J.

of Supercritical Fluids 27 (2003) 239
/245

www.elsevier.com/locate/supflu

Solubility of three hydroxycinnamic acids in supercritical

carbon dioxide
Ruth Murga, Mara Teresa Sanz, Sagrario Beltran *, Jose Luis Cabezas
Departamento de Ingeniera Qumica, Universidad de Burgos, Pl. Misael Banuelos s/n, Burgos 09001, Spain

Received 27 May 2002; received in revised form 21 November 2002; accepted 4 December 2002

Abstract

The solubility of three trans -hydroxycinnamic acids, 4-hydroxycinnamic acid (p -coumaric acid), 3,4-dihydroxycin-
namic acid (caffeic acid), and 4-hydroxy-3-methoxycinnamic acid (ferulic acid), in supercritical carbon dioxide (SC-
CO2) has been determined at different pressures up to 50 MPa and temperatures from 313 to 333 K. These
hydroxycinnamic acids are phenolic compounds contained in grape seeds and other natural substrates, and the data
presented in this work are valuable to know the possibility of their separation from such natural matrices by extraction
with SC-CO2. Experimental data have been modeled by using the Peng
/Robinson equation of state (PR-EOS) to
determine the fugacities of the fluid phase. Two semiempirical density dependent correlations have also been used for
data correlation, specifically, the Chrastil model, and a model that assumes a linear correlation between the
enhancement factor and the density of the solvent.
# 2003 Elsevier Science B.V. All rights reserved.

Keywords: Solubility; Supercritical carbon dioxide; Hydroxycinnamic acids; Natural phenolic compounds; Grape seeds

1. Introduction properties. Supercritical fluid extraction (SCFE)

The search of natural products having antiox-
idant properties, that could replace the use of
synthetic antioxidants, or could be used directly
for therapeutics, is being the subject of research of
many groups in the last years. Nature is rich of
such type of products, but they need to be
separated from their natural matrices as selectively
as possible and without loosing their antioxidant
* Corresponding author. Tel.: /34-947-25-8810; fax: /34-
947-25-8831.
E-mail address: beltran@ubu.es (S. Beltran).
0896-8446/03/$ - see front matter # 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0896-8446(02)00265-6
has been shown to be a suitable technique to do
such a job efficiently [1]. However, if an efficient
SCFE process is to be designed, the solubility of
the compounds to be extracted needs to be known.
The work presented in this paper is part of a
project whose aim is to study the possibility of
using supercritical carbon dioxide (SC-CO2) to
selectively extract some complex phenols found in
grape seeds [2]. In a previous publication [3],
solubilities for protocatechuic acid, methyl gallate,
and protocatechualdehyde were reported. In this
work, the solubilities of three trans -hydroxycin-
namic acids, p-coumaric acid, caffeic acid, and
ferulic acid in pure SC-CO2, have been measured
240 R. Murga et al. / J. of Supercritical Fluids 27 (2003) 239
/245
using the dynamic analytical method. The Trolox
equivalent antioxidant activity (TEAC) of these
compounds has been reported to be 2.22 mM for
of p -coumaric acid, 1.26 mM for caffeic acid, and
1.9 mM for ferulic acid [4]. The TEAC is defined
as the concentration of Trolox (the water-soluble
vitamin E analog) solution with an antioxidant
potential equivalent to a 1 mM concentration of
the compound under consideration.
Choi et al. [5] have previously reported the
solubility of p -coumaric acid covering a range of
pressures from 8.5 to 25 MPa and a range of
temperatures from 308 to 323 K. In this work both
ranges have been widened reaching pressures up to
50 MPa and temperatures up to 333 K. Correla-
tion of the experimental data has been made using
two types of models: a rigorous thermodynamic
method [6] where the Peng
/Robinson equation of
state (PR-EOS) was used to determine the fuga-
cities of the fluid phase, and two semiempirical
density dependent correlations: a linear correlation
between the enhancement factor and the density of
the solvent, as suggested by Schmitt and Reid [7],
and the Chrastil correlation [8]. These semiempi-
rical correlations are widely used and, although
they are not capable of predicting unknown phase
equilibria, they are useful tools for experimental
data correlation.
2. Experimental
2.1. Chemicals
The solutes used in this work, p -coumaric acid,
caffeic acid, and ferulic acid were supplied by
Sigma, (
/97% purity). The molecular weight
(MW) and normal melting point (mp) of these
Table 1
Physical properties of the solutes
Solute Formula R MW
p -Coumaric acid C9H8O3 /H 164.16
Caffeic acid C9H8O4 /OH 180.16
Ferulic acid C10H10O4 /OCH3 194.19
a
Sigma data base (http://www.sigma-aldrich.com/saws.nsf/msdshelp).
b
Estimated by the Joback method implemented in PE [9].
solutes are reported in Table 1 together with the
estimated normal boiling point (Tb), critical tem-
perature (Tc), critical pressure (pc), and acentric
factor (v ). The chemical structure of these hydro-
xycinnamic acids is given below. The R group for
each compound is also given in Table 1.
The carbon dioxide used as supercritical solvent
(SFC/SFE quality) was supplied by Air Liquide.
HPLC-grade methanol and acetonitrile (Lab-
Scan) and puriss. p.a. (98%) formic acid (Fluka)
were used as solvents for sample analyses. Glass
beads 30/60 mesh (Phase Separations) were used to
distribute the solute in the equilibrium cell, and
glass wool (Panreac) was placed at the top of each
of the three stages of the equilibrium cell to
prevent solid entrainment.
2.2. Apparatus and procedure
The solubility of the solids was experimentally
determined by the dynamic analytical method in
an apparatus that has been previously described
and used in our laboratory to obtain solubility
data of solids in SC-CO2 [3]. The equilibrium cell
was filled with the solute ( :/200 mg) distributed in
glass beads in order to improve the solute
/solvent
contact, and to avoid channeling when the SC-
solvent passes through. The solvent flow was
adjusted to be as low as possible in order to
achieve solvent saturation, and thus, equilibrium.
The equilibrium cell was kept at the desired
temperature (9/0.1 K) by placing it inside of an
oven provided with a thermal regulator. A back
pressure regulator (Tescom Co., model 26-1721)
mp (101.33 kPa) (K)a Tb (101.33 kPa) (K)b Tc (K)b pc (MPa)b vb
4889/2 662.3 905.50 4.685 0.9682
4699/2 742.9 993.05 5.809 1.2881
4439/2 712.6 947.97 3.661 1.1103
 R. Murga et al. / J. of Supercritical Fluids 27 (2003) 239
/245
allowed the pressure to be controlled with an
accuracy of 9/1% of relief pressure range. A 5 mm
in-line filter and a column packed with stainless
steel balls were used for solute trapping after CO2
depressurization. Filter and trap were rinsed
repeatedly with an 80% methanol aqueous solu-
tion in order to collect the solute quantitatively
and have the sample ready for analysis. The
depressurized CO2 was quantified with a totalizer
flow meter.
Analysis of the solutes was carried out off-line
by using an HPLC equipped with a diode array
detector (DAD) (Hewlett
/Packard 1100 series).
The coupling of chromatography and DAD would
allow to detect the compound degradations that
could take place [10], or some of the impurities of
the pure solute that could be preferentially solved
by the SC-CO2. None of those was the case with
the solutes studied in this work. Quantification
was made at a wavelength of 280 nm for p -
coumaric acid, and at 320 nm for caffeic and
ferulic acids [11], with the calibration curves
previously obtained for each compound in the
range where the Lambert
/Beer law was valid.
Solids solubility was determined by carrying out
at least five experiments in which the total amount
of SC-CO2 that flowed through the cell was varied,
hence varying the total amount of solute dissolved.
The absolute amount of SC-CO2 used depended
on the solubility of the solute; it was large when
solubilities were low, and viceversa. The plot of the
amount of solute solved versus the amount of CO2
used to dissolve it, should be a linear function
whose slope is the solubility at the temperature
and pressure at which the operation was carried
out. The standard error (S.E.) given next to every
solubility datum in Table 2 indicates the quality of
the linear fit.
3. Results and discussion
The experimental solubility data of p-coumaric
acid, caffeic acid, and ferulic acid in SC-CO2, at
different conditions of pressure, p , and tempera-
ture, T , are listed in Table 2 and plotted in Figs.
1
/3, together with their S.E. The temperatures
studied range from 313 to 333 K and the pressures
241
from 8.5 to 50 MPa. The density of CO2 at the
given pressure and temperature, obtained from the
IUPAC international thermodynamic tables [12],
has also been included in Table 2.
The results show that the solubility increased
with pressure, at constant temperature, in all cases.
The effect of temperature is more complex and
retrograde solubility (crossover pressure) behavior
[7] can be observed for ferulic acid and for p -
coumaric acid. At pressures above the crossover
pressure, the solubility increases with temperature
while at pressures below the crossover pressure,
the solubility decreases with increasing tempera-
ture. Caffeic acid does not show this behavior in
the range of pressures studied in this work, where
its solubility in SC-CO2 increases with tempera-
ture.
Under the same conditions of temperature and
pressure, ferulic acid shows a higher solubility in
SC-CO2 than p-coumaric acid whose solubility is
higher than that of caffeic acid. This behavior
could be expected since, as has been previously
reported [13], the addition of a hydroxyl group
generally results in a decrease of the solubility of
the new molecule (caffeic acid is formed from p -
coumaric acid by addition of a hydroxyl group),
and partial etherification (ferulic acid is formed by
etherification of caffeic acid) generally causes a
large increase in solubility.
The most rigorous methods for modeling equili-
brium data consist of calculating the fugacity of
the components in the different phases. The fluid
phase in the type of systems studied in this work
can be considered as a dense gas or as an expanded
liquid, although it is most commonly considered as
a dense gas. In this case, the solubility of the solid
solute in the SCF can be calculated by means of
Eq. (1), which considers a solid phase, formed by
the pure solute (2), in equilibrium with a fluid
phase formed by a mixture of solvent (1) and
solute (2) [6].
ps2
y2  E (1)
p
where
242 R. Murga et al. / J. of Supercritical Fluids 27 (2003) 239
/245

Table 2
Experimental solubility data of caffeic acid, ferulic acid and p -coumaric acid in SC-CO2: pressure, p ; temperature, T , SC-CO2 density
at each p and T , rCO2, solubility expressed as solute mole fraction, y , and standard error of each solubility datum, S.E.

T (K) P (MPa) rCO2 (kg m 3) p -Coumaric acid Caffeic acid Ferulic acid
7 7 7 7
y /10 S.E. /10 y /10 S.E. /10 y /107 S.E. /107

313 8.5 442 0.069 0.004

10 615 0.286 0.016 8.225 0.233
15 781 1.545 0.073 0.008 0.001 36.680 0.889
20 841 2.031 0.064 0.020 0.002 65.480 2.014
25 881 2.730 0.098 0.052 0.001 88.290 2.894
30 911 3.890 0.055 0.092 0.007 112.806 2.167
35 936 4.879 0.214 0.118 0.004 138.776 3.664
40 957 6.032 0.241 0.133 0.002 151.293 4.960
45 976 7.131 0.295 0.154 0.005 169.392 1.127
50 992 8.629 0.380 0.182 0.007 188.078 5.067
323 8.5 254 0.059 0.007
10 408 0.079 0.011 1.119 0.075
15 701 1.663 0.123 0.011 0.000 40.256 0.967
20 785 2.628 0.331 0.042 0.001 84.355 2.052
25 835 3.629 0.051 0.118 0.009 134.271 4.052
30 872 5.259 0.143 0.189 0.004 155.990 4.877
35 900 8.241 0.252 0.231 0.011 186.018 12.589
40 924 10.576 0.563 0.270 0.015 208.383 6.694
45 945 12.171 0.590 0.298 0.013 239.222 7.748
50 963 13.804 0.158 0.324 0.013 266.124 9.712
333 10 295 0.032 0.001 0.653 0.028
15 607 1.886 0.238 0.032 0.000 38.310 0.870
20 725 6.914 0.198 0.106 0.000 108.507 2.397
25 788 10.648 0.471 0.182 0.015 182.259 3.987
30 831 13.028 0.246 0.265 0.010 244.847 11.595
35 864 15.895 0.655 0.333 0.011 314.725 3.918
40 891 20.296 0.563 0.393 0.018 368.282 13.582
45 914 22.949 1.072 0.436 0.016 400.097 7.487
50 935 25.523 0.131 0.472 0.018 433.239 14.571

Fig. 1. Experimental solubility of p -coumaric acid in SC-CO2; Fig. 2. Experimental solubility of caffeic acid in SC-CO2; (^)
(^) 313 K, (I) 323 K, (k) 333 K. The error bars represent the 313 K, (I) 323 K, (k) 333 K. The error bars represent the S.E.
S.E. of each solubility data. Continuous lines represent the of each solubility data. Continuous lines represent the solubility
solubility isotherms calculated with the PR-EOS. isotherms calculated with the PR-EOS.
 R. Murga et al. / J. of Supercritical Fluids 27 (2003) 239
/245
Fig. 3. Experimental solubility of ferulic acid in SC-CO2; (^)
313 K, (I) 323 K, (k) 333 K. The error bars represent the S.E.
of each solubility data. Continuous lines represent the solubility
isotherms calculated with the PR-EOS.


fs2 ys2 (p  ps2 )
E exp ; region, while larger errors are found close to it.
f2 RT
ps2 is the saturation (vapor) pressure of the pure
solid, f2 the fugacity coefficient at the system
pressure p , fs2 the fugacity coefficient at saturation
pressure ps2, and ys2 the solid molar volume, all at
the system temperature T . The enhancement
factor (E ) provides a measure of the extent that
pressure enhances the solubility of the solid in the
gas (see Eq. (1)) and it is always greater than unity.
The PR-EOS [6], whose parameters for the
mixture have been calculated according to the
quadratic mixing rules [6], has been used for
calculating the fugacity coefficients.
The PR-EOS gives good quantitative fits for a
wide variety of systems, but critical properties (Tc
and pc) and the acentric factor (v ) of solute and
solvent are needed for calculating its parameters
for the pure components. These properties are
available for CO2, but not for the solids subject of
this work. Therefore, they have been estimated by
the Joback group contribution method [9] and are
listed in Table 1. This group contribution method
has also been used for prediction of the normal
boiling temperature of the solutes.
The interaction binary parameters, k12 and l12,
of the PR-EOS have been obtained by fitting the
experimental solubility data to Eq. (1). The
regression procedure was carried out by minimiz-
ing the average absolute relative deviation
(AARD) between experimental (yexp) and calcu-
243
lated (ycal) solubility data, (see Eq. (2)) using the
Phase Equilibria 2000 (PE 2000) program devel-
oped by Professor Brunners group [9].
 
100 X yexp  ycal
AARD (%)  (2)
n yexp
In this equation, n is the number of solubility data
used for obtaining the parameters.
The values of the interaction parameters k12,
and l12 for the different temperatures are listed in
Table 3 together with the corresponding AARD
values. The solubility curves as predicted by Eq.
(1) and the PR-EOS are represented by continuous
lines in Figs. 1
/3, where it may be observed that
the model reproduces fairly well the experimental
solubilities far from the retrograde solubility
This may be due to the large scatter that experi-
mental solubility data usually present in the
vicinity of the critical point as a consequence of
the high sensitivity that density presents to pres-
sure.
Two semiempirical models, have also been used
for data correlation. The first one, proposed by
Schmitt and Reid [7] assumes a linear functionality
between the enhancement factor, as calculated
from Eq. (1), and the density of the solvent; that is,
ln E ABr (3)
The vapor pressure of the solid, necessary to
calculate the enhancement factor using Eq. (1), has
been estimated according to the method proposed
by Grain [14]. Eq. (3) is very useful for data
correlation because of its simplicity and good fits.
The results of data correlation to this model are
reported in Table 4 together with the AARD
between experimental and calculated solubility.
The quality of the linear correlation is indicated
by s -squared. It may be observed that the model
reproduces much better the experimental data
when the parameters are obtained without includ-
ing the solubilities obtained at pressures in the
vicinity of the critical point of carbon dioxide.
The second empirical correlation used, was the
Chrastil correlation [8] that assumes the formation
of a solvato complex between molecules of the SC-
solvent and the solute at equilibrium. This model
244 R. Murga et al. / J. of Supercritical Fluids 27 (2003) 239
/245

Table 3
Results of the solubility data correlation through the PR-EOS: number of data points used in the correlation, n , binary interaction
parameters, kij and lij , and percentage of AARD

Solute n T (K) Parameters of PR-EOS

kij lij AARD (%)

p -Coumaric acid 10 313 0.3732 0.2657 6.41

10 323 0.3842 0.2922 10.17
9 333 0.2880 0.0635 7.58
Caffeic acid 8 313 0.1512 /0.1189 6.03
8 323 0.0551 /0.4014 7.73
8 333 0.0498 /0.4416 1.17
Ferulic acid 9 313 0.0362 /0.1923 14.84
9 323 0.0019 /0.2717 20.24
9 333 /0.0306 /0.3299 21.37

leads to the following linear relationship between
the solubility of the solute, C , expressed as (g of
solute)/(L of solvent), and the density of the
solvent, r , in g/L, for a given temperature (in K):
ln C  k ln ra=T b
a , b and k are the adjustable parameters of the
model. The Chrastil equation has the advantage of
having only three parameters to fit all the experi-
mental data, no matter at which temperature they
were obtained. Moreover, this equation does not
require the estimation of the properties of the pure
components.
The parameters a , b, and k were obtained by
nonlinear regression of the experimental data
using the Marquardt algorithm, and are reported
(4) in Table 5 together with the AARD between
experimental and calculated solubility, C . The
quality of the correlation is indicated by s -
squared. As with the previous models used, a
better correlation of the experimental solubility
data to the Chrastil equation was found when the
solubilities obtained in the retrograde solubility

Table 4
Results of the solubility data correlation considering a linear dependence between the enhancement factor and the density of the
solvent; number of data points used in the correlation, n , parameters of Eq. (3), A and B , percentage of AARD, and s -squared, s2

Solute n Pressure range (MPa) T (K) Parameters of Eq. (3)

A B /102 AARD (%) s2

p -Coumaric acid 10 8.5
/50 313 5.09 1.204 17.63 0.9905

10 8.5
/50 323 5.30 1.064 32.95 0.9791
9a 10
/50 323 4.02 1.212 16.85 0.9920
9 10
/50 333 0.83 1.284 14.97 0.9949
8a 15
/50 333 2.05 1.139 5.57 0.9963
Caffeic acid 8 15
/50 313 3.29 2.101 16.10 0.9842
8 15
/50 323 5.14 1.783 18.41 0.9798
8 15
/50 333 9.00 1.201 4.64 0.9981
Ferulic acid 9 10
/50 313 12.74 1.276 4.39 0.9983
9 10
/50 323 11.47 1.271 8.88 0.9976
8a 15
/50 323 12.39 1.166 2.82 0.9981
9 10
/50 333 10.50 1.262 13.23 0.9954
8a 15
/50 333 11.69 1.121 3.08 0.9990
a
Correlations excluding the solubility data obtained at pressures in the vicinity of the critical point of carbon dioxide.
 R. Murga et al. / J. of Supercritical Fluids 27 (2003) 239
/245 245

Table 5
Results of the solubility data correlation by the Chrastil model: number of data points used in the correlation, n , parameters of Chrastil
equation, a , b and k , percentage of AARD, and s -squared, s2

Solute n Pressure range (MPa) Parameters of the Chrastil equation

k a b AARD (%) s2

p -Coumaric acid 29 8.5
/50 6.101 /8675 /20.55 28.57 0.9658

p -Coumaric acida 27 10
/50 6.956 /9292 /24.39 13.83 0.9905
Caffeic acid 24 15
/50 9.953 /11 317 /42.25 30.33 0.9166
Ferulic acid 27 10
/50 7.016 /7297 /27.69 9.65 0.9951
a
Correlations excluding the solubility data obtained at pressures in the vicinity of the critical point of carbon dioxide.

region were not considered, as can be observed in
Table 5.
Acknowledgements
Financial support provided by the CICYT
(QUI96-0691), and the EU, CICYT and JCyL
(1FD97-1471-QUI) is gratefully acknowledged.
References
[1] M.A. Ribeiro, M.G. Bernardo-Gil, M.M. Esquvel, Me-
lissa officinalis L.: study of antioxidant activity in super-
critical residues, J. Supercrit. Fluids 21 (2001) 51
/60.
[2] R. Murga, R. Ruiz, S. Beltran, J.L. Cabezas, Extraction of
natural complex phenols and tannins from grape seeds by
using supercritical mixtures of carbon dioxide and alcohol,
J. Agric. Food Chem. 48 (8) (2000) 3408
/3412.
[3] R. Murga, M.T. Sanz, S. Beltran, J.L. Cabezas, Solubility
of some phenolic compounds contained in grape seeds, in
supercritical carbon dioxide, J. Supercrit. Fluids 23 (2002)
113
/121.
[4] C.A. Rice Evans, N.J. Miller, G. Paganga, Structure
/
antioxidant activity relationships of flavonoids and phe-
nolic acids, Free Radic. Biol. Med. 20 (7) (1996) 933
/956.
[5] E.S. Choi, M.J. Noh, K.P. Yoo, Solubilities of o -, m - and
p -coumaric acid isomers in carbon dioxide at 308.15
/
323.15 K and 8.5
/25 MPa, J. Chem. Eng. Data 43
(1998) 6
/8.
[6] J.M. Prausnitz, R.N. Lichtenthaler, E. Gomes de Azevedo,
Molecular Thermodynamics of Fluid Phase Equilibria,
third ed, Prentice Hall, Englewood Cliffs, NJ, 1999.
[7] W.J. Schmitt, R.C. Reid, The influence of the solvent gas
on solubility and selectivity in supercritical extraction, in:
J.M.L. Penninger, M. Radosz, M.A. McHugh, V.J.
Krukonis (Eds.), Supercritical Fluid Technology, Elsevier,
Amsterdam, 1985, pp. 123
/137.
[8] F.J. Chrastil, Solubility of solids and liquids in super-
critical gases, Phys. Chem. 86 (1982) 3016
/3021.
[9] S. Petkov, O. Pfohl, G. Brunner, PE */A Program to
Calculate Phase Equilibria, Herbert Utz Verlag, Munchen,
2000.
[10] M.J. Cocero, S. Gonzalez, S. Perez, E. Alonso, Super-
critical extraction of unsaturated products. Degradation of
b-carotene in supercritical extraction processes, J. Super-
crit. Fluids 19 (2000) 39
/44.
[11] B. Bartolome, M.L. Bengoechea, M.C. Galvez, F.J. Perez-
Ilzarbe, T. Hernandez, I. Estrella, C. Gomez Cordoves,
Photodiode array detection for elucidation of the structure
of phenolic compounds, J. Chromatogr. A 655 (1993)
119
/125.
[12] S. Angus, B. Armstrong, K.M. de Reuck, IUPAC,
International Thermodynamic Tables of the Fluid State,
Carbon Dioxide, vol. 3, Pergamon Press, Oxford, UK,
1976.
[13] D.K. Dandge, J.P. Heller, K.V. Wilson, Structure solubi-
lity correlations: organic compounds and dense carbon
dioxide binary systems, Ind. Eng. Chem. Prod. Res. Dev.
24 (1985) 162
/166.
[14] W.J. Lyman, W.F. Reehl, D.H. Rosenblatt, Handbook of
Chemical Property Estimation Methods, ACS, Washing-
ton, DC, USA, 1990.