carotene and lipids from pressed
2.2 Solute mass balance for solid phase palm oil fibers are described by three steps in analogy to
drying curves: the first linear portion is a period of constant
For both types of operations, differential mass balance for extraction rate and is characterized by convective mass
solute in the solid phase provides the following equation: transfer between the solid surface and the fluid phase.
The second part of the extraction curve represents the
qq falling rate period; at this step both convection and diffu-
qs ¼ ÿaNA ð3Þ sion in the solid must be considered. For the third step,
qt
which is a diffusion-controlled rate period, diffusion in the
Where q is the mass fraction of solute in the solid and qs is solid controls the rate of mass transfer. In another publica-
the density of the solids. The flux, NA, is a sink term here tion [22], the author demonstrated that for some systems,
whereas in Eqs. (1) and (2) it is a source term. Equation an SFE model based on a desorption step provides a bet-
(3) is valid for the two types of applications, i. e. for indus- ter fit for experimental SFE curves than typical mass
trial and analytical vessels. For analytical tests, exact deri- transfer models.
vatives are used in Eq. (3) since q and c are only time
In general, there are two types of approaches for account-
dependent. In certain applications, the mass fraction of
ing for the sink-source term (NA). These include (1) the
solute in the solid phase may vary with the internal posi-
mass transfer approach, which can be based on internal,
tion inside the solids. In such cases, an internal diffusion
external, or overall mass transfer resistance, and (2) des-
model would have to be solved for the local mass fraction
orption kinetics. All these approaches are applicable for
(q r) within the solid (see Section 2.3.1.2), from which the
SFE is both industrial and analytical vessels, depending
average bulk mass fraction of solute in solids (q) is deter-
on the characteristics of the two phases (SF and solids).
mined.
The following sub-sections establish the rate equations
for these various cases.
2.3 Solute transfer across the interface
(sink-source term) 2.3.1 Mass transfer approach
The rate of transfer of solute across the interface between 2.3.1.1 External mass transfer
the solid and SF phases, i.e. the sink-source term (NA) in
Eqs. (1) – (3), is governed by the physics of the transport As in conventional mass transfer operations, the flux of
process. There are three steps, which may control such solute from the interface with a concentration c i into the
transport processes. These are bulk of SF with a concentration c is given by

1. Solute diffusion inside the solid particles (internal


mass transfer). NA ¼ kc ðc i ÿ cÞ ð4Þ
2. Solute dissolution from solid into SF phase at the
Where k c is the mass transfer coefficient in the SF phase.
solid-SF interface (desorption kinetics).
Several correlations for mass transfer coefficients for var-
3. Solute convection in SF from the interface to the bulk ious types of solid geometries and flow conditions can be
of SF (external mass transfer). found in books on mass transfer.
In addition to these, other mechanisms may be involved in
more complicated systems. For example, when the poros-
ity of the solid is pronounced, the rate of outward solute 2.3.1.2 Internal mass transfer
transport from the solid is governed by osmotic flow in the The rate of solute transfer across the solid particles may
pores (fluid mechanics) as well as diffusion in the fluid that be obtained from internal diffusion models. Such models
fills the voids in addition to diffusion in the solid phase are based on mass balance of solute across a differential
480 Al-Jabari J. Sep. Sci. 2002, 25, 477–489

shell within a given single solid particle. The solid particles NA ¼ kco ðc  ÿ cÞ ð8Þ
may be considered as individual spherical particles [34],
semi-infinite bodies [12, 6] when the diffusion path is using the SF variables, or
much shorter than the characteristic dimension of the par-
ticles, or as a slab when the solid particles are thin and
wide. The resulting mathematical equations are similar to NA ¼ kso ðq ÿ q  Þ ð9Þ
those describing conductive heat transfer through similar
geometries for which solutions are available in heat con- Using the solid variables, where k co and k so are overall
duction literature. mass transfer coefficients, c * is the concentration of
solute in the SF that is in equilibrium with the average
As an example, the general equation for diffusion in sphe-
(bulk) mass fraction of solute in the solid phase and q * is
rical particles is presented. For a bed of homogeneous
the mass fraction of solute in the solid that is in equilibrium
spherical particles, a solute mass balance on a differential
with the average (bulk) solute concentration in the SF
shell of thickness (dr) within a single particle leads to the
phase.
following PDE:

 2  2.3.2 Desorption kinetics


qqr q qr 2 qqr
¼ Ds þ ð5Þ In most of previous SFE models, desorption kinetics was
qt qr 2 r qr
assumed to be instantaneous, and thus the local equili-
Where r is the radial position within the sphere, Ds is the brium concept was implemented at the interface between
diffusion coefficient in the solid and q r is the local mass the two phases. However, for some systems where the
fraction of solute in the solid particle. The solution of interactions between the solute and the solid surface are
Eq. (5) requires two boundary conditions and one initial considerable, such a step could be the rate-limiting step
condition (see Section 2.5). and thus there is a need to describe such kinetics by a cer-
tain rate equation based on molecular exchange at the
After solving Eq. (5) for q r in the solid, the rate of transfer surface. The mechanism of molecular transport can be
of solute outward across the surface of solid particle (with visualized as follows: the molecules are usually attached
radius r0) can be obtained from the solution of Eq. (5) and to a solid surface by secondary forces like van der Waals
using Fick’s law. This affords forces. As a result of vibrational energy of the adhered
(adsorbed) molecules, part of these molecules will have
qq r sufficient energy to overcome the bonding energy with the
aNA ¼ ÿqs Ds ð6Þ
qr solid surface and thus will leave the surface. On the other
hand, those molecules that are already in the supercritical
where the derivative is estimated at r0. phase are in continuous random motion with rotational,
As an alternative to this approach, a mass flux expression kinetic, and vibrational energies and thus part of them will
may be used for solute transport in the solid phase similar be continually striking the surface and consequently a
to that used in conventional leaching processes. The flux fraction of these molecules will adhere back onto the sur-
equation is face. Hence, the process of solute transport is a reversible
process in a fashion similar to conventional molecular
adsorption processes. In such a case, the overall solute
NA ¼ ks ðq ÿ q i Þ ð7Þ desorption process consists of two elementary processes:
(1) desorption of solute molecules from the solid surface
Where k s is the mass transfer coefficient in the solid phase
into the supercritical phase and (2) re-adsorption of solute
(which may account for the various types of internal resis-
molecules onto the surface at the available free sites. An
tances as mentioned above), q s is the mass fraction of
equation for desorption kinetics consists of two terms for
solute in the solid at the interface with SF (which is in equi-
these two elementary steps.
librium with c i when the desorption kinetics is negligible
according to local equilibrium theory). Langmuir kinetics was used in chemical reaction literature
as the rate equation to describe the adsorption or desorp-
tion processes [48, 49]. In analogy to the molecular
2.3.1.3 Overall mass transfer resistance adsorption process, Langmuir kinetics was also used, in
The mass flux equation using the bulk concentration (c *) recent studies, to describe the rate of particle deposition/
rather than the interfacial concentration may be preferred detachment onto and from solid surfaces [43, 45].
for cases when the equilibrium isotherm is considered to Recently, the author developed an SFE model based on
be linear. Using the overall mass transfer coefficient, the Langmuir kinetics, which was the first attempt to base
flux is given by SFE models on desorption kinetics [22].
J. Sep. Sci. 2002, 25, 477–489 Kintic models of supercritical fluid extraction 481

According to Langmuir kinetics, the rate of the elementary state-of-the-art equations of state are still not capable of
molecular desorption is basically proportional to the num- adequately and quantitatively predicting SF behavior [25,
ber of molecules attached to the solid surface, or their sur- 14]. Details of such thermodynamic modeling are beyond
face concentration expressed as mass fraction, q. The the scope of this paper as mentioned earlier.
rate of re-adsorption onto the solid surface is proportional The equilibrium concentration of solute in SF is equal to
to the rate of molecular collisions with the surface, which the solubility (S ) when the solute has no affinity to the solid
is proportional to solute concentration in the supercritical matrix. In such a case, the concentration in the SF phase
phase, c. The surface capacity for a solute can not exceed is independent of the mass fraction in the solids:
that corresponding to a monomolecular layer (q max) since
the primary attraction forces that attach the molecules to
the surface diminish rapidly with the distance from the sur- c ¼ S m f ðqÞ ð12Þ
face. Hence, only those molecules colliding with unoccu-
pied sites on the surface will be adsorbed, and thus the Solute is attracted to and adsorbed onto the solid phase
rate of adsorption is also proportional to the fraction of through physical forces, and an equilibrium isotherm is
unoccupied sites. Accordingly, Langmuir kinetics estab- required to describe these interactions. Various types of
lishes that the net rate of desorption in SFE is given by the equations including linear and non-linear equilibrium iso-
difference between these two competing rates as follows therms can be used to describe the equilibrium distribu-
tion of solute between the solid phase and the SF. The lin-
  ear equilibrium isotherm is given by
q
NA ¼ kd q ÿ ka c 1 ÿ ð10Þ
qmax
q ¼ Kc ð13Þ
Where k d and k a are the desorption and adsorption coeffi-
cients, respectively. Both k a and k d are dependent on the where K is the equilibrium distribution coefficient. Non-lin-
temperature and the molecular energy according to the ear equilibrium isotherms may also be used, such as the
Arrhenius equation. Langmuir isotherm, which can be obtained from Eq. (9) by
setting the rate of desorption to zero (equilibrium). Upon
rearranging, one obtains:
2.3.4 Instantaneous solute transfer
In some practical SFE applications, the two mass transfer
1 1 1 1
resistances (internal and external) can be negligible and ¼ þ ð14Þ
desorption kinetics can be extremely rapid. In such cases, q K c qmax
the concentration in the SF phase (c) reaches equilibrium
instantaneously. Thus, modeling SFE becomes simpler Where K (the equilibrium constant) is defined as the ratio
since the rate equation (sink-source term) is not needed in (k a/k d). The dimensionless equilibrium constant is defined
this case, while Eq. (3) is transformed by estimating the as follows:
derivative of q from the applicable equilibrium relation
(see Section 2.4 below). Then, Eq. (3) is replaced by ð1 ÿ eÞ
Ke ¼ qs K ð15Þ
e
qq dq qc
¼ ð11Þ
qt dc qt
2.5 Initial and boundary conditions
Solving various combinations of the above general differ-
2.4 Local equilibrium concept ential equations requires various combinations of bound-
When desorption kinetics is not rate-limiting, equilibrium is ary and initial conditions, which are summarized in this
achieved instantaneously at the interface. This is the section.
basis for local equilibrium theory. Thus, the concentration In both types of applications, the solution of ODE’s for the
of solute in an SF at the interface may be governed by its static stage, requires the following two initial conditions
solubility in the SF, which, in turn, correlates with the den- referring to fresh initial SF phase and to the initial mass
sity of the SF. For example, a simple semi-log plot of fraction of solute on the solids, q 0:
solute solubility as a function of SF density can provide an
excellent correlation of experimental equilibrium data over
a relatively wide range of temperature. Modeling of solubi- at t = 0, c = 0 (16)
lities in SFE has followed the approach based on the
known equations of state. Despite extensive studies, at t = 0, q = q 0 (17)
482 Al-Jabari J. Sep. Sci. 2002, 25, 477–489

While the initial conditions for the dynamic stage are at r = R, q r = 0 (23)
obtained from the solution of the transport equations for
the static stage defined as cs and qs for the concentrations where R is the radius of the particle. On the other hand,
in SF and solid phase, respectively, and estimated at the when the external mass transfer contributes to the solute
end of the static stage. Consequently, for the dynamic transport, a surface balance is required for the solute at
stage, the initial conditions are: the interface in order to specify the boundary condition.
The surface balance assures that the rate of solute diffu-
at t = 0, c = c s (18) sion in the solid according to Fick’s law is equivalent to
that transferred across the SF by convection. In this case
at t = 0, q = q s (19) the following boundary condition is used

In previous models developed for SFE in packed beds, qqr


at r ¼ R; ð1 ÿ eÞqs Ds ¼ ekc ðci ÿ cÞ ð24Þ
some researchers [52] have mistakenly replaced the initial qr
conditions in Eqs. (18) and (19) with those in Eqs. (16)
and (17) although the model was for fitting data from two
stage operations (see static stage below). Equations (16) 3 Conceptual scheme for developing SFE
and (17) are applicable initial conditions for the dynamic models
stage if no static stage is involved.
For models describing analytical vessels, Eqs. (16) – (19) Combining and arranging the above general equations
are the only required conditions for solving the transport required for any considered SFE system leads to specific
equations (ODE). However, for the industrial type of models for certain practical systems. This approach is
operation in packed beds, boundary conditions are also based on an understanding the constraints of the physical
needed in addition to the above same initial conditions. system, which are then translated into a list of assump-
For solving the axial dispersion model, two boundary con- tions enabling a researcher to specify the equations
ditions are needed at the bed inlet and outlet. Such bound- required for modeling SFE.
ary conditions, for mass transfer operations in packed Prior to starting the process of modeling SFE, it is very
beds in general, have been the subject of numerous important to specify the type of operation, i. e. whether it is
reports in the literature [49]. The Danckwerts’ boundary a two-stage operation (dynamic after static) or a one stage
conditions are (dynamic) operation in order to determine whether a static
model is needed as well as the dynamic model. In cases
Da qc where a static stage is involved, a separate model for no-
at x ¼ 0; c ¼ ð20Þ flow conditions has to be developed and solved prior to
U qx
modeling the dynamic stage. The static model is similar
qc for both types of processes (packed beds and CSTR).
at x ¼ L; ¼0 ð21Þ The solution of the static model would provide the initial
qx
conditions for the dynamic model.

For models that handle internal diffusion equations, other Now the first step in developing a dynamic SFE model is
sets of boundary conditions are needed, together with an to identify the type of flow field in the process, which will
initial condition similar to that in Eq. (17). For example, for determine what type of equation is to be used for the dif-
models of internal diffusion in spherical particles, Eq. (5), ferential mass balance (PDE or ODE). As summarized
the first boundary condition is above, two types of flow fields are usually encountered in
mass transfer operations, viz. those including the axial
dispersion model and the continuous stirred tank reactor
qq r
at r ¼ 0; ¼0 ð22Þ (CSTR) model. The axial dispersion model is chosen
qr when the concentration of the solute in SF is position and
time dependent as in the packed bed. However, when the
The second boundary condition depends on the relative SF is considered to be a well-mixed phase, the CSTR
importance of various rates of solute transfer as well as model is used and the concentration of the solute in the
the rate of refreshment of the SF phase (wash out rate). SF is only time dependent. In addition to this categoriza-
When the rate of desorption kinetics, external mass trans- tion, it is also helpful to compare the rate of purging of
fer, and the rate of SF wash out are all very rapid, the SF solute out of the SFE vessel (fluid displacement) with the
phase can be considered as a perfect sink for the solute, rate of its desorption into the SF. When the fluid displace-
i. e. ment rate is faster than solute desorption rate, the con-
J. Sep. Sci. 2002, 25, 477–489 Kintic models of supercritical fluid extraction 483

Table 1. Types of SFE models with sample references.

Rate limiting Mass transfer limiting Desorption Instantaneous


Flow field Internal External Two resistance kinetics solute transfer
Diffusion Flux Eqn. (Langmuir)

Packed beds Rahme et al. Reis-Vasco et al. Rahme et al Montero Fullana Analogy with Reis-Vasco et al.
(1995) [25] (2000) [27], (1995) [25], et al. (2000) [33], Al-Jabari et al. (2000) (27),
Reverchon and Reverchon and Dooley (1995) [12] (1994) [43] Reverchon and
Marrone (1997) [35] Marrone (1997) [35] Marrone (1997) [35]

CSTR Future research Similar to Al-Jabari and Weber Al-Jabari and Al-Jabari Developed in
(1997) [23] Weber (1999) [23] (in press) [22] this paper

Fast SF Solid particles:


flow rate Sphere: Esquivel et al. (1999) [34] and Rincon et al. (1998) [6]
(Zero c-value) Thin plate: Bartle (1990)
Semi infinite slab: Dooley et al. (1995) [12]

centration in the SF is considered to be constant or zero in with reversible adsorption or particle deposition models.
some cases [34]. In such a case, the mass fraction of The following section provides some examples.
solute in the solid phase is considered to be dependent on
time and internal position in the solid phase and the non-
steady state diffusion equation (example: Eq. (5)) is 4 Modeling results and discussion
solved for the transfer of the solute across the solid phase.
Models of this type usually express results as percentage 4.1 Static stage
extraction (or average percentage remaining in solid) as a
function of time. For both types of process vessels (packed beds and
CSTR's), the model for the static stage is the same since it
The second step is to identify the rate-limiting step(s) in describes no flow conditions. Setting the third (dispersion)
the transfer of solute from the solid phase into the SF term and the second (convective) term in Eq. (1) to zero
phase. The controlling mechanism can be desorption leads to the same ODE as is obtained after setting the first
kinetics or mass transfer which in turn can be internally or (wash out) term in Eq. (2) of CSTR to zero. The resulting
externally controlled, or both phases may have consider- ODE is then solved simultaneously with Eq. (3) after sub-
able resistance. When the mass transfer process is inter- stituting the proper sink-source term as given by Eqs. (4),
nal, it is necessary to specify the type of mass transfer (6) – (10). Closed form solutions are available for the
modeling approach, i. e. whether it should be based on resulting system of ODE’s for cases with convective mass
mass transfer flux equations, like Eqs. (4), (7) – (9), or on transfer and Langmuir kinetics. Such solutions are
solving a diffusion equation and then estimating the flux obtained using Mathematica software (Wolfram
from Fick’s law. In addition, it may be necessary for some Research) which provides a useful tool in solving such
systems to consider both desorption kinetics and mass models. Figure 1 illustrates the static SFE curves for the
transfer rates. Such considerations have not been treated case with mass transfer controlling and linear equilibrium
in the literature. For certain systems, all these steps are isotherm with K e = 0.5 The solution is expressed in terms
extremely rapid and thus there is no need for a flux equa- of dimensionless concentrations where q is non-dimen-
tion since equilibrium is approached instantaneously. The sionalized using q 0 and c using q 0/K. The fractional recov-
third step, for mass transfer models, is to specify the type ery (Y ) during the static stage is then defined as follows
of equilibrium isotherm at the interface.
e c
Y ¼ ð25Þ
At the end of the modeling process, one of two types of ð1 ÿ eÞqs q0
model is reached. The model for packed beds may consist
of a single or coupled PDE's while models for CSRT con- The dimensionless time (s) during the static stage is
sist of a single or coupled ODE's. defined as follows

Table 1 summarizes these possible modeling ð1 ÿ eÞ


approaches and lists sample references to models that s¼ k i at ð26Þ
e
already exist in the literature. There is still room for the
development of new models. This can be accomplished Where k i is the mass transfer coefficient, the subscript (i)
by adjusting some of the existing models or by analogy is to be replaced with the approriate subscript according
484 Al-Jabari J. Sep. Sci. 2002, 25, 477–489

value. This modification is still insufficient since the actual


initial recovery can be at any value between these two lim-
its as shown in Figure 1 depending on the allowed time
prior to the dynamic stage.

4.2 Fluid displacement models for full SFE static


recovery
The plateau reached by the curve of fractional extraction
during the static stage (Figure 1) is dependent on the
equilibrium constant, which is strongly dependent on the
operating pressure. If the operating pressure is high
enough and if sufficient time is allowed during the static
stage, the fractional recovery during the static stage can
Figure 1. SFE curves for a static stage with mass transfer approach a value that is close to full recovery. Conse-
control and linear equilibrium isotherm with K e = 0.5.
quently, the subsequent dynamic stage is governed by
to the type of mass flux equation. Similar general curves the fluid displacement process, which is basically charac-
are obtained with Langmuir kinetics [22]. The static model terized by the residence time distribution curves in the pro-
with external mass transfer resistance provided a good fit cess vessel used. For CSTR, such a RTD curve can be
with experimental data for static SFE tests, as demon- obtained by setting the source term in Eq. (2) to zero and
strated by Al-Jabari and Weber [23]. solving the resulting equation with the initial condition in
Eq. (18), noting that c s in this case is equivalent to full
As mentioned above, developing a model and obtaining a recovery. The author demonstrated experimentally that
solution for the static stage is one important step since it the extraction of metal complexes with SF CO2 was well
specifies the type of initial condition that would be needed described by the RTD of CSTR when enough static time
for the model of the dynamic stage. Some researchers was permitted [51].
omitted this step when they used an invalid initial condi-
Similarly, for packed beds the RTD curve can be obtained
tion, and thus obtained misleading results. For example,
from the solution of the PDE in Eq. (1) after setting the
Montero et al. [52] implemented a stage with no flow con-
source term to zero, together with the initial and boundary
ditions when studying the SFE of hazardous contami-
conditions in Eq. (18), (20), and (21), respectively. Such a
nants from soils. However, in the model used to describe
model was solved by Brenner [50] and used by Al-Jabari
the resulting dynamic curves, the mass transfer during
et al. [43] to describe the experimentally obtained RTD of
such a static stage was not taken into account and the
water flow in packed beds. A simplified solution of the axial
initial conditions used were of the type given in Eqs. (16)
dispersion models is also given by Smith [49].
and (17).
The first attempt to consider a separate model for the sta-
tic stage was made by Al-Jabari and Weber [23] for SFE 4.3 SFE models for packed beds
in analytical tests. It is essential to reconsider previous Models for SFE operations in packed beds consist of the
models of packed beds and resolve the PDE’s with the two PDE's of Eqs. (1) and (3) together with the proper
resulting initial condition from the proper static stage equation for the sink-source term. The required initial and
model. Such new solutions will provide better fits for boundary conditions are given in Eqs. (16) – (21), depend-
experimental two-stage SFE processes. The solution will ing on the type of operation. The existing models for
not require any additional mathematical work since closed packed beds did not account for cases with initial static
form solutions of the static models exist and can be imple- stage.
mented.
Usually, no closed form solutions are possible for the
However, Reverchon and Marrone [35] had considered PDE’s of SFE in packed beds and thus numerical solu-
two types of initial conditions for their dynamic model, one tions are obtained using, for example, the Crank-Nicolson
is when the SF phase is free of solute (no static stage) and finite difference scheme. In some cases, the axial disper-
the other is for SF that is fully saturated with solute. In their sion (third) term in Eq. (1) may be neglected when the
work, the latter initial condition was used to account for Peclet number estimated on the basis of the axial disper-
solute recovery into SF during loading the extraction ves- sion coefficient is very low. In such a case, plug flow [49]
sel with CO2 and pressurizing it to the required pressure. characteristics are obtained. Reverchon and Marrone [35]
Obviously, they used the two terminal states of the static used such an assumption in modeling their SFE work on
curve (Figure 1), i. e. the initial value and the asymptotic the extraction of clove bud essential oil. They also com-
J. Sep. Sci. 2002, 25, 477–489 Kintic models of supercritical fluid extraction 485

pared the results with those obtained from a model includ-


ing the axial dispersion term and found that adding the
axial dispersion term did not improve the fit.
In such an approach, the mass transfer process was con-
sidered to be either internally or externally controlled,
while other models considered the two mass transfer
resistances. In these previous models, desorption kinetics
was not considered to any significant extent, even for
cases where interactions between solid matrix and ana-
lytes were reported to have a substantial effect on the rate
of SFE. In addition to these general mass transfer models,
other simplified models were also developed in the litera-
ture for cases with instantaneous solute transfer step (i. e.
rapid desorption and mass transfer). Such models are
referred to as equilibrium models [35, 27]. The model con-
sists of Eqs. (1) and (11) after substituting the applicable Figure 2. Dynamic SFE curves for analytical tests with mass
equilibrium relation for the derivative of q with respect to c transfer control and linear equilibrium isotherm and with
K e = 1.0.
(Section 2.4). Previous models of this type considered
basically the linear equilibrium isotherm. depressurized and the extracted solute is recovered while
Models of packed beds based on axial dispersion to be gas is bubbled out, referred to as the dynamic stage. Pre-
found in the existing literature include vious characterization of such SFE tests had been per-
formed in the analytical chemical literature based on
– Axial Dispersion models with external mass transfer
experimental observations. For example, Wang and Mar-
resistance and linear equilibrium.
shal [16] described the dynamics of such tests by an
– Axial Dispersion models with external mass transfer
empirical biphasic exponential decay.
resistance and non-linear equilibrium.
– Axial Dispersion models with internal mass transfer As described above, the flow field in such analytical tests
resistance using a flux equation. is of CSTR type due to contraction-expansion at the inlet
– Axial Dispersion models with internal mass transfer and outlet of the test vessel (which is small). Conse-
resistance, where the mass transfer rate is obtained quently, models for such tests usually consist of two
from non-steady internal diffusion equation. ODE’s as given in Eq. (2) and (3) after implementing the
sink-source term according to the controlling mechanism.
– Equilibrium models with linear equilibrium.
A closed form solution for the dynamic SFE with external
Other models that are of future interest and suggested to
mass transfer and linear equilibrium isotherm are
be considered by researchers include:
obtained using Mathematica software (Wolfram
– Axial dispersion models with Langmuir kinetics as the Research). In the case of Langmuir kinetics, however, no
rate equation (see Section 4.6 below). closed form could be obtained for the dynamic stage and
– Resolving the axial dispersion models listed above the system of ODE's was therefore solved numerically by
after developing a static model accounting for the the author [22]. Figure 2 shows the dimensionless curves
initial static stage. for c, q, and Y obtained from the mathematical solution of
– Axial dispersion equilibrium models with a non-linear a mass transfer model with external resistance and linear
equilibrium isotherm. local equilibrium. The dynamic SFE curves shown are for
the case with no static stage. The fractional recovery dur-
ing the dynamic stage is estimated from the solute bal-
4.4 Modeling analytical SFE tests (CSTR)
ance around the test vessel:
The SFE tests are usually performed in three steps:
(1) Pressurizing the test vessel containing the sample to Z t
be extracted with the supercritical fluid. This step is
Kc
Y ¼ dh ð27Þ
usually fast; changes in concentrations between the two 0 q0
phases can be neglected during this stage. (2) Soaking
the sample in the supercritical phase for certain time in The time during the dynamic stage is non-dimensiona-
order to allow the solute desorption into the supercritical lized with respect to the average residence time of the SF
phase, referred to as the static stage. (3) Displacement of in the vessel (i. e. eV/Q) and referred to as h.
the supercritical fluid phase out of the test vessel into a The mass fraction of the solute on the solid surface
collection vessel containing a liquid solvent where it is decreases continuously while the concentration of the
486 Al-Jabari J. Sep. Sci. 2002, 25, 477–489

solute in the SF phase increases and then decreases as a


result of the two competing rates (wash-out of SF in the
vessel and desorption). The fractional recovery, Y,
increases until it approaches the full recovery in a time
period that is dependent on the mass transfer coefficient,
the length of the static stage, and the equilibrium con-
stant.
The solutions obtained for CSRT models provided good
fits to different types of experimental results, including two
stage SFE of metal complexes [23] as well as for SFE of
organics [22].

4.4.1 New equilibrium model for analytical SFE Figure 3. Dynamic SFE curves obtained from equilibrium
tests model, mass transfer model, and Langmuir kinetics without
In this section, a new CSTR model is presented which uti- an initial static stage and all at the same K e = 0.5. Also
shown is the SFE curve from the fluid displacement model.
lizes a modeling approach similar to that used by
Reverchon and Marrone [35] and Reis-Vasco et al. [27]
less time (h). Figure 3 also compares theoretical dynamic
for packed beds. In their work, it was assumed that when
SFE curves obtained from the various models including
the solute is accessible at the solid surface, the kinetics of
cases with fluid displacement, mass transfer, and deso-
the process is governed by equilibrium. This assumption
rption kinetics at the same dimensionless equilibrium con-
is not adequate for assuming instantaneous solute trans-
stant (Ke = 0.5).
fer between the two bulk phases. It is also essential to
assume that desorption kinetics is instantaneous and that Obviously, the fastest dynamic SFE process is obtained
the external mass transfer resistance is negligible. In this with the fluid displacement model since all the solute is
case, the mass fraction of solute on the solid is in equili- pre-extracted into SF during the static stage. This
brium with its concentration in the bulk of the SF phase. response can also be obtained without a static stage but
only if the solvating capacity of the SF phase exceeds the
Making such assumptions, combining Eqs. (2) and (3),
requirements of all the solute present in solids and all
and substituting Eq. (11) in the resulting equation after
solute transport steps are instantaneous. The second
estimating the derivative of q with respect to c from the lin-
fastest dynamic SFE is with the equilibrium model since in
ear equilibrium isotherm in Eq. (13) results in the following
this case the solute transport process is instantaneous,
ODE:
the delay of the SFE curve relative to that of the fluid dis-
placement model is dependent on the value of the equili-
Q dc ð1 ÿ eÞ dc brium constant. When the solute transport process is sig-
cþ ¼ÿ qs K ð28Þ nificant, a further delay is observed due to additional times
eV dt e dt
required for transport or kinetic processes. The slowest
which requires one initial condition given by Eq. (18), not- dynamic SFE is obtained with mass transfer control.
ing that for this case, c s is in equilibrium with q 0, i. e.

at t = 0, c = q 0/K (29)
4.5 Internal diffusion models
When the rate of refreshment of SF is much faster than
since the solute transfer process is instantaneous, thus
the rate of the solute release from the solid surface and
equilibrium is reached regardless of the length of the initial
the external mass transfer resistance is negligible, internal
(static) stage. The solution of Eq. (28) with the initial con-
diffusion in the solid controls the whole SF process. In
dition given in Eq. (29) is
such a case, there will be no need for a differential mass
balance in the SF phase (i. e. no axial dispersion or CSTR
 
c 1 equations). The solute concentration in the bulk of the SF
¼ Exp ÿ h ð30Þ
ðq0 =K Þ 1 þ ð1 ÿ eÞ=eqs K is usually taken to be negligible (zero) since the SF phase
is considered to be displaced with fresh SF at a rate that is
Where h is the dimensionless time during the dynamic faster than the rate of solute transfer from solids into the
stage. The fractional recovery can be estimated from SF phase. In models of this type, the solid phase is
Eq. (27). The resulting solution for the fractional recovery assumed to be made up of uniform particles and the inter-
(Y ) is plotted in Figure 3 as a function of the dimension- nal non-steady state diffusion equation is solved for one
J. Sep. Sci. 2002, 25, 477–489 Kintic models of supercritical fluid extraction 487

particle. The solution is usually expressed as solute resi- packed beds developed by Al-Jabari et al. [43] and the
dual on the solid particle (q) as a function of time. corresponding model for stirred vessels also developed
The solid particles may be considered as individual sphe- by Al-Jabari et al. [45]. The solution obtained for particle
rical particles [34], semi-infinite bodies [12] when the diffu- deposition in stirred vessel [45] is similar to that for static
sion path is much shorter than the characteristic dimen- SFE after modifying the initial condition [22].
sion of the particles, or as a slab when the solid particles Based on such similarity, a new model for SFE in packed
are thin. The resulting PDE’s are similar to those describ- beds, for the case where solute desorption from the solid
ing conductive heat transfer through a similar geometry, is the rate-limiting step, is derived from the previous
for which solutions are available in literature. Each type of deposition model published by Al-Jabari et al. [43]. The
model consists of a single PDE, which is solved with one previous model was developed to describe the particle
initial condition and two boundary conditions. deposition in packed beds of collectors using Langmuir
For example, the spherical particle model is obtained from kinetics. The model consisted of two coupled partial differ-
the solution of PDE in Eq. (5) together with the initial and ential equations similar to Eqs. (1), (3), and (10). The axial
boundary conditions in Eqs. (17), (22), and (23). The solu- dispersion term was neglected compared to the convec-
tion of this model is available in the literature and was tion term. This assumption is similar to that used by
used by Esquivel et al. [34] to describe the SFE of olive Reverchon and Marrone in their mass transfer model in
husk oil, providing acceptable fit at high SF velocities. packed beds (see Section 4.3).
The resulting PDE's were solved numerically using the
Galerkin Finite Element method. The same solution algo-
4.6 Establishing new SFE models by analogy rithm can be used to describe the SFE. A Fortran program
was developed by the author to solve the resulting PDE's
For many transport phenomena, the general transport
of deposition in packed beds; the same software is applic-
equations may have the same mathematical form since
able after modifying the initial condition to account for de-
the transport mechanisms are based on similar physical
sorption rather than adsorption. Results from such a
backgrounds. For example, the mechanism of mass
model are to be presented in a subsequent paper.
transfer diffusion is similar to that for heat conduction.
Also, desorption into SF is similar to the conventional
molecular adsorption (or desorption) into liquids. When
the geometry of the systems is also identical, similar 5 Concluding remarks
mathematical solutions are obtained for different transport
There is a common theoretical basis for modeling SFE
processes. Thus, existing solutions for conventional pro-
processes. Most of the existing models can be con-
blems similar to those in SFE can be used in new SFE
structed on the basis of a group of general equations pre-
models. The previous section introduces the concept of
sented in this paper. Directions and a scheme for develop-
analogy where SFE models were similar to those in pre-
ing new SFE models are suggested. Previous SFE mod-
vious conductive heat transfer models. The same mathe-
els for packed beds have not considered the case with
matical solutions could therefore be used, thus saving
static stage operation prior to dynamic stage operation;
considerable solution time.
new models of this type are still needed. A new model is
Such a concept of analogy can be implemented for obtain- presented in this paper for analytical tests assuming
ing “new” models for SFE. The literature of chemical engi- instantaneous solute transfer and equilibrium desorption
neering has a plethora of transport models that describe and compared to other mass transfer, desorption kinetic,
the transfer of molecules (or fine particles) from a fluid into and fluid displacement models. The concept of analogy is
a solid surface and vice versa. By analyzing the physical demonstrated for the purpose of developing new models
behavior of the SFE process and identifying the transport for SFE based on previously existing models for similar
similarity with conventional transport processes, many of phenomena.
the previous models can be used for SFE processes.
In this section, it is shown that a previously developed par-
ticle deposition model [43] can be used for SFE in packed Nomenclature
beds after merely changing the initial conditions. The phy-
sical behavior of particle deposition systems is similar to a specific surface area of the solid particles
that of molecular adsorption systems; the Brownian (cm2 N cm – 3)
motion of fine particles resembles molecular diffusion. c concentration of the solute in the bulk supercritical
The same Langmuir kinetics used for molecular adsorp- phase (g cm – 3)
tion onto solid surfaces, Eq. (10), was used in a model to c* concentration of the solute in the bulk of supercritical
describe the reversible particle deposition process in phase that is in equilibrium with q (g cm – 3)
488 Al-Jabari J. Sep. Sci. 2002, 25, 477–489

ci concentration of the solute at the interface in the References


supercritical phase (g cm – 3)
cs concentration of the solute in the supercritical phase [1] R.M. Smith, J. Chromatogr. A 1999, 856, 83 – 115.
at the end of the static stage [2] R. Marra, T. Gamse, Chem. Eng. Process. 2000, 39,
Da axial dispersion coefficient (cm2 s – 1) 19 – 28.
Ds diffusion coefficient of solute in solids (cm2 s – 1) [3] K.W. Hutchenson, N.R. Foster, Innovations in Supercri-
K equilibrium constant (cm3 g – 1) tical Fluid Science and Technology, Ch. 1 in: Innovations
Ke dimensionless desorption equilibrium constant in Supercritical Fluids Science and Technology, K.W.
Hutchenson, N.R. Foster, eds., American Chemical
defined in Eq. (15) Society, Washington, DC, 1995.
ka adsorption coefficient (cm s – 1)
[4] M.D. Luque de Castro, M.M. JimØnez-Carmon, V.
kd desorption coefficient (g cm – 2 s – 1) Fernµndez-PØrez, Tr. Anal. Chem. 1999, 18, 708 – 716.
kc mass transfer coefficient in SF phase (cm s – 1)
[5] T.L. Chester, J.D. Pinkston, D.E. Raynie, Anal. Chem.
k co overall mass transfer coefficient in SF phase
1994, 66, 106R-130R.
(cm s – 1)
[6] J. Rincon, A. de Lucas, M.A. Garcia, A. Garcia, A.
ks mass transfer coefficient in solid phase (g cm – 2 s – 1)
Alvarez, A. Carnicer, Sep. Sci. Technol. 1998, 33, 411 –
k so overall mass transfer coefficient in solid phase 423.
(g cm – 2 s – 1)
[7] Luiz F. de Franc, M. Angela, A. Meireles, J. Supercrit.
L bed length (cm) Fluids 2000, 18, 35 – 47.
NA flux of solute from solids into SF phase (g cm – 2 s – 1)
[8] Lixiong Li, Erdogan Kiran, I & EC Research 1988, 27,
Q volumetric flow rate (cm3 s – 1) 1301.
q mass fraction of the solute in the solid sample (g/g)
[9] A.J. Sequeira, L.T. Taylor, J. Chromatogr. Sci. 1992, 30,
q* mass fraction of the solute in the solid sample which 405 – 408.
is in equilibrium with c (g/g)
[10] E.C. McDonald, J. Howard, B. Bennett, Fluid Phase
qmax maximum mass fraction of the solute that can be Equilibria 1983, 10 , 337 – 344.
adsorbing on the solid sample for monolayer cover-
[11] E. Kiran, H. Balkan, J. Supercrit. Fluids 1994, 7, 75 – 86.
age (g/g)
q0 initial mass fraction of the solute in the solid sample [12] K.M. Dooley, D. Launey, J.M. Becnel, T.L. Caines, Mea-
surement and Modeling of Supercritical Fluid Extraction
(g/g)
from Polymeric Matrices, Chapter 18 in: Innovations in
qs mass fraction of the solute in the solid sample at the Supercritical Fluids Science and Technology, K.W.
end of the static stage. Hutchenson, N.R. Foster, eds., American Chemical
qr local mass fraction of solute in solids (g/g) Society, Washington, DC, 1995.
qi mass fraction of solute at the interface in the solid [13] Z.G. Rahme, R.G. Zytner, W.H. Stiver, Kinetic Model for
phase (g/g). the Supercritical Extraction of Contaminants from Soil,
r radial position in a spherical solid particle (cm) Chapter 20 in: Innovations in Supercritical Fluids
Science and Technology, K.W. Hutchenson, N.R. Foster
R radius of the particle (cm). American Chemical Society, Washington, DC, 1995.
S solubility of solute in SF (g cm – 3)
[14] S.J. Macnaughton, P. Alessi, A. Cortesi, I. Kikic, N.R.
t time (s) Foster, Predictive and Experimental Methods for the
V total volume of the extraction vessel (cm3) Choice of Cosolvent in the Supercritical Fluid Extraction
U superficial SF velocity in the extraction vessel of Pesticides, Chapter 8 in: Innovations in Supercritical
(m s – 1) Fluids Science and Technology, K.W. Hutchenson, N.R.
Foster, American Chemical Society, Washington, DC,
x the linear position in the extraction vessel measured
1995.
from SF inlet (m)
Y fraction of the solute recovered [15] K.E. Laintz, C.M. Wai, C.R. Yonker, R.D. Smith, Anal.
Chem. 1992, 64, 2875 – 2878.
[16] Jin Wang, W.D. Marshall, Recovery of Metals from Aqu-
eous Media by Extraction with Supercritical Carbon
Greek symbols Dioxide, Anal. Chem. 1994, 66, 1658–1663.
e volume fraction which is occupied by the supercriti- [17] K. Thurbide, M. Al-Jabari, M. Kowalchuk, Chem. Eng.
cal phase in the extraction vessel. Commun. 2002, 189, 676–684.
qs density of the solid sample (g N cm – 3) [18] C.M. Wai, Anal. Sci. 1995, 11, 165 – 167.
s dimensionless time during the static stage [19] M. Ashraf-Khorassani, M.T. Combs, L.T. Taylor, J.
h dimensionless time during the dynamic stage Chromatogr. A. 1997, 774, 37 – 49.
J. Sep. Sci. 2002, 25, 477–489 Kintic models of supercritical fluid extraction 489

[20] K.E. Laintz, C.M. Wai, C.R. Yonker, R.D. Smith, J. [37] S. Oiao, X. Hu, Sep. Purif. Technol. 1999, 16, 261 – 271.
Supercrit. Fluids 1991, 4, 194 – 198. [38] Y. Ding, E. Alpay, Chem. Eng. Sci. 2000, 55, 3461 –
[21] C.M. Wai, S. Wang, J. Chromatogr. A 1997, 785, 369 – 3474.
383. [39] G. Dünnebier, K.-U. Klatt, Chem. Eng. Sci. 2000, 55,
[22] M. Al-Jabari, Chem. Eng. Commun., in press. 373 – 380.
[23] M. Al-Jabari, M.E. Weber, Sep. Sci. Technol. 1999, 34, [40] D.E. Cherrak, S. Khattabi, G. Guiochon, J. Chromatogr.
2853 – 2863. A 2000, 877, 109 – 122.
[24] E. Björklund, M. Järemo, L. Mathiasson, J.Š. Jönsson, [41] K. Wang, D.D. Do, Sep. Purif. Technol. 1999, 17, 131 –
L. Karlsson, Anal. Chim. Acta 1998, 368, 117 – 128. 146.

[25] Z.G. Rahme, R.G. Zytner, W.H. Stiver, Kinetic Model for [42] R. Subramanian, V. Lakshminarayanan, Electrochimica
the Supercritical Extraction of Contaminants from Soil, Acta 2000, 45, 4501 – 4509.
in: Innovations in Supercritical Fluids Science and Tech- [43] M. Al-Jabari, A.R.P. van Heiningen, T.G.M van de Ven,
nology. ACS Symposium Series 608, K.W. Hutchenson, J. Pulp Paper Sci. 1994, 20, 249.
N. R. Foster, Eds., American Chemical Society,
[44] H. Mousa, M. Al-Jabari, I. Al-Khateeb, Chem. Eng. J.
Washington, DC, 1995, Ch. 20.
2001, 83, 79 – 83.
[26] M. Goto, B.C. Roy, A. Kodama, T. Hirose, J. Chem. Eng.
[45] M. Al-Jabari, H. Mousa, I. Al-Khateeb, J. Sep. Sci. Tech-
Jpn. 1998, 31, 171 – 177 [in Japanese].
nol. 2002, 37, 431 – 449.
[27] E.M.C. Reis-Vasco, J.A.P. Coelho, A.M.F. Palavra, C. [46] M. Kamiti, T.G.M. van de Ven, J. Pulp Pap. Sci. 1994,
Marrone, E. Reverchonba, Chem. Eng. Sci. 2000, 55, 20, 199 – 205.
2917 – 2922.
[47] T.G.M. van de Ven, Nordic Pulp Paper Res. J. 1993, 8,
[28] E. Reverchon, A. Kaziunasb, C. Marrone, Chem. Eng. 130 – 147.
Sci. 2000, 55, 2195 – 2201.
[48] H.S. Fogler, Elements of Chemical Reaction Engineer-
[29] I. Goodarzni, M.H. Eikani, Chem. Eng. Sci. 1998, 53, ing, 2nd Ed. Prentice Hall, New Jersey, USA 1992.
1387 – 1395.
[49] J.M. Smith, Chemical Engineering Kinetics, 3rd Ed.
[30] B. Ramchandran, J.B. Riggs, H.R. Heichelheim, A.F. McGraw-Hill, New York, USA 1981.
Seibert, J.R. Fair, I & EC Research 1992, 31, 281.
[50] H. Brenner, Chem. Eng. Sci. 1962, 17, 229.
[31] N.T. Dunford, M. Goto, F. Temelli, J. Supercrit. Fluids
[51] M. Al-Jabari, Characterizing Supercritical Fluid Extrac-
1998, 13, 303 – 309.
tion of Metal Complexes using Spectrophotometric
[32] E. Reverchon, G. Lambertia, P. Subraa, Chem. Eng. Technique, Proceedings of First International Chemical
Sci. 1998, 53, 3537 – 3544. Engineering Conference, University of Jordan, Septem-
ber 2001.
[33] M. Fullana, F. Trabelsi, F. Recasens, Chem. Eng. Sci.
2000, 55, 79 – 95. [52] G.A. Montero, T.D. Giorgio, K.B. Schnelle, Jr., Removal
of Hazardous Contaminants from Soils by Supercritical
[34] M.M. Esquivel, M.G. Bernardo-Gil, M.B. King, J. Super- Fluid Extraction, Ch.19 in: Innovations in Supercritical
crit. Fluids 1999, 16, 43 – 58. Fluids Science and Technology. K.W. Hutchenson, N.R.
[35] E. Reverchon, C. Marrone, Chem. Eng. Sci. 1997, 52, Foster, Eds., American Chemical Society, Washington,
3421 – 3428. DC, 1995.
[36] S. Oiao, X. Hu, Chem. Eng. Sci. 2000, 55, 1533 – 1544. [JSS 1155]

You might also like