EPA/600/A-97/055
EVALUATION OF A SAMPLING METHOD
FOR ACETONITRILE EMISSIONS FROM STATIONARY SOURCES
Larry D. Johnson and Robert G. Fuerst
National Exposure Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Joette L. Steger and Joan T. Bursey
Eastern Research Group
P.O. Box 2010
Morrisville, NC 27560-2010
ABSTRACT
A method for measurement of acetonitrile emissions from stationary sources of air pollution
has been badly needed. Acetonitrile is a widely used industrial solvent and reaction medium, and
appears in numerous hazardous waste streams. Acetonitrile is one of the most difficult non-
halogenated compounds to incinerate, and has been suggested as an excellent compound for use as a
hazardous constituent spike during trial burn tests of hazardous waste combustors. Lack of an
effective sampling and analysis method has prevented its utilization.
This paper describes successful laboratory development and field evaluation of an effective
method for sampling acetonitrile from stationary sources. The acetonitrile sampling train uses
Modified Method 5 sampling procedures and hardware nearly identical to those described in US EPA
Method 0010, but employs Carboxen-1000 sorbent rather than Amberlite XAD-2. A field evaluation
conducted according to US EPA Method 301 demonstrated RSDs of 17% and 13% for the 20
unspiked and dynamically spiked samples, respectively. The bias was statistically insignificant, so no
bias correction factor was required. The estimated detection limit for the method is 60 ppbv (100
Mg/m3), using quantitation by flame ionization detector.
INTRODUCTION
A method for measurement of acetonitrile emissions from stationary sources of air pollution
has been badly needed for a number of years. Acetonitrile is a component of many industrial
hazardous waste streams, especially from fiberglass and synthetic fiber manufacturing. Acetonitrile is
one of the most difficult non-halogenated compounds to incinerate, and has been suggested as an
excellent compound for use as a hazardous constituent spike during Resource Conservation and
Recovery Act (RCRA) Subpart-B trial burn tests of hazardous waste combustors. It is important to
have a difficult-to-burn spike compound which doesn't produce halogens or halogen acids upon
combustion, in order to test combustors not equipped with scrubber systems. Lack of an effective
sampling and analysis method for acetonitrile has prevented its utilization as a principal organic
hazardous constituent (POHC) indicator compound. In addition to its importance in the field of
hazardous waste combustion, acetonitrile is one of the 189 hazardous air pollutants listed in the Clean
Air Act Amendments of 1990. Acetonitrile is a widely used solvent and reaction medium, and is
employed in production of synthetic fibers, Pharmaceuticals, petroleum products, perfumes, and
fiberglass.
The 82°C boiling point of acetonitrile along with its high polarity and high water-solubility,
all make the compound difficult to sample from stationary sources and to recover and analyze
1
-------
quantitatively. Unconfirmed reports have been received that the compound has been successfully
sampled from low moisture emission or process streams, using a midget impinger train with water as
the collection medium. In theory, a modified Method 0030/ 5041a (VOST) procedure1 should also
work for sources dry enough that no moisture drops are present in the stack, and no condensate is
formed during sample collection. The sample would be collected as usual, using Method 0030, but
would be analyzed by a modified 504la where the water filled purge-and-trap chamber would be
omitted. This modification would avoid problems caused by the poor efficiency and precision of
purging of acetonitrik. No written reports of application of either of these approaches have been
identified. In any case, many of the emissions from sources of interest contain moderate to high
moisture levels, which greatly complicate the process of collecting and analyzing a representative
sample for acetonitrile. The first complication occurs very early in the sampling process. If water
droplets larger than 2 fim are present in the emission stream, then isokinetic sampling is necessary in
order to avoid non-representative sampling of dissolved acetonitrile. This immediately rules out the
use of most methods usually used for volatile organics, including Method 0030, Method 0031',
Method 00401, and Method 182. Each of these methods also contains the potential for serious
acetonitrile recovery problems during analysis, caused by the combination of high water-solubility and
volatility. Method 0010, a.k.a. MM5 or SemiVOST, samples isokinetically but is designed for
collection of compounds with boiling points above 100°C. Lower boiling compounds bind less
strongly to the XAD-2 sorbent, and are swept off the end of the column before sampling is
completed.
Considerable time and effort was expended trying to turn the high water solubility of
acetonitrile into a sampling asset rather than a liability.3 A project was carried out to evaluate
collection of acetonitrile in water-filled impingers using standard Method 5 hardware2. The intent
was to demonstrate a train similar in principle to the reported midget impinger train, but capable of
isokinetic sampling. The project was abandoned after a condensate collector followed by 6 sequential
water impingers only collected 72% of the acetonitrile. Apparently the relatively higher gas flow
rates through the isokinetic train make quantitative trapping of acetonitrile in water impractical.
Attempts to increase the collection efficiency by cooling the impingers and by placing oily barrier
layers on the water were unsuccessful.
EXPERIMENTAL
It was decided that an experimental train configuration with a back-up sorbent trap after water
impingers would be inherently inferior to one with the sorbent ahead of the impingers. Assuming the
sorbent had adequate "stopping power," the latter design would concentrate most of the acetonitrile on
the sorbent, while the former configuration would result in the compound of interest being distributed
through a series of impingers and a sorbent bed.
Final Method
Investigation of revisions to the Method 0010 technology4-5'6 produced a method capable of
isokinetic sampling and analysis of acetonitrile, which performed better than needed to meet Method
301 acceptance criteria.7 The sampling train configuration for the Acetonitrile Sampling Method is
shown in Figure 1. Readers familiar with Method 0010 will recognize that the only change in the
equipment is the substitution of Carboxen 1000 sorbent for the XAD-2 that is usually used in Method
0010. The analytical recovery method employs a reverse gravity or "backflush" extraction with
dichloromethane rather than the usual Soxhlet extraction. Determinative analysis, for this study, was
by Gas Chromatography with Flame lonization Detection (GC/FID). Other detector systems, such as
the Nitrogen Phosphorus Detector (NPD) could be used with this method, but Mass Spectrometry
(MS) usually performs poorly for acetonitrile
-------
Laboratory Phase
Sorbent Selection. The first step in modifying Method 0010 for acetonitrile collection was to
identify and evaluate a sorbent with a sufficiently high volumetric breakthrough capacity for
acetonitrile that collection would be quantitative after two hours of sampling. It was further reasoned
that the extremely high water solubility of acetonitrile might result in a tendency for liquid condensate
to strip the compound from the sorbent bed during sampling. The initial screening tests for sorbent
performance measured the ability of the sorbent to remove acetonitrile from water and to release the
acetonitrile upon extraction.
Selection of the eight sorbents for initial screening was based on a literature survey,
discussions with sorbent suppliers, and previous experience of the investigators. Table 1 shows the
eight sorbents, suppliers of each sorbent, and the results of initial screening tests. "%ACN Retained"
is the portion of the acetonitrile retained on a sorbent column when an aqueous solution of 250 ppm
concentration was passed through it. The percentage retained was acquired by analysis of the eluant
for acetonitrile, followed by subtraction of the amount of acetonitrile in the eluant from the amount
introduced to the column and reduction to a percentage. No attempt to recover the acetonitrile from
the resin was made during this phase of the test. The negative retention figure for Amberlite XAD-7
was a result of sorbent contamination. The eluant contained more acetonitrile, or another interfering
compound, than was contained in the original solution passed through the column. The test was
repeated with similar results. Since several other compounds showed excellent retention, it was not
worthwhile developing a cleaning procedure for the XAD-7. Amberlite 200 and Porapak N both
exhibited unacceptable retention for acetonitrile, and were eliminated from further consideration along
with XAD-7.
The remaining five sorbents were tested to determine recovery of sorbed acetonitrile after
extraction with several solvents. A small column, containing 4-6 g of each sorbent was spiked with
25 mg of acetonitrile in 100 mL of water and then extracted by simple column elution. The column
in Table 1 labeled "%ACN Recovered" lists the recoveries from each sorbent with dichloromethane
elution. Similar recoveries were obtained by elution with mixed carbon disulfide/dimethylformamide
and with mixed dichloromethane/butanol. Ease of handling and simplicity of dealing with a single
solvent rather than a mixture dictated that the solvent of choice be dichloromethane. Porapak T was
dropped from further consideration because of its extremely poor recovery performance. It can easily
be seen that Carboxen 1000 was far superior to the other sorbents in these initial screening tests, and
could have been chosen as the sole sorbent for further evaluation based on the results in Table 1.
Because of the high cost and questionable availability in bulk of Carboxen 1000, the other three top
sorbents in Table 1 were carried throughout the further extraction experiments, described below, in
hope that one of them would exhibit acceptable behavior as the extraction process was optimized. All
three did show improved results, but never reached performance levels of Carboxen 1000.
Extraction Studies. Initial attempts to scale up the extraction procedure for use with full size
Method 0010 sorbent modules produced highly variable recoveries. A reverse gravity elution
procedure as shown in Figure 2 was considerably more successful. Reverse gravity elution of 70 mL
of dichloromethane through the sorbent module resulted in better than 90% recovery of acetonitrile
from Ambersorb XEN-563, Carboxen 1000, and Anasorb 747. The acetonitrile recovery from
Carboxen 569 was less than 40% with dichloromethane. Recovery of acetonitrile from this sorbent
was improved to 80% by use of 1:1 carbon disulfide/dimethylformamide as the extraction solvent.
This level of performance was adequate for use in full scale sampling train tests.
Fpll Scale Sampling Train Tests. The ability of the four sorbents to remove acetonitrile from
simulated stack gas containing high levels of moisture was evaluated using sampling trains containing
-------
multiple sorbent modules in series. Because a limited supply of Carboxen 1000 was available, only
two Carboxen 1000 traps were used in series. Three sequential modules were used in testing the
other three sorbents. The trains were dynamically spiked with an aqueous solution of aeetonitrile.
Each train was rinsed with 1:1 methanol/dichloromethane for the front half and methanol for the
condensor and condensate collector. The results in Table 2 were obtained using simulated stack gas
containing 20-30% moisture and dynamic spike levels of 32-45 ppmv of aeetonitrile. Each entry
represents a single experimental run. Although all four sorbents performed well from an overall
recovery standpoint, the ability of the Carboxen 1000 to collect virtually all of the aeetonitrile on the
first sorbent module made it a clear winner and the sorbent selected for field evaluation.
Carboxen 1000 is a carbon molecular sieve sorbent available from Supelco Inc. in 60/80 mesh size. It
is slightly hydrophobic, which tends to be helpful in sampling high moisture stack gas.
Approximately 48 g of the sorbent is needed to fill a typical Method 0010 sorbent module. At the
time of testing, the cost of a module full of the sorbent was approximately $400.
Field Test
The newly developed aeetonitrile sampling and analysis procedures were field tested using an
experimental design consistent with guidance outlined in EPA Method 301, "Protocol for the Field
Validation of Emission Concentrations from Stationary Sources." The field test included ten "quad
train" runs at a single hazardous waste incinerator emission source. For each quadruple run, four
independent flue gas samples were collected. Two of the sampled gas streams for each quad run were
dynamically spiked with known concentrations of aeetonitrile equivalent to approximately 45 ppmv in
the stack gas. The dynamic spiking procedure and equipment have been described in previous
publications.8-9 The precision of the test method was estimated from the variation in results obtained
for pairs of spiked and unspiked samples. Accuracy (bias) was determined from the differences
between the spiked and measured quantities of aeetonitrile.
Ten quad runs (40 sample trains) were scheduled during the testing program. All
40 independent trains were completed and accepted during the test period. This completion rate
exceeded the minimum requirement of at least six quad runs (24 independent trains) for statistical
analysis by Method 301. This number of runs provided a sample population large enough to produce
credible data quality assessments as described in the project report.6 The static pressure in the stack
was positive, and remained constant at approximately 6.35 mm (0.25 inches) of water during all test
runs. The average sample volume collected was 0.959 ± 0.041 dry standard cubic meters
(33.9 ±1.5 dry standard cubic feet). The sampling time was 60 minutes. Moisture values ranged
from 15 to 28% by volume. Moisture values were low (15 %) for one run because the process was
interrupted during the run. The process interruption did not affect the test data. The source did not
contain aeetonitrile so aeetonitrile levels in the unspiked trains were not reduced by the interruption.
The samples were collected in seven fractions: the probe rinse, the rinse of the front half of
the filter housing, the filter, the rinse of the back half of the filter and the condenser rinse, the
sorbent, the condensate, and the impinger contents. The probe rinse was collected at the end of each
day. The other fractions were collected for each train. Runs 4 and 5 had two sequential sorbent
modules, which were collected and analyzed separately in order to gain additional information about
breakthrough. The high cost and limited bulk availability of the Carboxen 1000 prevented equipping
all ten of the quad runs with two sorbent modules. All of the fractions for Runs 4 and 5, except for
the impinger fraction, were analyzed. Runs 1 through 3 and Runs 6 through 10, had one sorbent
fraction. Only the sorbent and condensate fractions were analyzed for these runs. The impinger
components of the trains were not analyzed and were archived. No aeetonitrile was detected in any of
the sample fractions except the sorbent and some of the condensates.
The percentage of aeetonitrile recovered in all of the analyzed components of each spiked
-------
sampling train ranged from 74 to 119% for the 20 spiked trains. The average recovery was 100%.
The relative standard deviation was 13%. Recovery results for individual spiked runs are shown in
Table 3. Data for unspiked runs are in the field report,6 but show no significant acetonitrile content.
The second sorbent module in Runs 4 and 5 were analyzed, in order that breakthrough of
acetonitrile into the second sorbent could be examined. Any amount of compound detected in the
second sorbent was classified as having broken through the first sorbent module. For the four spiked
double sorbent module trains, breakthrough ranged from 2 to 8 percent. The average breakthrough
was 4 percent. The relative standard deviation was 90 percent. Breakthrough of acetonitrile was
inconsistent. Three of the trains exhibited 2% breakthrough, while one train exhibited 8%
breakthrough. No reason was identified to explain why breakthrough was higher in the one train.
The condensate fraction was analyzed for Runs 1 through 3 and Runs 6 through 10, in order
that breakthrough of acetonitrile into the condensate could be estimated. Since acetonitrile is not
quantitatively collected in water, some of the acetonitrile that broke through the sorbent may not have
been collected, thus causing low bias in the breakthrough estimates for the single sorbent trains. Any
amount of acetonitrile detected in the condensate was classified as having broken through the sorbent
module. No acetonitrile was detected in the condensate for the unspiked single sorbent module trains.
Thus, no breakthrough analysis was possible using these samples. For the 16 spiked double sorbent
module trains, breakthrough ranged from 0 to 11 percent. The average breakthrough was 5 percent.
The relative standard deviation was 73 percent. Two of the trains exhibited 0% breakthrough. These
were the two spiked trains collected when the process went down. Less moisture was collected
during this run than during the other runs, so it may be speculated that the amount of moisture in the
source may contribute to the amount of acetonitrile that breaks through the sorbent. One train
exhibited 11% breakthrough. Calculated breakthrough for all of the other trains was less than 10
percent. Again, breakthrough of acetonitrile was inconsistent. No explanation of why breakthrough
was higher in some trains was identified. Breakthrough was < 10% for 95% of the spiked trains.
For 50% of the spiked trains, breakthrough was <5%. Use of two sorbent modules in series may be
necessary when sampling sources containing higher levels of moisture.
CONCLUSIONS
Based on laboratory studies and one field test, the acetonitrile train consisting of a Method
0010 train with 48 g of Carboxen 1000 in the sorbent module, is adequate for sampling and collection
of acetonitrile from hazardous waste incinerators, and probably other combustion sources. Extraction
using a reverse gravity elution with dichloromethane, followed by analysis with GC/FID was
effective.
The bias for acetonitrile (0.07 mg of acetonitrile at the 74 mg level) was calculated and shown
to be insignificant using Method 301 statistical procedures. Thus, no bias correction factor is
needed.
• The relative standard deviations were 13% for spiked trains and 17% for unspiked trains.
These standard deviations are within the Method 301 criteria of < 50%.
• The mean recovery of 100% and relative standard deviation of 13% for the spiked trains is
within the EPA's Quality Assurance Handbook9 requirements of 50 to 150% recovery and less
than 50% relative standard deviation.
• The detection limit, estimated according to Method 301, for the method is 60 ppbv (100
/ug/m3), using quantitation by flame ionization detector.
-------
• , Greater than 90% of the recovered acetonitrile was collected on the Carboxen-1000.
Essentially no acetonitrile was collected in the probe rinses, in the rinse of the front half of the
filter holder, or on the filters.
• For the four spiked trains containing dual sorbent modules, less than 2% of the acetonitrile
broke through to the second module for three of the trains and less than 8% broke through in
the fourth train.
• For the 16 spiked trains containing single sorbent modules, less than 5% of the acetonitrile
broke through to the condensate for eight of the trains and less than 9% broke through for 15
trains.
STATUS
A draft method and report are currently under review, and are expected to be available by July
1997. The draft acetonitrile method will be transmitted to EPA's Office of Solid Waste (OSW) and to
EPA's Office of Air Quality Planning and Standards (OAQPS). It is expected that OSW will
eventually publish the method as part of a future update to the SW-846 Methods Manual, and that
OAQPS will likely include it in their collection of Provisional Test Methods.
ACKNOWLEDGEMENTS
The following people contributed to one or more phases of the work reported here: Merrill
Jackson, EPA (retired), Ray Merrill, David Epperson, Amy Bederka, Danny Harrison, Steve
Hoskinson, Cheryl Klassa, Jim Howes and Mark Owens, all with Radian Corp./Eastern Research
Group at the time of their contributions. Because of the long time span involved in the multiphase
project, it is inevitable that others who deserve recognition have been left off this list. To them, we
apologize.
NOTICE
The information in this document has been funded wholly or in part by the United States
Environmental Protection Agency under Contracts 68-D1-0010 and 68-D4-0022 to Radian corp. and
Eastern Research Group. It has been subjected to the Agency's peer and administrative review, and it
has been approved for publication. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
REFERENCES
1. Test Methods for Evaluating Solid Waste, Physical/Chemical Methods, SW-846 Manual, 3rd
ed. Document No. 955-001-0000001. Available from Superintendent of Documents, U.S.
Government Printing Office, Washington, DC, November 1986.
2. Code of Federal Regulations, Title 40, Part 60, Appendix A, U.S. Government Printing
Office, Washington, DC, 1994.
3. Steger, J.L. and Hoskinson, S., Development of a Method for Determination of Acetonitrile,
Draft Interim Report, Radian Corp. under Work Assignments 5 & 22, Contract 68-D1-0010 to
U.S. Environmental Protection Agency, Research Triangle Park, NC, September 1992,
4. Steger, J.L. and Klassa, C., Evaluation of Sorbents for Collecting Acetonitrile from Stationary
Sources, Draft Internal Report, Radian Corp. under Work Assignment 58, Contract 68-D1-
0010 to U.S. Environmental Protection Agency, Research Triangle Park, NC, October 1993.
-------
5. Steger, J.L., Acetonitrile Method Development and Field Test, Letter Report, Radian Corp.
under Work Assignment 4, Contract 68-D4-0022 to U.S. Environmental Protection Agency,
Research Triangle Park, NC, September 1995.
6. Steger, J.L., Bursey, J.T., and Epperson, D., Acetonitrile Field Test, Draft Report, Eastern
Research Group under Work Assignment 45, Contract 68-D4-0022 to U.S. Environmental
Protection Agency, Research Triangle Park, NC, September 1995.
7. Code of Federal Regulations, Title 40, Part 63, Appendix A, U.S. Government Printing
Office, Washington, DC, 1993.
8. McGaughey, J.F.; Bursey, J.T.; Merrill, R.G., Field Test of a Generic Method for
Halogenated Hydrocarbons: SemiVOST Test at a Chemical Manufacturing Facility, EPA-
600/R-96/133, PB97-115349, U.S. Environmental Protection Agency, Research Triangle
Park, NC, February 1997.
9. McGaughey, J.F.; Bursey, J.T.; Merrill, R.G., Field Test of a Generic Method for
Halogenated Hydrocarbons, EPA-600/R-93/101, PB93-212181AS, U.S. Environmental
Protection Agency, Research Triangle Park, NC, September 1993.
10. Handbook: Quality Assurance/Quality Control (QA/QC) Procedures for Hazardous Waste
Incineration, EPA/625/6-89/023, U.S. Environmental Protection Agency, Cincinnati, OH,
January 1990.
-------
Table, 1. Results of initial sorbent retention and recovery tests.
Sorbent
Ambersorb XEN-563
Anasorb 747
Carboxen 569
Carboxen 1000
Porapak T
Porapak N
Amberlite 200
Amberlite XAD-7
Table 2. Laboratory train
Sorbent
Ambersorb XEN-563
Carboxen 1000
Carboxen 569
Anasorb 747
Supplier % ACN Retained
Supelco 95
SKC 85
Supelco 88
Supelco 99
Supelco 92
Supelco 16
ICN 8
Supelco -17
recoveries.
Tube 1 Tube 2 Tube 3
85 15 4
76 11 5
89 1 0
99 1 0
75 18 7
88 7 2
75 17 4
81 14 5
%ACN Recovered
66
53
33
94
6
Not tested
Not tested
Not tested
Liquid Total, %
1 105
1 93
0 90
0 100
1 101
1 98
1 97
1 101
-------
Table. 3. Field test recoveries.
Run
1
2
3
4
5
6
7
8
9
10
Acetonitrile
Train A
105
89.3
95.6
99,2
101
73.8
83.9
109
103
109
Recovery3 (%)
Train B
111
100
114
97.2
104
108
114
87.6
74.9
119
"Spike Recovery = (100)(Amount Recovered in Train)/(Amount Spiked)
-------
Figure 1. Acetonitrile sampling train.
Seorsor, Rmj
n«avid
MTTMg^
Ikttig *
V
N.K.:^ \
Figure 2. Reverse gravity sorbent extraction.
10
-------
TECHNICAL REPORT DATA
1. REPORT NO.
EPA/600/A-97/055
2.
4. TITLE AND SUBTITLE
Evaluation of a Sampling Method for Acetonitrile Emissions from
Stationary Sources
5.REPORT DATE
6.PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Larry D. Johnson & Robert G. Fuerst, U.S. EPA
Joette Steger & Joan Bursey, Eastern Research Group
8.PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
National Exposure Research Lab
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Eastern Research Group
P.O. Box 2010
Morrisville, NC 27560-2010
10.PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
68-D4-0022
12. SPONSORING AGENCY NAME AND ADDRESS
National Exposure Research Lab
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
13.TYPE OF REPORT AND PERIOD COVERED
Symposium Proceedings
Measurement of Toxic and
Related Air Pollutants, RTP
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
16. ABSTRACT
A method for measurement of acetonitrile emissions from stationary sources of air pollution has been badly
needed. Acetonitrile is a widely used industrial solvent and reaction medium, and appears in numerous hazardous waste
streams. Acetonitrile is one of the most difficult non-halogenated compounds to incinerate, and has been suggested as an
excellent compound for use as a hazardous constituent spike during trial burn tests of hazardous waste combustors. Lack
of an effective sampling and analysis method has prevented its utilization.
This paper describes successful laboratory development and field evaluation of an effective method for sampling
acetonitrile from stationary sources. The acetonitrile sampling train uses Modified Method 5 sampling procedures and
hardware nearly identical to those described in US EPA Method 0010, but employs Carboxen-1000 sorbent rather than
Amberlite XAD-2. A field evaluation conducted according to US EPA Method 301 demonstrated RSDs of 17% and 13%
for the 20 unspiked and dynamically spiked samples, respectively. The bias was statistically insignificant, so no bias
correction factor was required. The estimated detection limit for the method is 60 ppbv (100 Mg/m3), using quantitation by
flame ionization detector.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/ OPEN ENDED
TERMS
c.COSATI
18. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This
Report)
Unclassified
21.NO. OF PAGES
20. SECURITY CLASS (This
Page)
Unclassified
22. PRICE
------- |