Table 2. Comparison of results from analysis of duplicate samples.
Relative percent difference*
Sample
HMX RDX TNB DNB DNA TNT 24D
4A
2A
4
Dir
0
24
SOE
0
15
6
100
8
18
11
SPE-C
1
12
0
45
8
5
SPE-M
3
8
0
17
2
1
29
Dir
SOE
SPE-C
26
SPE-M
7
B (spike)
Dir
1
0
0
1
1
SOE
4
4
4
3
3
SPE-C
6
1
7
6
6
SPE-M
5
7
7
13
6
* Relative percent difference calculated by subtracting the two values obtained
from duplicates, dividing by the mean value, and multiplying by 100.
tion was then filtered through a 0.45-m nylon
Analytical precision was assessed by comput-
66 membrane (Supelco) into clean brown glass
ing the relative percent difference (RPD) between
bottles and stored at room temperature. Test solu-
duplicate samples based on a single analysis of
tions were made by diluting these stocks with
each (Table 2). In only four cases were the RPD
reagent-grade water.
values greater than 20%: sample 4, RDXdirect
The retention of HMX and RDX by the modi-
analysis; sample 4, DNASOE; sample 4, TNT
fied SDB membranes was tested by extracting a
SPE-C; and sample 29, RDXSPE-C. In all of these
2-L aliquot of reagent-grade water that had been
cases, the concentration values were near MDL
spiked with 100 g/L of HMX and RDX using
values where errors calculated on a percentage
aqueous stocks. Samples of the membrane run-
basis are magnified. Otherwise, analytical preci-
through were taken every 250 mL and analyzed
sion was excellent for both the direct and
by RR-HPLC. The results are plotted in Figure 1.
preconcentration methods, even at low concen-
No breakthrough occured for either analyte until
tration.
more than 1 L had been extracted. This is a sub-
Accuracy was assessed from spike recovery
stantial improvement in retention of these two
data. For spiked reagent-grade water (blank
polar analytes compared with that experienced
spikes), recoveries of all analytes, for both the di-
with the original SDB membranes.
rect and preconcentration methods, ranged from
78 to 102%, with the majority of the recoveries
Quality control
above 90% (Table 3). Recoveries of matrix spikes
QC spiking stock solutions were prepared in
for the direct analysis protocol were also excel-
acetonitrile. Spikes for the direct injection meth-
lent for all analytes in both sample 4 and 29 with
od were made to produce an added concentra-
recoveries ranging from 92.8 to 105.5% (Table 3).
tion about 5 times the average method detection
Recoveries for matrix-spiked sample 29 ranged
limit (MDL) at 100 g/L. Spikes for the samples
from 65 to 107% for the three preconcentration
to be preconcentrated using SOE and SPE were
methods, again indicating excellent recovery. Re-
made to produce an added concentration of 2.00
coveries for matrix-spiked sample 4 are quite vari-
g/L, approximately 100 times the average MDL
able, with values ranging from 23 to 191% for the
reported for the SOE method. Blanks and matrix
SOE method, 49 to 308% for the SPE-M method,
spikes were prepared in 1-L volumetric flasks,
and 91 to 351% for SPE-C (Table 3). However, to
then subsampled for the appropriate method. Two
interpret these results, one must remember that
groundwater samples from NSWC were collect-
spike recovery is calculated by subtracting the
ed in sufficient quantity to allow analysis of ma-
concentration obtained for the original sample
trix duplicates and spikes. In addition, two sets
from the concentration obtained from the spiked
of reagent-grade water blanks (one on each day
sample and expressing the result as a ratio of the
samples were preconcentrated), blank spikes, and
spike recovered relative to the spike added. This
blank spike duplicates were analyzed.
procedure is useful when the concentration of the
5
Previous Page
