- Email: [email protected]
Sample integrity of trace level volatile organic compounds in ambient air stored in SUMMA® polished canisters
At,,,o,+ric Printed
,‘huironnuat in Great
0@34-6981/%6 53.00+0.00 Pcqpmon Jod Ltd.
Vol. 20. No. 7, pp. 1403-141 I, 1986.
Britaitt.
SAMPLE INTEGRITY OF TRACE LEVEL VOLATILE ORGANIC COMPOUNDS IN AMBIENT AIR STORED IN SUMMA@ POLISHED CANISTERS KAREN D. OLIVER*
and JOACHIMD. PLEIL
Northrop Services,Inc.-EnvironmentalSciences, P. 0. Box 12313,ResearchTriangle Park, NC 27709, U.S.A.
and WILLIAM A. MCCLENNY Methods Development Branch, Environmental Monitoring Systems Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711, U.S.A. (First received 19 August 1985 and injIna1form 19 December 1985) Abstract-Sets of new and used SUMMA@ polished stainless steelcanisters were tested for storage stability of volatile organic compounds (VOCa). Evacuated canisters were filled at a controlled rate with ambient air containing added concentrations of 15 VOCs (14 chlorinated, one brominated) at <: 2 ppbv. Concentrations of VOCs in each canister were then periodically determined during 7day or 30day storage periods using simultaneous flame ionization and electron capture detection. No initial decreasesin concentrations of target compounds were observed, Statistical analysis of data showed that the relative standard deviation of concentrations of most VOCs in each canister set was 10 % or less during the storage periods. For the ‘I-day tests, the mean change in concentration per day was within f 3.2 %. These canisters appear suitable as an alternative to other sampling techniques, at least for most of the compounds tested here.
Key word index: Volatile organic compounds, ambient air, SUMMA’polished containers, stainless steel canisters.
INTRODUCTION The accurate determination
of trace level volatile organic compounds (VOCs) in ambient air requires sophisticated instrumentation. To avoid the cost, inconvenience and diBiculty of transporting such equipment to sampling sites, field samples are collected in stainless steel canisters and returned to a central laboratory for analysis. Stainless steel canisters are not subject to sample permeation or photo-induced chemical effects, and they can be reused after a simple cleanup procedure. Interior surfaces of these stainless steel canisters are passivated using the Molectrics SUMMA@ process. Various sample integrity studies of gases stored in SUMMA@ polished stainless steel canisters have been conducted in other laboratories. Harsch (1980) reported stability of a number of halocarbons stored in canisters at parts per trillion by volume levels. Cox (1983) has discussed the storage stabihties of certain hydrocarbons in canisters at concentrations greater than 2SOppbv. Also, Rasmussen and Khalil(l980) and
*Author to whom correspondence should be addressed.
Rasmussen and Lovelock (1983) have used stainless steel canisters extensively in the field and have reported the stability of halocarbons stored in canisters at high pressures for extended periods. Westberg et al. (1984) reported stability of parts per billion by volume levels of benzene and toluene in canisters, but observed losses of o-xylene. A comparison of sampling containers, including stainless steel canisters, was reported by Pellizmri et al. (1984). They observed an initial decrease in concentrations of halocarbons stored in canisters, which our tests did not co&m. The sample storage characteristics of stainless steel canisters were previously tested in this laboratory using a mixture of 15 VOCs in humid&d zero air. These results were documented in an EPA contract report from Battelle Columbus Laboratories (Holdren et al., 1984). This work reports more detailed experiments that have been conducted using ambient air spiked with less than 2 ppbv of each of 15 VOCs in both new and used stainless steel canisters. The mixtures were initially stored at m 30 psig in the canisters. These experiments were designed to test sample stability under anticipated field sampling conditions in which compounds such as H20, C02, OS, NO and NO1 would be present. 1403
KAREND. OLIVERet al
1404
EXPERIMENTAL Instrumentation
An automated cryogenic sampling and gas chromatographic (GC) system installed in a mobile laboratory was used for sample analysis. This system consisted of a Hewlett-Packard 588OALevel 4 GC equipped with an OV-1 (SO-mx 0.31-mm i.d. x 0.17~pm film thickness) fused silica high resolution capillary column (Hewlett-Packard, Avondale, PA), electron capture (ECD) and flame ionization (FID) detectors, and a modtied Nutech 32@01 cryogenic preconcentration unit (Nutech Corp, Durham, NC). A detailed discussion of the automated sampling and analysis system can be found elsewhere (McClenny et al., 1984). A Model MD-125-48(T) Perma Pure dryer (Perma Pure Products, Farmingdale, NJ) was used to remove water vapor from the gas stream prior to analyte preconcentration. To prevent excessive moisture build-up and any memory effects in the dryer, an automated clean-up procedure that involved periodically heating the dryer while purging with zero air was devised. Experiments were conducted to ensure that this analytical procedure did not alter sample integrity (Pleil and Oliver, 1985). Figure la shows the configuration for 6lling canisters from the sampling manifold. A stainless steel Model MB-151 Metal
-
A
Heated
Marxfold
+
40Umtn
I
Bellows pump (Metal Bellows Corp., Sharon, MA) was used to pressurize the canisters. The flow rate was regulated with a mass flow controller (Model 260, Tylan Corp., Carson, CA) set to deliver 500 ml min- I. All flow rates were verified with a soap-film bubble flowmeter. Figure lb depicts the configuration used during the analysis cycle, in which the sample was vented past the inlet to the system at a rate of 73 ml min _ * using a mass flow controller. This oversupplied the system inlet since the sample was collected at a rate of 34 ml min- I. A total sample volume of 476ml was collected into the cryogenic preconcentrator of the analytical system by sampling for 14 min. Canisters
All canisters used in this study were passivated using the SUMMA@ process. In the 7-day storage experiments with new canisters, four Model 0647 6-t stainless steel canisters (Demaray Scientific Instruments, Ltd., Pullman, WA) and four 3-t stainless steel canisters (Biospherics Research Corp., Hillsboro, OR) were tested. In a similar experiment with used canisters, five Model 0650 6-f stainless steel canisters (Demaray Scientific Instruments, Ltd.) which had been used extensively for 3 months were evaluated. The Model 0650 canisters had been used in a 3-month study involving 22 cities in the United States, so these canisters had been continually filled, shipped and cleaned during that time. A 3Oday storage experiment was also conducted using one Model 0650 and two Model 0647 6-L canisters which had been in use for a minimum of 3 months.
Standards L, 34 mUmin b
Through Perma Pure Dryer tO Analyrlr System
Two Scott Environmental Technology, Inc. pressurized cylinders containing a mixture of VOCs at nominal concentrations of 10 ppmv in nitrogen were used as working standards. These cylinders contained the following compounds: Cylinder 1 benzene toluene o-xylene Cylinder 2 -___ vinyl chloride vinylidene chloride 1,1,2-trichloro-1,2,2,tritluoroethane chloroform 1,Zdichloroethane methyl chloroform carbon tetrachloride hexachloro-l,3-butadiene
Heated
Marxfold
Fig. 1. (A) Configuration for filling stainless steel canisters. (B) Configuration for analysis of samples from stainless steel canisters.
trichioroethylene cis- 1,3-dichloropropene trans-1,3-dichloropropene 1,2-&bromoethane tetrachloroethylene chlorobenzene benzyl chloride.
For calibration runs, standards were diluted with zero air to nominal concentrations of 10 ppbv using mass flow controllers. Ail flows were audited daily, and precise dilution ratios, calibration concentrations, and instrument response factors were calculated. To obtain an indication of instrument stability, two calibration runs were performed each afternoon on consecutive workdays, and the results were averaged daily. Response factors from calibrations were treated statistically to determine precision. Cleaning procedure The cleaning procedure consisted of evacuating the cariisters while heating. Four canisters at a time were @aced in a Fisher Isotemp oven (Model 349, Fisher Scientific, Pittsburgh, PA), and were connected to a vacuum system with Teflon tubing. Canisters were heated to 100°C and were evacuated to less than 23~ Hg using a Wekh Duo-Seal vacuum pump (Model 1400, Sargent-Welch, Skokie, IL) and a liquid nitrogen trap. A Sargent-Welch thermocouple
Sample integrity of trace level volatile organic compounds in ambient air vacuum gauge was used for measuring vacuum This cleaning prccess usually required 5-6 h. Test procedures
Before beginning the sample integrity experiments, all new and used canisters were cleaned. Humidi6ed zero air was passed through the pump, mass flow controller, and tubing of the canister-filling apparatus and analyzed for comparison with chromatograms of instrument background to ensure that the filling apparatus would not contaminate samples. (Chromatograms of instrument background were obtained by passing zero grade air or nitrogen through a 5OOml impinger flask of water, through a mass flow controller, and into the analysis system.) The canisters were then pressurized to 15-20psig with humid&d zero gas. Samples from each canister were analyzed and compared with instrument background. AU canisters were cleaned again at least once before beginning the experiments. For the sample integrity experiments, spiked ambient air samples were obtained by pulling ambient air from a parking area through a glass ‘candy-cane’ inlet to the sampling manifold at a rate of4O/min-‘. The standard mixture of 15 VoCs in cylinder 2 was bled through tubing to the candycane inlet area at a rate of 12 mlmin-i during the canister filling period. This procedure was designed so that ambient air spiked with roughly 1 ppbv of each of the 15 compounds would be available in the sampling manifold. To test initial sample integrity, used 6-f canisters were pressurized individually with spiked ambient air from the sampling manifold. For these experiments, canisters were filled at a rate of 1.2 /mitt-’ for 14 min to exactly coincide with the 14min sample collection cycle of the analytical system. As soon as the analysis of the real-time sample was
completed, an air sample from thecanister was analyzed. This experiment was performed eight times. Four sets of sample integrity experiments were then performed at separate times using the same procedure. The foUowing sets of canisters were evaluated over a ‘I-day period in separate experiments: four new 6-/ canisters, four new 3-I canisters, and five used 6-I canisters. A fourth set of three used 6-f canisters was also evaluated over a 30day period. The instrument was calibrated before beginning each experiment. On Day 0 of each experiment, canisters were pressurized simultaneously to approximately 32 psig with the spiked ambient air mixture from the sampling manifold. Canisters were filled in the mornings during peak trathc activity. Control samples from the sampling manifold were analyzed continually during the filling cycles to provide an estimate of VDC concentrations placed in the canisters. Each analysis required 64 min, so control samples were taken approximately at hourly intervals. A graph of the 14min control sampling periods within the filling cycles for each canister set is shown in Fig. 2. Air samples from each canister in the ‘I-day test were analyzed on Days 0, 2,4 and 7 after filling the 6-1 canisters and on Days 0,4 and 7 after tilling the 3-1 canisters. Samples from canisters in the 30day test were analyzed on Days 1, 15 and 30 after filling. Calibrations were conducted on each day of the experiments.
RESULTS AND DRXXJSSION Canister
blank
tests
Before beginning the sample integrity study, each canister was cleaned and filled with humidified zero
4cIO
1
T-41 , Used 6.Liter C.WllSterS
N&V 3.Liter Canisters
I
N&V 6-Liter Camsterr
0700
0800
0900
1000
1405
1100
1200
1300
1400
Time of Day Canister Fllllng Period
•I
Direct Analysts Sampling Period
Fig. 2. Time periods for filling canisters and for simultaneous control analyses during the filling period.
1406
KAREND. OLIVERet al.
gas. FID and ECD chromatograms of humidified zero gas samples from new canisters revealed trace level contamination in the canisters, but none of the target compounds were present. One large, broad peak appeared early in FID chromatograms of new canister samples but was not evident in chromatograms of instrument background. This contaminant was removed by subsequent cleanings. Chromatograms of humid&d zero gas samples from used canisters were similar to the chromatograms of instrument background. Methylene chloride was identified as a contaminant in several used canisters but was removed by repeated cleaning. No target compounds were identified in humidified zero gas samples from used canisters except for negligible traces (c 0.08 ppbv) of benzene and toluene. These two compounds were also present at similar levels in instrument background samples because of slight contamination within the analytical system. Sample integrity
Comparisons of canister and real-time samples collected simultaneously showed no initial decrease in concentrations of sample compounds in the canisters.
The mean difference (n = 8) in concentrations of target compounds in the manifold and canister sampies ranged from - 5.17 y0 to + 6.24 %. These differences are attributed to the inherent scatter in the system, which is detailed below. Results for the four canister sets are presented in Tables 1,2,3 and 4. Although they were not spiked into the sample stream, benzene, toluene and o-xylene are included since these compounds were present in the ambient air. The mean concentration of each target compound in control runs is included in each table. However, these numbers do not necessarily equal the concentrations placed in canisters because control samples were taken only at intervals during the filling cycle as described previously. Canister data for vinyl chloride and truns-1,3_dichloropropene are not available in Table 4 since some low-level peaks were not integrated by the automated system software. Statistics on data were obtained by treating all canisters and all analyses in each set as equivalent. Tables l-4 list mean concentration, standard deviation and average daily drift in concentration for each compound in each canister set. With the exception of vinylidene chloride and chloroform stored in new
Table 1. Sample integrity results for four new 6-P canisters tested on Days 0, 2, 4 and I Control samples
:PTI (N = 16)
S.D.* (ppbv)
0.76 1.08 1.04 1.11 0.86 1.29 0.31 1.02 0.96 0.48 0.38 0.46 1.04 0.94 0.88 0.15s 1.07 1.10
0.76 2.35 1.17 2.47 0.92 1.38 0.36 0.98$ 1.01 0.50 0.37 0.62 1.10 0.95 0.92 0.09 1.19 1.21
0.04
1.01 1.04 1.21 1.23 1.06 1.12 0.93 O.%
1.05 1.00 1.23 1.23 1.08 1.09 0.94 1.00
M=$-“;J’ Compound FID vinyl chloride vinylidene chloride trichlorotrifluoroethane chloroform 1,2dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cia-1,3dichloropropene trans-1,3-dichloropropene toluene 1,Zdibromoethane tetrachloroethylene chlorobenzene ckxylene benxyl chloride hexachlorobutadiene ECD trichlorotritbmroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,Zdibromoethane tetrachloroethylene hexachlorobutadiene
Canister samples
0.85 0.05 0.92 0.03 0.05 0.01 0.07
Mean 7; change/dayt - 1.58 0.893 -0.256 0.769 -0.870 -0.652 -0.278 -0.102 -0.693 - 1.20 -2.16 0.322 - 1.45 -0.947 - 1.41 -2.22 1.68 -2.15 0.190 -0.300 -0.081 0.655 0.095 0.642 1.49 0.600
*Standard deviation, t mean y0 change/day = (linear regression slope+ mean concentration) x 100, $ N = 6, 8 N = 1.
Sample integrity of trace level volatile organic compounds in ambient air
1407
Table 2. Sample integrity results for four new 3-l canisters tested on Days 0,4 and 7 Control samples
M; Compound FID vinyl chloride vinylidene chloride trichlorotritluoroethane chloroform 1,2_dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-IJ-dichloropropene trans-1,Michloropropene toluene 1,2dibromoethane tetrachloroethylene chlorobenzene o-xylene benzyl chloride hexachlorobutadiene ECD trichlorottiuoroethane chloroform methyl chloroform
carbon tetrachloride trichloroethylene 1,2dibromoethane tetrachloroethylene hexachlorobutadiene
Canister samples
gy) (N = 12)
S.D. (ppbv)
Mean % change/day
0.71 0.98 0.98 1.05 0.88 1.27 0.31 0.99 0.97 0.48 0.37 0.47 1.07 1.21 0.87 0.12t 1.05 1.14
0.74 2.32 1.04 1.19 0.88 1.28 0.35 0.91’ 0.98 0.47 0.37 0.51 1.00 1.06 0.85 0.09 0.94 1.19
0.07 0.35 0.08 0.37 0.04 0.05 0.01 0.01 0.05 0.03 0.03 0.02 0.06 KG 0.03 0.10 0.15
-0.946 -2.20 -1.44 3.03 -1.02 -0.547 -0.286 -0.816 - 1.49 -1.35 -0.196 -1.40 -1.23 -1.76 1.11 -2.77 -277
1.01 1.04 1.21 1.23 1.06 1.09 1.21 1.01
1.02 1.00 1.20 1.24 1.05 1.05 1.16 1.02
0.02 0.02 0.01 0.04 0.02 0.06 0.08 0.05
0.294 -0.400 O.CNlO 0.645 0.190 -1.14 1.64 0.882
(p;bv) =
canisters, mean concentrations for control samples were typically within 0.05 ppbv of the mean concentrations for canister samples. Because these differences were of the same magnitude as the standard deviations of the canister data, the control data and the canister data were essentially identical. Daily drift values (given as a numerical indicator of temporal stability) were calculated by normalizing slopes of linear regression plots to obtain a percent-per-day change, i.e. slope divided by mean concentration. For the 7-day storage tests, the mean change was within f 3.2 % per day, and concentrations of most compounds changed within f 1% per day. With the exception of o-xylene, daily drifts for most compounds in the 30&y test were equivalent to those observed in 7-day tests. For each set .ofcanisters, the mean percentage change/day for compounds on one detector was consistently either slightly positive or negative, whereas the mean percentage change/day of compounds on the other detector was consistently the opposite. This is most likely due to a slight change in the FID/FCD split ratio rather than a canister storage problem. This phenomenon had been previously encountered during analyses that did not involve canisters. Typical concentrations of compounds in individual canisters are presented in Table 5. As examples, FID
and ECD data are tabulated for tetrachloroethylene stored in used 6-P canisters, and FID data are tabulated for chlorobenzene stored in new 3-P canisters. The data are representative of all experiments and show that concentrations of compounds were reproducible for individual canisters. A comparison of the relative standard deviation (R.&D.), i.e. standard deviation divided by the mean, for data from canister tests and calibrations is presented in Table 6. For the majority of target compounds in each canister set, the R.S.D. of concentrations was 10% or less when all results of canister sample analyses were averaged for each compound. A statistical evaluation of calibration response factors (ppbv/area counts) has shown the instrumental error to average approximately 3.6% R.S.D. for the ECD and 6.6 % R.S.D. for the FID over a 6-week period. The R.S.D.s of calibration response factors for individual compounds are presented in Table 6. In most cases the percent RS.D. for compounds in the canisters was similar to the percent R.S.D. for the calibrations. Approximately one-third of the compounds exhibited somewhat greater scatter in data during the 30&y tests than in the 7day tests. FID data for vinylidene chloride, chloroform, oxylene, benzyl chloride and hexachlorobutadiene were
KAREND. OLIVERet al.
1408
Table 3. Sample integrity results for five used W canisters tested on Days 0,2,4 and 7 Control samples
Compound FID vinyl chloride vinylidene chloride trichlorotritluoroethane chloroform 1,tdichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-1,3-dichloropropene trans-1,3-dichloropropene toluene 1,Zdibromoethane tetrachloroethylene chlorobenzene o-xylene betuyl chloride hexachlorobutadiene ECD trichlorotriIIuoroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,tdibromoethane tetrachloroethylene hexachlorobutadiene
Canister samples
Man (ppbv) (N = 5)
Mean (ppbv) (N = 19)
S.D. (ppbv)
Mean I’0 change/day
0.81* 1.11 1.13 1.19 0.90 1.41 0.57 0.946 1.01 0.49 0.3811 0.99 1.23 1.09 0.97 0.1411 0.75 0.96
0.71t 0.98$ 1.13 1.06 0.88 1.37 0.55 0.88$ 0.97 0.46 0.3217 0.89 1.09 1.00 0.87 0.12 0.64 0.91
0.07 0.07 0.09 0.38 0.04 0.05 0.03 0.06 0.06 0.02 0.02 0.05 0.10 0.04 0.06 0.02 0.08 0.10
0.845 0.000 0.708 2.26 0.227 0.073 -0.364 -1.59 0.309 0.435 1.88 1.Ol 0.000 0.100 0.115 0.000 -3.13 0.769
0.99 1.02 1.20 1.52 1.01 0.84 0.97 0.96
0.99 0.99 1.19 1.55 1.00 0.91 0.98 1.01
0.03 0.04 0.04 0.04 0.03 0.05 0.04 0.10
-0.909 -0.909 - 0.924 -0.129 -0.300 -0.879 -0.102 - 0.990
__-.-
*N=2.tN=14,$N=18,§N=4,IjN=3,~N=5.
generally more scattered than data for the other compounds. Ortho-xylene was present at concentrations lower than those for which the experiment was designed, so a large amount of scatter in the data was not unexpected. Historically, excess scatter in data on benzyl chloride and hexachlorobutadiene has been due more to the instrumentation than to a problem with the canisters. FID data on vinylidene chloride and chloroform were not consistent during the experiments. Low-level contaminants in new canisters co-eluted with vinylidene chloride and chloroform on the FID. This resulted in the measureme.nt of slightly higher concentrations of both compounds in samples from new canisters than those measured in control samples. However, in the experiments with used canisters, FID data on vinyl&me chloride did not vary as much as in the experiments with new canisters. Possibly, some tenacious contaminants were removed by repeated cleaning of the canisters. Since chloroform responds well on the ECD, data from the ECD can be used instead of the less reliable FID data when contamination is a problem. However, vinylidene chloride does
not respond on the ECD, so there is no simple way to correct for interference in this case.
CONCLUSIONS
Although initial blank tests on new SUMMA@ polished stainless steel canisters revealed some tracelevel contamination, subsequent cleaning removed many of the contaminants. Used canisters that had been subjected to many sampling/cleaning cycles did not exhibit such contamination. After new and used canisters were cleaned, samples were collected with no measureable cross-contamination from previous samples. One set ofexperiments was designed so that canister samples and real-time samples of ambient air spiked with 15 VOCs were collected simultaneously. Comparison of real-time and canister samples revealed no initial decreases in concentrations of target compounds stored in canisters. Ambient air spiked with low parts per billion by volume levels of 15 VOCs was also stored in new and
Sample integrity of trace level volatile organic compounds in ambient ai Table 4. Sample integrity results for three used 6-I canisters tested on Days 1, 15 and 30 Control samples
Canister samples
“‘(“N _@s5b” Compound FID vinyl chloride vinylidene chloride trichlorotrifluoroethane chloroform 1,2-dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-1,3-dichloropropne tram-1,fdichloropropcne toluene 1,2dibromoethane tetrachloroethylene chlorobenzene o-xylene benzyl chloride hexachlorobutadiene ECD trichlorotrilluoroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,2dibromocthane tetrachloroethylene hexachlorobutadiene
Mean (Ppbv) (iv = 9)
-
0.81* 1.11 1.13 1.19 0.90 1.41 0.57 0.94t 1.01 0.49 0.38$ 0.99 1.23 1.09 0.97 0.14$ 0.75 0.96
0.07 0.09 0.45 0.06 0.09 0.03 0.09 0.05 0.04 0.12 0.07 0.03 0.04 0.15
1.09 1.13 1.32 0.93 1.41 0.61 0.87 1.02 0.50 1.01 1.23 ;z 0.23 0.84 1.05
0.99 1.02 1.20 1.52 1.01 0.84 0.97 0.96
Mean% change/day
&k)
0.97 0.97 1.19 1.59 1.00 0.87 1.01 0.98
:;
-0-14 -0.44 1.14 -0.40 -0.37 0.18 -0.49 -0.38 0.12 0.67 0.10 -0.10 0.24 4.40 0.36 1.17
0.02 0.03 0.02 0.07 0.02 0.04 0.05 0.10
-0.18 -0.21 -0.14 -0.30 -0.07 0.11 0.34 0.53
Table 5. Typical examples of data for canister samples* a FID-tetrachIoroethylenet CAN 106 CAN 04 DaY 0
1.01 1.06 2 4 oa94t zzi 7 1:04 0.97 b. ECD-tetrachloroethylenet DaY CAN 106 CAN 04 20 1.00 0.97 4 0.99 7 0.98 c. FID-chlorobenzeneQ DaY CAN 5 0 4 7
0.89 0.80 0.81
CAN 11
CAN 15
CAN 190
1.00 1.00 0.95 1.00
1.01 1.00 ;;r
1.04 1.01 0.96 1.05
CAN 11
CAN 15
CAN 190
0.99 1.00 1.00 1.01
0.98 1.01 0.95 1.00
kg 2
CAN 6
CAN 7
CAN8
0.92 0.80 0.82
0.95 0.79 0.84
0.w 0.80 0.87
;: 685 0.99
* Values expressed as parts per billion by volum, t Stored in used 6-f canisters, $ Data for this run not available due to column blockage, #Stored in new 3-f canisters.
1409
1410
KAREN
OLIVER et al.
D.
Table 6. Comparison of scatter in calibration data and canister data (% R.S.D.*)
Compound FID vinyl chloride vinvlidene chloride tr~hlorotrifluoroethane chloroform 1,2_dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-1,3-dichloropropene tram-1,3-dichloropropene toluene 1,2dibromoethane tetrachloroethylene chlorobenxene o-xylene benxyl chloride hexachlorobutadiene ECD trichlorotrifluoroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,2dibromoethane tetrachloroethylene hexachlorobutadiene
Calibration Data (N = 33)t
Canister data, 7-day test _________ _ New 6P New 3/ Used 6P (N = 16) (N = 12) (N = 19)
6.37 7.24 6.96 6.21 5.91 5.38 5.30 4.84 5.44 6.86 6.81 5.11 8.08 6.38 8.64 6.88 15.8 17.5
4.00 8.11 1.61 5.45 3.16 4.35 22.2 12.6 9.09
3.94 3.08 3.00 1.38 2.31 6.21 3.54 4.95
1.90 2.00 1.63 2.44 1.85 4.59 5.32 5.00
5.26 36.2 4.27 37.3 3.26 3.62 2.78
9.46 15.1 7.69 31.1 4.55 3.91 2.86 l.lOT[ 5.10 6.38 8.11 3.92 6.00 5.66 1.06
33.3 10.6 12.6 1.96 2.00 0.83 3.23 1.90 5.71 6.90 4.90
Canister data, 3@day test
~~~ ._..___
Used 6 ! (N = 9) .--
9.86$ 7.14$ 7.96 35.9 4.55 3.65 5.45 6.824 6.19 2.17 6.25.. 5.62 9.17 4.00 6.90 16.7 12.5 11.0
5.98 7.59 33.8 6.53 6.06 5.52 10.8 5.29 7.63 .11.8 5.60 3.10 4.55 65.4 11.2 18.7
3.03 4.04 3.36 2.58 3.00 5.49 4.08 9.90
2.27 3.56 1.76 4.12 1.58 4.65 4.88 10.1
*Relative standard deviation, t Based on 33 pairs of daily calibrations, $N = 14, $N = 18, (1N = 6, 7N = 2, ‘*N = IS.
used canisters for ‘I-day and 30-day periods. Because a number of canisters were filled simultaneously for each experiment, the fill cycle was longer, and several realtime control samples were analyzed, a simultaneous real-time analysis of several hours was not possible. Of the compounds examined, none exhibited any signiftcant loss or gain in concentration during the test periods. Low-level contaminant peaks from new canisters sometimes co-eluted with chloroform and vinylidene chloride on the FID, creating a positive bias in the measurement of concentrations of these two compounds. This was not observed in used canisters
that had been cleaned repeatedly. However, for most target compounds the sample integrity results for all canister sets were similar to system precision. The good sample integrity observed in these tests suggests the routine use of canisters as a viable alternative to other sampling techniques, at least for the compounds tested. Acknowledgement-The authors thank Dr B. A. Rasmussen for helpful advice and assistance during the study. Members of the Methods Standardization Branch (MBB), Environmental Monitoring Systems Lsbonrtory (EMSL), Environmental Protection Agency (EPA) and Gas, Kinetics
and Photochemistry Branch (GKPB), Atmospheric Sciences Research Laboratory (ASRL), EPA also shared valuable experience. Disclaimer-Although the research described in this article has been funded wholly or in part by the U.S. Environmental Protection Agency through Contract 68-02-4035 to Northrop Services, Inc.-Environmental Sciences, it has not been subjected to agency review and therefore does not necessarily reflect the views of the agency. No official endorsement should be inferred. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.
REFERENCES
Cox R. D. (1983) Sample collection and analytical techniques
for volatile organics in air. Presented at APCA Specialty Conference, Chicago, IL, 22-24 March. Harsch D. E (1980) Evaluation of a versatile gas sampling container design. &rospheric Environment 1&1105-i 10% Holdren M.. Rust S., Smith R. and Koetx J. (19g4) Evaluation of cryogenic trapping as a means for coll&ing organic compounds in ambient air. Final report on EPA Contract 68-02-3487 to BatteBe Columbus Laboratories, Columbus, OH. McCknny W. A., Pleil J. D., Ho&en M. W. and Smith R. N. (1984) Automated cryogenic preconcentration and gas
Sample integrity of trace level volatile organic compounds in ambient air chromatographic determination of volatile organic compounds in air. Analyt. Chem. 56,2947-2951. Pelhzxari E. D., Gutknecht W. F., Cooper S. and Hardison D. (1984) Evaluation of sampling methods for gaseous atmospheric samples. Final report on EPA Contract No. 68-022991 to Research Triangle Institute, Research Triangle Park, NC. Pleil J. D. and Oliver K. D. (1985) Evaluation of various configurations of Nalion dryers: water removal from air samples prior to gas chromatographic analysis. Technical Note 85-03. Northrop Services, Inc.-Environmental Sciences, Research Triangle Park, NC.
1411
Rasmussen R. A. and Khalil M. A. K. (1980) Atmospheric halocarbons: measurements and analyses of selected trace gases. Proc. NATO ASI on Atmospheric Ozone, pp. 209-231. Rasmussen R. A. and Lovelock J. E. (1983) The atmospheric lifetime experiment-2. Calibrations. J. geophys. Res. 88, 83698378. Westberg H., Lonneman W. and Holdren M. (1984) Analysis of individual hydrocarbon species in ambient atmospheres: techniques and data validity. In Identijkation and Analysis of Organic Pollutants in Air (edited by Keith L. H.), pp. 323327. Butterworth, Woburn, MA.
,‘huironnuat in Great
0@34-6981/%6 53.00+0.00 Pcqpmon Jod Ltd.
Vol. 20. No. 7, pp. 1403-141 I, 1986.
Britaitt.
SAMPLE INTEGRITY OF TRACE LEVEL VOLATILE ORGANIC COMPOUNDS IN AMBIENT AIR STORED IN SUMMA@ POLISHED CANISTERS KAREN D. OLIVER*
and JOACHIMD. PLEIL
Northrop Services,Inc.-EnvironmentalSciences, P. 0. Box 12313,ResearchTriangle Park, NC 27709, U.S.A.
and WILLIAM A. MCCLENNY Methods Development Branch, Environmental Monitoring Systems Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711, U.S.A. (First received 19 August 1985 and injIna1form 19 December 1985) Abstract-Sets of new and used SUMMA@ polished stainless steelcanisters were tested for storage stability of volatile organic compounds (VOCa). Evacuated canisters were filled at a controlled rate with ambient air containing added concentrations of 15 VOCs (14 chlorinated, one brominated) at <: 2 ppbv. Concentrations of VOCs in each canister were then periodically determined during 7day or 30day storage periods using simultaneous flame ionization and electron capture detection. No initial decreasesin concentrations of target compounds were observed, Statistical analysis of data showed that the relative standard deviation of concentrations of most VOCs in each canister set was 10 % or less during the storage periods. For the ‘I-day tests, the mean change in concentration per day was within f 3.2 %. These canisters appear suitable as an alternative to other sampling techniques, at least for most of the compounds tested here.
Key word index: Volatile organic compounds, ambient air, SUMMA’polished containers, stainless steel canisters.
INTRODUCTION The accurate determination
of trace level volatile organic compounds (VOCs) in ambient air requires sophisticated instrumentation. To avoid the cost, inconvenience and diBiculty of transporting such equipment to sampling sites, field samples are collected in stainless steel canisters and returned to a central laboratory for analysis. Stainless steel canisters are not subject to sample permeation or photo-induced chemical effects, and they can be reused after a simple cleanup procedure. Interior surfaces of these stainless steel canisters are passivated using the Molectrics SUMMA@ process. Various sample integrity studies of gases stored in SUMMA@ polished stainless steel canisters have been conducted in other laboratories. Harsch (1980) reported stability of a number of halocarbons stored in canisters at parts per trillion by volume levels. Cox (1983) has discussed the storage stabihties of certain hydrocarbons in canisters at concentrations greater than 2SOppbv. Also, Rasmussen and Khalil(l980) and
*Author to whom correspondence should be addressed.
Rasmussen and Lovelock (1983) have used stainless steel canisters extensively in the field and have reported the stability of halocarbons stored in canisters at high pressures for extended periods. Westberg et al. (1984) reported stability of parts per billion by volume levels of benzene and toluene in canisters, but observed losses of o-xylene. A comparison of sampling containers, including stainless steel canisters, was reported by Pellizmri et al. (1984). They observed an initial decrease in concentrations of halocarbons stored in canisters, which our tests did not co&m. The sample storage characteristics of stainless steel canisters were previously tested in this laboratory using a mixture of 15 VOCs in humid&d zero air. These results were documented in an EPA contract report from Battelle Columbus Laboratories (Holdren et al., 1984). This work reports more detailed experiments that have been conducted using ambient air spiked with less than 2 ppbv of each of 15 VOCs in both new and used stainless steel canisters. The mixtures were initially stored at m 30 psig in the canisters. These experiments were designed to test sample stability under anticipated field sampling conditions in which compounds such as H20, C02, OS, NO and NO1 would be present. 1403
KAREND. OLIVERet al
1404
EXPERIMENTAL Instrumentation
An automated cryogenic sampling and gas chromatographic (GC) system installed in a mobile laboratory was used for sample analysis. This system consisted of a Hewlett-Packard 588OALevel 4 GC equipped with an OV-1 (SO-mx 0.31-mm i.d. x 0.17~pm film thickness) fused silica high resolution capillary column (Hewlett-Packard, Avondale, PA), electron capture (ECD) and flame ionization (FID) detectors, and a modtied Nutech 32@01 cryogenic preconcentration unit (Nutech Corp, Durham, NC). A detailed discussion of the automated sampling and analysis system can be found elsewhere (McClenny et al., 1984). A Model MD-125-48(T) Perma Pure dryer (Perma Pure Products, Farmingdale, NJ) was used to remove water vapor from the gas stream prior to analyte preconcentration. To prevent excessive moisture build-up and any memory effects in the dryer, an automated clean-up procedure that involved periodically heating the dryer while purging with zero air was devised. Experiments were conducted to ensure that this analytical procedure did not alter sample integrity (Pleil and Oliver, 1985). Figure la shows the configuration for 6lling canisters from the sampling manifold. A stainless steel Model MB-151 Metal
-
A
Heated
Marxfold
+
40Umtn
I
Bellows pump (Metal Bellows Corp., Sharon, MA) was used to pressurize the canisters. The flow rate was regulated with a mass flow controller (Model 260, Tylan Corp., Carson, CA) set to deliver 500 ml min- I. All flow rates were verified with a soap-film bubble flowmeter. Figure lb depicts the configuration used during the analysis cycle, in which the sample was vented past the inlet to the system at a rate of 73 ml min _ * using a mass flow controller. This oversupplied the system inlet since the sample was collected at a rate of 34 ml min- I. A total sample volume of 476ml was collected into the cryogenic preconcentrator of the analytical system by sampling for 14 min. Canisters
All canisters used in this study were passivated using the SUMMA@ process. In the 7-day storage experiments with new canisters, four Model 0647 6-t stainless steel canisters (Demaray Scientific Instruments, Ltd., Pullman, WA) and four 3-t stainless steel canisters (Biospherics Research Corp., Hillsboro, OR) were tested. In a similar experiment with used canisters, five Model 0650 6-f stainless steel canisters (Demaray Scientific Instruments, Ltd.) which had been used extensively for 3 months were evaluated. The Model 0650 canisters had been used in a 3-month study involving 22 cities in the United States, so these canisters had been continually filled, shipped and cleaned during that time. A 3Oday storage experiment was also conducted using one Model 0650 and two Model 0647 6-L canisters which had been in use for a minimum of 3 months.
Standards L, 34 mUmin b
Through Perma Pure Dryer tO Analyrlr System
Two Scott Environmental Technology, Inc. pressurized cylinders containing a mixture of VOCs at nominal concentrations of 10 ppmv in nitrogen were used as working standards. These cylinders contained the following compounds: Cylinder 1 benzene toluene o-xylene Cylinder 2 -___ vinyl chloride vinylidene chloride 1,1,2-trichloro-1,2,2,tritluoroethane chloroform 1,Zdichloroethane methyl chloroform carbon tetrachloride hexachloro-l,3-butadiene
Heated
Marxfold
Fig. 1. (A) Configuration for filling stainless steel canisters. (B) Configuration for analysis of samples from stainless steel canisters.
trichioroethylene cis- 1,3-dichloropropene trans-1,3-dichloropropene 1,2-&bromoethane tetrachloroethylene chlorobenzene benzyl chloride.
For calibration runs, standards were diluted with zero air to nominal concentrations of 10 ppbv using mass flow controllers. Ail flows were audited daily, and precise dilution ratios, calibration concentrations, and instrument response factors were calculated. To obtain an indication of instrument stability, two calibration runs were performed each afternoon on consecutive workdays, and the results were averaged daily. Response factors from calibrations were treated statistically to determine precision. Cleaning procedure The cleaning procedure consisted of evacuating the cariisters while heating. Four canisters at a time were @aced in a Fisher Isotemp oven (Model 349, Fisher Scientific, Pittsburgh, PA), and were connected to a vacuum system with Teflon tubing. Canisters were heated to 100°C and were evacuated to less than 23~ Hg using a Wekh Duo-Seal vacuum pump (Model 1400, Sargent-Welch, Skokie, IL) and a liquid nitrogen trap. A Sargent-Welch thermocouple
Sample integrity of trace level volatile organic compounds in ambient air vacuum gauge was used for measuring vacuum This cleaning prccess usually required 5-6 h. Test procedures
Before beginning the sample integrity experiments, all new and used canisters were cleaned. Humidi6ed zero air was passed through the pump, mass flow controller, and tubing of the canister-filling apparatus and analyzed for comparison with chromatograms of instrument background to ensure that the filling apparatus would not contaminate samples. (Chromatograms of instrument background were obtained by passing zero grade air or nitrogen through a 5OOml impinger flask of water, through a mass flow controller, and into the analysis system.) The canisters were then pressurized to 15-20psig with humid&d zero gas. Samples from each canister were analyzed and compared with instrument background. AU canisters were cleaned again at least once before beginning the experiments. For the sample integrity experiments, spiked ambient air samples were obtained by pulling ambient air from a parking area through a glass ‘candy-cane’ inlet to the sampling manifold at a rate of4O/min-‘. The standard mixture of 15 VoCs in cylinder 2 was bled through tubing to the candycane inlet area at a rate of 12 mlmin-i during the canister filling period. This procedure was designed so that ambient air spiked with roughly 1 ppbv of each of the 15 compounds would be available in the sampling manifold. To test initial sample integrity, used 6-f canisters were pressurized individually with spiked ambient air from the sampling manifold. For these experiments, canisters were filled at a rate of 1.2 /mitt-’ for 14 min to exactly coincide with the 14min sample collection cycle of the analytical system. As soon as the analysis of the real-time sample was
completed, an air sample from thecanister was analyzed. This experiment was performed eight times. Four sets of sample integrity experiments were then performed at separate times using the same procedure. The foUowing sets of canisters were evaluated over a ‘I-day period in separate experiments: four new 6-/ canisters, four new 3-I canisters, and five used 6-I canisters. A fourth set of three used 6-f canisters was also evaluated over a 30day period. The instrument was calibrated before beginning each experiment. On Day 0 of each experiment, canisters were pressurized simultaneously to approximately 32 psig with the spiked ambient air mixture from the sampling manifold. Canisters were filled in the mornings during peak trathc activity. Control samples from the sampling manifold were analyzed continually during the filling cycles to provide an estimate of VDC concentrations placed in the canisters. Each analysis required 64 min, so control samples were taken approximately at hourly intervals. A graph of the 14min control sampling periods within the filling cycles for each canister set is shown in Fig. 2. Air samples from each canister in the ‘I-day test were analyzed on Days 0, 2,4 and 7 after filling the 6-1 canisters and on Days 0,4 and 7 after tilling the 3-1 canisters. Samples from canisters in the 30day test were analyzed on Days 1, 15 and 30 after filling. Calibrations were conducted on each day of the experiments.
RESULTS AND DRXXJSSION Canister
blank
tests
Before beginning the sample integrity study, each canister was cleaned and filled with humidified zero
4cIO
1
T-41 , Used 6.Liter C.WllSterS
N&V 3.Liter Canisters
I
N&V 6-Liter Camsterr
0700
0800
0900
1000
1405
1100
1200
1300
1400
Time of Day Canister Fllllng Period
•I
Direct Analysts Sampling Period
Fig. 2. Time periods for filling canisters and for simultaneous control analyses during the filling period.
1406
KAREND. OLIVERet al.
gas. FID and ECD chromatograms of humidified zero gas samples from new canisters revealed trace level contamination in the canisters, but none of the target compounds were present. One large, broad peak appeared early in FID chromatograms of new canister samples but was not evident in chromatograms of instrument background. This contaminant was removed by subsequent cleanings. Chromatograms of humid&d zero gas samples from used canisters were similar to the chromatograms of instrument background. Methylene chloride was identified as a contaminant in several used canisters but was removed by repeated cleaning. No target compounds were identified in humidified zero gas samples from used canisters except for negligible traces (c 0.08 ppbv) of benzene and toluene. These two compounds were also present at similar levels in instrument background samples because of slight contamination within the analytical system. Sample integrity
Comparisons of canister and real-time samples collected simultaneously showed no initial decrease in concentrations of sample compounds in the canisters.
The mean difference (n = 8) in concentrations of target compounds in the manifold and canister sampies ranged from - 5.17 y0 to + 6.24 %. These differences are attributed to the inherent scatter in the system, which is detailed below. Results for the four canister sets are presented in Tables 1,2,3 and 4. Although they were not spiked into the sample stream, benzene, toluene and o-xylene are included since these compounds were present in the ambient air. The mean concentration of each target compound in control runs is included in each table. However, these numbers do not necessarily equal the concentrations placed in canisters because control samples were taken only at intervals during the filling cycle as described previously. Canister data for vinyl chloride and truns-1,3_dichloropropene are not available in Table 4 since some low-level peaks were not integrated by the automated system software. Statistics on data were obtained by treating all canisters and all analyses in each set as equivalent. Tables l-4 list mean concentration, standard deviation and average daily drift in concentration for each compound in each canister set. With the exception of vinylidene chloride and chloroform stored in new
Table 1. Sample integrity results for four new 6-P canisters tested on Days 0, 2, 4 and I Control samples
:PTI (N = 16)
S.D.* (ppbv)
0.76 1.08 1.04 1.11 0.86 1.29 0.31 1.02 0.96 0.48 0.38 0.46 1.04 0.94 0.88 0.15s 1.07 1.10
0.76 2.35 1.17 2.47 0.92 1.38 0.36 0.98$ 1.01 0.50 0.37 0.62 1.10 0.95 0.92 0.09 1.19 1.21
0.04
1.01 1.04 1.21 1.23 1.06 1.12 0.93 O.%
1.05 1.00 1.23 1.23 1.08 1.09 0.94 1.00
M=$-“;J’ Compound FID vinyl chloride vinylidene chloride trichlorotrifluoroethane chloroform 1,2dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cia-1,3dichloropropene trans-1,3-dichloropropene toluene 1,Zdibromoethane tetrachloroethylene chlorobenzene ckxylene benxyl chloride hexachlorobutadiene ECD trichlorotritbmroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,Zdibromoethane tetrachloroethylene hexachlorobutadiene
Canister samples
0.85 0.05 0.92 0.03 0.05 0.01 0.07
Mean 7; change/dayt - 1.58 0.893 -0.256 0.769 -0.870 -0.652 -0.278 -0.102 -0.693 - 1.20 -2.16 0.322 - 1.45 -0.947 - 1.41 -2.22 1.68 -2.15 0.190 -0.300 -0.081 0.655 0.095 0.642 1.49 0.600
*Standard deviation, t mean y0 change/day = (linear regression slope+ mean concentration) x 100, $ N = 6, 8 N = 1.
Sample integrity of trace level volatile organic compounds in ambient air
1407
Table 2. Sample integrity results for four new 3-l canisters tested on Days 0,4 and 7 Control samples
M; Compound FID vinyl chloride vinylidene chloride trichlorotritluoroethane chloroform 1,2_dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-IJ-dichloropropene trans-1,Michloropropene toluene 1,2dibromoethane tetrachloroethylene chlorobenzene o-xylene benzyl chloride hexachlorobutadiene ECD trichlorottiuoroethane chloroform methyl chloroform
carbon tetrachloride trichloroethylene 1,2dibromoethane tetrachloroethylene hexachlorobutadiene
Canister samples
gy) (N = 12)
S.D. (ppbv)
Mean % change/day
0.71 0.98 0.98 1.05 0.88 1.27 0.31 0.99 0.97 0.48 0.37 0.47 1.07 1.21 0.87 0.12t 1.05 1.14
0.74 2.32 1.04 1.19 0.88 1.28 0.35 0.91’ 0.98 0.47 0.37 0.51 1.00 1.06 0.85 0.09 0.94 1.19
0.07 0.35 0.08 0.37 0.04 0.05 0.01 0.01 0.05 0.03 0.03 0.02 0.06 KG 0.03 0.10 0.15
-0.946 -2.20 -1.44 3.03 -1.02 -0.547 -0.286 -0.816 - 1.49 -1.35 -0.196 -1.40 -1.23 -1.76 1.11 -2.77 -277
1.01 1.04 1.21 1.23 1.06 1.09 1.21 1.01
1.02 1.00 1.20 1.24 1.05 1.05 1.16 1.02
0.02 0.02 0.01 0.04 0.02 0.06 0.08 0.05
0.294 -0.400 O.CNlO 0.645 0.190 -1.14 1.64 0.882
(p;bv) =
canisters, mean concentrations for control samples were typically within 0.05 ppbv of the mean concentrations for canister samples. Because these differences were of the same magnitude as the standard deviations of the canister data, the control data and the canister data were essentially identical. Daily drift values (given as a numerical indicator of temporal stability) were calculated by normalizing slopes of linear regression plots to obtain a percent-per-day change, i.e. slope divided by mean concentration. For the 7-day storage tests, the mean change was within f 3.2 % per day, and concentrations of most compounds changed within f 1% per day. With the exception of o-xylene, daily drifts for most compounds in the 30&y test were equivalent to those observed in 7-day tests. For each set .ofcanisters, the mean percentage change/day for compounds on one detector was consistently either slightly positive or negative, whereas the mean percentage change/day of compounds on the other detector was consistently the opposite. This is most likely due to a slight change in the FID/FCD split ratio rather than a canister storage problem. This phenomenon had been previously encountered during analyses that did not involve canisters. Typical concentrations of compounds in individual canisters are presented in Table 5. As examples, FID
and ECD data are tabulated for tetrachloroethylene stored in used 6-P canisters, and FID data are tabulated for chlorobenzene stored in new 3-P canisters. The data are representative of all experiments and show that concentrations of compounds were reproducible for individual canisters. A comparison of the relative standard deviation (R.&D.), i.e. standard deviation divided by the mean, for data from canister tests and calibrations is presented in Table 6. For the majority of target compounds in each canister set, the R.S.D. of concentrations was 10% or less when all results of canister sample analyses were averaged for each compound. A statistical evaluation of calibration response factors (ppbv/area counts) has shown the instrumental error to average approximately 3.6% R.S.D. for the ECD and 6.6 % R.S.D. for the FID over a 6-week period. The R.S.D.s of calibration response factors for individual compounds are presented in Table 6. In most cases the percent RS.D. for compounds in the canisters was similar to the percent R.S.D. for the calibrations. Approximately one-third of the compounds exhibited somewhat greater scatter in data during the 30&y tests than in the 7day tests. FID data for vinylidene chloride, chloroform, oxylene, benzyl chloride and hexachlorobutadiene were
KAREND. OLIVERet al.
1408
Table 3. Sample integrity results for five used W canisters tested on Days 0,2,4 and 7 Control samples
Compound FID vinyl chloride vinylidene chloride trichlorotritluoroethane chloroform 1,tdichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-1,3-dichloropropene trans-1,3-dichloropropene toluene 1,Zdibromoethane tetrachloroethylene chlorobenzene o-xylene betuyl chloride hexachlorobutadiene ECD trichlorotriIIuoroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,tdibromoethane tetrachloroethylene hexachlorobutadiene
Canister samples
Man (ppbv) (N = 5)
Mean (ppbv) (N = 19)
S.D. (ppbv)
Mean I’0 change/day
0.81* 1.11 1.13 1.19 0.90 1.41 0.57 0.946 1.01 0.49 0.3811 0.99 1.23 1.09 0.97 0.1411 0.75 0.96
0.71t 0.98$ 1.13 1.06 0.88 1.37 0.55 0.88$ 0.97 0.46 0.3217 0.89 1.09 1.00 0.87 0.12 0.64 0.91
0.07 0.07 0.09 0.38 0.04 0.05 0.03 0.06 0.06 0.02 0.02 0.05 0.10 0.04 0.06 0.02 0.08 0.10
0.845 0.000 0.708 2.26 0.227 0.073 -0.364 -1.59 0.309 0.435 1.88 1.Ol 0.000 0.100 0.115 0.000 -3.13 0.769
0.99 1.02 1.20 1.52 1.01 0.84 0.97 0.96
0.99 0.99 1.19 1.55 1.00 0.91 0.98 1.01
0.03 0.04 0.04 0.04 0.03 0.05 0.04 0.10
-0.909 -0.909 - 0.924 -0.129 -0.300 -0.879 -0.102 - 0.990
__-.-
*N=2.tN=14,$N=18,§N=4,IjN=3,~N=5.
generally more scattered than data for the other compounds. Ortho-xylene was present at concentrations lower than those for which the experiment was designed, so a large amount of scatter in the data was not unexpected. Historically, excess scatter in data on benzyl chloride and hexachlorobutadiene has been due more to the instrumentation than to a problem with the canisters. FID data on vinylidene chloride and chloroform were not consistent during the experiments. Low-level contaminants in new canisters co-eluted with vinylidene chloride and chloroform on the FID. This resulted in the measureme.nt of slightly higher concentrations of both compounds in samples from new canisters than those measured in control samples. However, in the experiments with used canisters, FID data on vinyl&me chloride did not vary as much as in the experiments with new canisters. Possibly, some tenacious contaminants were removed by repeated cleaning of the canisters. Since chloroform responds well on the ECD, data from the ECD can be used instead of the less reliable FID data when contamination is a problem. However, vinylidene chloride does
not respond on the ECD, so there is no simple way to correct for interference in this case.
CONCLUSIONS
Although initial blank tests on new SUMMA@ polished stainless steel canisters revealed some tracelevel contamination, subsequent cleaning removed many of the contaminants. Used canisters that had been subjected to many sampling/cleaning cycles did not exhibit such contamination. After new and used canisters were cleaned, samples were collected with no measureable cross-contamination from previous samples. One set ofexperiments was designed so that canister samples and real-time samples of ambient air spiked with 15 VOCs were collected simultaneously. Comparison of real-time and canister samples revealed no initial decreases in concentrations of target compounds stored in canisters. Ambient air spiked with low parts per billion by volume levels of 15 VOCs was also stored in new and
Sample integrity of trace level volatile organic compounds in ambient ai Table 4. Sample integrity results for three used 6-I canisters tested on Days 1, 15 and 30 Control samples
Canister samples
“‘(“N _@s5b” Compound FID vinyl chloride vinylidene chloride trichlorotrifluoroethane chloroform 1,2-dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-1,3-dichloropropne tram-1,fdichloropropcne toluene 1,2dibromoethane tetrachloroethylene chlorobenzene o-xylene benzyl chloride hexachlorobutadiene ECD trichlorotrilluoroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,2dibromocthane tetrachloroethylene hexachlorobutadiene
Mean (Ppbv) (iv = 9)
-
0.81* 1.11 1.13 1.19 0.90 1.41 0.57 0.94t 1.01 0.49 0.38$ 0.99 1.23 1.09 0.97 0.14$ 0.75 0.96
0.07 0.09 0.45 0.06 0.09 0.03 0.09 0.05 0.04 0.12 0.07 0.03 0.04 0.15
1.09 1.13 1.32 0.93 1.41 0.61 0.87 1.02 0.50 1.01 1.23 ;z 0.23 0.84 1.05
0.99 1.02 1.20 1.52 1.01 0.84 0.97 0.96
Mean% change/day
&k)
0.97 0.97 1.19 1.59 1.00 0.87 1.01 0.98
:;
-0-14 -0.44 1.14 -0.40 -0.37 0.18 -0.49 -0.38 0.12 0.67 0.10 -0.10 0.24 4.40 0.36 1.17
0.02 0.03 0.02 0.07 0.02 0.04 0.05 0.10
-0.18 -0.21 -0.14 -0.30 -0.07 0.11 0.34 0.53
Table 5. Typical examples of data for canister samples* a FID-tetrachIoroethylenet CAN 106 CAN 04 DaY 0
1.01 1.06 2 4 oa94t zzi 7 1:04 0.97 b. ECD-tetrachloroethylenet DaY CAN 106 CAN 04 20 1.00 0.97 4 0.99 7 0.98 c. FID-chlorobenzeneQ DaY CAN 5 0 4 7
0.89 0.80 0.81
CAN 11
CAN 15
CAN 190
1.00 1.00 0.95 1.00
1.01 1.00 ;;r
1.04 1.01 0.96 1.05
CAN 11
CAN 15
CAN 190
0.99 1.00 1.00 1.01
0.98 1.01 0.95 1.00
kg 2
CAN 6
CAN 7
CAN8
0.92 0.80 0.82
0.95 0.79 0.84
0.w 0.80 0.87
;: 685 0.99
* Values expressed as parts per billion by volum, t Stored in used 6-f canisters, $ Data for this run not available due to column blockage, #Stored in new 3-f canisters.
1409
1410
KAREN
OLIVER et al.
D.
Table 6. Comparison of scatter in calibration data and canister data (% R.S.D.*)
Compound FID vinyl chloride vinvlidene chloride tr~hlorotrifluoroethane chloroform 1,2_dichloroethane methyl chloroform benzene carbon tetrachloride trichloroethylene cis-1,3-dichloropropene tram-1,3-dichloropropene toluene 1,2dibromoethane tetrachloroethylene chlorobenxene o-xylene benxyl chloride hexachlorobutadiene ECD trichlorotrifluoroethane chloroform methyl chloroform carbon tetrachloride trichloroethylene 1,2dibromoethane tetrachloroethylene hexachlorobutadiene
Calibration Data (N = 33)t
Canister data, 7-day test _________ _ New 6P New 3/ Used 6P (N = 16) (N = 12) (N = 19)
6.37 7.24 6.96 6.21 5.91 5.38 5.30 4.84 5.44 6.86 6.81 5.11 8.08 6.38 8.64 6.88 15.8 17.5
4.00 8.11 1.61 5.45 3.16 4.35 22.2 12.6 9.09
3.94 3.08 3.00 1.38 2.31 6.21 3.54 4.95
1.90 2.00 1.63 2.44 1.85 4.59 5.32 5.00
5.26 36.2 4.27 37.3 3.26 3.62 2.78
9.46 15.1 7.69 31.1 4.55 3.91 2.86 l.lOT[ 5.10 6.38 8.11 3.92 6.00 5.66 1.06
33.3 10.6 12.6 1.96 2.00 0.83 3.23 1.90 5.71 6.90 4.90
Canister data, 3@day test
~~~ ._..___
Used 6 ! (N = 9) .--
9.86$ 7.14$ 7.96 35.9 4.55 3.65 5.45 6.824 6.19 2.17 6.25.. 5.62 9.17 4.00 6.90 16.7 12.5 11.0
5.98 7.59 33.8 6.53 6.06 5.52 10.8 5.29 7.63 .11.8 5.60 3.10 4.55 65.4 11.2 18.7
3.03 4.04 3.36 2.58 3.00 5.49 4.08 9.90
2.27 3.56 1.76 4.12 1.58 4.65 4.88 10.1
*Relative standard deviation, t Based on 33 pairs of daily calibrations, $N = 14, $N = 18, (1N = 6, 7N = 2, ‘*N = IS.
used canisters for ‘I-day and 30-day periods. Because a number of canisters were filled simultaneously for each experiment, the fill cycle was longer, and several realtime control samples were analyzed, a simultaneous real-time analysis of several hours was not possible. Of the compounds examined, none exhibited any signiftcant loss or gain in concentration during the test periods. Low-level contaminant peaks from new canisters sometimes co-eluted with chloroform and vinylidene chloride on the FID, creating a positive bias in the measurement of concentrations of these two compounds. This was not observed in used canisters
that had been cleaned repeatedly. However, for most target compounds the sample integrity results for all canister sets were similar to system precision. The good sample integrity observed in these tests suggests the routine use of canisters as a viable alternative to other sampling techniques, at least for the compounds tested. Acknowledgement-The authors thank Dr B. A. Rasmussen for helpful advice and assistance during the study. Members of the Methods Standardization Branch (MBB), Environmental Monitoring Systems Lsbonrtory (EMSL), Environmental Protection Agency (EPA) and Gas, Kinetics
and Photochemistry Branch (GKPB), Atmospheric Sciences Research Laboratory (ASRL), EPA also shared valuable experience. Disclaimer-Although the research described in this article has been funded wholly or in part by the U.S. Environmental Protection Agency through Contract 68-02-4035 to Northrop Services, Inc.-Environmental Sciences, it has not been subjected to agency review and therefore does not necessarily reflect the views of the agency. No official endorsement should be inferred. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.
REFERENCES
Cox R. D. (1983) Sample collection and analytical techniques
for volatile organics in air. Presented at APCA Specialty Conference, Chicago, IL, 22-24 March. Harsch D. E (1980) Evaluation of a versatile gas sampling container design. &rospheric Environment 1&1105-i 10% Holdren M.. Rust S., Smith R. and Koetx J. (19g4) Evaluation of cryogenic trapping as a means for coll&ing organic compounds in ambient air. Final report on EPA Contract 68-02-3487 to BatteBe Columbus Laboratories, Columbus, OH. McCknny W. A., Pleil J. D., Ho&en M. W. and Smith R. N. (1984) Automated cryogenic preconcentration and gas
Sample integrity of trace level volatile organic compounds in ambient air chromatographic determination of volatile organic compounds in air. Analyt. Chem. 56,2947-2951. Pelhzxari E. D., Gutknecht W. F., Cooper S. and Hardison D. (1984) Evaluation of sampling methods for gaseous atmospheric samples. Final report on EPA Contract No. 68-022991 to Research Triangle Institute, Research Triangle Park, NC. Pleil J. D. and Oliver K. D. (1985) Evaluation of various configurations of Nalion dryers: water removal from air samples prior to gas chromatographic analysis. Technical Note 85-03. Northrop Services, Inc.-Environmental Sciences, Research Triangle Park, NC.
1411
Rasmussen R. A. and Khalil M. A. K. (1980) Atmospheric halocarbons: measurements and analyses of selected trace gases. Proc. NATO ASI on Atmospheric Ozone, pp. 209-231. Rasmussen R. A. and Lovelock J. E. (1983) The atmospheric lifetime experiment-2. Calibrations. J. geophys. Res. 88, 83698378. Westberg H., Lonneman W. and Holdren M. (1984) Analysis of individual hydrocarbon species in ambient atmospheres: techniques and data validity. In Identijkation and Analysis of Organic Pollutants in Air (edited by Keith L. H.), pp. 323327. Butterworth, Woburn, MA.