Analytica Chimica Acta 587 (2007) 89–96

Headspace–mass spectrometry determination of benzene, toluene

and the mixture of ethylbenzene and xylene isomers in
soil samples using chemometrics
F.A. Esteve-Turrillas, S. Armenta, S. Garrigues, A. Pastor, M. de la Guardia ∗
Department of Analytical Chemistry, Universitat de València, Edifici Jeroni Muñoz, 50th Dr. Moliner, 46100 Burjassot, València, Spain
Received 5 October 2006; received in revised form 5 January 2007; accepted 15 January 2007
Available online 21 January 2007

Abstract
A simple and fast method has been developed for the determination of benzene, toluene and the mixture of ethylbenzene and xylene isomers
(BTEX) in soils. Samples were introduced in 10 mL standard glass vials of a headspace (HS) autosampler together with 150 ␮L of 2,6,10,14-
tetramethylpentadecane, heated at 90 ◦ C for 10 min and introduced in the mass spectrometer by using a transfer line heated at 250 ◦ C as interface.
The volatile fraction of samples was directly introduced into the source of the mass spectrometer which was scanned from m/z 75 to 110. A partial
least squares (PLS) multivariate calibration approach based on a classical 33 calibration model was build with mixtures of benzene, toluene and
o-xylene in 2,6,10,14-tetramethylpentadecane for BTEX determination. Results obtained for BTEX analysis by HS–MS in different types of soil
samples were comparables to those obtained by the reference HS–GC–MS procedure. So, the developed procedure allowed a fast identification
and prediction of BTEX present in the samples without a prior chromatographic separation.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Benzene; Toluene; Ethylbenzene; Xylene isomers; Headspace-mass spectrometry; Chemometric analysis

1. Introduction
Volatile organic compounds (VOCs) are chemical products
based on carbon chains or rings, also containing hydrogen with
a vapour pressure greater than 0.27 kPa at 25 ◦ C excluding
methane. Measurement of VOCs in different kind of samples
has received significant attention over the past few years because
of direct and indirect impacts of individual VOCs on human
health and ecosystem. Many VOCs are either known or sus-
pected carcinogens and some have toxic effects [1–3]. The 1990
US Environmental Protection Agency (EPA) Clean Air Act [4]
includes 189 hazardous air pollutants that are mostly VOCs, and
it has been evidenced that these compounds also play a critical
role in formation of tropospheric ozone [5,6].
BTEX [benzene, toluene, ethylbenzene and the three xylene
isomers (ortho, meta and para)] are a subclass of VOCs with
boiling points between 80 and 150 ◦ C. Their solubility in water,
∗ Corresponding author. Tel.: +34 963544838; fax: +34 963544838.
E-mail address: miguel.delaguardia@uv.es (M. de la Guardia).
together with the chronic toxicity associated with the aromatic
ring present in their structure, make BTEX display a high
pollution potential. These aromatic hydrocarbons are widely
distributed in the environment due to natural sources; which
include superior plant wax, algae and plankton, and anthro-
pogenic sources of hydrocarbons; that comprise domestic and
industrial wastes, biomass and wood burning, incomplete fuel
oil combustion and urban runoff [7].
Analytical methods of reference for the determination of
BTEX residues in environmental samples are based on gas
chromatography–mass spectrometry (GC–MS), using either
static [8] or dynamic [9] headspace (HS) as sample introduction
modules. On the other hand, some other techniques like liquid
chromatography [10] or near infrared [11] have been already
employed.
In recent years, several novel techniques of sample intro-
duction and sample pre-treatment for GC analysis, like
purge-and-trap [12], membrane extraction [13], solid-phase
microextraction [14] and single-drop microextraction [15]
have been developed for the determination of BTEX and other
volatile compounds in different samples. On the other hand,

0003-2670/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.aca.2007.01.036
90 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96
it has been done a big effort to improve the BTEX extraction
by using the accelerated solvent extraction (ASE) in order to
achieve an efficient removal of analytes from different matrices
[16,17] prior to chromatographic analysis.
However, the use of chromatography means long time of anal-
ysis and, because of that, sample screening methodologies to
detect BTEX residues like the combination of headspace and
mass spectrometry detection have been explored providing a
yes/no response in order to reduce the amount of samples to
be analyzed [18]. The, HS–MS technique has been successfully
applied in qualitative applications such as the detection of hydro-
carbons in environmental matrices [19] and for classification of
the geographical origin of olive oils [20].
Multivariate calibration is a very useful tool for extract-
ing chemical information from instrumental signals. The most
commonly used multivariate methods for chemical analysis are
partial least squares (PLS) regression and principal component
regression (PCR), where factors that relate to variation in the
response measurements are regressed against the properties of
interest.
Y = Tq × f
where T is a matrix of latent variables and q contains the regres-
sion coefficients for the columns in T. The PLS factors (T) must
model most of the variation, both in the X- and Y-variable and
also attain fully adequate predictive ability. Ideally, each factor
added to the model would describe variation relevant for predict-
ing property values. To find the adequate number of PLS factors
to retain in the calibration model there are different methods.
One of the most employed is that which minimizes the predicted
residual error sum of squares (PRESS) of the model [21].
In summary, the relationships between analytical signals and
data to be determined are developed from the training set and
then applied to the set of unknowns. For unknown samples, the
concentrations of various constituents can often be predicted
with a good accuracy.
The aim of this study has been the development of a simple
and quick method capable to identify and to quantify BTEX
in soil samples using HS–MS combined with a partial least
squares (PLS) treatment of spectral data obtained, without any
chromatography separation.
2. Experimental
2.1. Chemicals and reagents
Toluene (99.8%, v/v), benzene (99.7%, v/v), hexane (96%,
v/v), iso-octane (99.8%, v/v) and petroleum ether (multisol-
vent grade) were supplied by Scharlau (Barcelona, Spain) and
the isomers of xylene, ortho, meta and para (99.0%, v/v)
and ethylbenzene (99%, v/v) were provided by Fluka (Buchs,
Switzerland). 2,6,10,14-Tetramethylpentadecane (98%, v/v)
used to prepare the standard solutions an added to the samples to
compensate the background was purchased from Sigma–Aldrich
(Madrid, Spain). Methyl tert-buthyl ether (MTBE) (99%, v/v),
used in the evaluation of interferents, was provided by Merck
(Darmstadt, Germany).
Stock standard solutions employed to build the PLS calibra-
tion models were prepared in 2,6,10,14-tetramethylpentadecane
at a concentration of 5.0 mL L−1 and stored in amber glass-
stoppered bottles at 4 ◦ C. Standard solutions containing BTEX
at individual concentration of 5, 25 and 50 ␮L L−1 of
benzene, toluene and xylene were also prepared in 2,6,10,14-
tetramethylpentadecane by appropriate dilution of the stock
ones. PLS calibration standards were prepared at three different
levels by adding 50 ␮L of the different dilutions, that corre-
sponds to 0.25, 1.25 and 2.5 nL of benzene, toluene and o-xylene,
inside glass headspace standard vials.
Spiked samples were prepared by adding 50 ␮L of each stan-
dard solution at 12.5, 25 and 37.5 ␮L L−1 level, that correspond
to 0.25, 0.625 and 1.875 nL of considered compounds, to 1 g of
soil.
Real soil samples were taken from the vicinity of petrol pump
stations in the area of Valencia and were directly heated in HS
glass vials after adding 150 ␮L of solvent.
3. Apparatus
3.1. The extraction–injection module
A static headspace autosampler Thermo Finnigan model
HS 2000 (Waltham, MA, USA) equipped with standard glass
vials of an internal volume of 10 mL (23 mm × 46 mm) was
employed. Once equilibrium was reached between the matrix
and the gaseous phase, 1 mL vapour sample was injected into
the system.
3.2. The transfer line
The transfer line ensures coupling of the extraction–injection
module to the mass spectrometer. This line must be designed to
maintain a high-quality vacuum in the spectrometer source while
allowing rapid transfer of the extracted molecules between the
two modules. A Scharlab fused silica tube of 30 m length with
an internal diameter of 0.32 mm was employed. The temperature
of this tube was held at 250 ◦ C to prevent condensation of the
considered compounds.
A Hewlett Packard HP5MS column (Palo Alto, CA, USA)
(30 m × 0.32 mm i.d., 0.25 ␮m film thickness) was used to
obtain the reference data by chromatography.
3.3. The mass spectrometer
A Thermo Finnigan Trace gas chromatograph, equipped
with a Thermo Finnigan ion trap mass spectrometer detector
Polaris Q was used. Ionization of the molecules introduced
into the source of the mass detector was carried out by elec-
tron ionization at 70 eV and an helium flow of 0.3 mL min−1
was used as damping gas. Data acquisition in full scan mode
was carried out over a mass range of m/z 75–110. The detec-
tor temperature was fixed at 250 ◦ C. Ion trap tests and mass
calibration were weekly performed with perfluorotributylamine
(PFTBA).
 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96
3.4. HS–GC–MS reference procedure
One gram of soil sample was accurately weighted inside
a 10 mL standard glass vial and 150 ␮L of 2,6,10,14-
tetramethylpentadecane were added. After that, the vial was
capped hermetically and heated in the headspace autosampler
at 90 ◦ C for 10 min with shaking. The syringe temperature was
selected at 100 ◦ C and 1 mL injection volume was introduced
into the system in split mode (1:10), employing a constant
flow of 1 mL min−1 helium as carrier. The oven temperature
program started from 40 ◦ C, held for 10 min, increased at a
rate of 20 ◦ C min−1 up to 200 ◦ C and finally held for 2 min.
The transfer line and source temperature were fixed at 300 and
250 ◦ C, respectively.
3.5. Headspace–mass spectrometry procedure (HS–MS)
Preparation of sample and headspace conditions were kept as
for the reference procedure. In order to measure the patterns of
volatile compounds in soil samples without any chromatography
separation (HS–MS methodology), the column was replaced by
a silica tube transfer line and the oven temperature was pro-
grammed to keep a constant temperature of 250 ◦ C.
3.6. Chemometric analysis
For instrumental, measurement control and data acquisition
it was employed the Xcalibur from Thermo Finnigan (Waltham,
MA, USA). Data were exported in excel format and the PLS cali-
bration models were developed by means of TurboQuant Analyst
6.0 software developed by Thermo Nicolet Corp. (Madison, WI,
USA).
The proposed PLS model was built using 27 standards for
calibration (using a classical 33 design) and it was validated
by means of a set of 10 spiked samples with known amounts
of BTEX. As it has been aforementioned, the calibration stan-
dards correspond to different mixtures of benzene, toluene and
o-xylene at three different concentration levels (33 ). High level
was fixed at 2.5 nL, medium value was evaluated at 1.25 nL and
low concentration level was fixed at 0.25 nL. Concentration of
the aforementioned components in spiked samples was predicted
employing the PLS calibration using the information of the mass
spectra between m/z 75 and 110.
To build and select PLS models, the optimum number of fac-
tors was chosen to minimize the PRESS, based on the criterion
of Haaland and Thomas [22].
4. Results and discussion
4.1. MS spectra of BTEX compounds
One nanoliter benzene, toluene, ethylbenzene and each
xylene isomer standards were measured by HS–GC–MS, in
order to obtain the retention time and mass spectra of each com-
pound. Mass spectra are shown in Fig. 1 in which it can be seen
the specific ions generated at m/z 77 and 78 for benzene, m/z 91
and 92 for toluene and finally m/z 91 and 106 for ethylbenzene
and the three xylenes. Benzene and toluene can be determined
91
employing their characteristic ions. However, ethylbenzene and
xylenes cannot be differentiated from each other because they
present the same characteristic ions and only could be quantified
after an appropriate separation.
Additionally, in Fig. 1, it can be seen the mass spectra of
2,6,10,14-tetramethylpentadecane. The specific ions generated
by this molecule at m/z 83–85 and 97–99 do not overlap with the
characteristic ions of the analytes and, thus this compound could
be used as internal reference for the determination of BTEX
compounds.
On the other hand, it is clear from Fig. 1 that ethylbenzene and
xylenes could be determined by MS without any interference of
benzene nor toluene.
4.2. HS–GC–MS recording of a BTEX mixture
A conventional chromatographic separation of a mixture of
1 nL of each one of BTEX compounds, using an HP-5 (5%-
phenyl)-methylpolysiloxane column and full scan acquisition
mode with selected ions at m/z 77, 78, 91, 92 and 106, is shown
in Fig. 2. As it can be seen in the figure, BTEX compounds
are well separated in a run time of 11 min, except meta- and
para-xylene that coelute. Thus, five peaks were obtained at 2.4,
4.1, 7.8, 8.4 and 10.0 min, corresponding to benzene, toluene,
ethylbenzene, (m,p)-xylenes and o-xylene, respectively.
4.3. HS–MS determination of benzene, toluene and
o-xylene
The main objective of this study was to establish a fast screen-
ing procedure based on the decrease of the run time and the
increase of the sample throughput for BTEX determination in
soils. So, the chromatography column was changed by a silica
empty tube as transfer line, in order to obtain one unresolved
peak and to use the mass spectra of this peak for BTEX quan-
tification.
In a first attempt, a silica tube of 5 m length and 0.32 mm
i.d. was assayed, but the vacuum of the mass spectrometer was
not acceptable, so a longer column of 30 m length and 0.32 mm
i.d. was used as transfer line and the vacuum problems were
avoided.
Using the aforementioned interface, a standard mixture of
1 nL of benzene, toluene and o-xylene was measured by HS–MS
in full scan mode in a mass range of m/z 75–110. Fig. 3 shows the
obtained MS time recording for selected ions of m/z 77, 78, 91,
92 and 106. Inset of Fig. 3 shows the corresponding mass spec-
trum of this wide peak from 1 to 4 min. Mass spectra obtained as
described for mixtures of benzene, toluene and o-xylene were
used for multivariate calibration based on partial least squares
(PLS) to predict the concentration of these compounds in soil
samples.
4.4. PLS models for HS–MS determination of benzene,
toluene and o-xylene
A classical design of 33 standards for the three analytes con-
sidered (benzene, toluene and o-xylene), at three concentration
levels was used. Table 1 shows the analytical features of the
92 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96

Fig. 1. Mass spectra of benzene, toluene, ethylbenzene, xylene and 2,6,10,14-tetramethylpentadecane obtained in the full scan mode for an individual amount of
1 nL of each one.

PLS model developed for the different compounds under study.
It has been indicated the number of factors required for predic-
tion of each analyte concentration with the minimum predicted
residual error sum of squares (PRESS), the coefficient of deter-
mination, R2 of the calibration model, the root mean square error
of calibration (RMSEC) and the mean absolute validation errors,
established for ten spiked soil samples previously analyzed by
the reference procedure and not included in the calibration set.
As it can be seen, in the best conditions, using all the infor-
mation of the spectral range, without any baseline correction nor
spectra pre-treatment, RMSEC of 12.1, 24.2 and 34.6 and mean
relative validation errors of 4.0, 5.4 and 5.9% were obtained for
benzene, toluene and xylene, respectively.
The principal component spectra (PCS) were obtained in
order to explain the contribution of factors on the analysis of
each volatile organic compound (see Fig. 4). On comparing
 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96 93

Fig. 2. Typical HS–GC–MS chromatogram obtained for a BTEX mixture of 1 nL of each compound measured in full scan mode at m/z 77, 78, 91, 92 and 106
selected ions.

Fig. 3. HS–MS time recording of a benzene, toluene and o-xylene mixture of 1 nL of each compound. Inset: Mass spectrum obtained for the recording between 1
and 4 min.

these signals with spectra in Fig. 1 it can be seen that the most
discriminant original variables correspond to the m/z ratios 85,
77, 78, 91, 92 and 106, which are the major mass fragments
for the 2,6,10,14-tetramethylpentadecane, used as internal ref-
erence, and BTEX compounds, and that the signals are well
modelized with the PCS from 1 to 5 being all other variables
characteristics of the background noise.
On the other hand, the general occurrence of positive and
negative peaks in PLS PC plots is due to the sign of corre-
lation between the component to be determined and the PLS
factor. The negative correlation between a PLS factor and an
analyte will provide a negative peak at its specific m/z. The load-
ings cannot be interpreted as spectra. Due to the orthogonality
forces, some loadings could have negative values that cannot
exist in spectra [21]. Fig. 5 shows the evolution of both, the
percentage of explained variance and that of the percentage of
cumulative variance for benzene determination as a function of
the number of PCs, that four components are enough for a good
modelization.
Calibration models obtained in different sessions provided
different detector responses in the whole m/z range used and thus,
in the following studies the calibration model was established
from measurements made in the same session than samples.
There are several circumstances that can introduce variations
in the new spectra that had not been included in the calibra-
tion step. The presence of these unmodeled variations in the
new data can lead to strongly biased predictions from multivari-
ate calibration models. The situation arises from changes in the

Table 1
Statistics for the PLS-MS calibration and prediction

Region Baseline Spectra pre-treatment Analyte Factorsa R2b RMSECc Mean accuracy errord (nL)

109-76 None None Benzene 4 0.9994 12.1 0.04

Toluene 4 0.998 24.2 0.05
Xylene 5 0.996 34.6 0.06
a The number of factors were chosen in order to obtain the minimum predicted residual error sum of squares.
b R2 is the coefficient of determination of the regression line between predicted and actual values of analyte concentration in the calibration set.
c Root mean square calibration error for a cross-validation.
d Mean accuracy error (%) found from the comparison of data obtained for the validation set of samples with results obtained by HS–GC–MS.
94 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96

Table 2
Effect of the use of o-xylene in the calibration set for PLS HS–MS determination
of ethylbenzene, different xylene isomers and a mixture of all of them in the
presence of benzene and toluene
Compound Amount added (nL) %Recovery

o-Xylene 1 101 ± 6
m-Xylene 1 102 ± 9
p-Xylene 1 101 ± 11
Ethylbenzene 1 93 ± 8
Mixa 1 99 ± 12

Results are the average of three independent measurements ± the corresponding

standard deviation.
a Mix: 0.25 nL of o-xylene, m-xylene, p-xylene and ethylbenzene.

ent, also, in the case of two different sessions, due to changes in

the environment of the instrument or in its response.
A solution to the calibration transfer problem is to apply
chemometric techniques to correct for instrumental and envi-
ronmental differences, thereby making the model transferable
and avoiding full recalibration. Various methods for calibration
transfer exist in the literature, and most have been discussed in
recent reviews [23,24].

4.5. Effect of the use of different xylene isomers in the

calibration set

Ethylbenzene and the three xylene isomers present exactly

Fig. 4. Principal component spectra diagnostic applied to verify the effect of
the number of factors on the signals of benzene, toluene and o-xylene. Vertical
lines represent the characteristic m/z for each BTEX compound. Note: Benzene
(B), toluene (T), ethylbenzene (E) and xylene (X).
instrumental response function. These changes can occur when
the data to be predicted are collected on an instrument differ-
ent from that used to build the model or simply in the same
instrument but in different measurement sessions. Although the
spectra may have the same basic shape or profile, the measured
intensity values from two instruments will generally be differ-
the same mass spectra in the m/z range monitored (see Fig. 1).
So, it seems impossible to do an individual determination of
all of them by HS–MS without any chromatography separation.
Due to this, the effect of the use of o-xylene in the calibration
set to determine the aforementioned compounds as total xylene
has been evaluated.
One nanoliter of ethylbenzene and the different xylene iso-
mers and a mixture of 0.25 nL of each one of the aforementioned
compounds were measured in the presence of 1 nL of both, ben-
zene and toluene. Table 2 shows the mean recovery data found
for the considered compounds in the assayed solutions expressed
as o-xylene. As it can be seen, the recovery of the different
Table 3
Effect of increasing amounts from 0.25 to 25 nL of different interferents on the
recovery of benzene, toluene and xylene at 1.25 nL level by PLS HS–MS
Interferent Compound %Recovery

Petroleum ether Benzene 105 ± 19

Toluene 104 ± 24
Xylene 107 ± 18
n-Hexane Benzene 91 ± 14
Toluene 98 ± 19
Xylene 106 ± 21
iso-Octane Benzene 104 ± 16
Toluene 105 ± 9
Xylene 96 ± 14
MTBEa Benzene 94 ± 13
Toluene 94 ± 14
Xylene 100 ± 9
Fig. 5. Effect of the number of PCs used in the PLS model on the explained a
variance and cumulative variance for benzene determination. MTBE: methyl tert-buthyl ether.
 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96 95

Table 4
BTEX concentration in soil samples obtained by both, reference HS–GC–MS and HS–MS developed procedures
Sample Soil concentration (nL g−1 ± s, n = 3)

HS–GC–MS HS–MS
 
Benzene Toluene E. Benzene (m,p)-Xylene o-Xylene E. Benzene + Xylenes Benzene Toluene E. Benzene + Xylenes

1 0.20 ± 0.10 1.5 ± 0.2 1.10 ± 0.08 1.68 ± 0.11 1.31 ± 0.10 4.09 –a 1.5 ± 0.2 4.2 ± 0.3
2 –a 2.50 ± 0.15 0.91 ± 0.07 –a –a 0.91 –a 2.4 ± 0.3 1.4 ± 0.2
3 –a 3.1 ± 0.2 –a –a –a –a –a 3.0 ± 0.4 –a
4 –a 1.72 ± 0.11 –a –a –a –a –a 1.5 ± 0.3 –a
5 –a 3.3 ± 0.2 0.20 ± 0.01 0.50 ± 0.03 0.53 ± 0.02 1.23 –a 3.4 ± 0.2 1.4 ± 0.2
6 –a 0.86 ± 0.10 1.4 ± 0.2 2.1 ± 0.10 1.16 ± 0.12 4.66 –a 0.5 ± 0.2 4.8 ± 0.2
7 –a –a 0.71 ± 0.08 1.81 ± 0.10 0.62 ± 0.03 3.14 –a –a 2.90 ± 0.10
8 –a 4.2 ± 0.3 –a –a –a –a –a 4.8 ± 0.3 –a
9 –a –a –a 1.52 ± 0.09 2.2 ± 0.2 3.72 –a –a 3.4 ± 0.4
10 –a –a 6.8 ± 0.4 –a –a 6.80 –a –a 7.0 ± 0.2
a Less than LOD (estimated in 0.1 nL g−1 ).

xylenes is around 100%. So, being acceptable, the use of o-
xylene in the calibration set for the prediction of ethylbenzene,
m- and p-xylenes.
4.6. Study of interferences
The main contamination sources of BTEX in soils include
the massive use of petroleum and its derivatives, and that of
solvents. So, it was evaluated the interference of hydrocarbons
like petroleum ether, n-hexane or iso-octane together with that
of methyl tert-buthyl ether, a common used additive of gasoline.
In order to assure the accuracy of the developed procedure,
it was evaluated the interference of increasing amounts of the
considered compounds from 0.25 to 25 nL for a fixed amount of
benzene, toluene and xylene of 1.25 nL of each one.
We can notice that whatever the interferent studied and the
concentration of that, the mean recovery values stay around
100%, as it can be seen in Table 3. So, the presence of these com-
pounds does not interfere on BTEX determination by HS–MS,
for interferent/analyte ratios from 0.2 to 20.
4.7. Real sample analysis
In a first attempt, 1 g of soil samples taken near petrol sta-
tions in the area of Valencia were introduced inside headspace
vials and measured by HS–MS, being the obtained results for
benzene, toluene and xylene unacceptables.
The main difference between soil samples and standards was
the presence of the solvent and, taking into consideration the
effect of the solvent in the PLS model which is well explained
by the PC1 (see Fig. 4), it was decided to add 2,6,10,14-
tetramethylpentadecane to the solid samples.
A 150 ␮L of solvent added to the soil samples avoided the
mentioned problem and results found for the PLS HS–MS deter-
mination of toluene and ethylbenzene plus xylenes shown in
Table 4 indicate the excellent comparability with data found by
HS–GC–MS. Benzene was detected only in a single sample by
GC–MS at a level of 0.2 ± 0.1 nL g−1 .
The regression between concentrations found by the recom-
mended procedure and those obtained by HS–GC–MS provide
equations CHS–MS = (−0.15 ± 0.15) + (1.08 ± 0.09)CHS–GC–MS
with a coefficient of determination R2 = 0.98 (n = 10) and
CHS–MS = (0.06 ± 0.11) + (1.00 ± 0.03)CHS–GC–MS with a coef-
ficient of determination R2 = 0.991 (n = 10) for toluene and
xylenes, respectively. In those equations, the intercept and slope
values were statistically comparables to 0 and 1, respectively,
for a probability level of 95% (being texp lower in all cases
than ttheoretical = 1.734 (n = 18)) thus indicating that the devel-
oped procedure provides an accuracy comparable with that of
the chromatography method.
The developed procedure allowed the individual identi-
fication and quantification of benzene and toluene and the
determination of the sum of ethylbenzene and xylenes. As it
can be seen, benzene was undetected in actual samples, proba-
bly as a result of the easy evaporation of this compound, as well
as its low concentration in petroleum derivatives (less than 5%,
v/v). Toluene and xylenes were found in the major part of soil
samples analyzed, at concentrations higher than those of other
analytes. These results are consistent with the high concentration
of these compounds in petroleum derivatives (mainly gasoline).
5. Conclusions
The present work describes the development of an HS–MS
method for the determination of aromatic hydrocarbons (BTEX)
in soil matrices by using a PLS multivariate calibration approach.
Considering the results, the HS–MS methodology could be
applied for the differentiation and classification of samples, and
also for the quantitative determination of BTEX in soil samples.
In fact, benzene, toluene and the mixture of ethylbenzene and
xylenes could be determined with an excellent accuracy. The dis-
advantage of the need of a multivariate calibration technique for
data treatment is compensated by the time saving (no chromatog-
raphy separation step is required). Results are comparables with
those obtained by HS–GC–MS and the sampling throughput is
considerably improved, from 5 samples per hour by HS–GC–MS
to 12 samples per hour by the developed procedure.
96 F.A. Esteve-Turrillas et al. / Analytica Chimica Acta 587 (2007) 89–96
Acknowledgements
The authors acknowledge the financial support of the Ministe-
rio de Educación y Ciencia (Project CTQ2005-05604, FEDER)
and Direcció General d’Investigació i Transferència Tecnològica
de la Generalitat Valenciana (Project 017609). F.A. Esteve-
Turrillas and S. Armenta also acknowledge the V segles
(Universitat de València) and FPU grant (Ministerio de Edu-
cación y Ciencia (Ref. AP2002-1874)), respectively.
References
[1] K. Saarela, T. Tirkkonen, J. Laine-Ylijoki, J. Jurvelin, M.J. Nieuwenhui-
jsen, M. Jantunen, Atmos. Environ. 37 (2003) 5563–5575.
[2] C.Y. Chan, L.Y. Chan, X.M. Wang, Y.M. Liu, S.C. Lee, S.C. Zou, G.Y.
Sheng, J.M. Fu, Atmos. Environ. 36 (2002) 2039–2047.
[3] K. Na, Y.P. Kim, Atmos. Environ. 35 (2001) 2603–2614.
[4] Amendments to 1990 Clean Air Act—List of 189 Hazardous Air Pollutants,
US Environmental Protection Agency, 1994.
[5] R. Atkinson, Atmos. Environ. 34 (2000) 2063–2101.
[6] B.J. Finlayson-Pitts, J.N. Pitts, J. Air Waste Manage. Assoc. 43 (1993)
1091–1100.
[7] A.L. Hinwood, H.N. Berko, D. Farrar, I.E. Galbally, I.A. Weeks, Chemo-
sphere 63 (2006) 421–429.
[8] Z. Wang, K. Li, M. Fingas, L. Sigouin, L. Menard, J. Chromatogr. A 971
(2002) 173–184.
[9] E. Martınez, S. Lacorte, I. Llobet, P. Viana, D. Barcelo, J. Chromatogr. A
959 (2002) 181–190.
[10] Y. Yang, A.D. Jones, C.D. Eaton, Anal. Chem. 71 (1999) 3808.
[11] J. Buerck, S. Roth, K. Kraemer, M. Scholz, N. Klaas, J. Hazard. Mater. 83
(2001) 11–28.
[12] M. Rosell, S. Lacorte, A. Ginebreda, D. Barceló, J. Chromatogr. A 995
(2003) 171–184.
[13] B. Hauser, P. Popp, J. Chromatogr. A 909 (2001) 3–12.
[14] J. Ji, C. Deng, W. Shen, X. Zhang, Talanta 69 (2006) 894–899.
[15] J. Cacho, V. Ferreira, P. Fernandez, Anal. Chim. Acta 264 (1992) 311–317.
[16] K. Li, M. Landriault, M. Fingas, M. Llompart, J. Hazard. Mater. 102 (2003)
93–104.
[17] H. Giergielewicz-Mozajska, L. Da browski, J. Namiesnik, Crit. Rev. Anal.
Chem. 31 (2001) 149–165.
[18] F. Peña, S. Cárdenas, M. Gallego, M. Valcárcel, J. Chromatogr. A 1052
(2004) 137–143.
[19] J.L. Perez-Pavon, A. Guerrero-Peña, C. Garcia-Pinto, B. Moreno-Cordero,
J. Chromatogr. A 1047 (2004) 101–109.
[20] C. Cerrato-Oliveros, R. Boggia, M. Casale, C. Armanino, M. Forina, J.
Chromatogr. A 1076 (2005) 7–15.
[21] P. Geladi, Spectrochim. Acta B 58 (2003) 767–782.
[22] D.M. Haaland, E.V. Thomas, Anal. Chem. 60 (1988) 1193–1202.
[23] X.L. Chu, H.F. Yuan, W.Z. Lu, Spectrosc. Spec. Anal. 21 (2001) 881–885.
[24] T. Fearn, J. NIRS 9 (2001) 229–244.