ed sediments. As a result, wide variations are rou-
Linearity
tinely observed in concentration estimates of sub-
To determine the feasibility of calibrating head-
samples from individual field-contaminated sed-
space SPME for water contaminated with white
iment samples. This heterogeneity complicates the
phosphorus, a series of aqueous calibration stan-
dards over the range 0.0072 to 7.2 g/L (Table 2)
comparison of different analytical methods.
One potential attribute of headspace SPME is
was prepared in duplicate. Linearity of response
the ability to detect the presence of an analyte with-
(peak heights) with respect to white phosphorus
out altering a sample. Provided that headspace
concentration was tested for water from four sourc-
SPME does not remove a significant amount of
es (a MilliQ purification system, a well water, a
analyte, the same subsample of sediment can be
pond water and a salt marsh) (Table 3). These aque-
ous standards were prepared by adding 25 L of
analyzed by headspace SPME followed by solvent
extraction and GC determination. Using the same
standards prepared in methanol to 25 mL of water,
subsample for both determinations, the results of
which resulted in a methanol concentration of 0.1%.
the two methods can be compared without the
Previous studies with BTEX showed that a metha-
confounding effects of heterogeneity. Since hetero-
nol of less than 1% did not effect peak areas ob-
geneity between subsamples of bulk water sam-
tained by SPME (Arthur et al. 1992c). Peak heights
ples has not been observed, we are primarily con-
obtained by headspace SPME-GC were converted
cerned with soil/sediment samples.
to mass (pg) of white phosphorus from response
To see if concentration estimates based on the
factors based on syringe injections of standards in
amount of white phosphorus in solvent extracts
methanol (Zhang and Pawliszyn 1993a).
would be biased if headspace SPME was per-
Table 2. Concentration of calibration
formed, we used the peak height data in Table 1
standards prepared for headspace SPME
and peak height data from direct injections of cal-
of water samples.
ibration standards to calculate the mass of white
phosphorus detected by SPME. Less than 0.4% of
Mass of white
Conc. of
Conc. of
phosphorus in each
the initial white phosphorus present was removed
methanol std.
aqueous std.
aqueous std.
by headspace SPME from each soil matrix, and less
(g/L)
(g/L)*
(ng)
than 1% from each water matrix. Thus the amount
of white phosphorus removed by one headspace
7.2
0.0072
0.18
14.4
0.0144
0.36
SPME extraction is negligible.
28.8
0.0288
0.72
72
0.072
1.8
Calibration
144
0.144
3.6
1440
1.44
36
has been achieved for other analytes by spiking a
7200
7.2
180
known amount of analyte, generally dissolved in
* Prepared by adding 25 L of methanol
a water-miscible solvent, into a clean water ma-
standard to 25 mL of water.
trix, then performing SPME as is done for regular
Table 3. Water quality measurements for
matrices used for aqueous white phosphorus
plex matrix such as soil is more complicated since
standards.
matrixanalyte interactions are much more pro-
nounced. While methods such as standard addi-
Redox
tions or internal standards have been suggested
Matrix
(mmho/cm)
pH
(mV)
and work well for some analytes (Zhang et al.
Reagent-grade water
0
1994), the difference in extractability between
Well water
0.277
7.6
163
spiked and native analyte can have a profound
Pond water
0.201
7.4
14
influence, unless the extraction is exhaustive. Re-
Salt marsh water
23
7.9
133
cently, exhaustive headspace SPME extraction of
some volatiles was achieved by simultaneous heat-
For data for each water matrix, a linear regres-
ing of the sample and cooling of the fiber coating
sion model with intercept (Fig. 3) and a linear re-
(Zhang and Pawliszyn 1995). Prior to that study,
gression model through the origin were tested for
quantitative results by SPME were reported pri-
lack of fit. In all cases, linear models adequately
marily for water samples, and results for soil sam-
described the data over the concentration range
0.0072 to 0.144 g/L (the F ratio for lack of fit was
ples have been predominantly qualitative.
5