Chemical Engineering and Processing 65 (2013) 58–67

Contents lists available at SciVerse ScienceDirect

Chemical Engineering and Processing:

Process Intensification
journal homepage: www.elsevier.com/locate/cep

Near critical and supercritical fluid extraction of Cu(II) from aqueous solutions
using a hollow fiber contactor
Hugo Valdés a,d , Rossana Sepúlveda a , Julio Romero a,∗ , Fernando Valenzuela b , José Sánchez c
a
Laboratory of Membrane Separation Processes (LabProSeM), Department of Chemical Engineering, University of Santiago de Chile (USACH), Chile
b
Department of Food Science and Chemical Technology, Faculty of Chemical and Pharmaceutical Sciences, University of Chile (UCH), Chile
c
Institut Européen des Membranes (UMR 5635 CNRS/UM2/ENSCM), Montpellier, France
d
School of Construction Engineering, Faculty of Engineering Science, Catholic University of Maule (UCM), Talca, Chile

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

Article history: The aim of this work is the experimental and theoretical characterization of the mass transfer in a
Received 20 August 2012 membrane-based dense gas extraction of metal ions from aqueous solutions using a hollow fiber contac-
Received in revised form tor. Extractions of Cu(II) were conducted in a single fiber membrane contactor operating under steady
13 December 2012
state conditions. Aqueous CuSO4 solutions were treated using a CO2 phase containing 1,1,1-trifluoro-
Accepted 16 December 2012
2,4-pentanedione (TFA) or 1,1,1,5,5,5-hexafluoro-2,4-pentanedione (HFA) at 40 ◦ C and pressures ranged
Available online 22 December 2012
between 70 and 90 bar.
Experiments show that the use of dense CO2 as extraction solvent of Cu(II) ions reaches extraction
efficiencies valued up to 98.7%. Simultaneously, a mass transfer model was proposed correlating an
effective rate function of the complex formation at the aqueous-CO2 interface.
The highest extraction efficiencies were observed at higher pressures and lower pH values, which could
confirm that a high content of protons is required to facilitate and stabilize the formation complex by
means of keto-enol tautomerism.
This work represents the first step in order to propose a novel intensified operation, which could be
applied for high valued metals or hazardous materials.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction commercially available organic solvents. Nevertheless, the knowl-

edge developed in the last six decades the studies about SCF
There is an immense body of literature on solvent extraction of extraction of metals from aqueous solutions are relatively few
metals using organic solvents, considering that solvent extraction [1–5], and further investigations in this field are necessary to lead to
is a relatively new technology compared to other unit opera- commercial processes using dense carbon dioxide in a wide variety
tions such as distillation and absorption. A surge in interest in of industries dealing with metals.
solvent extraction occurred during the 1940s, initiated by its appli- The extraction of metals based on a membrane contactor sys-
cation for uranium production and for reprocessing of irradiated tem with conventional solvents is a process widely studied using
nuclear materials. This first application became a stimulus for the different configurations, extractants and extraction solvents [6–9].
non-nuclear industry and solvent extraction was introduced as Generally, several authors [9,10] report studies of extracting metals
separation and purification process in numerous chemical and met- from aqueous solutions by means of conventional organic solvents
allurgical industries in the 1950s and early 1960s. Subsequently, such as hexane or kerosene, and commercial extractants, similar to
in the 1980s and 1990s, a significant number of studies have those used in SX extraction.
been reported in the literature on supercritical fluid (SCF) extrac- Here the extraction of Cu2+ from aqueous solutions by means of
tion of a wide variety of matrices with the objective of replacing dense gas extraction was achieved by using a hollow fiber mem-
brane contactor device. This application could be described by
means of the same phenomenological and thermodynamic prin-
ciples of conventional contactor operations. Nevertheless, to our
∗ Corresponding author at: University of Santiago de Chile (USACH), Department
knowledge no previous study is available relating to dense gas
of Chemical Engineering, Av. Lib. Bdo. O’Higgins 3363, Estación Central, Santiago,
Chile. Tel.: +56 2 718 18 21. extraction of metals from aqueous solutions using a membrane
E-mail address: [email protected] (J. Romero). contactor operation.

0255-2701/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cep.2012.12.005
 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67 59

Nomenclature

C concentration (mol kg−1 )

D diffusion coefficient (m2 s−1 )
d diameter (m)
e membrane thickness (m)
J flux of Cu2+ (mol m−2 s−1 )
K mass transfer coefficient in boundary layer (m s−1 ) Fig. 1. Chemical structure of extractants used in this study.
keff kinetic coefficient (mol1−˛−ˇ s−1 m3(˛+ˇ) )
L length (m) contact between two phases. On one side, an aqueous liquid solu-
N mass transfer flow of Cu2+ (mol s−1 ) tion is circulated and on the opposite side the extraction solvent
P pressure (Pa) is a near-critical or supercritical fluid. When the membrane used
Re Reynolds number is hydrophobic, the aqueous solution does not penetrate within
S surface area available for mass transfer (m2 ) the membrane porosity. A meniscus is formed in the pore mouth
Sc Schmidt number establishing a dense gas–liquid interface. The driving force of the
Sh Sherwood number mass transfer process through the membrane is a concentration
T temperature (K) gradient between both circulating phases. In this operation, per-
z stoichiometric ratio meability and selectivity are defined by the transport phenomena
at the proximities and through the membrane coupled to the phase
Greek letters equilibrium between the aqueous feed solution and the CO2 phase.
ε porosity The size and number of fibers in a membrane contactor will
 tortuosity be determined by the volume of the liquid to be treated. Previ-
 viscosity (kg m−1 s−1 ) ous studies on membrane-based dense gas extraction report higher
 dipole moment (Debyes) extraction efficiencies when the near critical or supercritical sol-
vent is circulated outside the fibers and the liquid feed solution is
Subscripts and superscripts circulated in the lumenside [16–18]. This process involves several
(1) refers to aqueous solution boundary layer advantages compared to conventional contactor devices used in
(2) refers to aqueous-dense gas interface supercritical fluid extraction like conventional contacting columns
(3) refers to membrane pores that disperse one fluid phase in another. In this case, the modularity
(4) refers to dense gas boundary layer of this membrane contactor application is very important consid-
A refers to extractant ering that in a typical porocritical extraction application large and
l refers to liquid expensive vessels are not used. The reduced complexity of process
M refers to metal ion and its comparative low cost will enable a broader industrial use
i liquid–dense gas interface of CO2 attributes as a non-toxic and an environmentally benign
0 bulk extraction solvent.
eq refers to equivalent diameter of the shell Although, the porocritical extraction is a commercial appli-
C critical point cation, there are only few studies analyzing the mass transfer
R reduced variable across the membrane and a physical model to scale-up the process
ID inner diameter [15,16,19–26].
OD outer diameter
SL saturated liquid
2.2. Supercritical fluid extraction of metal ions
VP vapor
p pore
In this work, the extraction of Cu2+ from aqueous solutions
using a cation exchanger as extractant was implemented by means
of a configuration based on a porocritical extraction process.
2. Membrane-based supercritical fluid extraction of metal The membrane contactor operation to remove Cu2+ from acid
ions aqueous solutions has been conducted by using 1,1,1-trifluoro-2,4-
pentanedione (TFA) or 1,1,1,5,5,5-hexafluoro-2,4-pentanedione
2.1. Supercritical fluids and membrane contactors (HFA) as extractant compound, since these ␤-diketones have been
previously tested in a dense CO2 medium [4], considering the
Although membrane contactor processes can be conducted favorable kinetics of complex formation, the thermodynamic prop-
using a number of different types of module geometry, including erties and a comparatively higher extraction capacity of Cu(II)
flat sheet, spiral wound, rotating annular and hollow fiber, this lat- ions reported in the literature [3]. The main characteristics of the
ter one receives the most attention, since it shows high specific extractants used in this work are reported in Fig. 1.
contact surface area per volume (m2 /m3 ). Hollow fiber modules tar- Murphy and Erkey [2,3] developed a thermodynamic model
geted for industrial applications (as opposed to medical ones, e.g., based on combined phase and reaction equilibriums for prediction
blood oxygenation) are available from a variety of sources, although of the extraction efficiencies by using dense CO2 with this type of
some are designed for pressure-driven filtration processes rather cation exchangers. In this model, the following aqueous phase reac-
than concentration-driven mass transfer. The most well-known tions and three phase equilibrium relations were considered for the
hollow fiber contactor (HFC) module designed for concentration- extraction of copper:
driven mass transfer is the Liqui-Cel® Extra-Flow module offered
Aqueous phase reactions:
by CELGARD LLC [11–13].
The PoroCrit® process or porocritical extraction [14,15] is a CO2(aq) + H2 O(l) ↔ H2 CO3(aq) (1)
commercial SFC extraction which uses a hollow fiber membrane
contactor. In this process a macroporous membrane allows the H2 CO3(aq) ↔ HCO3(aq) − + H(aq) + (2)
60 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67

HA(aq) ↔ H(aq) + + A(aq) − (3) aqueous feed and extraction streams, which are contacted in the
single fiber contactor. The flow rate and pressure of liquid feed solu-
2A(aq) − + Cu(aq) 2+ ↔ CuA2(aq) (4) tion is controlled by using a HPLC isocratic pump Jasco PU-2080 and
Phase equilibrium relationships: the pressure of the CO2 phase is controlled by ISCO 500D syringe
pump with a temperature control jacket connected to a thermo-
KCO
2
CO2(SC) ←→CO2(aq) (5) statized bath. The flow rate of the dense CO2 phase is controlled by
means of a micrometering valve AE 10VRMM2812.
KHA In Fig. 3, the black lines represent the aqueous feed solution
HA(SC) ←→HA(aq) (6)
streams; and the gray ones, the extraction gas streams. In this case,
KCuA
CuA2(SC) ←→CuA
2
2(aq) (7) the extractant agent was dissolved in pressurized CO2 . The concen-
tration of extractant in the system was controlled by feeding CO2 at
where HA is the cation exchanger and CuA2 is the complex
52 bar and 20 ◦ C (CO2 = 152.3 kg m−3 ) in a 100 mL view cell (mix-
extracted in the CO2 phase.
ture vessel in Fig. 3) containing 3 mL of extractant during 2 h. The
Erkey [5] reported the chemical equilibrium constants for Eqs.
initial volume of extractant putted in the mixture vessel was cho-
(1)–(4) and distribution coefficients (KCO2 , KHA , KCuA2 ) of Eqs.
sen in order to introduce an excess amount of extractant, obtaining
(5)–(7). Thus, reactions (1) and (2), and reactions (3) and (4) can
a saturation concentration in the CO2 phase that was injected in
be combined as:
the experimental setup. The concentration of extractant in the CO2
K =K ·K
CO2(aq) + H2 O(l) I←→
1 2
H(aq) + + HCO3(aq) − (8) phase has been estimated through the difference between the ini-
tial and final mass of extractant contained in the total flow of CO2
K =K 2 ·K4
2+ II injected to the experimental setup. The extraction phase was cir-
←→ 2H(aq) + + CuA2(aq)
3
2HA(aq) + Cu(aq) (9)
culated at pressures between 70 and 90 bar and 40 ◦ C by using the
Erkey [5] reports the equilibrium constants, distribution syringe pump. The extraction solution (CO2 + ␤-diketone) circu-
coefficients as well as the experimental and predicted extraction lates through the shellside. A previous theoretical study developed
efficiencies for different chelating agents. by our group [18] about the mass transfer in porocritical extrac-
tion process shows larger values of extraction efficiency when the
2.3. Experimental process is implemented using this configuration.
A single macroporous polytetrafluoroethylene (PTFE) fiber
Fig. 2a shows the circulation configuration of solutions inside membrane with an outside diameter of 1.8 mm and a wall thickness
and outside of the hollow fiber during the experiments. The mass of 0.4 mm was used to carry out the extraction experiments. The
transfer of Cu2+ through the membrane was quantified using a sin- pore size and porosity of the fiber are 2.0 ␮m and 55%, respectively.
gle fiber membrane contactor. The liquid feed solution containing Modules with an active length of 0.353 m were prepared potting
Cu2+ was circulated into the lumenside, and dense CO2 phase con- fibers into a 316L stainless steel shell with an internal diameter
taining the extractant agents was circulated on the shellside in a of 3.6 mm by means of a solvent-resistant epoxy. The experimen-
countercurrent configuration. tal setup described in Fig. 3 is miniaturized hollow fiber contactor
In Fig. 2b the mass transfer of metal ion through the membrane system, but this apparatus shows the same fluid dynamic condi-
is schematically represented. Concentration profiles of metal ion tions of a pilot or industrial scale system. The process itself does
(Mz+ ), extractant (A) and formed complex (M(A)z ) at the proxim- not require conventional solvents and the use of dense CO2 allows
ities of the membrane are represented by dotted lines. A global the easy separation of the metal complex formed in the CO2 phase
flow of metal, NM (mol s−1 ), is observed from the aqueous solu- by means of its depressurization.
tion circulating into the fiber to the dense gas phase circulating in On the other hand, the feed solution is an acid CuSO4 solution.
the shellside. The mass transfer of metal at the proximities of the Temperature and concentrations of Cu2+ were 200 ppm and 40 ◦ C,
membrane may be described by the integration of the four steps respectively; pH was ranged between 2.0 and 4.0. These values have
indicated on the figure: been chosen in function of the conditions available in the literature
about the thermodynamic and transport properties of the aque-
1) Mass transfer of the metal ion (Mz+ ) through the boundary layer ous solutions [5]. The pH values in the aqueous feed solution were
of the aqueous feed solution. adjusted by addition of H2 SO4(aq) .
2) Mass transfer due to the formation of complex M(A)z at the The raffinate obtained from the circulation of liquid solution
interface. through the single fiber membrane contactor was collected by
3) Mass transfer of the complex M(A)z through the membrane means of a micrometering valve (AE 10VRMM2812). The concen-
pores filled with gas. tration of Cu2+ in the raffinate was quantified using a GBC (SensAA)
4) Mass transfer of the complex through the boundary layer of the atomic absorption spectrophotometer with a hollow cathode lamp
extraction solution. for Copper. The liquid feed solution was circulated with flow
rates ranged from 0.1 mL min−1 to 1.4 mL min−1 , and the dense
The mass transfer in the first step is represented by the migra- gas extraction was circulated with a flow rates varying between
tion of metal ion, Mz+ , through the boundary layer of the aqueous 1.7 × 10−3 and 3.7 × 10−3 mol min−1 .
solution; the second step is the reaction of complex formation at the The transmembrane flux of metal ion (mol m−2 h−1 ) and extrac-
interface established at the pore mouth. Thus, the global reaction tion percentage were calculated from the concentration values of
may be represented by the following equation: metal ion in the raffinate. Extraction efficiency has been defined by
the following equation:
zHA(SC) + M(aq) z+ → zH(aq) + + M(A)z(SC) (10)
C − C 
The third step is represented by the transport of complex M(A)z f R
EE (%) = Extraction efficiency = × 100 (11)
into the membrane pores, where the membrane pore is filled with Cf
dense gas, and the fourth step is the transport of M(A)z through the
dense gas boundary layer. where Cf is the concentration of metal M in the aqueous feed solu-
An outline of the experimental device used in this work is tion and CR is the concentration of metal M in the raffinate. Thus,
presented in Fig. 3. Two independent circuits are considered for the transmembrane flux of metal ion can be estimated by means
 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67 61

Fig. 2. Outline of the process studied in this work: (a) circulation of the solutions inside and outside the hollow fiber, and (b) schematic representation of the mass transfer
of metal ion Mz+ (Cu2+ ) through the membrane. The figure shows the concentration profile of the cation Mz+ , the extractant compound A, and the complex M(A)z (Cu(A)2 ) at
the proximities of the membrane.

of the mass balance in the membrane contactor according to the 3. Mass transfer model: resistances-in-series approach
following equation:
3.1. Theoretical considerations and mass transfer equations
 
Faq (Cf − CR )
NM = (12) Few assumptions and considerations have been taken into
SID
account to apply a resistances-in-series approach to describe the
overall mass transfer through the membrane, which is controlled
where Faq is the aqueous feed solution flow rate and SID is the sur- by the four steps detailed in Fig. 2. This model was based on the fol-
face area available for mass transport on the lumenside. This surface lowing assumptions: (a) the aqueous feed solution is diluted and
area value was chosen because it corresponds to the surface area of hydrodynamic properties are very close to the values for water and
aqueous-dense gas interface, which is considered practically equal its physical properties have been considered constants during the
to the surface area of the fiber’s internal wall. extraction process; (b) the system operates under steady state con-
The mass transfer through the membrane depends on the simul- ditions; (c) reaction is limited at the aqueous-dense gas interface;
taneous effects of the fluid dynamic conditions in the contactor, the (d) circulation rates of the solutions in the membrane contactor are
diffusion within the membrane pores and the complex formation under laminar flow conditions.
kinetics at the interface. In order to evaluate these three aspects, According to assumptions a, b, and d, the flow of Cu2+ (NM ,
experiments have been carried out to determinate the values of mol s−1 ) through the boundary layer of the aqueous solution cir-
the transmembrane flow and extraction efficiency as a function of culating on the lumenside can be estimated by Eq. (13):
the flow rates of solutions, the operating pressure and pH of the
(1)
aqueous feed solution. NM = k1 · SID · (CM,0 − CM,i ) (13)

Fig. 3. Outline of the experimental setup based on a single hollow fiber contactor used in this study.
62 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67

where k1 is the mass transfer coefficient for the boundary layer transport through the shellside. The mass transfer coefficient, k4 ,
of aqueous feed solution at the proximities of the membrane in corresponds to the fourth transfer step indicated in Fig. 2b and it
the lumenside; SID is the surface area available for mass transport can be calculated by using the correlation proposed by Gawronski
on the lumenside; CM,0 is the bulk concentration of Cu2+ in the and Wrzensika [28] for the shellside when the Reynolds number is
aqueous feed solution and CM,i is its concentration at the interface lower than 100:
(see Fig. 2b). The equation of Sieder and Tate [27] may be used to (0.8−0.16) 1/3
predict the individual mass transfer coefficient into the fiber for the Sh4 = 0.09 · (1 − ) · Re4 · Sc4 (20)
first transfer step when the Reynolds number is lower than 2100: where  is the packing density of the module and Sh4 , Re4 and Sc4
 din
1/3 are the dimensionless Sherwood, Reynolds and Schmidt numbers,
Sh1 = 1.86 Re1 · Sc1 · (14) respectively.
L
where Sh1 , Re1 and Sc1 are the dimensionless Sherwood, Reynolds 3.2. Determination of chemical kinetics of complex formation
and Schmidt numbers, respectively, L is the length of the fiber, and
din is its internal diameter. Experimental values of the mass transfer flow of Cu2+ , NM
In order to fit an effective relationship to estimate the reaction (mol s−1 ), have been used to estimate the concentrations of metal
rate of the complex formation at the aqueous-dense gas interface ion, CM,i , and extractant, CA,i , at the liquid–gas interface from Eqs.
described by Eq. (9), a simple kinetics equation can be proposed: (1) (2) (3)
(13), (18) and (19) under steady state condition (NM = NM = NM ),
(2)
NM = ˛
SID · keff · CM,i
ˇ
· CA,i

·C + (15) since the bulk concentrations of Cu2+ , CM,0 , and extractant, CA,0 ,
H ,i
are known. These concentration values were correlated in order
(2)
In Eq. (15), NM represents the overall reaction rate (mol s−1 ); to obtain the parameters keff , ˛, ˇ and  in Eq. (15) applying the
CM,i and CA,i are the concentrations of Cu2+ and extractant method described in a previous work [29] using the LabFit soft-
(mol m−3 ) at the interface, respectively. Considering the hydropho- ware to predict the transfer of Mo(VI) in a similar membrane
bic condition of the membrane, the reaction interface would be extraction process using a organic extraction phase. Thus, an effec-
placed at the pore entrance on the internal side of the fiber. Thus, tive chemical kinetics relationship can be obtained to quantify the
SID will be the surface area available for this reaction. rate of complex formation at the liquid–dense gas interface. The
At steady state conditions, the flow of metal ion through the kinetic model of Eq. (15) obtained from these calculations does not
(1)
liquid feed phase boundary layer (NM ) is equal to the flow of metal represent the real reaction mechanism at the proximities of the
ion-TFA complex through the pores. Thus, the flow of complex must membrane. However, it involves a first approach to extrapolate and
be equal to the stoichiometric relationship of the counterflow of scale-up the extraction capacity of the studied system under other
TFA from the bulk of the extraction phase to the interface at the pore fluid dynamic conditions.
entrance (NA ), which is represented by the following equation:
3.3. Physical and transport properties of the solutions
NA
NM = − (16)
z
Density values of the dense gas extraction phase were obtained
In this case, z = 2. from empirical information reported by Reynolds [30] and verified
If the mass transfer is driven by molecular diffusion, the transfer using Peng Robinson EOS. Viscosity of the CO2 phase was calculated
of extractant TFA in the third transfer step is represented by Eq. from the method proposed by Lucas [31], while the diffusion coef-
(17): ficient of extractant in dense CO2 was obtained from data reported
ε · DA–CO2 by Yang and Matthews [32].
(3)
NA = − · SAV · (CA,p − CA,i ) (17) On the other hand, density values of aqueous CuSO4 solutions
e·
were obtained from empirical information reported by Klein and
In this equation, DA–CO2 is the diffusion coefficient of the extrac- Alvarado [33] and its viscosity was calculated from the method
tant agent in the dense CO2 medium; SAV is the average surface proposed by Lucas [31].
area available for mass transport through the membrane pores Density, viscosity and diffusion coefficient values for both solu-
((SID + SOD )/2); CA,p and CA,i are the concentrations values of TFA at tions are summarized in Table 1.
the pore entrance on the shellside and at the interface, respectively
(see Fig. 2a); ε,  and e are the porosity, tortuosity and thickness of
4. Results and discussion
the membrane, respectively.
Eqs. (16) and (17) can be combined to estimate the flow of the
4.1. Effect of the fluid dynamics on the Cu(II) extraction
metal ion in the third transfer step, considering that the transfer of
performance
each mol of metal ion Mz+ through the aqueous solution boundary
layer requires z moles of TFA transferred to the interface under
In the first part of this study, the effect of the fluid dynamics
steady state conditions. Thus, the molar flow of Cu2+ is calculated
on the mass transfer rate in the membrane extraction process has
by Eq. (18). In this study, z = 2.
been evaluated. In this way, experimental runs according to the
(3) 1 ε · DA–CO2 procedure described in Section 2.3 were done in order to quantify
NM = · SAV · (CA,p − CA,i ) (18)
z e· the extraction percentage of Cu2+ as a function of the liquid feed
Considering these same assumptions, the flow of metal ion solution and dense CO2 phase flow rates. From these experiments
through the dense gas boundary layer in the shellside could be esti- the mass transfer fluxes (mol h−1 m−2 ) were calculated by means of
mated as a function of the TFA transfer by means of the following a mass balance described in Eq. (12). Fig. 4 shows these results for
equation: the extraction of Cu2+ from aqueous solutions at 90 bar and 40 ◦ C
when the liquid feed flow into the lumenside was ranged between
(4) k4 0.1 and 1.0 mL min−1 .
NM = · SOD · (CA,0 − CA,p ) (19)
z Fig. 4a shows the extraction efficiencies obtained from these
where CA,0 is the bulk concentration of extractant agent in the experiments, obtaining high extraction values, which are compara-
extraction phase; SOD is the surface area available for mass ble to the solvent extraction process using organic solvents [29,34].
 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67 63

Table 1
Properties of aqueous and CO2 phases for 70 and 90 bar.

Property Aqueous phase Dense phase Unit

Pressure: 90 bar Pressure: 70 bar Pressure: 90 bar Pressure: 70 bar

Density 996.08 995.22 483.81 198.19 kg m−3

Viscosity 6.645 × 10−4 6.620 × 10−4 3.615 × 10−5 2.017 × 10−5 kg m−1 s−1
Diffusion coefficient 2.695 × 10−9 a 2.706 × 10−9 1.406 × 10−8 b 1.489 × 10−8 m2 s−1
a
Diffusion coefficient of Cu2+ in water.
b
Diffusion coefficient of TFA in CO2 .

Higher residence time obtained from lower liquid feed solution Moreover, the average solvent/feed mass ratio obtained in this
flow rates across the hollow fiber contactor increases the extrac- work at 70 and 90 bar was 0.20 and 0.49 kg CO2 /kg of feed solution,
tion efficiency up to 98.7% in this work. The experimental mass respectively. These values could be considered relatively low if they
transfer fluxes obtained in this study cannot be directly compared are compared to a typical value of organic/aqueous volumetric ratio
to other systems described in the literature, since in our knowl- near unity in the mixing compartments of liquid–liquid extraction
edge, there is no previous reports on a similar configuration of plants [35].
membrane-based supercritical fluid extraction of metal ions from A clear influence of the liquid feed boundary layer on the mass
aqueous solutions. However, Fig. 4a showed extraction percent- transfer flux may be observed in the Fig. 4a and b. Predictions
ages close to 98.7% for the lowest feed solution flow rates, which obtained from Eq. (13) to estimate the Cu2+ flux supposing that
represent the highest residence times in the membrane contactor the concentration of the metal ion is practically equal to zero at
(2.78 min). These results represent extraction percentages close to the liquid–gas interface in the pore entrance (CM,i ) are included in
the values reported in literature [5] for supercritical fluid extrac- Fig. 4b. The theoretical curve obtained from Eq. (13) reports the
tion of metal ions in batch configuration, despite these last ones same behavior of the experimental transfer fluxes in function of
required close to 1 h of operating time. the liquid feed flow rate, which proves the significant effect of the
liquid feed boundary layer on the overall transfer of Cu2+ through
the membrane. In this first estimation, both theoretical calculations
overestimate experimental data even if they are of the same order
of magnitude, but the additional mass transfer resistances (at the
membrane pores as well as in the CO2 phase boundary layer) and
the complex formation must be considered for an accurate descrip-
tion of the process. This description is developed in the following
sections.
On the other hand, extraction experiments were done for dif-
ferent CO2 phase flow rate values, which were ranged between
40 and 90 mL min−1 (measured at 1 atm, 20 ◦ C) and a constant
aqueous feed flow rate of 0.7 mL min−1 . Transmembrane Cu(II) flux
obtained under these conditions was practically constant with a
value of 0.058 ± 0.0054 mol m−2 h−1 . Fig. 5 shows the transmem-
brane fluxes of Cu(II) obtained from these experiments. In this
case, no significant influence of this variable was observed for
the conditions applied in the experiments. This behavior is coher-
ent considering that the liquid feed boundary layer can represent

Fig. 4. (a) Extraction efficiency and b) Mass transfer flux values of Cu2+ (mol h−1 m−2 )
as a function of the aqueous feed solution flow rate (mL min−1 ) at 40 ◦ C and 90 bar. Fig. 5. Experimental values of mass transfer flux of Cu2+ (mol h−1 m−2 ) as a function
Initial concentration of CuSO4 in the feed phase = 3.147 mol m−3 , pH 2.0. Initial con- of the CO2 phase flow rate at 40 ◦ C and 90 bar. Initial concentration of CuSO4 in the
centration of TFA in the extraction CO2 phase = 17.716 mol m−3 in the extraction CO2 feed phase = 3.147 mol m−3 , pH 2.0, initial concentration of TFA in the extraction CO2
phase, CO2 phase flow rate = 2.079 × 10−3 mol min−1 . phase = 17.716 mol m−3 , aqueous feed flow rate = 0.7 mL min−1 .
64 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67

an important fraction of the overall mass transfer resistances and

the complex formation kinetic at the liquid–gas interface and the
complex transport through the membrane pores could represent
significant barriers on the mass transfer rate through the mem-
brane.

4.2. Chemical aspects of the extraction mechanism: the effect of

the extractant compound and pH

Extraction experiments of Cu2+ were carried out as a function of

liquid feed flow rate for two different extractant compounds. These
extractants were two different ␤-diketones: 1,1,1-trifluoro-2,4-
pentanedione (TFA) and 1,1,1,5,5,5-hexafluoro-2,4-pentanedione
(HFA), which were used considering the experimental data
reported in the literature about its high solubility in pressurized
CO2 and low solubility in water as well as its application in Cu(II)
extraction experiments [36,37]. Previous to the extraction experi-
ments, ␤-diketones were dissolved in CO2 at 50 bar and 25 ◦ C in a
view cell before to be injected in the extraction device. Thus, similar
mass of diketones was injected in the extraction system for each
experimental run. From this procedure, the concentration of TFA
and HFA in the dense CO2 phase before the extraction process was
17.7 mol m−3 and 18.4 mol m−3 , respectively. Thus, similar concen-
trations allow the same stoichiometric availability for the complex
formation.
TFA seems to be a better extractant compound than HFA. Fig. 6a
and b reports the extraction efficiencies and transmembrane fluxes
obtained from different experimental runs with both extractants.
The best extraction capacity of TFA could be related with the struc-
ture of both diketones compounds. The difference is the number
and position of fluorine atoms. TFA poses three fluorine atoms
located in the same carbon atom in one extreme of its structure
whereas HFA has six fluorine atoms positioned in the opposite
extreme of the molecule. It is well known that fluorine atom is
clearly electronegative attracting electrons and generating a posi-
tive pole in other atom of the compound. Diketones as extractants
act as cationic exchanger since due a tautomeric equilibrium, which
Fig. 6. (a) Extraction efficiency and (b) mass transfer flux values of Cu2+
originate around their structures an enolic isomer with an hydroxyl
(mol h−1 m−2 ) as a function of the aqueous feed solution flow rate (mL min−1 ) and
group which provides the acid hydrogen-ion which is exchangeable type of extract agent at 40 ◦ C and 90 bar. Initial concentration of CuSO4 in the feed
by a metallic ion, the Cu(II) ions in this study. In the case of TFA, the phase = 3.147 mol m−3 , pH 2.0, initial concentration of TFA in the extraction CO2
three fluorine atoms attract electrons towards its position facilitat- phase = 17.716 mol m−3 and HFA in the extraction CO2 phase = 18.37 mol m−3 , CO2
phase flow rate = 2.079 × 10−3 mol min−1 .
ing the keto-enol tautomerism around the carbon atom neighbor
generating the OH− group with the acid proton used in the cation
exchange equilibrium with Cu(II). In turn, HFA structure with three
fluorine atoms in opposite extreme of its structure bring about the
annulment of this fact because all F− atoms tend to attract electrons done under the same operation conditions. Fig. 7 shows that the
towards their opposite side preventing the keto-enol tautomeric extraction efficiency at pH 2.0 is 3.5 times higher than the obtained
equilibrium and making difficult the formation of hydroxyl-group one at pH 4.0. Chou and Yang [38] suggest that the pH of the solu-
with the acid-proton. In the HFA extractant molecule, it is not pos- tion should be a critical parameter in the extraction efficiency of
sible the occurrence of a kind of a double keto-enol tautomeric indium(III) ions using a ␤-diketone (AcAcH) as extractant com-
equilibrium to each side of the molecule since would require two pound. A previous study [39] showed that when water is in contact
double-bonds in a same central carbon atom. Two double bonds are with supercritical CO2 , the pH value at the equilibrium is less than
possible in a conjugate structure as in the aromatic-ring containing 3.0 due to the formation and dissociation of an important amount
compounds. of carbonic acid in water. Therefore, the optimum pH values for
In Fig. 6a is clear that a higher aqueous flow rate decreases the dense CO2 extraction should be ranged from 2.0 to 3.0.
residence time of the copper-containing solution inside the reactor, The influence of the acidity of the aqueous solution on the
which avoid obtaining higher equilibrium concentration of metal extraction performance showed in Fig. 7a and b would confirm
complex in the CO2 phase. that is required a high content of protons (lower pH values) to
On the other hand, Fig. 6b reports that higher aqueous feed flow facilitate and stabilize the formation of the keto-enol tautomerism.
rate values involve higher metal fluxes through the membrane, A higher pH could revert the equilibrium towards the keto-form,
which show that the overall transport rate could be determined which clearly has lower extraction capacity.
by the chemical kinetic step, since a greater concentration of cop- In this work, the maximum operation pressure was 90 bar and
per ions available at the reaction interface would be increasing the the highest extraction efficiencies (98.7%) were obtained at pH
complex formation rate. 2.0. Under these conditions, corrosion was not observed in the
On the other hand, a significant effect of the pH of the liquid experimental setup during the 18 weeks spent to achieve these
feed solution can be observed from different extraction experiment experimental runs.
 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67 65

Fig. 8. Experimental data of metal ion (Cu2+ ) transfer flux as a function of the aque-
ous feed flow rate for different pH values and its prediction using the mass transfer
model proposed in this work. Operating pressure = 70 bar; initial concentration of
CuSO4 in the feed phase = 3.147 mol m−3 ; initial concentration of TFA in the extrac-
tion CO2 phase = 17.716 mol m−3 and CO2 phase flow rate = 2.079 × 10−3 mol min−1 .

improvement of the extraction efficiency when HFA was used at

90 bar.
The effect of the pressure on the extraction efficiency and the
mass transfer flux could be explained by the solubility on CO2 and
the subsequent availability of carbonic acid in the aqueous phase.
Fig. 7 shows that the extraction efficiency and the mass transfer
rate are improved at lower pH values. The pH reported in this
figure represent its initial value in the feed aqueous solution cir-
culated through the membrane contactor. The real pH value in the
aqueous phase during the extraction process could have decreased
due to the solubility of pressurized CO2 . Thus, the effect of pH and
pressure could be coupled. In this case, the extraction mechanism
Fig. 7. (a) Extraction efficiency and (b) mass transfer flux values of Cu2+
(mol h−1 m−2 ) as a function of the aqueous feed solution flow rate (mL min−1 )
and physicochemical nature of CO2 generate a coupled effect of
for different pH values at 70 bar. Initial concentration of CuSO4 in the pressure and pH value on the mass transfer rate.
feed phase = 3.147 mol m−3 . Initial concentration of TFA in the extraction CO2
phase = 17.716 mol m−3 , CO2 phase flow rate = 2.079 × 10−3 mol min−1 .
4.4. Phenomenological approach for determination of complex
formation kinetic
4.3. Influence of the pressure on the mass transfer rate
From the phenomenological Eqs. (13), (18) and (19) as well as
The operation pressure should be an important variable on the the experimental metal ion transfer flow values (NM ), the interfa-
extraction efficiency, since the pressure can have a significant influ- cial concentrations of Cu2+ , CM,i , and extractant, CA,i , were directly
ence on the solubility of ␤-diketones and its metal complexes in the estimated in order to correlate the parameters of Eq. (15). The con-
CO2 phase, as well as the availability of dissolved CO2 and the pH centration of protons at the interface, CH+ ,i , can be considered much
in the aqueous phase. Fig. 6 shows the effect of the pressure value higher than the metal ion and the extractant ones and its value
on the extraction efficiency and transmembrane flux of Cu(II) ions will be similar to bulk concentration. Thus, Eq. (15) represents the
obtained from experiments as a function of the liquid feed flow rate effective complex formation rate at the interface as a function of
and two different extractants. the concentrations of Cu2+ , extractant and protons.
It is possible to observe that the pressure increasing from 70 to Several simple partial orders of reaction ˛, ˇ and  were tested
90 bar improves the extraction efficiency in all the cases. However, using LabFit. The best correlation was obtained when reaction
the magnitude of this improvement is different depending on the orders respect to Cu2+ (˛), extractant (ˇ) and protons () in Eq.
extractant compound. The results reported in Fig. 6 consider simi- (15) were 2.0, 1.0 and 0.642, respectively. keff is the effective kinetic
lar concentration (mol m−3 ) for both extractants (TFA and HFA) in constant of the complex formation and its value was estimated in
the CO2 phase that is fed in the extraction system. Higher extrac- 1.4313 × 10−7 (mol2.642 m8.926 s−1 ).
tion efficiencies were obtained using TFA, which has a chemical Fig. 8 shows the experimental transmembrane Cu ions fluxes
structure favorable to the cationic exchange mechanism explained (mol m−2 h−1 ) compared to the values obtained by the correlated
in Section 4.2. model. Good agreement between the model and experimental data
On the other hand, the increasing of pressure in the extraction is expected except for predictions at very low aqueous feed solution
experiments represents a significant density increase in the CO2 flow rate. This could be explained by limited prediction capacity
phase, from 198.2 kg m−3 at 70 bar to 483.8 kg m−3 at 90 bar. The of the effective reaction equation or an inaccurate prediction of
density increase can be associated to an important increasing of the the mass transfer coefficients at the proximities of the membrane
complex solubility in the CO2 phase, which can explain the large when flow rates or Reynolds number are very low. The Reynolds
66 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67

Table 2
Contribution of local transfer steps to the global mass transfer resistance. Results obtained from experimental data obtained at 40 ◦ C and 70 bar. Initial concentration of CuSO4
in the feed phase = 3.147 mol m−3 ; initial concentration of TFA in the extraction CO2 phase = 17.716 mol m−3 and CO2 phase flow rate = 2.0785 × 10−3 mol min−1 .

pH Feed flow rate (mL min−1 ) Resistance

Contribution of the aqueous Contribution of kinetic of Contribution of the gas gap Contribution of the CO2
feed boundary layer (%) the complex formation (%) in the membrane pores (%) phase boundary layer (%)

0.1 46.29 6.67 5.26 41.77

0.4 35.19 8.05 6.35 50.41
2.0
0.7 31.06 8.57 6.76 53.62
1.0 28.57 8.87 7.00 55.55

0.1 20.88 57.90 2.37 18.84

0.4 14.26 62.75 2.57 20.42
4.0
0.7 12.13 64.31 2.64 20.93
1.0 10.91 65.19 2.67 21.22

number into the lumenside of the single fiber contactor is lower aspects such as pH, pressure and structure of the extractant com-
than 6.5 when the aqueous feed solution flow rate is lower than pound are different from what is usually expected in solvent extrac-
0.2 mL min−1 . tion using cation exchangers. In this way, the influence of the acidity
of the aqueous solution on the extraction performance could con-
firm that is required a high content of protons to facilitate and stabi-
4.5. Assessment of the local mass transfer resistances
lize the formation of a keto-enol tautomerism during the complex
formation. The effect of pH and pressure may be coupled in the reac-
The relevance of every local transport resistances can be quan-
tion mechanism, since the solubility of CO2 in the aqueous phase
tified from the mass transfer model coupled to the complex
could modify the initial content of protons in the aqueous solution.
formation rate equation. In a previous work, Valdés and coworkers
From the experimental results and phenomenological equations
[29] estimated the contributions of different mass transfer steps on
an effective complex formation rate was estimated at the aqueous-
the basis in the local transport resistances for the boundary layers,
CO2 interface. This effective kinetic equation allows correlating the
membrane pores and reaction rate in the membrane-based sol-
transfer of Cu2+ through the porous membrane and it could be used
vent extraction of Mo(VI). According to the resistances-in-series
as a tool to scale-up this process. Nevertheless, this kinetic model
model, the rate-controlling steps can be identified by means of the
is not useful to understand the real mechanism of the complex for-
quantification of the fractional resistances. These calculations can
mation. Thus, the fit of correlated data is poor at low circulation
be adapted to the system studied in this work, since it represents
rates when the complex formation reaction is highly influenced by
the same process configuration. Table 2 summarizes the fractional
the transfer of reactants from the bulk of both phases to the prox-
transfer resistances for each contribution as a function of pH and
imities of the membrane. Future researches should be focused on
aqueous feed flow rate.
the introduction of the mechanistic aspects in the transfer model
From the results reported in Table 2 two mass transfer behaviors
in order to improve its prediction capacity and the design of new
influencing the extraction process can be observed. When the pH
intensified extraction processes for other metal ions or hazardous
value of the aqueous solution is 2.0 fractional mass transfer resis-
materials.
tances are concentrated on the boundary layers at the proximities
of the membrane. The effect of the complex formation rate and the
complex transfer through the membrane pores can be considered
Acknowledgments
not significant on the overall resistance. Thus, Fig. 4b shows the the-
oretical curve of transmembrane metal ion flux predicted by means
This research has been developed in the framework of the
of Eq. (13) when pH is 2.0, supposing that the concentration of Cu2+
Projects FONDECYT 1100305 (CONICYT Chile) and ECOS-CONICYT
at the aqueous-CO2 interface, CM,i , is equal to cero. The predicted
C10E05. The financial support for the PhD student Hugo Valdés is
flux value represents the maximum value that could be observed
gratefully acknowledged.
through the aqueous feed boundary layer.
On the other hand, when pH value is 4.0 fractional mass trans-
fer resistances of boundary layers continue contributing more than
References
30% of the overall resistance. However, the complex formation rate
increases its contribution, becoming the rate-controlling step of [1] S.M. Walas, Phase Equilibria in Chemical Engineering, Butterworth-
this process. The reaction rate can represent 65.19% of the overall Heinemann, Boston, 1985.
mass transfer resistance under this condition. [2] J. Murphy, C. Erkey, Copper(II) removal from aqueous solutions by chelation
in supercritical carbon dioxide using fluorinated ␤-diketones, Industrial and
Engineering Chemistry Research 36 (1997) 5371.
[3] J. Murphy, C. Erkey, Thermodynamics of extraction of copper(II) from aqueous
5. Conclusions solutions by chelation in supercritical carbon dioxide, Environmental Science
& Technology 31 (1997) 1674.
In this work, dense gas extraction of Cu2+ from aqueous solu- [4] K. Laintz, C. Hale, P. Stark, C. Rouquette, J. Wilkinson, A comparison of liquid and
supercritical carbon dioxide as an extraction solvent for plating bath treatment,
tions has been implemented in a continuous configuration using a
Analytical Chemistry 70 (1998) 400–404.
hydrophobic hollow fiber contactor. The extraction phase was pres- [5] C. Erkey, Supercritical carbon dioxide extraction of metal ions from aqueous
surized CO2 containing ␤-diketones. High extraction efficiency up solutions: a review, Journal of Supercritical Fluids 17 (2000) 259–287.
[6] K. Soldenhoff, M. Shamieh, A. Manis, Liquid–liquid extraction of cobalt with
to 98.7% can be obtained under steady state operation. The mass
hollow fiber contactor, Journal of Membrane Science 252 (2005) 183–194.
transfer fluxes through the membrane seems to be higher than [7] Y.J. Wang, Y. Wang, F. Chen, G.S. Luo, Y.Y. Dai, Mass transfer characteristics of
similar membrane-based extraction processes reported in the lit- cadmium(II) extraction in hollow fiber modules, Chemical Engineering Science
erature using organic solvents, since transport properties of the 58 (2003) 3223–3231.
[8] R.S. Juang, H.C. Kao, Extraction separation of Co(II)/Ni(II) from concentrated HCl
extraction phase are closer to the gas phase. However, the response solutions in rotating disc and hollow-fiber membrane contactors, Separation
of the studied system to the operation variables related to chemical and Purification Technology 42 (2005) 65–73.
 H. Valdés et al. / Chemical Engineering and Processing 65 (2013) 58–67 67

[9] F.J. Alguacil, Facilitated transport and separation of manganese and cobalt by a [24] S. Bocquet, A. Torres, J. Sanchez, G. Rios, J. Romero, Modeling the mass transfer
supported liquid membrane using DP-8R as a mobile carrier, Hydrometallurgy in solvent-extraction processes with hollow-fiber membranes, AIChE Journal
65 (2002) 9–14. 51 (2005) 1067–1079.
[10] F. Valenzuela, M. Vega, M. Yáñez, C. Basualto, Application of a mathematical [25] A. Gabelman, S. Hwang, Experimental results versus model predictions for
model for copper permation from a Chilean mine water through a hollow fiber- dense gas extraction using a hollow fiber membrane contactor, Journal of
type supported liquid membrane, Journal of Membrane Science 204 (2002) Supercritical Fluids 35 (2005) 26–39.
385–400. [26] A. Gabelman, S. Hwang, A theoretical study of dense gas extraction using hollow
[11] A. Gabelman, S. Hwang, Hollow fiber membrane contactors, Journal of Mem- fiber membrane contactor, Journal of Supercritical Fluids 37 (2006) 157–172.
brane Science 159 (1999) 61–106. [27] R. Bird, W. Stewart, E. Lightfoot, Fenómenos de Transporte, 1st ed., Reverté S.A.,
[12] J.G. Crespo, I.M. Coelhoso, R.M.C. Viegas, Membrane contactors: membrane México, 1998.
separations, in: Encyclopedia of Separation Science, 2000, pp. 3303–3311. [28] R. Gawronski, B. Wrzesinska, Kinetics of solvent extraction in hollow fiber
[13] K. Sirkar, Membrane separation technologies: current developments, Chemical contactors, Journal of Membrane Science 168 (2000) 213–222.
Engineering Communications 157 (1997) 145–184. [29] H. Valdés, J. Romero, J. Sanchez, S. Bocquet, G. Rios, F. Valenzuela, Character-
[14] J. Robinson, M. Sims, Method and system for extracting a solute from a fluid ization of chemical kinetics in membrane-based liquid–liquid extraction of
using dense gas and a porous membrane, U.S. Patent 5,490,884, 1996. molybdenum(VI) from aqueous solutions, Chemical Engineering Journal 151
[15] M. Sims, Porocritical fluid extraction from liquids using near-critical fluids, (1–3) (2009) 333–341.
Membrane Technology 97 (1998) 11–12. [30] W. Reynolds, Thermodynamic Properties in SI: Graphs, Tables and Computa-
[16] S. Bocquet, J. Romero, J. Sanchez, G. Rios, Membrane contactors for the extrac- tional Equations for Forty Substances, Department of Mechanical Engineering,
tion process with subcritical carbon dioxide or propane: simulation of the Stanford University, Stanford, CA, 1979.
influence of operating parameters, Journal of Supercritical Fluids 41 (2) (2007) [31] B. Poling, J. Prausnitz, J. O’Connell, The Properties of Gases and Liquids, 5th ed.,
246–256. McGraw Hill, 2001.
[17] H. Estay, Modeling and simulation of the mass transfer and selectivity of mem- [32] X. Yang, M. Matthews, Diffusion of chelating agents in supercritical CO2 and
brane contactor systems using supercritical fluids. Master Thesis, Chemical a predictive approach for diffusion coefficients, Journal of Chemical and Engi-
Engineering, University of Santiago de Chile, 2005. neering Data 46 (2001) 588–595.
[18] H. Estay, S. Bocquet, J. Romero, J. Sanchez, G. Rios, F. Valenzuela, Modeling [33] Y. Çengel, M. Boles, Thermodynamics, McGraw Hill, 2006.
and simulation of mass transfer in near-critical extraction using a hollow fiber [34] F. Valenzuela, C. Basualto, C. Tapia, J. Sapag, C. Paratori, Recovery of copper from
membrane contactor, Chemical Engineering Science 62 (21) (2007) 5794–5808. leaching residual solutions by means of a hollow fiber membrane extractor,
[19] G. Bothun, B. Knutson, H. Strobel, S. Nokes, E. Brignole, S. Díaz, Compressed Minerals Engineering 9 (1) (1996) 15–22.
solvents for the extraction of fermentation products within a hollow fiber [35] K.C. Sole, Solvent extraction in the hydrometallurgical processing and purifi-
membrane contactor, Journal of Supercritical Fluids 25 (2003) 119–134. cation of metals: process design and selected applications, in: M. Aguilar, J.M.
[20] M. Sims, E. McGovern, J. Robinson, Porocritical fluid extraction application: Cortina (Eds.), Solvent Extraction and Liquid Membranes: Fundamentals and
continuous pilot extraction of natural products from liquids with near-critical Applications in New Materials, CRC Press, Boca Raton, 2008, pp. 141–200.
fluids, in: Proceeding of 5th Meeting on Supercritical Fluids, Materials and [36] W. Cross, A. Akgerman, C. Erkey, Determination of metal-chelate complex solu-
Natural Processing, Nice, France, 1998. bilities in supercritical carbon dioxide, Industrial and Engineering Chemistry
[21] M. Sims, E. Estigarribia, Continuous sterilization of aqueous pumpable food Research 35 (1996) 1765–1770.
using high pressure carbon dioxide, in: 4th International Symposium on High [37] R. Gupta, J. Shim, Solubility in Supercritical Carbon Dioxide, Taylor & Francis
Pressure Process Technology and Chemical Engineering, Venice, Italy, 2002, pp. Group, LLC, 2007.
22–25. [38] W. Chou, K. Yang, Effect of various chelating agents on supercritical carbon
[22] M. Sims, E. Estigarribia, Membrane carbon dioxide sterilization of liquid food: dioxide extraction of indium(III) ions from acidic aqueous solution, Journal of
scale up of a commercial continuous process, in: 6th International Symposium Hazardous Materials 154 (1–3) (2008) 498–505.
on Supercritical Fluids, Versailles, France, 2003, pp. 28–30. [39] K. Toews, R. Shroll, C. Wai, N. Smart, pH-defining equilibrium between water
[23] M. Sims, Simultaneous reaction and extraction with near-critical carbon diox- and supercritical CO2 . Influence on SFE of organics and metal chelates, Analyt-
ide using a membrane contactor, in: International Symposium on Supercritical ical Chemistry 67 (1995) 4040–4043.
Fluids, ISSF 2005, Orlando, USA, 2005, pp. 1–4.