Journal of Chromatography A, 1033 (2004) 193–203

Determination of fuel dialkyl ethers and BTEX in water using

headspace solid-phase microextraction and gas
chromatography–flame ionization detection
Idoia Arambarri, Maitena Lasa, Rosa Garcia, Esmeralda Millán∗
Departamento de Quı́mica Aplicada, Facultad de Quı́mica, Universidad del Paı́s Vasco, Apdo. 1072, 20080-San Sebastián, Spain

Received 25 July 2003; received in revised form 18 January 2004; accepted 26 January 2004

Abstract

A simple procedure for the determination of methyl tert-butyl ether (MTBE), ethyl tert-butyl ether (ETBE), ethyl butyl ether (EBE), tert-amyl
methyl ether (TAME), benzene, toluene, ethylbenzene, and xylenes (BTEX) in water using headspace (HS) solid-phase microextraction
(HS-SPME) was developed. The analysis was carried out by gas chromatography (GC) equipped with flame ionization detector (FID) and
100% dimethylpolysiloxane fused capillary column. A 27–4 Plackett–Burman design for screening and a central composite design (CCD) for
optimizing the significant variables were applied. Fiber type, extraction temperature, sodium chloride concentration, and headspace volume
were the significant variables. A 65 ␮m poly(dimethylsiloxane)-divinylbenzene (PDMS-DVB) SPME fiber, 10 ◦ C, 300 g/l, and 20 ml of
headspace (in 40 ml vial) were respectively chosen for the best extraction response. An extraction time of 10 min was enough to extract the
ethers and BTEX. The relative standard deviation (R.S.D.) for the procedure varied from 2.6 (benzene) to 8.5% (ethylbenzene). The method
detection limits (MDLs) found were from 0.02 (toluene, ethylbenzene, and xylenes) to 1.1 ␮g/l (MTBE). The optimized method was applied
to the analysis of the rivers, marinas and fishing harbors surface waters from Gipuzkoa (North Spain). Three sampling were done in 1 year
from June 2002 to June 2003. Toluene was the most detected analyte (in 90% of the samples analyzed), with an average concentration of
0.56 ␮g/l. MTBE was the only dialkyl ether detected (in 15% of the samples) showing two high levels over 400 ␮g/l that were related to
accidental fuel spill.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Water analysis; Headspace analysis; Solid-phase microextraction; Methyl tert-butyl ether; Dialkyl ethers; BTEX

1. Introduction (1%, v/v). Although these ethers concentrations are similar

Because of the extensive use of fuels, their components
are ubiquitous environmental contaminants. Of primary
concern are the volatile aromatics, benzene, toluene, ethyl-
benzene, and the three xylene isomers (BTEX). In addition
to these compounds the dialkyl ethers, as oxygenates and
octane enhancers, are present in gasoline. Methyl tert-butyl
ether (MTBE) is the most common oxygenate added to
gasoline. Smaller amounts of other ethers have been used
for gasoline oxygenation, including ethyl tert-butyl ether
(ETBE), and tert-amyl methyl ether (TAME). European di-
rective 98/70/EC allows concentration of oxygenate ethers to
15% (v/v) and limits the aromatics (42%, v/v) and benzene
∗ Corresponding author. Tel.: +34-943015419; fax: +34-943212236.
E-mail address: qapmimae@sq.ehu.es (E. Millán).
to that used in reformulated gasoline in USA, European fuel
blends mostly contain concentrations lower than 5%. An
increase of this percentage is expected in 2005 due to more
restrictive legislation on the aromatic content of gasoline [1].
The presence of these compounds in both ground- and
surface water are related to fuel spills, leaking underground
storage tanks, and the release of unburned fuel directly
into the atmosphere and surface waters [2–4]. However,
non-point sources like vehicle exhausts or evaporation of
fuel compounds at gasoline stations may be responsible for
low surface and rain water concentrations of MTBE [5]. The
occurrence of BTEX in the environment is a great con-
cern because of their toxicity. Among them, benzene is the
most dangerous; since is recognized as carcinogenic agent.
A maximum level of 1 ␮g/l is fixed by European legisla-
tion in drinking water [6], and a maximum contaminant
level (MCL) of 5 ␮g/l is established by US Environmental

0021-9673/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2004.01.046
194 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203
Protection Agency (USEPA) [3]. Among dialkyl ethers,
MTBE is the most widespread used and studied. It is a
persistent water contaminant due to its high water solubil-
ity and slow rate of degradation [2]. Although MTBE is a
suspected human carcinogen, MCL has not been done for
it. However, the USEPA has set an advisory level for taste
and odor at 20–40 ␮g/l [3].
Analytical methods used to alkyl ethers and BTEX de-
termination include purge and trap (PT), headspace (HS)
or direct aqueous injection (DAI) onto gas chromatography
(GC) [3,4,7–10]. Solid-phase microextraction (SPME) has
been used as an alternative. This type of extraction does
not require solvents and it can be carried out directly from
the liquid phase or from headspace over the liquid samples
(HS). The HS sampling is more advisable when matrix
could affect the determination of target analyte. On the
other hand, HS versus immersion extraction often shows
an important reduction of extraction time [11]. Most of the
SPME applications within the fuel compounds and oxy-
genates have been devoted to MTBE determination. Some
of them used direct immersion coupled with GC and MS
detector [12–14], the others used headspace followed by
GC–MS [15–17] or GC with flame ionization detector (FID)
[18]. Both extraction modes have been used in the determi-
nation of BTEX [19]. The simultaneous determination of
MTBE, ethyl butyl ether (EBE), and BTEX using headspace
solid-phase microextraction (HS-SPME) has been presented
using two-dimensional gas chromatography [20].
There are several experimental variables affecting the
HS-SPME procedure such as type of fiber, stirring rate,
temperature, extraction time, and addition of salt. The study
considering one by one the variables to get the best possi-
ble conditions have been showed in several works [16,20].
However, this procedure requires a high number of runs and
also is time consuming. An experimental design, that could
take into account simultaneously several variables, seems
the most convenient approach searching for the optimal op-
erational conditions in a reasonable number of runs [21,22].
This kind of methodology has been used for optimization
of PT–GC–MS conditions to determine MTBE, tert-butyl
alcohol (TBA), and BTEX in water [23]. Also, has been
applied to obtain the best conditions of HS-SPME–GC–FID
for MTBE determination [18].
Since the environmental implications of fuel oxygenates
and BTEX, the determination of both types of analytes is
now considered and will be more often required. Screening
methods that allow simultaneously monitoring both contam-
inants would be a helpful way to identify the contaminated
samples.
This work mainly focused on obtaining a screening
method for simultaneous determination of alkyl ethers and
BTEX using HS-SPME–GC–FID. The experimental design
was applied to know the significant variables and to opti-
mize the HS-SPME extraction of analytes from water. The
influence of high amounts of BTEX in adsorption process
with mixed coating fibers was also checked. Finally, the
procedure was applied to the analysis of marine and river
surface water samples from Gipuzkoa (North Spain).
2. Experimental
2.1. Chemicals and materials
Methyl tert-butyl ether (99.8%), ethyl butyl ether (99%),
ethyl tert-butyl ether (99%), and tert-amyl methyl ether
(97%) were purchased from Aldrich (Sigma-Aldrich Chemie
GmbH, Steinheim, Germany). Benzene (>99.5%), ethyl-
benzene (>99%), o-xylene, m-xylene, p-xylene (>99%)
were from Fluka (Sigma-Aldrich Chemie GmbH, Stein-
heim, Germany). Toluene (99.5%) and methanol (>99.8%)
were from Panreac (Panreac Quı́mica S.A., Barcelona,
Spain). Sodium chloride (99.5%) was purchased from
Merck (Merck, Damstadt, Germany). SPME holders and
fibers [100 ␮m thickness poly(dimethylsiloxane) (PDMS)
and 65 ␮m poly(dimethylsiloxane)-divinylbenzene (PDMS-
DVB)], sample vials (40 ml, amber glass), and PTFE-silicone
septa were obtained from Supelco (Bellefonte, PA, USA).
Single standard solutions (1000 mg/l) of ethers and BTEX
compounds were prepared by spiking each compound in
methanol. Then, were stored at 4 ◦ C and used within 4
weeks. Working aqueous solutions, prepared just before use,
were made from the stock methanolic solutions by spiking
and mixing them with 20 ml of mineral organic free water
previously put in 40 ml vials. The spring mineral water was
chosen after its comparison with deionized and bidistilled
water regularly used in the laboratory. In the latter there
were other organics besides the target analytes that compli-
cate the blank signal. This could be due to the disinfection
byproducts present in the drinking water used as source for
distilled and deionized water.
2.2. Equipment
A HP 6890 gas chromatograph equipped with split–
splitless injector and FID detector (Agilent Technologies,
Wilmington, DW, USA) was used in all measurements. The
injection port fitted with a 0.75 mm i.d. injection liner (Su-
pelco) was operated in the splitless mode, with the split–
splitless purge valve opened at 1 min after injection. The
injection port temperature was 250 ◦ C and helium served as
carrier gas with a flow-rate of 2.2 ml/min. Chromatographic
separation was accomplished with a fused-silica capillary
column (HP-1, 30 m × 0.32 mm × 0.25 ␮m film thickness,
Agilent Technologies). The temperature program used was:
40 ◦ C for 1 min, 20 ◦ C/min to 100 ◦ C, hold 1 min, then
50 ◦ C/min to 200 ◦ C, and hold 2 min. The detector temper-
ature was set at 280 ◦ C. A PC interfaced to the GC using
Chemstation software (Agilent Technologies) was used for
data acquisition and processing.
A Heidolph MR 3003 magnetic stirrer (Heidolph Elek-
tro GmbH & Co KG, Kelheim, Germany) was used. PTFE
 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203
coated stir bars of 25 mm were put in the 40 ml vials just
before runs.
The experimental matrix designs were performed and
evaluated using the STATISTICA software package (Stat-
Soft, Tulsa, USA).
2.3. Water sampling
In total 18 sampling sites were chosen in the province
of Gipuzkoa (North Spain). Sixteen of the sampling lo-
cations were selected based in their proximity to gasoline
service stations where fuel oxygenates and BTEX com-
pounds expected to be occurred. Nine of these sites were
located in marinas or fishing harbors, and other seven sites
were in river waters. Another two sampling points cho-
sen as control sites were in pristine rivers. Three sampling
were done in 1 year (June 2002, February 2003, and June
2003).
Surface water samples were collected in 100 ml glass vials
that were completely filled with no headspace, covered with
aluminum foil, closed with a metallic cap, and then trans-
ported to the laboratory in a cooler with ice. All the sam-
ples maintained in the cooler were analyzed within 48 h of
sampling. In order to check a potential contamination from
containers, shipping, storage, sample preparation, and mea-
surements a trip blank was included and maintained as the
same conditions of the collected water samples.
2.4. Analytical procedure
Adequate standards volumes were added immediately be-
fore use to the vials containing the salty water (6g/20 ml) to
achieve the corresponding micrograms per liter concentra-
tion. The vials with salty water had been previously put in
water bath to acquire 10 ◦ C.
Sampling extraction was performed by HS mode exposing
the PDMS-DVB fiber over stirred samples. The extraction
was done for 10 min at 10 ◦ C (using a constant-temperature
water bath). For all the runs a stirring rate of 1000 rpm was
used. After sampling the fiber was withdrawn into the needle
of the holder and was immediately placed in the GC injector.
The desorption temperature was 250 ◦ C and 1 min was the
desorption time for all the runs. No carryover was observed
after this desorption time.
River water samples were prepared in a similar manner.
For practical reasons in the marine waters the amount of
sodium chloride added was as much as the saturation level.
A great variability of the salinity in these samples, checked
by the conductivity measured in situ, was found. Therefore,
the addition of salt to get 300 g/l in each of the sea samples
was considered an impracticable task for a desirable simple
procedure. In order to quantify the ethers and BTEX com-
pounds in the real samples the standard addition procedure
was used. Three mixed standards with closer concentrations
to levels of the analytes present in the sample were used for
quantification.
195
3. Results and discussion
The different properties of the ethers (MTBE, TAME,
EBE, and ETBE) and BTEX aromatic hydrocarbons affect
the HS-SPME extraction process. Both ethers and BTEX are
volatile as indicated by the moderate to high vapor pressure
(from 6.6 to 251 mm Hg, at 25 ◦ C). However, the oxygenates
do not rapidly partition from water to air because of their
high aqueous solubility (for MTBE is 50,000 mg/l). There-
fore, they present moderately low Henry’s Law constant (H).
This constant gives the ratio of the partial pressure of the
analyte in the gas phase to the concentration in the water
that is in equilibrium with the partial pressure (expressed in
atmosphere cubic meters per mol). When H is divided by
the product of the gas constant (R, in atm m3 /mol ◦ K) and
the temperature (T, in ◦ K), the resulting H/RT is referred to
as the dimensionless Henry’s Law constant. A compound
with an H/RT value of 0.05 or larger is volatile from water;
and with a low value, tends to remain in water phase [2].
At 25 ◦ C, H/RT for MTBE is 0.02 and for TAME is 0.05;
in contrast, H/RT for benzene is 0.22 and for ethylbenzene
is 0.35 [2,3].
Taking into consideration these different properties, in
all the runs in the screening and optimization designs the
following concentrations were used. For ethers the aqueous
solution contained 500 ␮g/l of each MTBE, EBE, ETBE, and
TAME. For benzene the concentration was 100 ␮g/l, and for
the rest of aromatics (toluene, ethylbenzene, and xylenes)
the concentration was 20 ␮g/l.
3.1. Screening design
Screening is the first step in the efficient assessment of
the factors involved in the studied analytical system. If a
large number of factors are involved, reduced factorial de-
signs are employed. A particular type of those designs is
the Plackett–Burman design, which assumes that the inter-
actions can be completely ignored and so the main effects
are calculated with a reduced number the experiments. With
this design k factors are studied in k + 1 runs, where the
total number or runs must be a multiple of 4.
On the basis of the literature and the experience of the lab-
oratory [15,16,18–20] seven variables were selected to de-
fine the experimental field. The variables considered were:
type of fiber, pH value, concentration of sodium chloride in
water, headspace volume, extraction temperature, extraction
time, and desorption time. Six variables were continuous and
one (fiber type) was qualitative or categorical. In the latter,
either (1) PDMS fiber or (2) PDMS-DVB fiber option was
used. The two coatings were chosen in order to contrast the
PDMS general phase with the PDMS-DVB mixed phase that
is a semipolar coating and presents complementary proper-
ties. The pH value was between 3 and 9. The concentration
of salt ranged from 0 to 300 g/l. The HS volume varied from
10 (1/4 of total volume in 40 ml vial) to 30 ml (3/4 of to-
tal volume in 40 ml vial). The temperature, maintained by
196 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203
water bath, was from 10 to 40 ◦ C. The extraction time was
from 2 to 20 min. And, the desorption time ranged from 0.5
to 3 min. The variables and their code, and the low and high
levels considered are shown in Table 1.
A 27–4 Plackett–Burman design was applied to evaluate
the main effects. Three replicates per run were carried out,
to obtain a value for experimental error. In total, the design
matrix had 24 runs randomly carried out trying to nullify
the effect of extraneous or nuisance variables. The design
matrix and the response for the studied analytes (peak area
in arbitrary units) are also given in Table 1.
The data obtained were evaluated by ANOVA test (not
included here) and visualized by using main effects Pareto
charts (Fig. 1). The Pareto charts showed in Fig. 1 belong to
three of the ethers (MTBE, ETBE, and TAME) and three of
the aromatics (toluene, ethylbenzene and o-xylene). In the
charts, the bar lengths are proportional to the absolute value
of the estimated main effects. These charts also include a
vertical line corresponding to 95% confidence interval. An
effect, which exceeds this reference line, may be consid-
ered significant as regard to the response. The sign of the
main effects showed that the response would be improved
or not on passing a given factor from the lower to the high
level. In this study, the fiber type was the most important
variable, with a positive effect for the PDMS-DVB coating.
The next most influential factor was sodium chloride con-
centration with a positive effect. Headspace volume (HSvol )
and extraction temperature (Textr ) were other two significant
effects in the ethers results.
The better effectiveness of mixed coatings (carboxen-
PDMS, DVB-carboxen-PDMS) compared to PDMS for
these type of analytes using HS-SPME have been also
showed in several works [15,20,24]. The positive effect of
the addition of sodium chloride have been remarked by
other authors [15,16,19,20]; varying the amount of added
salt, 10% (w/w), 5 g/20 ml or even saturation.
Headspace volume is often a variable not considered in the
optimization procedure, and usually 0.5 volume ratio (ml of
solution/ml of vial) is used. The latter is the headspace vol-
ume ratio chosen for single determination of MTBE [15,18];
for BTEX [19,24], and for MTBE, EBE and BTEX [20].
Therefore, the extraction temperature presents a wider range.
For MTBE determination goes from room temperature to
35 ◦ C [15,16,18], and for BTEX varies from low to room
temperature [19,24]. In the MTBE, EBE, and BTEX si-
multaneous determination a 40 ◦ C temperature was chosen
[20].
The results of this screening design led to use the PDMS-
DVB fiber and to eliminate three variables for the follow-
ing optimization step: pH, extraction time, and desorption
time. Hence, the actual pH value of the salty aqueous dilu-
tion (around 5.8), 10 min for extraction time, and 1 min for
desorption time was chosen for the following step.
The time of 10 min for extraction, in headspace mode,
was also used for determination of MTBE [15], and MTBE,
EBE, and BTEX [20]. In other works the extraction time
used was from 5 to 30 min for MTBE [16,18], and from 4
to 30 min for BTEX [19,24].
3.2. Optimization design
The second step is concerned with improving the output
in an analytical system, as a function of several experimental
factors. Many designs for modeling are based on the cen-
tral composite design (CCD, sometimes called a response
surface design). It is assumed that the central point for each
factor is 0, and the design is symmetric around this. A CCD
is constructed as several superimposed designs.
In this study the three variables and its low, central,
and high levels were the followings: extraction tempera-
ture (Text , 15–22–30 ◦ C), sodium chloride concentration
(NaCl, 120–185–250 g/l), and headspace volume (HSvol ,
15–20–25 ml). Those values are showed in Table 2.
CCD consisted of the points of factorial design (2N ) aug-
mented with (2N + 1) star points. In this work, 23 aug-
mented with (2 × 3 + 1). The star points were located at
+α and −α from the center of the experimental domain. An
axial distance α was selected with a value of 1.6818 in or-
der to establish the rotatability condition. With the inclusion
of this condition, the design generates information equally
in all directions; i.e., a rotation of design about the origin
does not alter the variance contours. The runs at the center
of the experimental field were performed three times more.
Therefore, in total the matrix of CCD design involved 18
experiments. The values corresponding to every variable in
each experiment are shown in Table 2. The experiments were
randomly carried out, and each run was done with three in-
dependent samples. The average values of the three data (in
arbitrary units of peak area) are shown in Table 2.
Taking into consideration the above results, the follow-
ing step was to find the select conditions of the independent
variables (Text , NaCl, and HSvol ) that maximize the desir-
ability of the responses on the dependent variables (the an-
alytes MTBE, ETBE, EBE, TAME, and BTEX). This could
be done by response–desirability profiling from central com-
posite designs in STATISTICA program. Firstly, adequate
models (i.e., prediction equations) have to be found to pre-
dict responses of the dependent variables based on the levels
of the independent variables; and second, the levels of the
independent variables that simultaneously product most de-
sirable predicted responses on the dependent variables have
to be determined.
The model used in STATISTICA program for standard
central composite design is a second degree polynomial, re-
sponse surface model. The regression coefficients obtained
are used in computing predicted values for the dependent
variables at different combinations of the independent vari-
ables levels. Then, the desirability function for each depen-
dent variable has to be fixed. This was done by assigning
desirability values of 0.0 (for undesirable, lowest result
in this work), 0.5 (medium), and 1.0 (for very desirable,
highest result in this work). After doing the desirability
Table 1
Experimental variables, levels, 27–4 Plackett–Burman design matrix, and results (in peak area arbitrary units) for MTBE, ETBE, EBE, TAME, and BTEX determination with HS-SPME
Variable Coded Level
Low High

Fiber type Fiber PDMS (1) PDMS/DVB (2)

pH value pH 3 9
NaCl-Salt concentration (g/l) NaCl 0 300
Headspace volume (ml) HSvol 10 30
Extraction temperature (◦ C) Textr 10 40

I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203

Extraction time (min) textr 2 20
Desortion time (min) tdesor 0.5 3
Run Fiber pH NaCl HSvol Textr textr tdesor MTBE ETBE EBE TAME Benzene Toluene Ethyl benzene m,p-Xylene o-Xylene

1 1 3 0 30 40 20 0.5 20 75 225 126 42 33 66 144 73

14 2 3 300 10 40 2 0.5 226 1233 6913 1892 797 1007 1734 3964 1868
6 2 3 300 10 40 2 0.5 204 1005 4388 1592 481 564 871 1995 944
20 2 9 0 30 10 2 0.5 167 558 2451 785 1230 747 818 1668 609
15 1 9 300 10 10 20 0.5 39 142 620 246 75 94 235 534 248
12 2 9 0 30 10 2 0.5 142 606 3235 927 1030 872 1069 2224 853
9 1 3 0 30 40 20 0.5 32 95 296 162 55 46 95 205 101
23 1 9 300 10 10 20 0.5 68 270 1131 429 146 172 391 870 397
21 1 3 300 30 10 2 3.0 144 440 1559 740 199 181 357 817 437
18 2 3 0 10 10 20 3.0 64 300 2061 576 332 496 999 2176 913
4 2 9 0 30 10 2 0.5 160 551 2366 780 1049 677 735 1504 559
22 2 3 300 10 40 2 0.5 235 1173 5372 1784 603 741 1413 3291 1554
16 2 9 300 30 40 20 3.0 319 920 3555 1901 399 477 945 2273 1209
10 2 3 0 10 10 20 3.0 68 326 2528 630 512 788 1361 2898 1155
7 1 9 300 10 10 20 0.5 54 203 823 325 102 117 278 625 283
2 2 3 0 10 10 20 3.0 63 284 2082 558 348 555 1020 2210 915
3 1 9 0 10 40 2 3.0 8 25 90 44 17 16 42 86 37
11 1 9 0 10 40 2 3.0 12 34 111 63 19 18 42 86 40
5 1 3 300 30 10 2 3.0 116 377 1492 607 195 191 386 891 457
8 2 9 300 30 40 20 3.0 329 928 3460 1921 402 480 1000 2439 1277
24 2 9 300 30 40 20 3.0 420 1210 4639 2456 559 670 1321 3237 1720
17 1 3 0 30 40 20 0.5 24 66 177 118 29 25 56 115 61
13 1 3 300 30 10 2 3.0 152 477 1843 787 231 227 458 1060 557
19 1 9 0 10 40 2 3.0 6 16 55 30 10 10 23 49 21

197
198 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203

Fig. 1. Pareto charts of the main effects obtained from 27–4 Plackett–Burman design for MTBE, TAME, ETBE, toluene, ethylbenzene, and o-xylene.

specifications, for each dependent variable is possibly to ob-
tain predicting responses at several levels of single inde-
pendent variable holding the remainder variables constant
at their current setting. There is also the option that allows
obtaining the profile of overall response desirability at each
level of the independent variable, holding the levels of all
other independent variables constant at their current values.
These can be visualized in prediction profiles and desirabil-
ity graphs for each analyte, and overall response desirability
graphs for each independent variable (graphs not showed
here). The overall desirability was computed obtaining a
value of 0.9115; with the settings of Text , HSvol , and NaCl
at 9.9 ◦ C, 20 ml, and 294.3 g/l, respectively. The comparison
of those conditions with the obtained in the single deter-
mination of MTBE [18] showed the similar values for salt
concentration and headspace volume. However, a higher ex-
traction temperature value (20 ◦ C) was best for individual
MTBE determination.
In Fig. 2 is presented the desirability surface response de-
veloped by the model in 3D plots. Those graphs are useful
for interpreting graphically the effect on overall response
desirability of each pair of independent variables. In Fig. 2a
the variables were NaCl concentration and headspace
volume, and in Fig. 2b salt concentration and extraction
 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203 199

Table 2
Experimental variables, levels, matrix of central composite design (CCD), and results (in peak area arbitrary units) for MTBE, ETBE, EBE, TAME, and
BTEX determination with HS-SPME
Variable Coded Level

Low Centre High

Extraction temperature (◦ C) Textr 15 22 30

NaCl- Salt concentration (g/l) NaCl 120 185 250
Headspace volume (ml) Hsvol 15 20 25
Run Textr NaCl HSvol MTBE ETBE EBE TAME Benzene Toluene Ethyl m,p-Xylene o-Xylene
benzene
1 15 120 15 241 1012 5989 1554 1097 993 1635 4007 1701
11 22 185 12 268 1178 6738 1783 1090 1054 2312 5890 2499
5 30 120 15 251 1087 5507 1680 858 707 1704 4279 1954
16 (C) 22 185 20 313 1300 7274 1972 1198 1068 2226 5600 2356
12 22 185 28 359 1313 6024 2107 793 692 1440 3731 1748
9 10 185 20 278 1199 7821 1835 1486 1490 2505 6247 2571
7 30 120 25 247 920 4168 1442 557 491 1198 3123 1399
17 (C) 22 185 20 291 1183 6717 1814 917 872 1833 4713 2063
15 (C) 22 185 20 317 1331 7240 2009 112 1048 2210 5688 2483
13 22 76 20 191 797 4400 1230 727 733 1620 4063 1642
3 15 120 25 242 1022 6015 1585 1038 1010 1875 4689 1944
18 (C) 22 185 20 330 1343 6794 2056 985 935 2099 5472 2419
10 35 185 20 341 1318 5871 2042 778 721 1751 4650 2150
6 30 250 15 390 1586 7294 2416 952 841 1981 5236 2439
4 15 250 25 362 1522 8760 2316 1311 1206 2225 5396 2388
2 15 250 15 354 1527 9038 2417 1336 1228 2499 6755 2840
8 30 250 25 436 1471 6097 2480 750 718 1555 4170 1995
14 22 294 20 524 1955 8733 3210 1093 901 2043 5386 2561
C, central point.

temperature. The best global response was reached when
the headspace volume was around 20 ml (in vial of 40 ml),
salt concentration was close to 300 g/l, and the extraction
temperature near to 10 ◦ C. For practical reasons, the two
latter were the values chosen instead of the previously
indicated (294.3 g/l and 9.9 ◦ C).
Resulting from this study, the working conditions to ob-
tain the best simultaneous response for the studied analytes
were: PDMS-DVB coating fiber, 300 g/l sodium chloride
concentration, 10 ◦ C extraction temperature, and 20 ml (in
a 40 ml glass vial) HS volume. The other fixed conditions
for no significant variables were: pH of the solution, 10 min
extraction time, and 1 min desorption time.
3.3. Method performance and BTEX influence
Procedure repeatability, expressed as relative standard de-
viation (R.S.D.), was obtained from the results of five in-
dependent samples in the same day. The concentrations
of the analytes were the same as previously used in the
screening and optimization runs: 500 ␮g/l for the ethers,
100 ␮g/l for benzene, and 20 ␮g/l for toluene, ethylbenzene

Fig. 2. Response surfaces for global desirability estimated from the central composite design. (a) NaCl concentration and headspace volume; and (b)
NaCl concentration and extraction temperature.
200 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203
and xylenes. R.S.D. was from 2.6 (benzene) to 8.5% (ethyl-
benzene). The values are similar to those found (<8.5%)
working with PT–GC–MS [4]. And also, very close to the
values obtained for the BTEX compounds (3–7%) when
HS-SPME–GC–FID, with PDMS fiber, was used [19].
The method detection limit (MDL) was considered as the
lowest concentration of the analyte that in the whole pro-
cess can differentiate from the background levels. It should
be measured from a real sample that has been treated as
stated in the method procedure [25]. Some works used LOD
calculated by a signal-to-noise ratio of 3 [3]. Other works
obtained it from the calibration plot [23].
In this study, the MDL of each compound was calculated
by the concentration corresponding to the intercept in the
calibration plot of the y-average blank plus three times the
standard deviation from the five independent blank runs.
Five point calibration curves were used with a range of
2.0–20 ␮g/l for ethers and 0.25–2.5 ␮g/l for BTEX. The
calibration equations obtained and used for MDL calcula-
tions were the followings: MTBE (y = 10.144x + 1.867,
R2 = 0.998); ETBE (y = 53.748x + 39.431, R2 =
0.992); TAME (y = 61.007x + 46.549, R2 = 0.994);
EBE (y = 101.986x + 180.225, R2 = 0.996); ben-
zene (y = 224.286x + 3.500, R2 = 0.9996); toluene
(y = 561.029x + 30.800, R2 = 0.9991); ethylbenzene
(y = 359.314x + 6.800, R2 = 0.9994); m,p-xylene
(y = 729.486x + 16.600, R2 = 0.9999); o-xylene (y =
361.143x + 8.000, R2 = 0.9997).
From those calibration equations the MDLs obtained
were: 0.2 (EBE), 0.3 (ETBE), 0.5 (TAME), and 1.1 ␮g/l
(MTBE). For the aromatics those were the MDL values:
0.02 (toluene, ethylbenzene, and xylenes), and 0.07 ␮g/l
(benzene). The comparison with other works that deter-
mined several analytes in the same matrix showed that the
MDLs obtained were similar or slightly better than oth-
ers found using the same technique (HS-SPME–GC–FID).
The differences with those works were mainly in the type
of fiber (PDMS or PDMS-carboxen) and in the extraction
temperature (room temperature or 40 ◦ C) [19,20]. The re-
sults were similar to those indicated using ASTM Method
D4815 (flame ionization detection) [7]. Also, they were
better than the obtained (≤2 ␮g/l for most of the analytes)
working with DAI–GC–MS technique [10]. However, the
values were higher than the showed levels (ng/l magnitude
order) by the researches where SPME or PT techniques
were followed by GC and the more specific MS detector
[3,13,23,26].
The comparison with works devoted only to determination
of MTBE, showed very close values of MDL using static
headspace GC–MS [27], and PT–GC–PID [28] (1.2–2.0 ␮g/l
and 1.0 ␮g/l, respectively), and slightly higher working
with PT–GC–FID (0.2 ␮g/l) [29]. The use of selective MS
detector provided lower MDLs than FID did. The level
of 10 ng/l was obtained working with SPME–GC–MS and
HS-SPME–GC–MS [12,16], and 50 ng/l using PT–GC–MS
[9]. In a previous work, using HS-SPME–GC–FID tech-
nique, with the single MTBE determination as goal, the
MDL value obtained was 0.45 ␮g/l [18]. The increase of the
MDL level is consequence of the cross competition with the
other analytes (BTEX and ethers) present in the sample ma-
trix that affects the adsorption process in this type of fiber.
Adsorption plays an important role in extraction mech-
anism of a mixed coating with porous solid-phase like
PDMS-DVB, while absorption occurs in PDMS. Compo-
sition of matrix can greatly affect the amount of analyte
extracted when compounds with higher affinity for the
PDMS-DVB phase replace compounds with less affinity
[30]. Competitive extraction during solid-phase microex-
traction among BTEX [24], and the influence of BTEX
on MTBE and tert-butyl alcohol have been remarked [8].
In the latter two works the mixed carboxen-PDMS fiber
was used. Working with BTEX have been reported that the
relatively affinity of each component differs according to
the total amount of analytes, and the general affinity to the
fiber followed the order xylene > ethylbenzene > toluene >
benzene [24].
The next experiments were planned in order to known
the influence of the BTEX concentration not only on the
MTBE but also on the rest of the studied ethers (ETBE,
EBE, TAME) working with PDMS-DVB fiber in HS-SPME
mode.
In all the runs for the extraction process the optimized con-
ditions were used. The concentrations of each of the ethers
were fixed at 500 ␮g/l. The BTEX concentrations added
were: 0, 20 (benzene, 10 ␮g/l; toluene, etylbenzene and each
of xylenes, 2 ␮g/l), 100 (benzene, 50 ␮g/l; toluene, etylben-
zene and each of xylenes, 10 ␮g/l), 200 (benzene, 100 ␮g/l;
toluene, etylbenzene and each of xylenes, 20 ␮g/l), 400 (ben-
zene, 200 ␮g/l; toluene, etylbenzene and each of xylenes,
40 ␮g/l), and 2000 ␮g/l (benzene, 1000 ␮g/l; toluene, etyl-
benzene and each of xylenes, 200 ␮g/l). From the data, the
two general affinities for the fiber could be indicated. For the
ethers followed the order EBE > TAME > ETBE > MTBE;
and for the aromatics the affinity order was m,p-xylene >
o-xylene > ethylbenzene > toluene > benzene. The latter
was similar to the order showed using carboxen-PDMS fiber
[24].
The normalized peak areas considering as reference the
system without BTEX were calculated. In Fig. 3 are repre-
sented the normalized values for MTBE, ETBE, EBE, and
TAME versus the BTEX concentration. It can be seen that
the higher the concentration of BTEX the lower the ether
peak areas were, and hence the normalized values repre-
sented. At 20 ␮g/l of BTEX, the areas decrease around 25%,
and at 400 ␮g/l, the peak areas decrease more than 60%
compared with of the highest peak areas. This decrease in
response working with the mixing coating carboxen-PDMS
has been showed [8]. The authors justified the decrease in
MTBE signal by the replacement of less polar aromatic
compounds, especially when total aromatics amount is
above 1 mg/l. In those cases they recommended dilution to
minimize the adverse matrix effect.
 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203 201

1. 2

1. 0
Normalized peak area

0. 8

0. 6

0. 4

0. 2

0. 0
0 500 1000 1500 2000 2500
Total BTEX concentration g/l)

Fig. 3. Normalized peak area of 500 ␮g/l of each ether as a function

of the total concentration of BTEX using 65 ␮m PDMS/DVB fiber with
headspace sampling. (䉬), MTBE; (䊏), ETBE; (×), EBE; (䉱), TAME.

The co-occurrence of ethers, mainly MTBE, with high

levels of BTEX has been registered in several contaminated
samples. For instance, in ground water wells contaminated
by underground storage tank leakage or gasoline spillage
[4,9]. In those samples the dilution is the advisable option
trying to reduce the competition displacement in mixed coat-
ing fiber like PDMS-DVB or carboxen-PDMS.
However, in most of the surface water samples BTEX are
at low concentration or even not detectable. In the results of
the study of Dutch surface water in the period 1992–1997
the maximum level was for toluene in the Rhine river with
an amount of 1.1 ␮g/l [26]. The analysis of a 249 storm wa-
ter runoff samples gave median contaminant concentrations
(when detected) from 0.07 for ethyl benzene to 0.11 ␮g/l
for toluene [3]. In the water analysis in marinas showed that
the most common compound of BTEX found with MTBE
was toluene where concentrations ranged from 1 to 2.0 ␮g/l
[28].
In those samples with low amounts, quantification can
be done using the standard addition method. The levels of
the standards have to be as close as possible to the actual
amounts of the analytes in the sample. With this method
the use of internal standard is avoid, since I.S. should be
Fig. 4. (a) Chromatogram of the standard mixture (500 ␮g/l of each ether,
100 ␮g/l of benzene, and 20 ␮g/l of toluene, ethylbenzene and xylenes),
(b) chromatogram of a marina water sample. Compound identification
number: 1, methanol; 2, MTBE; 3, ETBE; 4, benzene; 5, TAME; 6, EBE;
7, toluene; 8, ethylbenzene; 9, m,p-xylene; and 10, o-xylene.
considered as another interfering analyte that could enter in
competition with the target analytes for the fiber.
3.4. Application
The optimized method was applied to the analysis of 18
superficial water samples. The sampling was done three
times in June 2002, in February, and in June 2003. In
all the samples the standard addition method was used
for quantification of detected analytes. The ranges of the
BTEX compounds are showed in Table 3. Fig. 4b shows

Table 3
Range of BTEX concentrations (␮g/l) in superficial waters of Gipuzkoa (North Spain)
Samples Benzene Toluene Ethylbenzene m,p-Xylene o-Xylene

June 2002
River waters nd to 8.87 nd to 3.38 nd to 0.36 nd to 0.94 nd to 0.39
Marinas/fishing harbors nd to 9.64 0.06 to 2.44 nd to 0.14 0.02 to 0.16 nd to 0.18
February 2003
River waters nd to 0.13 nd to 0.53 nd to 0.07 nd to 0.12 nd to 0.06
Marinas/fishing harbors <ld to 0.23 0.05 to 1.82 <ld to 0.19 nd to 0.34 <ld to 0.27
June 2003
River waters nd to 0.07 nd to 0.26 nd to 0.06 nd to 0.05 nd to 0.05
Marinas/fishing harbors nd to 1.21 0.03 to 6.41 nd to 0.66 0.03 to 1.58 0.03 to 1.28
nd, not detected; <ld, below method detection limit.
202 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203
the chromatogram from one marina water sample obtained
with the proposed analytical procedure. For comparison, a
chromatogram of the standard mixture has been included in
the Fig. 4a.
MTBE was the only alkyl ether present in 15% of the
analyzed samples, all of them in June 2002. Two high MTBE
levels found (more than 400 ␮g/l) were related to accidental
fuel spill, one in a marina recreational area and the other
in a fuel service station. In six samples the MTBE obtained
was higher than the USEPA advisory level for taste and odor
(20–40 ␮g/l). The highest amounts of MTBE often were
found concurrently with the highest levels of BTEX. Respect
to BTEX, those were in a great number of samples in all the
sampling period. Toluene was the analyte most detected (in
90% of the total number of samples analyzed), followed by
xylenes and ethylbenzene. Five samples give concentrations
of benzene higher than 1 ␮g/l (level of EU drinking water
standard). Average concentration of toluene varied from 0.44
(continental waters) to 0.63 ␮g/l (marine waters). More than
80% of the samples contain BTEX amount less than 1 ␮g/l.
Only 5% of the samples presented BTEX values higher than
10 ␮g/l, being the maximum BTEX amount found 14 ␮g/l.
The comparison of alkyl ethers (mainly MTBE) and
BTEX in continental surface water from different areas
showed similar values to most of the samples analyzed in
this work. Concentrations of MTBE and toluene found in
Rhine river respectively were 1.5 and 1.1 ␮g/l [26]. MTBE
and BTEX were detected in several samples of stormwa-
ter runoff, where elevate contaminant concentrations were
related to direct runoff from gas station [3]. In Europe the
studies about MTBE in the aquatic environment have been
extensively done in Germany [16,17]. The ranges varied
depending of the water type; i.e., from river water was
from < 10–1674 ng/l, and for industrial waste waters from
<10–28422 ng/l [17]. The MTBE has been also determined
in marine and recreational areas in California [14,31] and
in South Spain coast [32]. In the latter study the MTBE
concentration ranged from below LOD to 1842 ␮g/l.
4. Conclusions
A screening HS-SPME method for the simultaneous de-
termination of fuel alkyl ethers and BTEX in water was de-
veloped. It was also shown the usefulness of experimental
design taken in the optimization of extraction conditions for
HS-SPME method.
The main advantages of the method were based on the
simplicity of the equipment and rapidity. A PDMS-DVB
fiber for the extraction and GC–FID for the determination
was the equipment required. A time of 10 min was enough
for the extraction of the studied analytes.
Considering the results, the proposed method could be
suitable for screening, and also for determining the content
of MTBE and other alkyl ethers simultaneously with BTEX
in surface contaminated waters. Because the competency
among the analytes for the adsorption process, the method
could apply better in samples with levels of total BTEX
below 20 ␮g/l. Also, the addition standard procedure was
advisable for quantification. The limiting concentration was
for MTBE among the ethers (1.1 ␮g/l) and for benzene
among the BTEX (0.07 ␮g/l). However, these levels were
well enough to satisfy the European standard for drink-
ing water in the case of benzene (1 ␮g/l), and the USEPA
advisory levels for taste an odor in the case of MTBE
(20–40 ␮g/l). Samples with low levels could be analyzed
by means of more sensitive detectors such as MS coupled
to GC.
The method was applied to the analysis of the rivers,
marinas and fishing harbors surface waters from Gipuzkoa
(North Spain). Three sampling were done in one year (June
2002–June 2003). Toluene was the most detected analyte
(in 90% of the samples), with an average concentration of
0.56 ␮g/l. MTBE was the only dialkyl ether detected (in
15% of the samples) showing two high levels over 400 ␮g/l
that were related to accidental fuel spill.
Acknowledgements
The authors acknowledge the financial support from the
Diputación Foral de Gipuzkoa.
References
[1] Directive 98/70/EC of the European Parliament and of the Council
of 13 October 1998 relating to the quality of petrol and diesel fuels
and amending Council Directive 93/12/EEC, Official Journal L 350,
28/12/1998, 0058–0068.
[2] P.J. Squillace, J.F. Pankow, N.E. Korte, J.S. Zogorski, Environ. Tox-
icol. Chem. 16 (1997) 1836.
[3] R.C. Borden, D.C. Black, K.V. McBlief, Environ. Pollut. 118 (2002)
141.
[4] M. Rosell, S. Lacorte, A. Ginebreda, D. Barceló, J. Chromatogr. A
995 (2003) 171.
[5] C. Achten, A. Kolb, W. Püttmann, Atmos. Environ. 35/36 (2001)
6337.
[6] Council Directive 98/83/EC of 3 November 1998 on quality of water
intended for human consumption, Official Journal L 330, 5/12/1998,
0032–0054.
[7] R.U. Halden, A.M. Happel, S.R. Schoen, Environ. Sci. Technol. 35
(2001) 1469.
[8] L. Black, D. Fine, Environ. Sci. Technol. 35 (2001) 3190.
[9] J. Klinger, C. Stieler, F. Sacher, H.-J. Brauch, J. Environ. Monit. 4
(2002) 276.
[10] L. Zwank, T.C. Schmidt, S.B. Haderlein, M. Berg, Environ. Sci.
Technol. 36 (2002) 2054.
[11] J. Pawliszyn, Solid Phase-Microextraction: Theory and Practice,
Wiley-VCH, New York, 1997.
[12] C. Achten, W. Püttmann, Environ. Sci. Technol. 34 (2000) 1359.
[13] D.A. Cassada, Y. Zhang, D.D. Snow, R.F. Spalding, Anal. Chem. 72
(2000) 4654.
[14] J.L. Zuccarello, J.A. Ganske, D.B. Green, Chemosphere 51 (2003)
805.
[15] F. Piazza, A. Barbieri, F.S. Violante, A. Roda, Chemosphere 44
(2001) 539.
 I. Arambarri et al. / J. Chromatogr. A 1033 (2004) 193–203
[16] C. Achten, A. Kolb, W. Püttmann, Fresenius J. Anal. Chem. 371
(2001) 519.
[17] C. Achten, A. Kolb, W. Püttmann, P. Seel, R. Gihr, Environ. Sci.
Technol. 36 (2002) 3652.
[18] J. Dron, R. Garcia, E. Millán, J. Chromatogr. A 963 (2002) 259.
[19] J.C. Flórez, M.L. Fernández, J.E. Sánchez, E. Fernández, A.
Sanz-Medel, Anal. Chim. Acta 415 (2000) 9.
[20] R.B. Gaines, E.B. Ledford, J.D. Stuart, J. Microcolumn Sep. 10
(1998) 597.
[21] P.W. Araujo, R.G. Brereton, Trends Anal. Chem. 15 (1996) 26.
[22] P.W. Araujo, R.G. Brereton, Trends Anal. Chem. 15 (1996) 63.
[23] F. Bianchi, M. Careri, E. Marengo, M. Musci, J. Chromatogr. A 975
(2002) 113.
[24] H.-J. Cho, K. Baek, H.-H. Lee, S.-H. Lee, J.-W. Yang, J. Chromatogr.
A 988 (2003) 177.
203
[25] P. Bruce, P. Minkkinen, M.-L. Riekkola, Mikrochim. Acta 128 (1998)
93.
[26] C.J.H. Miermans, L.E. van der Velde, P.C.M. Frintrop, Chemosphere
40 (2000) 39.
[27] D.T. O’Neill, E.A. Rochette, P.J. Ramsey, Anal. Chem. 74 (2002)
5907.
[28] Y.-J. An, D.H. Kampbell, G.W. Sewell, Environ. Pollut. 118 (2002)
331.
[29] D.P. Lince, L.R. Wilson, G.A. Carlson, A. Bucciferro, Environ. Sci.
Technol. 35 (2001) 1050.
[30] T. Górecki, X. Yu, J. Pawliszyn, Analyst 124 (1999) 643.
[31] J.S. Brown, S.M. Bay, D.J. Greenstein, W.R. Ray, Mar. Pollut. Bull.
42 (2001) 957.
[32] M. Mezcua, A. Agüera, M.D. Hernando, L. Piedra, A.R.
Fernández-Alba, J. Chromatogr. A 999 (2003) 81.