Development and Validation of a Canister Method for Measuring

C. HERNDON WILLIAMS. URS, P.O. Box 201088, Austin, Texas 78720-1088. L. WADE BONTEMPO. Air Toxics Ltd. Laboratory, 180 Blue Ravine Road, Suite B,...
0 downloads 0 Views 91KB Size
Environ. Sci. Technol. 2004, 38, 4200-4205

Development and Validation of a Canister Method for Measuring Ethylene Oxide in Ambient Air BART M. EKLUND* AND C. HERNDON WILLIAMS URS, P.O. Box 201088, Austin, Texas 78720-1088 L. WADE BONTEMPO Air Toxics Ltd. Laboratory, 180 Blue Ravine Road, Suite B, Folsom, California 95630 MOLLY ISBELL Signature Science LLC, 8328 N. Mopac Boulevard, Austin, Texas 78759 KARL R. LOOS Shell Global Solutions (US) Inc., P.O. Box 4327, Houston, Texas 77210

A sampling and analytical method for measuring ethylene oxide (EO) in ambient air was developed and evaluated. The method is based on the use of evacuated canisters and gas chromatography-mass spectrometry (GC-MS). The objectives of this work were to characterize the performance of the method with respect to the following: (1) stability/ recovery of ethylene oxide in a canister over a 15-day holding time; (2) detection capability; and (3) measurement of EO in an ambient air matrix. Both electropolished and silicalined stainless steel canisters were evaluated in this study. The method evaluation involved both laboratory and field tests. The recovery of the EO was evaluated both on an absolute basis and relative to a spiked internal standard of toluene. EO spiked at levels of 2 ppbv and 20 ppbv was found to be stable for holding times of up to 15 days at 25 °C in both a humidified nitrogen matrix and in ambient air. The detection limit of the method was found to be 0.25 ppbv using EPA’s traditional approach of seven replicate analyses of a low-level standard and 0.20 ppbv using a probability-based approach. EO recoveries in the laboratory stability study generally were 100 ( 25%, and did not vary by canister type, nor did the EO recoveries decrease with holding time. Field studies demonstrated that the method is capable of detecting EO (as well as benzene and toluene) in an ambient air matrix.

Introduction Ethylene oxide is a reactive gas that is produced in large quantities for industrial purposes. It is regulated in both the U.S. and Canada as a human health hazard. Current regulatory programs likely will include assessments of potential exposures of persons living or working near facilities that produce or use EO. A major focus of exposure assessments could be outdoor air exposures at or near the facility * Corresponding author phone: (512) 419-5436; fax: (512) 4548807; e-mail: [email protected]. 4200

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

fenceline. Prior to this study, however, there were no sampling and analytical methods validated for determining ambient air EO concentrations in the low ppbv range. The USEPA’s Compendium of Toxic Organic Methods does not include any method for sampling and analyzing ethylene oxide in ambient air.1 NIOSH Method 1614 does address EO, but it has a stated working range for EO of 50 to 4600 ppbv and therefore is best suited for measuring EO in the occupational exposure (workplace) range.2 NIOSH Method 3702 has a reported working range of 0.001 to 1000 ppm, but only for “relatively noncomplex atmospheres.”3 Prior stability studies have shown that EO was relatively stable in Summa canisters. Kelly and Holdren reported that tests of nine canisters containing an 11 ppbv EO concentration exhibited a 26% change in concentration over 33 days.4 The objective of this study was to evaluate an air monitoring method for EO based on the use of evacuated stainless steel canisters and gas chromatography-mass spectrometry (GC-MS). The method was intended to be capable of detecting EO in ambient air at concentrations in the low (or sub) ppb range, while also providing measurements for other selected volatile organic compounds (VOCs)(e.g., benzene and toluene). The method was designed to be similar to EPA Method TO-15, which was developed as a whole air canister sampling technique for VOCs.5 The analysis of ethylene oxide, however, required analytical conditions different from those of Method TO-15 for VOCs to avoid interference to EO by acetaldehyde. The tasks of the study were to characterize the performance of the method with respect to the following: (1) stability/recovery of ethylene oxide in a canister over a 15day holding time; (2) detection capability (sensitivity); and (3) measurement of EO in an ambient air matrix. Both electropolished (Summa) and silica-lined stainless steel canisters were evaluated in this study. Summa stainless steel canisters have been used for over 20 years in measuring VOCs in ambient air by EPA Method TO-15 and are widely available. The newer silica-lined stainless steel canisters have been used for reactive/unstable air contaminants (e.g., sulfurcontaining VOCs) and could provide an alternative if EO were found to have a short lifetime in the Summa canisters.

Experimental Section Design. The technical approach followed in this study consisted of three sequential phases or tasks. The study design is summarized in Table 1. All of the canister cleaning, blanking, spiking, and analysis was done by Air Toxics Ltd. (ATL) laboratory in Folsom, California. In the first task, a total of 96 analyses were performed on 18 spiked canisters to evaluate the stability of EO in both electropolished (Summa) and silica-lined stainless steel canisters for a hold time of up to 15 days, in a humidified nitrogen matrix. Two ages of Summa canisters were used: new (1 year in service). All silica-lined canisters used in this task, however, were new. Canisters were spiked at either 2 or 20 ppbv with both EO and toluene, and analyzed at five time intervals. The recovery of the EO was evaluated both on an absolute basis and relative to the toluene internal standard. The second task was to evaluate the detection capability of the method. The detection capability study was designed to determine the method sensitivity by two approaches. One approach was based on the traditional EPA approach for computing method detection limits (MDLs), i.e., seven replicate analyses of a low-level standard.6 The other approach used analyses at five different EO concentrations and 10.1021/es049861o CCC: $27.50

 2004 American Chemical Society Published on Web 06/16/2004

TABLE 1. Overview of EO Spiking and Sampling Conditions task EO stability method sensitivity initial field test a

canister type

na

matrix

RH %

Summa silica Summa silica Summa silica

12 6 40 40 2 2

humid N2 humid N2 humid N2 humid N2 ambient Air ambient Air

50 50 25-50 25-50 49 49

range of parameter temp (°C) O3 (ppbv) 22 22 22 22 10.7 10.7

0 0 0 0 31 31

EO (ppbv)

hold/analysis time(days)

0, 2, 20 0, 2, 20 0.1-10 0.1-10 2.0 2.0

0, 1, 2, 7, 15 0, 1, 2, 7, 15 1 1 1, 7, 15 1, 7, 15

n ) number of canisters.

a statistical logistic regression analysis to determine the probability of EO detection as a function of concentration.7 Tests were performed at two levels of relative humidity (25 and 50%) in nitrogen. In total, 80 canisters were analyzed once each for the detection capability task. The results of the preceding stability study showed that there was no detectable difference in EO stability between new and old Summa canisters, so the age was not specified in the detection capability study and the Summa canister age distribution was mixed. The two tasks described above were performed in the laboratory using a humidified nitrogen matrix. The third task involved a first demonstration of the applicability of the method in the field. A set of four spiked samples was collected at 12:30 p.m. on January 24, 2002 at an existing Sacramento Air Quality Management District (AQMD) monitoring site located at 50 Natoma Street in Folsom, CA. This location was expected to have negligible ambient concentrations of EO and served to provide ambient air of known ozone content. The canisters were spiked with 2 ppbv each of ethylene oxide and toluene immediately prior to filling with ambient air. The samples were collected as grab samples and were filled in less than two minutes at a height of five feet above ground level. No sampling canes or flow regulators were used. A total of 12 analyses were performed on these four canisters. Canister Cleaning, Blanking, and Spiking. All of the canisters used in this study were owned by ATL and were subjected to ATL’s normal procedure for cleaning and blanking for EPA Method TO-15 for VOCs. For this project, ATL used a complete certification procedure for the blanking, in which 100% of the cleaned canisters were pressurized with dry, ultrahigh purity (UHP) nitrogen and analyzed by GCMS for VOCs. The acceptance criterion used was 0.2 ppbv for VOC species on the TO-15 analyte list. Canisters were spiked with EO and toluene in every phase of this project. The high-pressure, high-concentration sources of these two chemicals were certified cylinder standards prepared in nitrogen: EO at 5 ppm and toluene at 20 ppm. An intermediate concentration standard mixture (with EO ) toluene) was prepared in an evacuated Summa canister. A gastight syringe was used to transfer a gas aliquot (at one atm) from the cylinder standards into a pre-humidified intermediate 6-L canister (under vacuum) through a septum. The intermediate canister was pressurized with UHP nitrogen, and an aliquot was taken for verification analysis of the EO and toluene concentrations in the Agilent GC-MS analytical system. Once the expected concentration was verified, an aliquot was transferred under vacuum to the canister to be spiked. For the stability and sensitivity studies, the spiked canister was pressurized to 5 PSIG with humidified UHP nitrogen. For a field or blank spike, the canister (spiked with EO, toluene, and water) was sent into the field with its residual vacuum of less than -28 in Hg, to be filled with either ambient air or UHP air, respectively. Analytical Methods. Table 2 gives a list of ATL’s analytical equipment and the GC and MS operating conditions used for the EO and toluene measurements. The Agilent GC-MS

TABLE 2. Analytical Equipment and GC-MS Operating Conditions Analytical Equipment description manufacturer and model gas chromatograph mass spectrometer capillary column (60 m × 0.32 mm × 1.8 um) sample inlet 6-L stainless steel canister

Agilent Model A6890 Agilent Model A5973 Restek Model Rtx-624

Air Toxics, Ltd. Scientific Instrumentation Specialists, Summa 6-L silica-lined canister Entech Silonite 8-hour flow controller Air Toxics, Ltd sampling canes, stainless steel Air Toxics, Ltd ethylene oxide standard (5 ppm) Scott-Marin ethylene oxide permeation tube VICI Metronics (0.22 µg/min @35 °C) permeation dilution device Air Toxics, Ltd toluene standard (20 ppm) Scott-Marin GC/MS (can certification) Varian Saturn 2000 purge station (can cleaning) Air Toxics, Ltd evacuation station (can cleaning) Air Toxics, Ltd GC-MS Operating Conditions description carrier gas injector temperature injector mode split ratio initial inlet pressure oven temperature (initial) initial hold time ramp rate, initial oven temperature hold #2 hold time #2 ramp rate, final oven temperature final hold time final MS transfer line heater MS autotune standard MS mass range MS acquisition mode solvent delay electron multiplier voltage setting threshold setting MS quadrupole temperature MS source temperature

setting

helium 100 °C split 5:1 7 psig -20 °C 4 min 10 °C/min 40 °C 1 min 40 °C/min 230 °C 5 min 200 °C BFB 35-350 amu scan mode at 0.61 scans/sec 6 min. 1447 150 150 °C 230 °C

analytical system was calibrated for EO and toluene with a 5-point calibration curve ranging from 0.5 to 50 ppbv. This curve was generated using a spiked canister standard containing 50 ppbv of EO, and an equal concentration of toluene. A 500-mL aliquot was injected in replicate to yield the calibration points at 50 ppbv. Smaller aliquots (down to 5 mL) were introduced to generate the other concentration values on the calibration curve. The calibration points were used to calculate an average response factor which was then VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4201

FIGURE 1. Total ion current chromatogram for a typical ambient air sample. used to calculate sample concentrations. A typical air sample chromatogram is shown in Figure 1. Two single-point calibration verifications were done on a daily basis. A continuing calibration verification (CCV) standard in a canister at 20 ppbv was analyzed at least once daily. In addition, as required by EPA’s National Environmental Laboratory Accreditation Program (NELAP), a secondsource calibration verification was done daily.8 The secondsource for EO was a calibrated EO permeation tube, the diluted output of which was introduced through the GCMS inlet. The single-point verifications had to agree with the multi-point calibration within (25%, or a new multi-point calibration curve would be generated. The GC-MS procedures and conditions used by ATL for the analysis of EO and toluene generally paralleled those recommended in EPA Method TO-15 for VOCs. The major difference between the method used in this project for EO and ATL’s version of TO-15 for VOCs was in the GC conditions (i.e., column type and temperature programming). The GC conditions in this project were selected to clearly resolve EO from acetaldehyde. This was important because acetaldehyde is often found in ambient air at low ppbv levels and its mass spectrum is practically identical to that of EO.

Results Stability Study. EO stability and recovery were evaluated at two spike levels (2 and 20 ppb in humidified N2) in three canister types over a 15-day period. To support this evaluation, graphical data displays were constructed and formal statistical evaluations were performed. Analysis of variance (ANOVA) was performed to characterize the rate at which concentrations decreased over time, and to assess whether recoveries as a function of hold time differed systematically by canister type or spike level. Average absolute percent recoveries were computed for each canister type, spike level, and canister type and spike level combination. These recoveries are shown in Table 3. This table shows that there were small differences in the average recoveries among canister types, and a relatively large difference in average recovery by spike level. Surprisingly, the lower spike level (2 ppbv) resulted in the higher absolute recovery, but this is thought to be due to experi4202

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

TABLE 3. Average Absolute Percent Recoveries for EO in the Stability Study canister group new Summa old Summa silica-lined 2 ppb spikesb 20 ppb spikesb

average percent recovery (%)a standard error (%) 102.5 102.8 104.5 112.0 96.3

1.6 1.6 1.6 1.4 1.2

a The estimated average percent recoveries and the standard errors shown in this table are based on the ANOVA model fit to the data.b The matrix was 50% relative humidity nitrogen for all canisters.

mental error; i.e., the difficulty in preparing the lower level EO spike. Table 3 shows the differences in recovery by spike level, which were consistent across all three types of canisters. The plots in Figure 2 of EO and toluene absolute recovery versus hold time indicate that, in general, concentrations did not decrease significantly over a 15-day period for any of the canister types or spike levels. Absolute EO recoveries differed significantly by concentration and there was more scatter at the 2 ppbv level, but all recoveries were relatively high (greater than 94% in all cases). Detection Capability Study. The objective of the detection capability study was to characterize the sensitivity of the EO method by two different approaches. Both of these sensitivity estimates were based on a 500-mL aliquot of a 6-L sample pressurized to +5 psi. The first approach for characterizing the detectability of the EO method was to compute the MDL following the EPA MDL methodology. This approach is based on analytical results for seven samples spiked at an initial estimate of the detection limit. The variability (standard deviation) of these seven results is computed and then multiplied by the 99th percentile from the Student’s t distribution with n - 1 (i.e., six) degrees of freedom to obtain the MDL. In this study, seven samples were spiked at 1.0 ppbv and the EPA MDL calculation resulted in an estimate of 0.25 ppbv. This can be interpreted to mean that detected EO concentrations as low as 0.25 ppbv can be definitively distinguished from zero. The results are shown in Table 4. As a complement to the EPA approach, a probabilitybased approach was used to evaluate the likelihood of

FIGURE 2. Stability study results as percent recoveries over time: open symbols, sample spiked at 2 ppbv; closed symbols, sample spiked at 20 ppbv.

TABLE 4. Method Detection Limits for EO canister type silica-lined Summa

TABLE 5. Detection Probability for EO

spike mean probability relative concn. n measured EO MDLa of humidity (ppbv) (ppbv) (ppbv) detectionb 25% 50% 25% 50%

1 1 1 1

7 7 7 7

1.10 1.25 0.93 0.96

0.22 0.29 0.43 0.28

a Based on the standard deviation of replicate samples. number of detects divided by number of analyses.

100% 100% 100% 100% b

Based on

detection as a function of spiked concentration. These results are shown in Table 5. The study showed that EO concentrations as low as 0.20 ppbv can be consistently detected (at 80% probability), in either silica-lined or Summa canisters, at either a 25 or 50% humidity level in nitrogen. Initial Field Demonstration. Following the sequence of laboratory tests, an initial field demonstration was performed

canister type

relative humidity

all types both levels combined combined

mean spike probability measured MDLa concn of (ppbv) n EO (ppbv) (ppbv) detectionb 0 0.1 0.5 1 2 10

4 12 12 28 12 12

ND 0.08 0.61 1.06 1.98 8.95

NA 0.28 0.26 0.25 0.54 2.03

a Based on the standard deviation of replicate samples. number of detects divided by number of analyses.

0% 42% 100% 100% 100% 100% b

Based on

to assess the effects of ambient air on the ability to collect, recover, and quantitate EO. This field demonstration involved spiking four canisters (two Summa and two silica-lined) with EO and toluene at 2 ppbv each and then filling the canisters with urban ambient air at a state ozone monitoring site near VOL. 38, NO. 15, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4203

TABLE 6. Average Percent Recoveries for EO in the Initial Field Demonstration canister type Summa silicab

average percent recovery (%)a day 1or 2 day 7 day 15 87.5 88.5

78.6 97.0

77.2 108

a Two canisters of each type were spiked with EO and toluene, filled with ambient air, and analyzed at several time intervals. The RPDs varied from 6% to 20%. b The value shown is for only one sample from the pair. The second silica-lined canister was nondetect for EO, but the toluene spike was recovered quantitatively (about 100%).

the ATL laboratory in Folsom, California. This spiking and ambient air collection was performed on January 24, 2002. The measured environmental conditions included temperature (51.3 °F), relative humidity (49%), and ambient ozone (31 ppbv). The results of this field test are shown in Table 6. For one of the silica-lined canisters, nondetects for spiked EO were observed at each time interval. For the three other canisters in which EO was detected, the recoveries were slightly lower than had been seen in the canisters in the laboratory tests. The stability of the recoveries with hold time, however, was very consistent with what was seen in the laboratory stability study. That is, for both canister types tested, EO recoveries did not decrease significantly over a 15-day period, even in the presence of an ambient air matrix. Only absolute EO recoveries are presented here, but the toluene-normalized EO recoveries showed less variability, especially at the 2 ppbv level. No specific cause was identified for the EO nondetect results in one of the silica-lined canisters. In the subsequent field sampling program at an EO production facility conducted during March 2002 (see below), one false negative value was found (in a Summa canister) from among more than 20 canister samples. These data suggest that an ambient air matrix may result in false negative measurements at some infrequent rate. However, neither the type of canister nor the specific ambient air conditions appear to be a predictor for when false negatives may occur. Data Quality Assessments. In addition to evaluating the results of each of the studies separately, a statistical analysis of the overall method capability for EO was performed using data from all of the studies combined. This involved first evaluating measurement accuracy and precision as a function of spike level, canister type, presence or absence of ambient air, and time between collection (or spiking) and analysis. In addition, the presence of EO in blank, or unspiked, canisters was assessed. The results of this data evaluation indicate that the EO method is a valid method. The analytical variability is low (less than 10%) for EO levels at or above 1 ppb. There is some indication that the absolute precision may be worse for levels of EO less than 1 ppbv. In addition, although there appears to be some potential for false negative results and slightly reduced recoveries in the presence of ambient air background, recoveries tended to be 100 ( 25%. The recoveries did not vary significantly by canister type and did not appear to decrease over time (up to 15 days). Finally, the blank results indicated that there are no large contamination or false positive problems with the method. Field Demonstration Studies. Two field demonstration studies were conducted at the fenceline of an EO production facility in the United States. The full results of these studies are outside the scope of this paper, but the results are briefly summarized here. A total of 7 spiked and 15 unspiked samples were collected over 8-hour sampling periods during 3 days of field sampling 4204

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 15, 2004

in March 2002. The study demonstrated that EO could be consistently detected at low part-per-billion concentrations in an ambient air matrix. Subsequently, a total of 15 spiked and 21 unspiked samples were collected over 1-hour or 8-hour sampling periods during 3 days of field sampling in July 2003 at the same facility. This study demonstrated that EO could be consistently detected at low part-per-billion concentrations in an ambient air matrix at air temperatures of 24-33 °C and relative humidities of 54-89%. A subsequent laboratory study was performed to evaluate the effect of water vapor and elevated temperature on spiked blank samples. Twenty canisters were spiked with EO and toluene at 5 ppbv and held under various conditions for up to 15 days. The experiments showed that spiked samples must be humidified for the EO to be stable. Addition of 100 µL of water (to the spiked canister still under vacuum) resulted in good EO spike recoveries. The recoveries of the toluene were generally quantitative (85-115%) and were not affected by the presence of water. Summary of Findings. The results of the various tasks to develop and validate a method for ambient ethylene oxide showed the following: (1) the sampling and analytical method employed in this study was successfully used to measure ambient concentrations of ethylene oxide at concentrations well below 1 ppbv (2) canisters containing ethylene oxide in an ambient air matrix are stable for periods up to 15 days (3) the analytical sensitivity of the method for EO and other volatile organic compounds (i.e., benzene and toluene) exceeded expectations (4) the accuracy and reproducibility of the method were very good in the range 0.2-20 ppbv (5) the positive identification of EO in an ambient air matrix by GC-MS required at least three characteristic ions, e.g. 44, 43, and 42. The sub-ppbv sensitivity improved when manual processing of the GC-MS data was performed (6) blank contamination was not a problem (7) false negatives were not a significant problem, but there were infrequent occurrences when low level spikes of EO were not recovered from an ambient air matrix (i.e.,