United States
                    Environmental Protection
                    Agency
Environmental Monitoring
Systems Laboratory
Las Vegas NV89114
                    Research and Development
EPA-600/S4-84-027 June 1984
^EPA         Project Summary
                    Inter laboratory  Comparison
                    Study:  Methods for Volatile and
                    Semivolatile  Compounds
                    Donald F. Gurka
                      A program is underway within the
                    U.S. Environmental Protection Agency
                    (EPA) to improve the reliability of its
                    analytical  methods. As part of this
                    program, test protocols for the deter-
                    mination of volatile and semivolatile
                    organic compounds in hazardous wastes
                    were developed and subjected to a nine-
                    laboratory intercomparison study. Vola-
                    tileswere determined by the purge-and-
                    trap gas chromatography/mass spec-
                    trometry (GC/MS) analysis  of a tetra-
                    glyme extract. Semivolatiles were
                    determined by fused silica capillary
                    column (FSCC) GC/MS analysis of a
                    methylene chloride extract obtained
                    from a neutralized dried sample. Detailed
                    quality assurance/quality control (QA/
                    QC) guidelines were provided  to each
                    laboratory. Seven  waste samples and
                    one standard reference material (SRM)
                    were  each analyzed  in triplicate by
                    GC/MS for a target list of  200 com-
                    pounds,  including 45 sample spikes.
                    The protocols were shown to work best
                    for the less polar  aprotic compounds.
                    The between-laboratory analytical
                    precision component for both volatiles
                    and semivolatiles was approximately
                    twice  that of the within-laboratory
                    component. The effect of  excluding
                    outlier data on precision calculations is
                    dramatically illustrated on  analytical
                    data derived from a standard reference
                    material.
                      This Project Summary was developed
                    by EPA's Environmental Monitoring
                    Systems Laboratory. Las Vegas, NV. to
                    announce key findings of the research
                    project that is fully documented in a
                    separate  report of the same title (see
                    Project Report ordering information at
                    back).
Introduction
  The U.S.  Environmental  Protection
Agency (EPA) is currently involved in a
broad program to assess and improve the
reliability of the  data generated using
recommended hazardous waste analyti-
cal protocols. This program is essential to
maintain the credibility of the EPA as a
regulatory agency. Included in this broad
program is  the  submission of those
analytical protocols  in widest usage to
interlaboratory comparison testing. Such
interlaboratory comparison testing is one
of the most  important elements of any
methods validation plan.  Although sub-
stantial expenditure  of both  time and
money is required, multi-laboratory
testing of many EPA methods is currently
underway. The full report describes the
evaluation of analytical protocols for the
determination of volatile (boiling up to ca.
200°C) and semivolatile compounds
(boiling up to ca. 500°C) in hazardous
wastes by an interlaboratory comparison
test.
  The workplan for this project consisted
of three distinct phases:
  1. Selection, evaluation and optimiza-
   tion of the best available methods for
   determining volatile and semivolatile
   organic compounds in solid wastes.
  2. Rigorous single laboratory evaluation
   of the optimized methods and method
   revision where necessary. Prepara-
   tion of all materials for distribution to
   the participating laboratories. Peer
   review of the selected methods.
  3. Interlaboratory comparison testing.
  The guiding workplan philosophy was to
severely challenge the selected methods.
Thus,  the target analytes and  sample
matrices were selected to represent  a
broad spectrum of hazardous waste

-------
analytical  situations. In addition, each
waste sample contained significant
quantities of endogenous compounds.
This ensured the presence of background
interference  problems and provided a
"real  world"  matrix background for the
spike  compounds.
  Two analytical test procedures were
evaluated  in  the  interlaboratory study.
One of the procedures was a GC/MS
analysis procedure for the determination
of volatile organic  compounds. This
procedure involved  the extraction of a
waste sample with  tetraethyleneglycol
dimethyl ether (tetraglyme), the addition
of a  portion  of the  extract to  reagent
water, and analysis according  to the
purge-and-trap  technique  of  EPA Test
Method  624. The  other method,  a
GC/MS analysis method for the determi-
nation of semivolatile organic compounds,
involved the methylene chloride extraction
of a waste sample by sonification under
neutral anhydrous conditions and analysis
of the extract using a fused silica capillary
column.

Experimental Section
Preparation of Calibration
Solutions
  All standard solutions required for the
interlaboratory comparison study  were
purchased  or prepared in sealed  glass
ampoules and distributed to the partici-
pating laboratories. The standards included
calibration solutions, surrogate recovery
solutions, and internal standard solutions.
The standard solutions distributed for the
determination of  volatile  compounds
included six calibration solutions contain-
ing a total  of 53 representative volatile
compounds, a surrogate solution contain-
ing four fluorinated compounds as recovery
standards, and an  internal standard
solution .containing p-bromofluoroben-
zene (BFB) as a mass spectrometer tuning
standard and three deuterated  internal
standards. The standard solutions distri-
buted  for the determination of semi-
volatile compounds included four calibra-
tion  solutions containing  a total of 140
representative semivolatile compounds,
a  surrogate solution containing  three
fluorinated recovery standards, an  inter-
nal standard solution containing decaflu-
orotriphenylphosphine (DFTPP) as a mass
spectrometer tuning standard, 12 deuter-
ated  internal standards  and GC/MS
performance standards.

Preparation of  Spiked Wastes
   The wastes used in the study (Table 1)
were selected to challenge the sample
Table 1.    Interlaboratory Comparison Samples
Sample
Number
2
3
4
5
6
7
8
9
Waste Name
Spiked latex paint waste
Spiked ethanes spent catalyst
Spiked coal gasification tar
Spiked oxych/orination spent catalyst
Spiked POTW sludge
Spiked herbicide acetone-water
Spiked chlorinated ethanes waste
Contaminated river sediment fJS/7/W]
Physical
Description
Semi-so/id
Oily powder
Tar
Pellet/zed solid
Wet filter cake
Liquid
Liquid
Oily powder
Analyses
Performed*
s.v
S.V
S.V
S.V
S.V
V
V
S
"S = determination of semivolatile compounds.
 V - determination of volatile compounds.
workup procedure and to represent a
broad range of waste types. These wastes
were shown to contain a significant
number of naturally incorporated com-
pounds, and were available in sufficient
quantities for a  multi-laboratory study.
The wastes were spiked with the repre-
sentative volatile and semivolatile com-
pounds, these spike concentrations were
chosen to be similar to the levels of other
volatile and  semivolatile components
present in the unspiked wastes. Prelimi-
nary screening  was  used to show that
most of the spike  compounds were
different from those naturally incorporated
within the unspiked wastes.
  To reduce  analytical  costs, a spiking
scheme was devised to  minimize the
number of GC/MS analyses required for
each waste. The scheme involved spiking
each  waste  with  pairs of chemically
compatible compounds. The two com-
pounds in each pair were selected on the
basis  of similar properties (volatility,
solubility, polarity, or acidity) that would
be expected to provide similar recovery
efficiencies.  A previous intralaboratory
study  had demonstrated that the two
compounds in each pair did yield com-
parable recoveries. One compound from
each pair was, spiked at a relatively low
level and the other at an approximately
five-fold higher level. With this approach,
analytical data, for naturally incorporated
components and  recovery data for
different classes of compounds  at two
spike levels,  were obtained  simultane-
ously in a single GC/MS run.
  The  high  and low spike levels used
corresponded to  those levels that would
give approximately 50 ng and 250 ng of
volatile compounds or 10 ng and 50 ng of
semivolatile compounds on the GC
column during analysis if 100% extraction
recovery were achieved. Since the degree
of dilution or concentration required for
each extract varied widely from waste to
waste,  the actual spike level used also
varied widely from waste to waste.
Design of Collaborative Test
  Each participating laboratory was sent
detailed  descriptions of the analysis
protocols, ampoules  of  the standard
solutions, samples of  the seven spiked
wastes and a sample  of NBS Standard
Reference Material No.  1645. The
standard reference material, a contami-
nated river sediment, was  included to
obtain  reference data for future work.
Each laboratory was requested to analyze
three replicates of each spiked waste to
determine volatile compounds and three
replicates of five of the spiked wastes and
the standard reference material for
semivolatile compounds. The laboratories
were requested to search for 53 specific
volatile compounds and 140 specific
semivolatile compounds, and to quantify
the amounts  of any of the compounds
found. These compounds, which  include
both spike compounds and naturally
incorporated compounds, were selected
as representative of various classes of
compounds to challenge the method and
assess its broad applicability. In addition,
the laboratories were requested to report
and  quantify  up to 10  major unlisted
compounds found in each sample.
Statistical Design
  The data reported by the laboratories
participating in this collaborative study
include the amount of compound found
(AF) in the sample, the response factor
(RF) and the retention time(RT) or relative
retention time (RRT). These data were
reported by nine laboratories for 53
volatile compounds in each of seven
sample types,  and for 140 semivolatile
compounds in each of six sample types.
  Since each  of  the nine laboratories
conducted three replicate analyses, there
were potentially 27 values to be reported
for each compound in each sample type.
However, due to various causes, data were
not always reported for each replication.

-------
  The percent of compound recovered by
each laboratory was calculated by
                    I AF „
      % Recovery =  '- J      x 1 nn
                  IAF+SLQ
where IAF is the  initial amount-found in
the sample by intralaboratory analysis,
and SLQ is the spike level quantity. The
summation is  taken  over  all AF data
reported. The percent of AF data reported
was calculated by dividing the number of
data  reported by  27, the  maximum total
number of available data  points.
  For each combination of sample type
and compound, the data  were analyzed
using the random componentsanalysis of
variance model
   Amount-Foundi, (AFij) = fj + U  + dj
        i = 1,2 ..... 9   j = 1,2,3
where AFij is the amount of  compound
found on the  j'h replication  by  the  ith
laboratory.  U is the random  systematic
error  due to laboratory i and E\\ is the
random  within-laboratory error.  It  is
assumed that Li  is distributed normally
with  mean zero  and variance crl2. The
within-laboratory  error term, Cij,  is
assumed to be normally distributed with
mean zero and variance af.
  The total variance for  the  AF  data is
given by:
           
-------
GC/MS response factor  performance
criteria to be recommended for the
interlaboratory comparison study. A  list
of 20 response factors was prepared, and
a response factor  precision of ±40
percent was  recommended to ensure
data quality.  The recommended ±40
percent range was based  on twice the
percent relative standard  deviation  (%
RSD) for compounds with % RSD up to 20
percent in order to provide at least a 95
percent confidence level.

Homogeneity Study
  A study was undertaken to determine if
the spiked and naturally occurring com-
pounds were uniformly distributed among
the sample aliquots  to be distributed to
the interlaboratory test participants.
Aliquots taken from replicate vials of the
homogenized spiked waste samples were
extracted  following the test protocol
procedures. The extracts were analyzed
in duplicate by GC to quantify the GC-
resolvable components, and in triplicate
by a microcoulometric method to deter-
mine total  purgeable organic halogen.
Analysis of variance  treatment indicated
that the samples in  the individual vials
were homogeneous within the precision of
the methods used, with the single excep-
tion of the semivolatile compounds
determined in sample number 6. The net
vial-to-vial variability of the compounds
in that  sample was about 15 percent. The
study indicated that the spike compounds
were as homogeneously distributed with-
in each sample as the naturally incorpor-
ated compounds.

Recovery, Precision,  and
Detectability of  Volatile
Organics
  The average percent recoveries and the
total between-laboratory and within-
laboratory  relative  standard  deviations
for the'spiked and naturally incorporated
volatile compounds in the seven wastes
were determined.  If  data were reported
for fewer than 20 percent of the required
analytical  runs the percent RSDs were
not determined. The variability of volatile
organics results exhibited a dependence
on sample matrix.
  The  within-laboratory variability  for
analysis of volatiles was generally less
than 30 percent but ranged from 5 to 300
percent depending onthecompound.The
total variability generally ranged from 20
to 80  percent for most of the waste
samples. The between-laboratory varia-
bility was usually  less than  70 percent
but ranged from 5 to 300 percent. The
highest variabilities, both  total and
component, were reported for non-spike
compounds that are common background
contaminants such as methylene chloride,
dichloroethane, chloromethane,  and
chloroform. Only in the case of sample 4
were  high percent RSD's reported for
spiked compounds.
  The effect of laboratory on detectability
of specific volatile compounds was
ascertained from the  reported data.
Variability dependent on the laboratories
was evident and was not reproducible on
a per-sample  basis, i.e.,  a certain
laboratory could report a larger number of
data  for one sample than  the other
laboratories and report the  fewest
number for another sample.

Recovery, Precision and
Detectability of Semivolatile
Organics
  The sample-to-sample variability noted
for the volatile compounds was  also
evident  for the semivolatile compound
analyses. The differences in amount of
data  reported  were  greater for the
determination of semivolatiles than for
the determination  of volatiles.  For
example, for sample 3, the  number of
compounds reported varied from 1 to 27,
with  20 being the average  number of
compounds reported.  The  total RSD
generally ranged from 30 to 80 percent.
The  RSD for  the within-laboratory
component was less than 30 percent and
the between-laboratory variability  was
about twice that value. The ranges of both
components  were the  same as those
reported for volatile  organics, 5 to 300
percent. No difference in ranges of values
is apparent for  the spiked versus non-
spiked compounds; however, some of the
poorer precision may be attributable to
polarity and to background contaminants,
such as phthalates.

Effect of Compound, Sample
and Spike Level on the
Detectability of Volatile and
Semivolatile Organics
  A  presentation  of  percent detection
data  for each  waste sample, arranged
according to specific compounds, is
provided in Table 2 for volatile compounds
and in Table 3 for semivolatile compounds.
These two tables provide information
regarding the  effect of compound and
sample  type on the method detection
capability. In  addition, the data  are
identified by  either  low  spike  level
(underlined values) or high  spike  level.
The  average percentage of reported
compounds was higher for both volatile
and semivolatile compounds spiked at the
high level (73 and70 percent, respectively).
The low  level spikes were reported 57
percent of the time for the determination
of volatiles and 47 percent  for the
determination of semivolatiles. The more
difficult waste samples are also apparent
from these data, namely samples 5 and 6
for volatiles and samples 3, 5 and 6 for
semivolatiles analysis.

Total Volatile and Semivolatile
Organic Compounds Reported
  It was instructive to examine the total
number  of volatile  and semivolatile
compounds reported by each laboratory
for each  sample. Such  an examination
could  identify  difficult samples and
differences in  laboratory performance.
The two samples for which the poorest
VOA results were obtained were sample
3 (oxychlorination spent catalyst) and 6
(POTW sludge).  These two solid sample
matrices might be the least retentive for
volatile compounds, and appreciable
amounts  of the  spike or naturally incor-
porated volatile components may have
been lost during sample storage. The best
results were obtained with samples 2, 3,
and 8 while 4 and 7 provided intermediate
results. The differences  in laboratory
performance are more difficult to assess.
However, the  VOA performance  of
laboratory no.  8 was definitely poorer
than that of the other laboratories. The
poorest semivolatile results were reported
by  laboratory  no.  7 while the best
performance was noted for laboratory no.
2. The problem  samples for semivolatile
determinations were samples  5 and 6,
closely followed by sample 3.  The best
semivolatiles results  were obtained for
samples 4 (except by laboratory no. 7) and
2. Laboratory no. 4 experienced consider-
able difficulty  with sample  3 while
laboratory  nos.  1  and 5 experienced
unusual detection problems with semi-
volatiles in sample 5.

Outlier Effect on Standard
Reference Material Results
  The results of all reported data for the
National Bureau of Standards SRM were
surprisingly variable. The total variabilities
ranged from  28 to  338 percent and
averaged  160 percent. The quantities of
compounds reported  were lower  than
those of most of the other test samples,
however,  the quantities of compounds
found in samples 5 and 6 were compara-
ble. The  homogeneity  of this  sample
should be better than  the other test
samples since NBS prepared this material.

-------
 Table 2.    Effect of Compound, Sample and Spike Level on Volatile Compound Detectability

    r mnnunrl                                   Percent of Data Reported for Given Sample3"
                                                          Average Percent
Number
12.
24.
25.
33.
44.
46.
47.

50.
51.

Compound
Propionitrile
2-Chloroa cry I on it rile
1, 1.1-Trichloroethane
1, 2-Dichloropropane
Bromoform
2-Hexanone
1, 1,2.2-Tetrachloro-
ethane
Chlorobenzene
Ethylbenzene
Average
2
11
100
100
100
100
89
89

96
100
87
3
0
96
93
96
85
52
93

96
96
79
4
22
4
78
100
56
37
67

100
89
61
5
11
11
11
56
67
11
100

89
41
44
6
0
4
7
48
4
26
4

WO
89
31
7
11
4
78
100
56
52
100

100
89
66
8
26
78
89
78
89
89
100

100
89
82
Lowc
4
24
65
75
67
38
62

97
77
57
High6
18
67
65
88
63
68
92

97
95
73
Total*
12
42
65
83
65
51
79

97
87
64
"The maximum number of analytical runs for which data could be reported for each sample was 27 (3 replicates x 9 laboratories).
 ThB r»rr*nt nf ri*ta rvpnr,*^ total number of analytical runs for which data were reported\ 10Q
                                               27
^Percent of data reported for compounds spiked at the low level are underlined.
^Average of the percents of data reported for samples in which the compound was spiked at the low level.
aAverage of the percents of data reported for samples in which the compound was spiked at the high level.
"Average of the percents of data reported for all samples.
  A closer inspection of the data reported
revealed that most of the variability could
be attributed to outlier values reported by
one laboratory. In some cases a significant
portion of the variability  was caused by
outlier values reported  by a  second
laboratory. The improvement in the total
variabilities obtained by omitting the data
from one laboratory or two laboratories in
the calculations are shown in Table 4. By
omitting the  data from two  laboratories
the total variabilities ranged from only 13
to 90 percent and averaged  31 percent.
An average total variability of 31  percent
for the analysis of complex sample is  un-
expectedly good. The  within-laboratory
component  of  the  variability would be
much less. Amount-found values for 8 of
12 compounds reported in Set C of Table
4 are within a factor of 2.2 of the values
reported for this SRM in a previous study.
This is surprisingly good  agreement for
the determination of naturally incorporated
compounds  by two  different analytical
methods. One of the four'compounds with
values  outside  the  factor of 2.2 was
obtained for di-n-butyl phthalate, a com-
mon laboratory background contaminant.


Conclusions and
Recommendations
  The  non-aqueous neutral extraction
followed by fused silica capillary column
GC/MS  analysis  eliminates  separate
acid/base extractions  and reduces the
required number of GC/MS runs. The
analytical data  reported for all samples
will be investigated in greater detail to
determine the causes  of  statistical
outliers. If these  outliers result from
laboratory deviations from  the test
Methods and quality control  protocols
they can  be excluded from the pooled
precision results. However, the validity of
discarding outlier results is still a subject
of disagreement among statisticians and
chemists. If outliers are shown  to be a
result of protocol ambiguities, the
appropriate  protocol sections should be
clarified. These steps may  reduce the
number of outliers generated in future
applications of these methods. The
between-laboratory to within-laboratory
ratio of about two for the precision of
these test methods compares favorably
with the recently compiled results of over
fifty  AOAC  interlaboratory comparison
tests.
  The wide variation in individual labora-
tory performance demonstrates the need
for strong QA/QC  monitoring  of labora-
tories performing routine environmental
analyses. Although each participating
laboratory had some  prior experience
with the test protocols, it is anticipated
that further experience will lead to
improved analytical performance.
  Further examination of the  analytical
results and  observations of the  interla-
boratory comparison study for evaluation
of hazardous waste  methodology will
significantly enhance the value of this
complex  study. The  volume of  data
generated far exceeds data obtained for
any  previous  interlaboratory method
evaluation. The scope of the study thus
far included statistical evaluation of only
a small  portion of this data base. In
addition, no provision  was included for
statistical reanalysis excluding outlier
results. Based on the interpretation of the
analytical data  the following specific
recommendations are presented:
  O Expand the statistical treatmentof the
    interlaboratory  comparison  data to
    include analysis of data  to identify
    sources of observed variabilities.
  O Repeat statistical analysis of percent
    recovery data excluding obvious
    outliers.
  O Investigate  sources of significant
    data variability  in  the  following
    method steps:
   1) Prescreening
   2) Extraction (Investigate reasons for
      the  wide  variation in  extract
      aliquot size chosen for  GC/MS
      analysis.)
   3) Instrumental Analysis
   4) Response factor determination
      (Are  anomalous GC/MS response
      factors correlated?)
   5) Confirmation of compound identi-
      fication and quantification proce-
      dures
   6) Calculation
  O Conduct interviews of participating
    laboratories to evaluate interpretation
    of instructions (both  excellent and
    poor performances  would provide
    valuable insight).
  O Modify method  instructions and
    quality control protocols to eliminate
    potential sources of analytical varia-
    bility.
  The completion of these recommended
tasks will  result  in a  more thorough
evaluation  of the data generated by this
collaborative study. It  is anticipated that
the revised statistics will be significantly
improved,  over  those  reported  in this
study, in terms of precision and accuracy.
Further, the resulting revisions in descrip-
tion  of the methods  for  volatile and
semivolatile organic analysis  will result
in more  uniform application of these
methods  in the future.

-------
Table 3.    Effect of Compound, Sample and Spike Level on Semivolatile CompoundDetectability
Compound
Number
5.
12.
13.
15.
16.
22.
25.
28.
33.
36.
43.
50.
51.
56.
57.
64.
65.
66.
69.
74.
76.
77.
80.
84.
85.
98.
100.
117.
118.
120.
121.
123.
125.
126.
132.
137.

Percent of Data Reported
by Given Laboratory***
Compound
4-Chhrotoluene
Bis(2-chloroethyl)ether
2-Chlorophenol
2,4. 6- Trimethylpyridine
1 , 4 -Dichlorobenzene
Acetophenone
Hexachloroethane
4-t-Butylpyridine
2, 4 -Dimelh yl phenol
Propiophenone
4-Chloroaniline
Quinoline
Bisf2-chloroethoxyjethane
4 • Chloro -3-methylaniline
4-Chlorobenzoic acid
1 -Chloronaphthalene
4-Methylquinoline
2-Ethylnapthalene
4-Bromobemoic acid
1 ,3-Dinitrobenzene
2, 6-Dinitrotoluene
3-Nitroaniline
2, 4 -Dinitrophenol
4-Nitrophenol
Pentachlorobenzene
2-Chloro-4-nitroanit/ne
Hexachlorobenzene
Anthraquinone
Fluoranthene
2-Methylanthraquinone
Pyrene
4,4' -ODD
4,4' -DDT
Triphen yl ph osphate
Tri-(p-tolyl)phosphate
Dibenzocarbazole
A verage
2
96
WO
100
59
70
89
89
44
56
89
74
100
56
41
33
33
63
100
22
63
44
44
0
11
74
22
100
100
100
78
WO
78
85
89
63
56
67
3
89
67
67
78
89
89
85
56
4
78
11
78
WO
22
0
0
67
11
11
19
56
7
0
11
78
15
89
78
89
67
22
89
74
89
41
11
51
4
89
96
89
37
100
89
100
89
89
89
56
89
89
78
78
44
89
89
22
78
78
78
56
11
100
59
89
89
100
89
100
89
89
78
89
11
77
5
11
78
0
22
78
67
78
67
0
89
Q
89
56
0
44
44
67
89
0
78
52
0
44
11
78
11
78
81
93
78
100
81
89
59
70
0
52
6
33
22
44
11
93
44
0
85
89
89
11
44
78
44
22
59
78
78
96
0
0
0
0
0
100
0
44
33
100
89
100
89
0
33
85
52
48
Average Percent
LowQ
44
45
44
23
80
72
43
50
30
78
22
61
67
32
11
40
73
45
15
10
58
4
0
8
76
31
70
68
98
73
22
84
74
57
52
22
47
High"
93
91
84
69
90
89
89
80
59
89
43
93
89
41
52
30
73
93
54
23
28
41
33
11
93
8
95
89
89
85
100
89
66
89
81
32
70
Totaf
64
73
60
41
86
76
70
68
48
87
30
80
76
37
35
36
73
73
30
48
46
26
20
9
86
21
80
76
96
80
84
85
67
70
70
26
59
'The maximum number of analytical runs for which data could be reported for each sample was 27
 13 replicates x 9 laboratories).
The percent of data r^nnPri=total number of analytical runs for which data were reported^ wo
                                                   27
^Percent of data reported for compounds spiked at the low level are underlined.
cAverage of the percents of data reported for samples in which the compound was spiked at the low
 level.
d/4 verage of the percents of data reported for samples in which the compound was spiked at the
 high level.
'Average of the percents of data reported for all samples.

-------
Table 4.    Effect of Removing Outliers on Precision for the Determination ofSemivolatile Compounds in NBS Standard Reference Material No. 1645
                                                 Number of Data
Amount-Found, ug/g
Percent Relative
Standard Deviation
Compound
Number
41.
72.
78.
90.
107.
109.
114.
117.
118.
121.
127.
130.
134.
136.
138.
139.
140.
Compound
Naphthalene
Acenaphthylene
Acenaphthene
Fluorene
Phenanthrene
Anthracene
Di-n -butyl phthalate
Anthraquinone
Fluoranthene
Pyrene
Benzofajanthracene
Chrysene
Benzolkjfluoranthene
Benzo(a)pyrene
lndeno(1,2,3-cd)pyrene
Dibenzofa. h)anthracene
Benzolg, h. i)perylene
using
A
18
18
15
14
18
18
15
17
24
24
24
24
21
21
21
21
20
• (jiven uati
B
15
15
12
11
15
15
12
14
21
21
21
21
18
18
18
18
17
3 set
C
12
12
9
8
12
12
9
11
18
18
18
18
15
15
15
15
14
Using
A
4
45
1
14
28
10
22
15
53
93
40
70
40
14
14
10
18
' (jiven uat.
B
1
13
1
2
3
2
4
10
34
69
33
62
27
9
9
7
13
a net
C
1
17
1
1
3
2
5
9
32
61
30
60
26
8
8
5
11
using 0
A
263
278
28
305
338
287
218
102
154
107
79
44
127
95
127
87
93
riven uati
B
20
111
27
42
26
87
48
30
23
23
38
18
28
33
48
73
48
) ier
C
18
90
24
23
13
74
30
15
18
15
32
17
28
18
36
48
29
"Data set A includes all reported data. The maximum number of data possible is 24 (3 replicates x 8 laboratories). One of the 9 laboratories did not
 analyze this sample.
 Data set B includes all reported data from 7 laboratories. One laboratory has been excluded.
 Data set C includes all reported data from 6 laboratories. Two laboratories have been excluded.
   The Project Report was prepared by Battelle-Columbus Laboratories, Columbus,
     OH 43201; the Project Summary was prepared by Donald F. Gurka (also the
     EPA Project Officer, see below) with the Environmental Monitoring Systems
     Laboratory, Las Vegas, NV 89114.
   The complete report, entitled "Interlaboratory Comparision Study: Methods for
     Volatile  and Semivolatile Compounds," (Order  No. PB 84-178 482; Cost:
     $29.50, subject to change) will be available only from:
          National Technical Information Service
          5285 Port Royal Road
          Springfield, VA2216J
          Telephone: 703-487-4650
   Environmental Monitoring Systems Laboratory
   U.S. Environmental Protection Agency
   P.O. Box 15027
   Las Vegas, NV89114

-------