Talanta
ELSEVIER Talanta 52 (2000) 285-291
www.elsevier.com/locate/talanta
Microscale extraction of perchlorate in drinking water with
low level detection by electrospray-mass spectrometry
Matthew L. Magnuson *, Edward T. Urbansky, Catherine A. Kelty
US Environmental Protection Agency, Office of Research and Development, National Risk Management Research Laboratory,
Water Supply and Water Resources Division, Treatment Technology Evaluation Branch, 26 W Martin Luther King Drive,
. Cincinnati, OH 45268 USA
Received 27 October 1999; accepted 14 February 2000 '
Abstract
Improper treatment and disposal of perchlorate can be an environmental hazard in regions where solid rocket
motors are used, tested, or stored. The solubility and mobility of perchlorate lends itself to ground water
contamination, and some of these sources are used for drinking water. Perchlorate in drinking water has been
determined at sub-jig 1 ~ ' levels by extraction of the ion-pair formed between the perchlorate ion and a cationic
surfactant with electrospray-mass spectrometry detection. Confidence in the selective quantification of the perchlorate
ion is increased through both the use of the mass based detection as well as the selectivity of the ion pair. This study
investigates several extraction solvents and experimental work-up procedures in order to achieve high sample
throughput. The method detection limit for perchlorate based on 3.14o-n_, of seven replicate injections was 300 ng
I"1 (parts-per-trillion) for methylene chloride extraction and 270 ng I"1 for methyl isobutyl ketone extraction.
Extraction with methylene chloride produces linear calibration curves, enabling standard addition to be used to
quantify perchlorate in drinking water. Perchlorate determination of a contaminated water compared favorably with
results determined by ion chromatography. © 2000 Elsevier Science B.V. All rights reserved.
i.
Keywords: Microscale extraction; Perchlorate; Drinking water; Low level detection; Electrospray-mass spectrometry
1. Introduction , gestion has potential health effects, related to its
ability to interfere with the proper functioning of
Ammonium perchlorate is used in solid rocket ^ thyroid gland Thereforej perchlorate has been
motors, and their maintenance may result in added to the ug Environmental Protection Agen-
perchlorate infiltrating watersheds by leaching cy's Drinking Water Contaminant Candidate List
and/or groundwater recharge. Perchlorate in- (CCL) ^ ^ ^ from ^^ futoe regulated
- compounds will be selected. The CCL has iden-
d': + 1-513'5697321: fax: + '- tified treatment research as a research priority for
E-mail address: magnuson.matthew@epa.gov (M.L. Magnu- perchlorate. Perchlorate is also subject to the Un-
son) regulated Contaminants Monitoring Regulation
0039-9 140/00/S - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: 80039-9140(00)00342-8
-------
M.L. Magnuson et al /Talanta 52 (2000) 285-291
286
(UCMR) [2]. In order to obtain reliable treatment
data, it is necessary to accurately quantify
perchlorate. The state of California (USA) has set
a maximum drinking water action level at a con-
centration of 18 ug I-1 [3,4], which is a useM
reference concentration in considering analytical
techniques. .
Several methods exist for the analysis of
perchlorate and have been reviewed elsewhere
[3 4] Among techniques for low level determina-
tion (<20 ug I"1), ion selective electrodes can
achieve a detection limit of 10 jig 1~' but require
combination with capillary electrophoresis to re-
duces interfering ions [5]. The extraction of an
ion-parr between perchlorate and an organic dye
[67] can achieve a 3 ug 1~' detection limit, but
because the detection is spectrophotometric, co-
extraction of interfering species such as nitrate
may produce an interference. Ion chromatogra-
phy [8] is popular for perchlorate determination,
but interference can be caused by ions commonly
found with perchlorate, such as iodide. Therefore,
the identification of perchlorate solely on the basis
of its retention time may not withstand legal
challenges [3,4]. Mass based detection has recently
been used to increase confidence in perchlorate
identification. Horlick directly observed perchlo-
rate-with electrospray mass spectrometry^ (ES1-
MS) with a report limit of 5 \ig 1 [9].
Lyophilization for sample concentration has been
followed by direct detection by ESI-MS of
perchlorate to achieve a 2 ug I"1 detection limit
[10]. In this laboratory, we recently investigated
electrospray mass spectrometry analysis to com-
plement the ion chromatography analysis [11,12].
Detection limits of 100 ng I"1 were achieved
through selective extraction of perchlorate from
drinking water as an ion-pair with a catiomc
surfactant (quaternary ammonium salt) [12]. Se-
lectivity for perchlorate in this analysis technique
is enhanced in two ways. First, the extraction was
selective for perchlorate-surfactant ion-parr. Sec-
ond, the ion-paring agent also forms a selective
surfactant-perchlorate complex hi the ESl-Mb
analysis with a higher mass than perchlorate itself.
This provides greater selectivity than observing
the perchlorate mass directly. The perchlorate
mass is in the 'chemical noise' region of the
electrospray mass spectrum, which contains spec-
troscopic interferences from small mass ions in
the sample as well as small mass ions formed by
the electrospray process, i.e. hydrated bromide.
The sample work-up procedure in the previous
study [12] involved extraction with a large volume
of organic solvent, removal of the solvent, and
reconstitution in a different solvent. In our studies
with larger numbers of samples, it was found to
be advantageous to develop a different sample
work-up in order to reduce analysis time and
minimize the generation of excess organic solvents
as hazardous waste. The results of the investiga-
tion into alternate sample work-up procedures are
presented here. The resulting procedure was used
to analyze perchlorate in several water matrices.
r
2. Experimental
2.1. Reagents
Decyl trimethyl ammonium bromide (C-10)
[2082-84-0] was used as received from Fluka
Chemical (Buchs, Switzerland) to make a 0.1 M
stock solution. Methylene chloride (Optima®) was
obtained from Fisher Scientific (Fairlawn, NJ),
1-butanol from EM Science (Gibbstown, NJ), and
methyl isobutyl ketone was obtained from Spec-
trum Chemical (Gardena, CA). Perchlorate for-
tifications were made with ammonium perchlorate
[7790-98-9] (Aldrich, Milwaukee, WI). Dilutions
were made with water deionized through reverse
osmosis.
2.2. Apparatus
Injections were made with a Rheodyne
(Rohnert Park, CA) model 7725 injector with a
200 ul loop. The pump for the carrier liquid was
a Waters 600 (Milford, MA). The mass spectrom-
eter was a Finnigan MAT TSQ-700 (San Jose,
CA) equipped with the standard, Finnigan electro-
spray interface. Mass spectra were acquired in the
negative ion mode by scanning Q3 over appropri-
ate mass ranges. Other experimental parameters
are listed in Table 1.
-------
M.L. Magnuson et al./Talanta 52 (2000) 285-291
287
2.3. Procedure
For comparing different extraction solvents,
the following steps were preformed. Sufficient C-
10 surfactant stock solution was added to make
the aqueous solution 1.0 mM in surfactant. The
C-10 is the ion-pairing agent, and this concentra-
tion was optimized previously [12]. For compar-
ing solvents, the 96.0 ml of the water sample and
5.00 ml solvent were combined in a 100 ml, class
A volumetric borosilicate flask. The flasks were
stoppered, inverted, and vigorously shaken. The
flasks were returned to upright and the phases
were permitted to partly separate. This process
was repeated four times to ensure > 1 min of
vigorous shaking, along with several minutes
during which the phases were partly mixed. The
combination of volumes forces the entire MIBK
layer into the neck of the flask where it may
easily be drawn off. For the MIBK, after the
final shaking, a period of 10-30 min was allowed
for adequate phase separation. The MIBK layer
was drawn off with a Pasteur pipette and placed
in a 1.8 ml screw cap glass vial with a PTFE-
lined septum. For the methylene chloride, after
the final shaking, solution was allowed to phase-
separate (w2 min) in a separatory funnel. The
Table 1
Summary of experimental conditions for 'microextraction' de-
termination of perchlorate by electrospray ionization mass
spectrometry
Acquisition mode
SIM scan window
Scan time
Applied ESI spray potential
(optimized)
Interface capillary
temperature
Sheath gas pressure
Injection mode/injection
volume
Carrier liquid/flow rate
Ion pairing agent
Extraction solvent
Negative ESI-MS on Q3
0.5 amu
0.5 s
4.0 kV
200°C
70 psi (480 kPa)
Flow injection/50.0 ul
Methanol at 0.3 ml min"1
1.0 mM
Decyltrimethylammonium
bromide
Dichloromethane
methylene chloride extracts were then placed in
1.8 ml screw cap glass vials with PTFE-lined
septa.
The alternate procedure for the methylene
chloride extraction was to place 38.0 ml of water
sample, an appropriate amount of stock C-10
solution, 1.8 ml of methylene chloride in a 40 ml
cylindrical glass screw cap vial with PTFE-lined
septa. This is an approximately proportional re-
duction in the volumes discussed above, and fit
conveniently in the vial. After shaking vigorously
for 2 min, the phases were allowed to separate
for 2 min. The vial could be tilted and sufficient
methylene chloride could be drawn from the bot-
tom edge of the vial with a Pasteur pipette. Be-
cause of the small amount of methylene chloride
used, this extraction will be referred to below as
the 'microextraction'.
3. Results and discussion
3.1. ESI-MS analysis
ESI-MS analysis using the flow injection mode
produces signal intensity versus time plots, as
shown in Fig. 1. Each peak in Fig. 1 represents a
separate 50 ul injection of extracts of deionized
water fortified with the perchlorate concentra-
tions indicated. Elution of the peak begins a few
seconds after injection, and the narrow peak
width provides for rapid sample throughput. The
intensity of the signal increases with increased
perchlorate concentration. Selected monitoring of
m/z 380 was used to prepare Fig. 1. This m/z is
derived from an association complex between the
surfactant, a perchlorate ion, and a bromide ion.
The bromide is present in high concentration as
the counter-ion for the surfactant. Detailed mass
spectrometric analysis is presented elsewhere [12].
The shape of the flow injection peak is sufficient
to allow quantification based on peak area. The
injection to injection reproducibility of the mea-
sured peak area averaged 3% for the concentra-
tions shown in Fig. 1. Optimized parameters for
the mass spectrometric signal are shown in Table
1. The optimization of extraction is described
below.
-------
288
M.L. Magnuson et al./Talanta 52 (2000) 285-291
too
80
60
.1
i
41
20
0 -I
50ug/L
20ng/L
lOugit,
5ug/L
10
15
• I '
25
• I '
30
elapsed time (min)
Fig. 1. Flow injection peaks for perchlorate extractions. Each peak represents a separate 50 ul injection of an extract of a perchlorate
solution with the concentration shown. The injection-to-injection error in the measured peak area averages 3% for the four
concentrations shown.
3.2. Choice of extraction solvents
Methylene chloride, 1-butanol, and methyl
isobutyl ketone (MIBK) were investigated. Be-
cause the aqueous solution contains a surfactant,
fairly stable emulsions are possible. Shaking
methylene chloride or MIBK with the aqueous
surfactant solution formed emulsions which were
stable for less than a few minutes. 1-Butanol
formed stable emulsions that did not breakdown
in less than 12 h. 1-Butanol was not investigated
further because of the much more rapid phase
separation possible with methylene chloride or
MIBK. Methylene chloride and MIBK were in-
vestigated in terms of analytical signal resulting
from the extraction.
For the ESI/MS analysis, methylene chloride
and MIBK behaved differently in several ways.
Fig. 2 plots the relative peak area versus fortified
perchlorate concentration. The response for ex-
traction solvents differ by around a factor of 2.
This response difference may be due to differences
in the extraction efficiency of the solvent for the
perchlorate-surfactant ion pair. Another cause of
the response difference may be the electrospray
process. Electrospray efficiency is related to a
number of factors, such as solvent viscosity, sur-
face tension, dielectric constant, and vapor pres-
sure of the solvent [13]. These factors influence the
ability of the perchlorate-surfactant complex to be
formed and ionized hi the electrospray interface
and enter the mass spectrometer. These interac-
tions are complex, fascinating, and beyond the
scope of this paper.
methyl hobutyl Kotone extraction
V-0.0037x2+0.88X+ 0.72, R1-0.933
perchlorate concentration (ugL )
Fig. 2. Peak area versus perchlorate concentration for extrac-
tion with methyl isobutyl ketone and methylene chloride.
Perchlorate solutions were prepared in deionized water. The
solid lines represent the best fit to the equations shown.
-------
M.L. Magnuson et al./Talanta 52 (2000) 285-291
289
0 10 20 30 40 50
perchlorate concentration G-igL"1)
Fig. 3. Peak area versus fortified perchlorate concentration for
three water matrices using the methylene chloride microextrac-
tion. The best fit linear lines are shown for distilled water and
ground water. The best fit line for surface water is not shown
since it falls almost on top of the distilled water line.
The other prominent feature of Fig. 2 is that
the best fit of the data for the MIBK extraction is
a second order polynomial (R2 = 0.993). The lin-
ear fit of the same data results in R2 = 0.959,
indicating less correlation for the linear fit. By
contrast, the best fit for the methylene chloride
extraction is linear (R2 = 0.994). The equations
for the fit are shown on Fig. 2. The cause of the
second order effects for MIBK is not entirely
clear. The leveling out hi the MIBK response
suggests an extraction/solubility effect may be
responsible, but it is possible that some electro-
spray-related phenomenon is partially responsible
as well. At higher perchlorate concentrations, sec-
ond order effects in the perchlorate analysis fol-
lowing the MIBK extraction may complicate data
analysis. It is worth noting that the MIBK extrac-
tion appears to provide a fairly linear fit (R2 —
0.985) for points <20 ug I"1.
Another difference between MIBK and
methylene chloride as extraction solvents was in
the choice of ions to monitor. In ESI/MS,
perchlorate complexes appear at m/z 380 and 400
[12], m/z 380 corresponds to the complex between
the surfactant, a perchlorate, and a bromide,
which is the counter-ion of the surfactant, m/z
400 corresponds to the complex with two perchlo-
rates. The choice of ions for quantification was
investigated for MIBK and methylene chloride.
For MIBK, the correlation coefficients for the
area versus concentration plots was higher when
the sum of the areas of m/z 380 and 400 (Fig. 2)
was used than when either mass was used sepa-
rately. In general (experiments not shown), it was
necessary to use the sum for quantification, i.e. it
was not sufficient to use either mass by itself. For
methylene chloride, a different case occurred. In
other experiments, such as the ones reported be-
low, it was sufficient to use m/z 380 by itself. For
example, the correlation coefficient, 0.994, calcu-
lated for the sum of the peak areas (Fig. 2) did
not differ much from the correlation coefficients
for the plots in Fig. 3 (discussed later), which were
aU > 0.990.
3.3. Method detection limit
The method detection limit (MDL) [14], as
defined in the US Federal Code of Regulations, is
a measure of the precision of replicate injections
of an analyte. The method detection limit for the
ESI/MS analysis of perchlorate using the proce-
dure above was calculated from 3.14ern_, of seven
replicate injections of a low level solution. For a 1
jig 1~' solution, the MDL was calculated to be
around 300 ng I"1 (300 part-per-trillion) for ex-
traction with methylene chloride and around 270
ng 1-1 for MIBK. These detection limits are
based on 100 ml of perchlorate containing water
being transferred through the extraction process
into 5.00 ml of solvent. The MDLs are similar
between the two solvents, possibly because as the
concentration decreases, the peak areas from the
two solvents become similar (Fig. 2).
In the large scale sample work-up [12], a vol-
ume of 500 ml, of water was extracted with 100
ml of methylene chloride and ultimately reconsti-
tuted in 1 ml of solvent. The MDL from this
sequence was around 100 ng I"1, which is only a
factor of three less, even though the nominal
extractive preconcentration factor was about 25
times less. This unexpected result allows for the
use of the small volume extractions in the present
study. The cause of this effect may be due to a
change in the electrospray efficiency. As solutions
become more ionic, the efficiency of the electro-
spray process typically drops. Thus, when the
-------
290
M.L. Magnuson el al./Talanta 52 (2000) 285-291
preconcentration factor is very large, the ionic
strength increases as the.surfactant concentration
increases. A smaller concentration factor may
therefore increase the analytic signal via a higher
electrospray efficiency, which partially makes up
for the loss in analyte concentration due to a
lower amount of analyte preconcentration in the
extraction step. The MDL of the small volume
extraction possibly may be lowered by careful
study of the extraction ratio. However, at 300 ng
I"1, the MDL is sufficient to accomplish the goal
of providing confirmation for ion chromatogra-
phy, which has a detection limit of around 3 ug
I-1 [8].
3.4. Standard addition analysis of drinking water
Based on the comparable MDLs of the MIBK
and methylene chloride and the more simple data
analysis resulting from the linearity of the re-
sponse (Fig. 1), methylene chloride was used for
further experiments. It was deemed inconvenient
to use separatory funnels for the methylene chlo-
ride extraction. Therefore, 40 ml, glass vials with
PTFE-lined septa were used as the 'microextrac-
tion' vessels, as described hi the procedure section.
For the analysis of actual drinking water sam-
ples, it is necessary to subtract out the 'blank
value' of the injection. For example, the blank
value causes the intercepts in Fig. 2 to not have a
value of zero, even for deionized water fortified
with perchlorate. The value may result from a
change in the noise of the instrument as the
solvent changes from the flow injection carrier
liquid, methanol, to the extraction solvent,
methylene chloride or MIBK. The extraction sol-
vent contains other substances extracted from the
water matrix. These substances may vary from
matrix to matrix and affect the electrospray pro-
cess, so each matrix produces a different blank
signal, typically <2 ug 1~', depending on the
water matrix. Because the C-10 surfactant is nec-
essary for the ion-pah- extraction of perchlorate,
when the extraction procedure is performed with-
out the surfactant, the blank value for the water is
obtained. For matrices not contaminated or for-
tified with perchlorate, the background was exper-
imentally determined to be the same with and
without the surfactant. Therefore, the blank value
can be subtracted for quantification of
perchlorate.
The 'microscale' extraction procedure with
methylene chloride was applied to several drink-
hag waters fortified with perchlorate. Calibration
plots for these different water matrices are shown
in Fig. 3. The correlation coefficient for the linear
regression of these calibration curve were all
greater than 0.990, indicating a high degree of
linearity. The slopes for these calibration plots are
somewhat different, especially for the ground wa-
ter, which is expected to have a higher ionic
content. It was shown previously that solutions
with higher ionic content result hi a suppressed
signal [12]. Fig. 3 indicates that perchlorate can be
determined hi these water matrices by the use of
standard additions.
In order to compare the perchlorate determined
by this extraction procedure with another tech-
nique, water contaminated with perchlorate was
obtained from a source in Nevada, USA. Using
standard additions, this water was determined to
contain 8.2 + 0.2 ug I"1 (w = 3) perchlorate. The
water utility determined the concentration to be
8-9 ug I"1 by ion chromatography. The same
water was determined to contain 8.4 ± 0.2 ug 1 ~ '
(n = 3) perchlorate by the large volume extraction
[12]. The ESI-MS determination of perchlorate is
more precise due its lower detection limits (0.3 ug
I"1) than ion chromatography («3 ug I"1). The
concentration of perchlorate determined with this
'microextraction' is the same as the large volume
extraction within experimental error, indicating
the viability of the microextraction. The agree-
ment between the ESI-MS results and the ion
chromatography increases confidence that the
peak quantified as perchlorate by ion chromatog-
raphy is in fact perchlorate and not an interfering
anion.
4. Conclusion
A microextraction of perchlorate has been de-
veloped which provides for the sample throughput
needed for large scale studies of perchlorate in
drinking water. The microextraction is performed
-------
M.L. Magnuson et al /Talanta 52 (2000) 285-291
291
with methylene chloride, and methyl isobutyl ke-
tone can be used for extraction, as well. The
results compare well to ion chromatography, as
well as a previously reported large scale extraction
technique. The detection limit for the microex-
traction is greater than the large scale extraction,
but the microextraction provides sufficient sensi-
tivity to complement ion chromatography. In ad-
dition to increasing the efficiency of the analysis,
the amount of hazardous waste is reduced 50-fold
in the process.
5. Notice
This paper is the work product of United States
government employees engaged in official duties.
As such, it is in the public domain and not subject
to copyright restrictions. Use of trade names or
specific manufacturers equipment or supplies does
not constitute endorsement by USEPA.
References
[1] (a) Fed. Reg. 63 (40) (1998) 10274; (b) US Environmental
Protection Agency, Drinking Water Contaminant List
1998, EPA Document No. 815/F-98-002.
[2] Fed. Reg. 64 (180) (1999) 50555.
[3] (a) E.T. Urbansky, Biorem. J. 2 (1998) 81-95 and refer-
ences therein; (b) Waterweek, Sept 1, 1997;
www.awwa.org/wtrwk901 .htm
[4] E.T. Urbansky, M.R. Schock, J. Environ. Mgmt. 56
(1999) 79-95.
[5] P.C. Hauser, N.D. Renner, A.P.C. Hong, Anal. Chim.
Acta 295 (1994) 181-186.
[6] A.A. Ensafl, B. Rezaei, Anal. Lett. 31 (1998) 167-177.
[7] D.T. Bums, N. Tungkananuruk, Anal. Chim. Acta 199
(1987) 237-240.
[8] (a) P. Harrington, Y. Shen, Proceedings of the American
Water Works Association Water Quality Technology
Conference, Nov. 1-4, San Diego, CA, 1998 (on CD-
ROM); (b) K. Wirt, M. Laikhlrnan, J. Rohrer, P.E.
Jackson, Am. Env. Lab. 10 (1998) 1-5; (c) A. Eaton, A.
Haghani, N. Cox, E. Wong, Proceedings of the American
Water Works Association Water Quality Technology
Conference, Nov. 1-4, San Diego, CA, 1998 (on CD-
ROM).
[9] D.A. Barnett, G. Horlick, J. Anal. Atomic Spectrom. 12
(1997) 497-501.
[10] R. Clewell, D.T. Tsui, American Chemical Society Divi-
sion of Environmental Chemistry Preprints of Extended
Abstracts, 39 (2) (1999) 71-74.
[11] E. T. Urbansky, M.L. Magnuson, D. Freeman, C. Jelks,
J. Anal. Atomic Spectrom. 12 (1999) 1861-1866.
[12] M.L. Magnuson, E.T. Urbansky, C.A. Kelty, Anal.
Chem. 72 (2000) 25-29.
[13] W.M.A. Niessen, Liquid Chromatography-Mass Spec-
trometry, Marcel Dekker, New York, 1999 (passim).
[14] (a) 40CFRI36.3; (b) J.A. Glaser, D.L. Forest, G.D. Mc-
Kee, S.A. Quave, W.L. Budde, Environ. Sci. Technol. 58
(1981)502-510.
-------
------- |