Journal of Chromatography A, 1217 (2010) 2768–2775

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Carbon nanotube reinforced hollow fiber solid/liquid phase microextraction: A

novel extraction technique for the measurement of caffeic acid in Echinacea
purpurea herbal extracts combined with high-performance liquid
chromatography
Zarrin Es’haghi a,∗ , Mazyar Ahmadi Golsefidi a,b , Ali Saify a , Ali Akbar Tanha a ,
Zohre Rezaeifar a , Zahra Alian-Nezhadi a
a
Department of Chemistry, Faculty of Sciences, Payame Noor University, Mashhad, Iran
b
Department of Chemistry, Faculty of Sciences, Islamic Azad University, Gorgan Branch, Gorgan, Iran

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

Article history: A new design of hollow fiber solid–liquid phase microextraction (HF-SLPME) was developed for the
Received 30 December 2009 determination of caffeic acid in medicinal plants samples as Echinacea purpure. The membrane extraction
Received in revised form 19 February 2010 with sorbent interface used in this research is a three-phase supported liquid membrane consisting of an
Accepted 22 February 2010
aqueous (donor phase), organic solvent/nano sorbent (membrane) and aqueous (acceptor phase) system
Available online 1 March 2010
operated in direct immersion sampling mode. The multi-walled carbon nanotube dispersed in the organic
solvent is held in the pores of a porous membrane supported by capillary forces and sonification. It is
Keywords:
in contact with two aqueous phases: the donor phase, which is the aqueous sample, and the acceptor
Solid/liquid phase microextraction
Multi-walled carbon nanotube
phase, usually an aqueous buffer. All microextraction experiments were supported using an Accurel
Hollow fiber Q3/2 polypropylene hollow fiber membrane (600 ␮m I.D., 200 ␮m wall thicknesses, and 0.2 ␮m pore
Echinacea purpurea size). The experimental setup is very simple and highly affordable. The hollow fiber is disposable, so
Caffeic acid single use of the fiber reduces the risk of cross-contamination and carry-over problems. The proposed
HPLC method allows the very effective and enriched recuperation of an acidic analyte into one single extract.
In order to obtain high enrichment and extraction efficiency of the analyte using this novel technique, the
main parameters were optimized. Under the optimized extraction conditions, the method showed good
linearity (0.0001–50 ␮g/L), repeatability, low limits of detection (0.00005 ␮g/L) and excellent enrichment
(EF = 2108).
© 2010 Elsevier B.V. All rights reserved.

1. Introduction were also described by Pederson-Bjergaard and Rasmussen as the

Biological and environmental sample analysis is often compli-
cated by low analyte concentrations, complex sample matrices and
limited sample volumes available for the determinations [1]. Thus,
a sample isolation and pre-concentration technique is necessary as
well.
Recent research activities are oriented towards the develop-
ment of efficient, economical, and miniaturized sample preparation
methods. As a result, solid-phase microextraction (SPME) [2] and
liquid phase microextraction (LPME) [3] have been developed,
among other microextraction techniques.
Some operating modes of LPME which were recently introduced
by Jeannot and Cantwell in 1996 [4] and He and Lee in 1997 [5]
∗ Corresponding author. Tel.: +98 511 8683003.
E-mail addresses: z eshaghi@pnu.ac.ir, zarrin eshaghi@yahoo.com (Z. Es’haghi).
hollow fiber liquid phase microextraction (HF-LPME) [6] in the
form of two and three-phase microextractions. Hollow fiber-based
liquid phase microextraction overcomes several disadvantages of
initial single drop LPME technique for example instability in com-
plex matrices like biological or environmental solutions as well as
some of SPME’s drawbacks; including SPME fibers are still com-
paratively expensive, the polymer coating is fragile and should be
handled carefully [7]. The chemical principle of hollow fiber LPME
was demonstrated by Jonsson and Mathiasson [8]. Although high
enrichment, clean-up and low solvent consumption are the major
advantages of HF-LPME, relatively long extraction times and low
selectivity are perhaps the major disadvantages of the mentioned
method [9].
In various fields of chemical analysis, there have been an increas-
ing number of applications of carbon nanotubes (CNTs). CNTs
exhibit an extraordinary adequacy of structural, mechanical and
electronic properties, which have made them potentially useful in

0021-9673/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2010.02.054
 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775 2769

nanotube-reinforced materials, as the sorbents for SPME and the

like [10]. In several works, sorption on MWCNTs has been exam-
ined for different organic analytes, such as phenol derivatives [11],
phthalate esters [12] and aniline compounds [13].
So, we decided to promote the HF-LPME technique by insert-
ing MWCNTs into the pores of polypropylene hollow fibers. Prior
to this, Basheer et al. used a small bag made of polypropylene
filled with MWCNTs. This micro-bag membrane was used as a pro-
tective barrier for solid-phase microextraction (SPME) [14]. They
found this ␮-SPE method robust, durable and that it is not easily
automated. In 2008, Hylton et al. explained this method and used
it for toluene and naphthalene extraction [15]. The authors cal-
culated some thermodynamic parameters like iso-steric heats of
adsorption for CNTs and compared it with similar parameters for a
non-tubular carbon.
The idea is therefore to have a membrane based on CNTs that
acts as an analyte trap, resulting in higher selectivity and enrich-
ment because the MWCNTs act as solid sorbents do in SPME fibers.
We have used this porous polypropylene membrane modified with
MWCNTs to pre-concentrate organic compounds from wastewater.
We have used caffeic acid (CA) as a target model for this purpose.
The results were promising with respect to HF-LPME convenience.
Caffeic acid, C9 H8 O4 , is a naturally occurring phenolic compound
(formerly called a carbolic acid), which is found as antioxidant in
many fruits, vegetables, and herbs, including coffee, although vary-
ing in amounts depending on the plant [16].

2. Experimental

2.1. Chemicals and materials

Caffeic acid (3,4-dihydroxy cinnamic acid), acetonitrile,

methanol, toluene, acetone and 1-octanol were purchased from
Merck (Schuchardt, Germany). Analyte, solvents, salts, acids, and
bases were of analytical grade.
The hollow fiber polypropylene membrane support Q3/2
Accurel PP (200 ␮m thick wall, 0.6 mm inner diameter and 0.2 ␮m
average pore size) was purchased from Membrana (Wuppertal,
Germany) (see Fig. 1). The multi-walled carbon nanotubes (MWC-
NTs) were purchased from the Research Institute of the Petroleum
Industry (Tehran, Iran). The mean diameter of the MWNTs was
10–15 nm, the length was 50–100 nm and purity >98%.
2.2. HPLC system
The HPLC system used in this work was a Knauer (Berlin, Zehlen-
darf, Germany) and consisted of a Knauer (K-2600) UV detector.
The column used was a Perfectsil Target RP-18 column (4.6 mm
diameter, 250 mm length, ODS-3 5 ␮m) from MZ-Analysentechnik
(Wohlerstrabe, Germany). An RP-18 guard column was fitted
upstream of the analytical column.
The mobile phase consisting of 0.1% (v/v) o-phosphoric acid-
acetonitrile (initially 85:15 and after 20 min 60:40) was filtered by
Milli-Q filtering system, and delivered by a Knauer K-1001 HPLC
pump. The flow rate of the mobile phase was 1.0 mL/min and the
UV detection wavelength was set at 330 nm.
2.3. Carbon nanotube treatment
One of the routes to increase the interaction between matrix and
reinforcement involves submitting CNTs to a process called func-
tionalization. Functionalization is a chemical process that inserts
functional groups on the sidewall of CNTs [17].
Functionalized CNTs are capable of adsorbing different inorganic
and organic chemicals. The addition of functional groups on CNTs is
Fig. 1. Scanning electron microscopy of a polypropylene hollow fiber. (a) P.P. hollow
fiber and (b) hollow fiber containing the dispersed functionalized MWCNTs solution
in 1-octanol.
commonly made by immersion in nitric acid or a mixture of sulfuric
acid (H2 SO4 ) and nitric acid (HNO3 ) in the range of 3:1 [18].
Acid treatment in a sonication bath has been reported to yield
short MWCNTs and introduce acidic groups onto the surface of the
MWCNTs. In these cases, the carboxylic CNTs formed a network like
a polymer and afforded the additional interactions for the analyte
molecules. The treated MWCNTs were terminated with carboxylic
acid groups and hydroxyl groups. The non-treated MWCNTs aggre-
gated in a disorderly fashion, but the functionalized MWCNT, in
contrast, was well dispersed because of the presence of the func-
tional groups that repel each other electrostatically. Therefore, the
treated MWCNTs were selected as the initial material for the sub-
sequent microextraction procedure.
We have examined acid treatment with nitric acid and
HNO3 /H2 SO4 mixture and found that the acid mixture showed the
better results in this case.
In the functionalization procedure, 0.2 g of the crude MWCNTs
were immersed in 100 mL mixture of concentrated H2 SO4 /HNO3
(3:1) at room temperature. They were ultrasonicated in a water
bath for 24 h at 40–50 ◦ C using a Branson Sonifier 450. After that,
CNTs were maintained at room temperature for 15 h. This solu-
tion was neutralized with ammonium hydroxide. The CNTs were
washed several times using deionized water until the pH 6.5 was
reached and filtered with a 0.22 ␮m nylon membrane filter (Milli-
pore, Bedford, MA, USA).
The acid-treated MWCNTs after drying were thoroughly dis-
persed in 1-octanol by ultrasonication at room temperature for
1.0 h (see Fig. 2).
2770 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775
Fig. 2. Photograph of a vial containing the functionalized MWCNTs solution in 1-
octanol.
2.4. HF-SLPME procedure
The membrane extraction with sorbent interface used in this
research is a three-phase supported liquid membrane consisting of
an aqueous: organic solvent/nano sorbent: aqueous system oper-
ated in direct immersion sampling mode. The multi-walled carbon
nanotubes dispersed in the organic solvent are held in the pores
of a porous membrane supported by capillary forces and sonifica-
tion. It is in contact with two aqueous phases, the donor phase and
the acceptor phase. The analyte from the aqueous sample diffuses
through the porous polypropylene membrane into the MWCNTs,
which were dispersed in the organic solvent and filled the hollow
fiber pores. The analyte is trapped in a sorbent (micro) trap and
an organic solvent simultaneously. In the back-extraction process,
the analyte is transferred to the small volume of aqueous acceptor
phase that was flowing on the inside of the membrane and are thus
enriched.
The polypropylene hollow fiber was cut into small segments
with a length of 2.5 cm. A disperse mixture (15 ␮L) of the function-
alized MWCNTs in 1-octanol was gradually injected into the fiber
manually by using an HPLC syringe so that the injection speed of
the mixture and the evaporation of the sample solvent from the
fiber were the same. The fiber pores were filled with the mixture
and sonicated until the mixture was dispersed and spread within
the pores. The excess amount of mixture was carefully removed
from the inside of the fiber. Subsequently, 6.5 ␮L of the acceptor
solution (NaOH, pH 11) was injected into the lumen of the hollow
fiber with a microsyringe. The fiber was then bent into a U-shape
and together with a small part of the supporting syringe needle was
submerged in the sample solution present in a proper vial (15 mL
volume). The vial was covered and stirred at 400 rpm for 25 min.
After the extraction/back-extraction, acceptor phase was removed
and the lumen of the hollow fiber washed two times (each time
with 100 ␮L of the acceptor solution) and the extracts were added
to the first portion which was transferred into a small glass vial.
A portion of the obtained solution (5.5 ␮L) was injected into the
HPLC. All experiments were performed in triplicate, and the means
of the results were used in the calculations.
2.5. HF-LPME procedure
The extraction procedure was carried out as described previ-
ously [19]. The optimization process was conducted using 1.0 ppm
standard solutions. The extraction and pre-concentration proce-
dure for target analyte in water samples was as follows: first of
all the hollow fiber was cut into segments with a length of 2.5 cm.
The fiber segment was cleaned with acetone by ultrasonication to
remove impurities and directly dried in air; then, the fiber was sub-
merged in the organic liquid for a few seconds to fill the membrane
pores of the hollow fiber wall. After that, the lumen of the fiber was
slowly flushed and completely filled with the 6.5 ␮L of the acceptor
solution (NaOH, pH 11). The looped fiber surface was washed with
water to remove excessive organic liquid. The prepared extraction
device was introduced into a proper vial with screw cap containing
15.0 mL of aqueous acceptor phase. The extraction and enrichment
was performed by shaking at 400 rpm at 25 ◦ C. At the end of the
extraction, the hollow fiber was taken out from the vial and the
analyte was transferred into a clean and dry glass vial. The hollow
fiber was washed as mentioned in Section 2.4. Then the 5.5 ␮L of
acceptor solution was injected for HPLC analysis.
2.6. Batch experiments
Meantime, during a series of initial experiments, batch extrac-
tion method on the caffeic acid solutions was also evaluated. In
this method a batch of feed is contacted with a batch of extracting
CNTs. The system is allowed to attain equilibrium and phases are
separated.
The experimental procedure involved dispersing 0.03 g of func-
tionalized MWCNTs into 3.0 mL caffeic acid solution (1.0 ppm). The
suspension was then stirred for 10.0 min under 400 rpm. After this
period, the suspension was filtrated. 10.0 ␮L of initial caffeic acid
solution and filtrate were injected into HPLC for analysis.
2.7. Adsorption processes
Different adsorption models that are commonly used for CNTs
are reported by Yang et al. [20]. Langmuir model is one of the most
common isotherm models which shows a rather linear band at
low relative concentrations and saturation behavior at relative high
concentrations.
The Langmuir model isotherm is valid for monolayer adsorp-
tion onto a surface containing a finite number of identical sites.
According to this model, a single adsorbate is retained in only one
molecular layer. The Langmuir isotherm model can be expressed as
qe = qmax · KL·Ce /1 + KL · Ce (1)
where qe is the adsorption capacity at equilibrium, Ce is the equi-
librium concentration, and qmax and KL are the Langmuir constants
related to the maximum adsorption capacity and the adsorption
energy, respectively. Maximum adsorption capacity qmax repre-
sents the monolayer coverage of sorbent with sorbate and KL
represents the enthalpy of adsorption (the adsorption energy or
the adsorption equilibrium constant, cm3 /mg).
The widely used linear expression of Langmuir isotherm is
Lineweaver–Burk equation. It is the double reciprocal of the Lang-
muir equation [21]:
1/qe = [1/(qmax · KL )] × (1/Ce ) + 1/qmax (2)
The data of the amount of caffeic acid adsorbed on oxidized
MWCNTs (q, mg/g) and the concentration of analyte remained
in solution (Ce , mmol/L) are described by this typical Langmuir
model, respectively. Langmuir model parameters and statistical fits
of experimental data to the above equation are given in Table 1.
 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775
Table 1
Linear Langmuir isotherm parameters.
Langmuir model (Lineweaver–Burk equation) qmax (mg/g)
MWCNTs/octanol (functionalized) 366.7 ± 0.806
The dependence between the equilibrium concentration of a
compound associated with the sorbent and its concentration in
the solution is commonly referred to as adsorption isotherm. Lang-
muir adsorption isotherm has well described equilibrium analyte
extraction by HF-SLPME.
3. Results and discussion
3.1. Optimization for the solid/liquid microextraction
In order to obtain high enrichment and extraction efficiency of
the analyte using this microextraction technique (HF-SLPME), the
main parameters were optimized.
3.1.1. Membrane solvent selection
The partition coefficient of caffeic acid as an organic molecule
in 1-octanol is moderate (pK = 1.15); therefore, extraction into 1-
octanol is relatively fast, and it extracts analyte via absorption.
On the other hand, caffeic acid is adsorbed on the surface of the
MWCNTs as well.
On the other hand, MWCNTs must be well dispersed in the
organic solvent. CNTs are insoluble in all solvents due to the large
inter-tube attraction energy [22] and therefore agglomerated in the
organic solvents. One approach is the functionalization of the sur-
face of CNTs using covalent chemistry. Thus, the dispersion of CNTs
into suitable organic solvent seems very attractive to preserve the
extended ␲ networks of the CNTs when it is used as modifier for
adsorption applications.
In this work, with respect to our last research, 1-octanol has been
selected as the organic solvent [19,23]. Moreover, CNTs’ diameter is
large enough to easily accommodate octanol molecules; however,
the hydrogen-bond breakage for transitioning an octanol molecule
from the bulk to the interior of the nanotube will be much smaller
than for water. It is reasonable to expect that since the nanotube is
wide enough to accommodate many octanol molecules, this effect
is also due to solvent molecules entering the cavity of the nanotube
and hydrogen bonds being broken resulting in a higher energy.
3.1.2. Extraction time profile
The extraction time is a very important parameter in an HF-
LPME procedure because it influences the partition of the target
analytes between the sample solution and the membrane (in the
pores of the fiber) and, subsequently, between the organic solvent
and acceptor phase. Different extraction times were tested, and the
corresponding results are provided in Fig. 3. The extraction effi-
ciency of the MWCNT/1-octanol filled fiber technique increased
when the extraction time increased from 10 to 60 min, but it did
not reach equilibrium. MWCNTs were a porous layer, in which
mass transfer was a process of diffusion through the pores. There-
fore, the porosity of MWCNTs should strongly affect the extraction
dynamics. Thus, nanometer-sized pores were found on the surface
of the MWCNTs, which might lead to a longer equilibrium time
for the extraction of the analytes compared with the conventional
HF-LPME mode.
Although LPME is not an exhaustive extraction technique, max-
imum sensitivity is attained at equilibrium conditions. It is not
necessary for a routine analysis to reach complete equilibrium as
long as the extraction time is kept constant. Therefore, 25 min was
chosen as the extraction time based on the consideration of sensi-
tivity and analysis speed.
2771
KL (cm3 /mg) R
0.297 ± 0.00634 0.9996
3.1.3. Effect of pH on the extraction
In the three phases, HF-LPME mass transfer is promoted by opti-
mal pH-conditions on each side of the supported membrane (pH
gradient). The pH value in the sample solution should suppress
analyte ionization, while the pH in the acceptor solution should
be adjusted to ensure total ionization of the analytes.
For acidic analytes, the pH in the sample should be low (prefer-
ably 3 units lower than the pKa value of the analytes), whereas the
acceptor phase should be basic (preferably 3 units higher than the
pKa value of the analytes), and vice versa for acidic analytes. Aque-
ous samples (1 mg/L for caffeic acid) with different pH values were
extracted with this method. It can be seen that the responses for
the target compound increase with the decrease of pH in the donor
aqueous phase.
Furthermore, carboxylic groups, which are polar groups and
might be ionized at high pH, are present on the MWCNT sur-
face. Therefore, besides hydrophobic interactions derived from the
graphite structure, an MWCNT fiber could have hydrophilic and ion
exchange interactions with target compounds. At low pH, the car-
boxylic groups on the pores and the caffeic acid in solution were
mainly in neutral form. The interactions between the MWCNT and
caffeic acid were hydrophilic and hydrophobic interactions, and the
ion exchange interaction was insignificant. The lower the pH was,
the higher the amounts of caffeic acid in solution and carboxylic
groups on the MWCNT fiber in their neutral forms. Thus, more caf-
feic acid could be extracted by the MWCNT through hydrophilic
and hydrophobic interactions. The pH was tested for donor phase
and acceptor phase in the ranges of 1–7 and 7–12 respectively. Con-
sequently, the highest extraction efficiency for the caffeic acid was
found at pH 2.5 (for donor phase). On the contrary, the optimal
acceptor pH value was 11.
3.1.4. Effect of the donor phase volume
As the analyte is extracted from relatively large sample vol-
umes into a very small volume of acceptor solution, most LPME
applications provide substantial analyte enrichment.
The enrichment factor in LPME is basically determined by the
analyte recovery and by the volume of the sample theoretically.
As the volume of the sample increases, the enrichment factor also
Fig. 3. Effect of the extraction time on the method performance. Conditions: analyte
concentration, 1 ␮g/L; dispersed CNTs in octanol, 2 mg/mL; donor phase volume,
10 mL with pH 2.50; acceptor phase volume 6.5 ␮L with pH 11; stirring speed
400 rpm; temperature 22 ± 0.5 ◦ C.
2772 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775
Table 2
Performance of the method.a .
Compound Enrichment factor RSD% (n = 3) Linear range (ng/L)
Caffeic acid 2618 3.9 0.1–50000
a
Method conditions: Hollow fiber membrane, dispersed CNTs in octanol (2 mg/mL); donor phase volume, 15 mL with pH, 2.50; acceptor phase volume 6.5 ␮L with pH, 11;
extraction time, 25.0 min.;stirring speed 400 rpm; temperature 22 ± 0.5 ◦ C.
increases [24].
1 Va
EF = + (3)
1/K Vd
Ca,eq
K= (4)
Cd,eq
In the above equations, EF is the enrichment factor, Va and Vd
are the volumes of acceptor and donor phases, and Ca,eq and Cd,eq
are the concentrations of these phases at equilibrium conditions,
respectively. According to the above equations, decreasing the vol-
ume ratio of the acceptor and the donor phases can lead to an
increase in extraction efficiency.
In HF-LPME, extraction is an equilibration process and, there-
fore the amount of analyte partitioning into the acceptor solution
becomes independent of the sample volume when this volume is
much higher than the product of the partition constant and the
volume of the acceptor solution. Obviously, this partition coeffi-
cient depends on the particular three-phase system. Furthermore,
a larger sample volume can even be disadvantageous due to poorer
mass-transfer kinetics, resulting in a worse extraction efficiency.
In the present work, the phase ratio of donor and acceptor solu-
tions was optimized by changing the volume of the donor phase
between 4 and 30 mL while the volume of acceptor phase was
kept constant at 6.5 ␮L. Generally, the extraction efficiency can
be improved by increasing the volume ratio of donor to acceptor
phase.
As seen in Fig. 4, however, the extraction results obtained for
the analyte was most favorable to suggest a phase ratio of 2307.7
(15.0 mL donor phase volume). Also, with an increase in the aque-
ous donor phase volume, organic phase dissolution may also be a
concern. This would lead to a decrease in the extraction efficiency.
Therefore, we selected a volume of 15.0 mL.
3.1.5. Effect of the stirring rate
The results indicate that the uptake increases with an increase
in the agitation speed. According to mass-transfer theory, in the
multiphase systems, the rate of agitation plays a dominant role.
The contact area of the phases can be increased with a higher
agitation rate. Fast agitation can increase the rate of extraction
through increasing the mass-transfer rate of the analyte to the
Fig. 4. Influence of the donor phase volume on the extraction. Conditions: analyte
concentration, 1 ␮g/L; dispersed CNTs in octanol, 2 mg/mL; donor phase pH 2.50;
acceptor phase volume 6.5 ␮L with pH 11; extraction time, 25.0 min; stirring speed
400 rpm; temperature 22 ± 0.5 ◦ C.
Squared regression coefficient (R2 ) LOD (ng/L) (n = 9) LOQ (ng/L) (n = 9)
0.997 0.05 0.1
membrane. Generally, an increase in stirring speed increases the
speed of extraction by reducing the thickness of the boundary layer
at the outer membrane surface. From the viewpoint of solid sorbent
MWCNTs, a high stirring rate helps the dispersion of nanotubes
across the extraction procedure.
The results showed that the peak areas of the analyte increase
with increasing stirring speed up to 400 rpm. It was also observed
that a stirring speed above 400 rpm resulted in a decrease in the
peak area, since the higher stirring speed led to mechanical stress
of the fiber [25]. Hence, a stirring speed of 400 rpm was chosen as
the optimum stirring rate.
3.1.6. Effect of the concentration of MWCNTs on the extraction
When two objects are closely in contact, Van der Waals inter-
actions, acid–base interactions, or both take place; therefore, an
intimate contact can increase the adsorption between two objects.
Physical adsorption includes the following stages: (1) molecules
diffuse from the bulk phase towards the interface; (2) molecules
diffuse inside the pores; (3) molecules diffuse on the surface of the
pores; and (4) adsorption/desorption processes occur at the surface
of the pores. In the second stage, diffusion takes place only when
the dimensions of the adsorbed molecules are slightly smaller than
the pore diameter [26].
The adsorption capacity is affected not only by the properties
of the adsorbents but also by the properties of the adsorbates. The
influences of the amount of MWCNTs on the extraction capacity
have been examined. 2.0 mg/mL was the optimal amount of the
CNTs (the range was between 0.8 and 10 mg/mL) in this research.
3.2. Evaluation of the method performance
The performance parameters of the HF-SLPME technique, such
as linearity, limits of detection and quantification, relative standard
deviations (RSD), extraction efficiencies and pre-concentration
factor, are listed in Table 2. The linear equations for analyte
are: Y = 339X − 168. The environmental samples were spiked
at 1 ␮g/L for the validation of this analytical method. The
enrichment factor which was 2108 (n = 3) for caffeic acid, deter-
mined experimentally from result of deviation of the analyte
peak area with the same concentration after and before the
extraction. The linearity in the samples was verified at con-
centrations ranging from 0.0001 to 50 ␮g/L by analyzing each
concentration in triplicate. All the analytes showed good lin-
earity with a squared regression coefficient (R2 ) of 0.997. A
detection limit (LOD) of 0.00005 ␮g/L was obtained for caffeic
acid.
The repeatability (three samples in one day) and reproducibil-
ity (nine samples in three consecutive days) in triplicate analysis
(each day) were tested in herb samples spiked at two concentra-
tions (1.0 ␮g/L and 10.0 ␮g/L). The mean RSD% of the experimental
procedure was 3.9% for repeatability. Thus, the method showed
good precision.
In order to make a comparison between ordinary HF-LPME, and
HF-SLPME or indeed, between the liquid phase and solid phase
microextraction and to assigning the portion of each mechanism
in this process, same amount of analyte was pre-concentrated and
extracted with the ordinary HF-LPME process over the same condi-
tions as was demonstrated in Section 2.5. The results have shown
 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775 2773

Table 3 It is may be due to many factors that are considered in the following
Relative recoveries (accuracy) for caffeic acid in the E. purpurea, spiked with four
paragraphs:
concentration levels of analyte.
The hexagonal arrays of sp2 -like carbon atoms in the tubu-
Concentration level Relative recovery RSD% (n = 3) lar graphene sheets on the surfaces of CNTs provide a favorable
(␮g/L) (%) (accuracy) (precision)
morphological structure for adsorbing aromatic compounds. The
1 88.9 5.21 small number of polar groups and the homogeneity of the sur-
0.1 94.6 3.59 face of MWCNTs resulted in faster adsorption kinetics for these
0.01 91.9 4.46
compounds. Thus, the MWCNTs exhibited excellent extraction
0.0005 89.1 5.37
capability and high recoverability for these analytes [27].
The highly developed surface of MWCNTs exhibits strong sorp-
that mean pre-concentration factor after extraction by HF-LPME tion properties towards caffeic acid. The strong binding is attributed
was 51.60 with linear range of 0.05–20 mg/L. The detection limit to the interaction of benzene ring of caffeic acid and the activated
was 0.01 mg/L (n = 5). surface of CNTs. It was deduced that not only outer surfaces but
In the batch experiments (Section 2.6) the peak area before also inner cavities and inter-layers in the structure of the MWCNTs
batch extraction was 8,045,677 and after it, changed to 91,921. Thus were responsible for removing the analyte. Has been proven that
87.528 portion of the initial concentration was reduced under the strong interaction can take place between the graphitic ring struc-
batch experiment. ture of the CNT sorbents and compounds having benzene ring in
The results of the mentioned experimental procedures mean their structures [28]. Since there was one benzoic ring,  bonds or
that “sorption” mechanism plays an important role in this process. lone pairs of electrons in the structure of caffeic acid, strong reten-

Fig. 5. Chromatograms of (a) hydro-alcoholic extract of the plant E. purpurea before microextraction. (b) E. purpurea extract after microextraction under optimal conditions.
2774 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775

tions are obtained for analyte on carbon sorbents, leading to high

References
breakthrough volumes of caffeic acid. There were anion-exchange
sites (such as –COOH, –OH) on the surface of the functionalized

[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[30]
CNTs, and it would be helpful for the adsorption of caffeic acid onto
these sorbents too.

98.4–101%
3.3. Study of the natural matrices

91–103%
Recovery

91–95%
82.30%

100.10%

92.53%
– In order to assess the applicability of the newly developed

–
method to natural samples, the proposed procedure was applied to
determine the caffeic acid in the Echinacea purpurea extract, which
is a medicinal herb.
1.12–2.68%

Extracts of the E. purpurea plant are widely used for medicinal

Less 10%
Less 10%

ECD: electron capture detection; CZE: capillary zone electrophoresis; SW Volt.: square wave voltammeter; CL: chemiluminescence; MEKC: micellar electrokinetic chromatography.
16.91%
1.31%

1.82%
5.76%
2.20%

3.30%
RSD%

purposes [29]. Effective quality control of these extracts requires

rapid methods to determine their chemical composition. A new
method for the analysis of caffeic acid from Echinacea extracts has
been developed. Quantitative analyses were carried out with this
0.9942
0.9999

0.9999
0.9994
0.9999

method on E. purpurea extracts.

0.995

0.999
0.999

15.0 mL of the extract were selected and examined under the

–
r

optimal conditions (n = 3). Relative recoveries were found in the

four concentration levels. These results illustrated that the matrix
0.0072–0.2456 mg/mL

effect was relatively low. This method was perfectly selective for
caffeic acid (see Table 3 and Fig. 5).
0.02–0.2 mmol/L

0.25–3.9 mg/L
0.05–10 mg/L

0.4–100 mg/L

25–225 mg/L
10–360 ␮g/L

0.1–1.5 ␮g/L
0.1–40 mg/L

4. Concluding remarks
DLR

A novel, fast and simple procedure based on equilibrium

hollow fiber solid–liquid phase microextraction (HF-SLPME) com-
1.345 mg/L

0.091 mg/L

bined with HPLC has been developed to extract caffeic acid

0.45 mg/L

from E. purpurea herb samples. The experimental setup is very

4 mg/L

4 ␮g/L
LOQ

simple and highly affordable. Among all microextraction tech-

–
–

niques reported, this technique is the most effective sample

preparation/pre-concentration technique. The hollow fiber is dis-
posable, so the single use of the hollow fiber reduces the risk
0.002 mmol/L
0.150 mg/L

0.028 mg/L

of cross-contamination and carry-over problems. The proposed

0.004 ␮g/L
0.13 mg/L

0.98 mg/L
0.1 mg/L

method allows the very effective and enriched recuperation of the

LOD

acidic analyte into one single extract. Moreover, the method was
–
–

applied to the analysis of natural herb samples giving good quali-

tative and quantitative results. This procedure can be successfully
used for the analysis of organic analytes in aqueous samples.
Detection

SW Volt.a
HPLC-UV
HPLC-UV

HPLC-UV

HPLC-UV
GC-FID

The method was compared with the other previous works

MEKCa
CZEa

(Table 4). In comparison with the other conventional sample-

CLa

preparation methods, the developed method has the merits of

Comparison of some methods which were used for determination of caffeic acid.

considerable analysis speed, good separation efficiency and ele-

vated pre-concentration, notable precision and high sensitivity.
Extraction method

Microdialysis

Acknowledgements
SPME

This work was supported by the Payame Noor University of Iran.

SPE
SPE

SPE
LLE

The authors would like to thank Giah Essence Phytopharm Co., Gor-
–
–

gan, Iran. This company has made important and highly appreciated
contributions.
Hydroalcoholic Extract
Hydroalcoholic Extract

Hydroalcoholic extract
Hydroalcoholic extract

References
Rabbit plasma

[1] K.T. Chung, Ecotoxicol. Rev. C18 (2000) 51.

Rat blood

Alcoholic
Aqueous
Aqueous

[2] W. Xie, J. Pawliszyn, W.M. Mullett, B.K. Matuszewski, J. Pharm. Biomed. Anal.
Matrix

45 (2007) 599.
[3] C. Ye, Q. Zhou, X. Wang, J. Chromatogr. A 1139 (2007) 7.
[4] M.A. Jeannot, F.F. Cantwell, Anal. Chem. 68 (1996) 2236.
[5] Y. He, H.K. Lee, Anal. Chem. 69 (1997) 4634.
[6] J.A. Jonsson, L. Mathiasson, Adv. Chromatogr. 41 (2001) 53.
1995
1999

2004
2006
2007
2007
2007
2008
2000

[7] K.E. Rasmussen, S. Pedersen-Bjergaard, Trends Anal. Chem. 23 (2004) 1.

Date

[8] J.A. Jonsson, L. Mathiasson, Trends Anal. Chem. 18 (1999) 318.

[9] J. Lee, H.K. Lee, K.E. Rasmussen, S. Pedersen-Bjergaard, Anal. Chim. Acta 624
(2008) 253.
Table 4

[10] L. Li, Y. Huang, Y. Wang, W. Wang, Anal. Chim. Acta 631 (2009) 182.
No.

[11] Y.Q. Cai, G.B. Jiang, J. Liu, Q. Zhou, Anal. Chem. 75 (2003) 2517.
1
2
3
4
5
6
7
8
9
a

[12] Y.Q. Cai, G.B. Jiang, J.F. Liu, Q.X. Zhou, Anal. Chim. Acta 494 (2003) 149.
 Z. Es’haghi et al. / J. Chromatogr. A 1217 (2010) 2768–2775 2775

[13] X. Liua, Y. Ji, H. Zhang, M. Liu, J. Chromatogr. A 1212 (2008) 10. [27] Q. Li, D. Yuan, Q. Lin, J. Chromatogr. A 1026 (2004) 283.
[14] C. Basheer, A.A. Alnedhary, B.S.M. Rao, S. Valliyaveettil, H.K. Lee, Anal. Chem. [28] Y.S. Ho, Polish J. Environ. 15 (2006) 81.
78 (2006) 2853. [29] B. Barrett, Phytomedicine 10 (2003) 66.
[15] K. Hylton, Y. Chen, S. Mitra, J. Chromatogr. A 1211 (2008) 43. [30] Y.S. Uang, F.L. Kang, K.Y. Hsu, J. Chromatogr. B: Biol. Med. Sci. Appl. 673 (1995)
[16] T. Nagaoka, A.H. Banskota, Y. Tezuka, I. Saiki, S. Kadota, Bioorg. Med. Chem. 10 43.
(2002) 3351. [31] T.H. Tsaia, Y.F. Chena, I.F. Chena, C.F. Chena, J. Chromatogr. B 729 (1999) 119.
[17] B. Vigoloa, V. Mamane, F. Valsaquea, T.N. Ha-Lea, J. Thabita, J. Ghanbajaa, L. [32] Y.K. Zhao, Q.E. Cao, H.T. Liu, K.T. Wang, A.X. Yan, Z.D. Hu, Chromatographia 51
Arandaa, Y. Fortb, E. Mc Raea, Carbon 47 (2009) 411. (2000) 483.
[18] J. Shen, W. Huang, L. Wu, Y. Hu, M. Ye, Mater. Sci. Eng. A 464 (2007) 151. [33] H. Wang, G.J. Provan, K. Helliwell, Food Chem. 87 (2004) 307.
[19] Z. Es’haghi, Anal. Chim. Acta 641 (2009) 83. [34] I. Citovı̌, R. Sladkovskı̌, P. Solich, Anal. Chim. Acta 573–574 (2006) 231.
[20] K. Yang, L. Zhu, B. Xing, Environ. Sci. Technol. 40 (2006) 1855. [35] S.C. Fernandes, I.R.W. Zwirtes de Oliveira, I.C. Vieira, Enzyme Microb. Technol.
[21] J. Hu, C. Chen, X. Zhu, X. Wang, J. Hazard. Mater. 162 (2009) 1542. 40 (2007) 661.
[22] Y. Tsai, J.D. Huang, Electrochem. Commun. 8 (2006) 956. [36] E.M. Risso, R.G. Peı̌res, J.A. Farfan, Food Chem. 105 (2007) 1578.
[23] A. Sarafraz-Yazdi, Z. Es’haghi, J. Chromatogr. A 1082 (2005) 136. [37] D.P. Rivelli, C.D. Ropke, V.V. Silva, D.V. Miranda, R.L. Almeida, T.C.H. Sawada,
[24] E. Psillakis, N. Kalograkis, Trends Anal. Chem. 22 (2003) 565. S.B.M. Barros, Brazil. J. Pharm. Sci. 43 (2007) 215.
[25] X. Wen, C. Tu, H.K. Lee, Anal. Chem. 76 (2004) 228. [38] E.N. Sieliwoniuka, J. Nazarukb, E. Antypiuka, A. Kojło, J. Pharm. Biomed. Anal.
[26] J.P. Fraissard, Physical Adsorption: Experiment, Theory and Applications, 48 (2008) 579.
Springer, New York, 1997.