Environ. Sci. Technol. 1994, 28, 238-245
Continuous Hourly Analysis of C*-C,o Non-Methane Hydrocarbon Compounds in Urban Air by GC-FID Charles 1. Farmer, Peter J. Mllne,' Daniel D. Rlemer, and Rod G. Zlka
Division of Marine and Atmospheric Chemistry, RSMAS, University of Miami, 4600 Rickenbacker Causeway, Miami, Florida 33149 ~~
Control strategies for the regulation of ozone levels in urban atmospheres are presently under active revision [National Research Council (I)]. Despite some 20 years of federally mandated regulation, the US. EPA estimates that some 67 million people in the United States are now routinely exposed to ozone levels exceeding that set under the Clean Air Act. Volatile organic compounds (VOCs), of which non-methane hydrocarbons are a significant fraction, are important precursor compounds that, in conjunction with nitric oxide and nitrogen dioxide, photochemically catalyze the formation of tropospheric ozone (2,3).It is likely that existing inventories of many reactive VOCs in urban atmospheres are inadequately known. This situation hampers the efforts of atmospheric modelers trying to understand and predict urban air quality. It further places constraints on control strategies for ozone attainment. Several existing analytical difficulties must be overcome before the knowledge base of urban VOC inventories can be expanded. The first of these is that there are so many individual compounds,literally hundreds in complex urban atmospheres. This diverse range of compounds arises from widely different sources. A second consideration is that many of the most important ozone precursor VOCs are reactive, so that their presence in air maybe ephemeral, or otherwiserestricted in space and time. The contribution
natural vegetation makes to urban VOC loads, especially in many heavily wooded urban areas, has recently been subject to renewed interest (4). Further considerations include the low concentrations (ppb to ppt) of these compounds and also the large amount of data that continuous speciated monitoring, desirable for detailed photochemical models, of even modest target lists of VOCs entails. As part of the Southern Oxidant Study, a multiinstitutional, multi-year study of the air quality of several sylvan cities in the southeastern U.S., we sought to develop and implement a reliable and verifiable method for the near-continuous profiling of a target list of 55 individual NMHCs for field use in the exploratory, intensive, and routine field-monitoring phases of the study. We report here aspects of the development of, and experience with, this method as employed during the intensive study held during 6 weeks over the summer of 1992 in Atlanta, as part of the Southern Oxidant Research Program on Ozone Non-Attainment (SORP-ONA). Analytical Approach. Some necessary steps in the NMHC analyses of ambient air by capillary GC, which is the most widely used method, usually include variants of the following: (a) sample introduction or collection, (b) enrichment or concentration step, (c) cryofocusing step, and (d) quantitation of detector response (FID,MSD, etc.) against known standards. A number of strategies have been used to achieve the first two steps. The first, and the most widely accepted of these, is the sampling of ambient air into stainless steel containers that have been carefully pretreated, cleaned, and evacuated (5-7). Electropolishing of all internal surfaces and inert gas welding of the canister connections ensure the optimum performance (i.e., chemical inertness) of these vessels. Providing that the uncontaminated collection of air samples by means of an all-metal bellows or other inert pump into the canister is achieved, canister sampling allows the collection and stable storage (8) of a large number of atmospheric trace components at ppb levels and below. Since the sample concentrator employed in this study (see below) had its own vacuum source, it was not strictly necessary to fill canisters to above ambient pressures. The collected air samples are transportable to a central laboratory site for analysis, facilitating the sampling of broad geographical study areas. A disadvantage to this scheme is the relative expense of the canisters themselves, together with the necessarily slow turn-around time associated with the tracking, analysis, and cleaning cycles. Other containers such as glass vessels or Teflon and Tedlar sample bags are still employed, although it is known that they may cause contamination effects measurable at the ppb range needed for such analyses. These vessels further suffer from the disadvantage that they m&t be shielded from ambient light in the analysis of photochemically sensitive components.
238 Envlron. Sci. Technol., Vol. 28, NO. 2, 1994
0013-938X/94/0928-0238$04.50/0
An integrated system that quantitatively measures speciated C2-Clo hydrocarbons and other volatile organic chemicals VOCs) down to the sub-parts-per-billion range is described. Ambient air was collected for up to 55 min of each hour into a holding canister, enabling a timeintegrated sample to be taken without the use of solid sorbent traps or excessive cryogen. Subsequent analysis, using both cryotrapping and cryofocusingsteps to produce high-resolution gas chromatograms, with either flame ionization (GC-FID) or mass selective (GC-MSD) detection, allowedfor quantitation and identification of sampled non-methane hydrocarbons (NMHCs) on a single capillary column. Under the chromatographic conditions used, the total analysis time of 1 h enabled concurrent sampling and analysis of the previously collected sample in a continuous, automated mode. The use of this system at a field site, during a 6-week field study in Atlanta, GA, as part of the Southern Oxidant Research Program on Ozone Non-Attainment is described. Consideration of the standard operating procedures developed, the calibration of the systems, and the validation of the data set is given. Illustration of advantages of this method is provided by some preliminary results of the NMHC analyses at a semiurban site (Lost Mountain, Marietta, GA), showing that the dominant single hydrocarbon for much of the study was isoprene.
Introduction
0 1994 American Chemical Society
A number of solid sorbent materials for the collection and enrichment of air samples have also been used (9-11). An ideal solid sorbent would selectively adsorb only the target compounds and not interact with any other atmospheric constituents. Thermal desorption and subsequent chromatographic analysis would allow separation and quantitation of the trapped VOCs. In practice, suboptimum results may occur for one or more of the following reasons: (i) sorbent bleed-off of interferant or target NMHCs [this may be alleviated by a thorough cleaning prior to the concentration step (12)],(ii) low breakthrough volumes for some volatile or polar NMHCs (131, (iii) incomplete desorption of high molecular weight NMHCs, (iv) chemical interaction of target NMHCs with sorbent, and (v) chemical interaction of other trace atmospheric components (e.g., 03, NO, NOz, halogens, etc.) with adsorbed VOCs leading to artifacts and losses (14-16). Athird approach combinesthe sampling and enrichment of NMHCs into one step. The compounds are frozen out (liquid nitrogen) in a cold trap containing an inert substance such as glass beads to provide increased surface area and then analyzed directly in the field. If correctly implemented, this 'on-line' approach should be subject to minimal sampling artifacts (17,18).The sampling strategies employed for the SOS studies used a combination of both canister and 'on-line' sampling of ambient air. The automated air concentrator used in this study, an Entech ELA 2000 (Simi Valley, CA), has previously been described for automated use in the analysis of canister samples (19). Modifications to the hardware and procedural details have demonstrated the utility of this device for continuous 'on-line' monitoring from a glass, airsampling manifold with subsequent capillary GC analysis in a mobile field laboratory. As described (20), by using an independent mass flow controller, a time-integrated sample was accumulated into an electropolished SUMMA canister over a 55-min period. At the end of this integration period, a subsample was automatically introduced into the ELA 2000 concentrator, and the analysis cycle begun. The integratingcanister was then automatically evacuated, and the integration-fill cycle was recommenced. During the ensuing 55 min, the GC analysis was completed, and the whole cycle was repeated. The sample-to-sample holdup within the integrating canister was shown to be insignificant (20). Materials and Methods Materials and Chemicals. Gases for the GC included the following: He, 99.999% purity (Liquid Carbonic, Chicago, IL), zero air generator (Type 75-85, Balston, Lexington, MA), and hydrogen generator (Model 8400, Packard, Downers Grove, IL). A 55-component retention time standard gas mixture was obtained commercially (Scott Specialty Gases, Plumsteadville, PA). The mass calibrant gas mixture, prepared and verified by the National Institute for Standards and Technology (NIST), was a two-component, butane-benzene mix distributed to the study by the National Center for Atmospheric Research (NCAR). Chromatography was performed on a 100-m, 0.25-mm, 0.5-km film thickness, DB-1 (Petro-DB1)capillary column (J&W Scientific, Folsom, CAI. These columns were specified throughout the study to enable a comparison of results from the five analytical systems used at the three different field sites of the overall study. Stainless steel or nickel (Supelco,Bellefonte, PA) tubing was used for gas supply lines, allowing samples and
standards to be introduced to the concentrator. Our preliminary studies had shown that, at room temperature, incomplete transfer of the higher weight NMHCs ( > C d along sample lines from cylinders and canisters was observed using stainless steel lines. This could be overcome by warming the stainless steel lines to approximately 40 "C or by using nickel lines. The use of deactivated fusedsilica tubing was also shown to be satisfactory in not retaining higher MW compounds. The mechanical strength of this material was a drawback, noticeably at points of swaged connections. All gas supply regulators were high-purity regulators (various suppliers). Stainless steel canisters were of the one- or two-valve (Nupro) type (BRC Rasmussen, Hillsboro, OR, or from Andersen Samplers, Atlanta, GA). The integrating canister used also had a dip entrance tube to ensure adequate mixing during sample collection. Vacuum pumps were all Teflonwetted surfaces (Model N726, KNF Neuberger, Princeton, NJ). Gas Chromatography. H P 5890 I1 (Hewlett Packard, Wilmington, DE) gas chromatographs with a cryogenic cooling option and FID detectors were used. An HP 5971A mass selective detector was also interfaced to a separate HP5890 I1 GC system. Data reduction was carried out using HP 3365 Chemstation I1 software on PC-DOS-based personal computers. The temperature program used was as follows: 3 min at -50 "C, ramp at 6 "C min-l to 175 "C, ramp at 25 "C min-1 to 250 "C, then hold for 3.5 min. The automated air concentrator used was an ELA 2000 (Entech Laboratory Automation, Simi Valley, CA). The integrating canister (6 L) used for manifold sampling was filled with an ELA 4510 realtime interface (Entech, Simi Valley, CA), which consisted of a mass flow controller and the necessary interfacing circuitry. Field Laboratory and Site. The field laboratory used for the study consisted of a standard 20 X 8 X 8 f t aluminum shipping container that had been previously outfitted as amobile laboratory enclosure. The field site was on private property located on a cleared ridge at Lost Mountain, Marietta, GA, approximately 27 km to the northwest of downtown Atlanta. The area immediately surrounding the site was lightly wooded. The nearest roads were located approximately 175 m below the site. A-10 m aluminum tower provided support for an 11-m all-glass sampling manifold (Pyrex, 5-cm i.d.1 which was constantly flushed by an approximate 2000 L min-l electric regenerative blower located at the base of the tower. A glass T piece, approximately 2 m from the base of the tower supported a nickel inlet line directly into the instrumentation. Measurement Conditions. Figure 1gives an overview of the analytical systems as configured for this study. Many of the 16 possible inlet ports to the concentrator were not used in this work, although four were dedicated to routine canister sampling from other sites. Dedicated lines were also connected to premixed cylinders (Scott Specialty Gases, Plumsteadville, PA) of a four-component internal standard, a 55-component retention time standard, and a NIST traceable calibration standard (n-butane). Important components of the ELA 2000 included the following: (i) the provision for a Nafion drying module for water removal, which for this entire study was replaced with a short length of nickel tubing. Under conditions of moderate ambient humidity or low sample volume, i.e., 5300 mL, it was not necessary to take special precautions to dry the ambient air sample. (ii)the cryotrappingmodule that was a nickel tube (0.3-cm o.d., 25 cm-long) filled with Environ. Scl. Technol., Vol. 28, No. 2, 1994
239
I l l
I I
I
lQ /,
am p I e
Vacuum Figure 1. Schematic of main gas flow paths for ambient air concentrator-GC anaiysls path. For this study, trapping module 1 was not utilized; trapping module 2 contained size-graded glass beads: and trapping module 3 consisted of a length of wide-bore (0.53 mm) deactivated silica tubing. The temperatures set in the traps and the gas flow rates through them, under mass flow control, were under software control. Dedicated input lines to a 16-port Inlet valve were connected to cylinders of standards, canister samples, or the integrating canister-glass manifold flow path.
a gradation of coarse to fine glass beads and then 5-cm of Chromasorb A. The packings were retained with silanized glass wool. (iii) a small volume cryofocusing trap (0.53mm 0.d. deactivated silica tubing) interfaced to the analytical column. This two-stage concentration enabled the NMHC content of a 300 mL air sample to be reduced to approximately 0.5 mL and then to approximately CO.1 p L before injection onto the GC column. The cryofocusing step on the concentrator, which took place after cryotrapping, was desirable to enable the smallest possible volume injection onto the head of the chromatographic column, ensuring the best possible chromatographic resolution. We have evaluated the retention time variability of systems that do not refocus after cryotrapping and have found it inferior to our current configuration. Because this arrangement was easily implemented with the Entech concentrator, we choose to use it to improve the resolution of low MW (Ca and CB)components. A recalculation of the observed retention times, perhaps using retention indices, is commonly made to yield the time precision we achieved. Additionally, this dual-step procedure provided for the purging, by the helium sweep gas that can be directed through the trapping modules, of associated oxidants present in the trapping module immediately prior to transfer to the cryofocusing trap and appeared to ameliorate water freeze-up often encountered in cryotrapping. All concentrator functions including the operation of a multiport sample introduction valve, external device (GC) signal recognition, cryogen delivery valves, heated zone temperature control, internal gas (purge and sweep) flow settings, and system bakeout functions were under software control implemented from a personal computer. Table 1 gives an overview of the steps of a typical analysis cycle. Standards and Blanks. Since the NMHC target compound list selected for the study was extensive, and given the nonselectivity of the response of a flame ionization detector to individual NMHCs, unequivocal identification of a given compound was entirely dependent upon the run-to-run reproducibility of the compound's retention time. A multicomponent retention time stan240
Environ. Scl. Technol., Vol. 28, No. 2, 1994
Table 1. Analytical Cycle of NMHC Analysis with ELA 2000 Concentrator
step
event
1 wait for GC ready 2 flush manifold and concentrator lines 3 set temperatures internally
trap preset volume of internal standard draw sample volume through cryotrap record initial and final sample pressure sweep lines to ensure quantitative transfer cool cryofocusing trap heat sample trap, back-flushing onto cryofocuser 10 inject VOCs onto capillary column; send start GC signal 11 bakeout traps; wait for set time before recycling 4 5 6 7 8 9
condition 0.5 min
cryotrap -180 OC 300 mL 2 min (100 mL) -185 O C cryotrap 120 "C
170 "C
dard was run daily to confirm the stability of the retention time windows of the target compounds. In addition, four internal standards (1,l-difluoroethane, neohexane (2,2dimethylbutane), 3-fluorotoluene, and n-decane) were added to each chromatographic run. This was achieved (see Table 1)by cryotrapping a preset volume (100 mL) of the internal standard gas mixture automatically, immediately prior to the trapping of the sample NMHCs, a provision allowed for in the design and control of the concentrator. These four compounds were chosen to bracket the chromatographic region of interest. Their registration as internal standards by the integration software relied on their recognition as the tallest peaks inside an expected retention time window of 0.4min. With the 100-m column used in this study, these limits were easily achievableand could even have been further reduced. A linear interpolation was performed between these marker peaks, enabling expected time windows of f6 s for individual target compounds to be established. The peak closest in time to the expected retention time found in these time windows was assigned the identity of the corresponding target compound. The target retention times were determined from the average retention times of the multicomponent retention time standard. The NMHC target list, in order of elution, is given in Table 2.
Table 2. SORP-ONAP Target NMHC List no.
compd
no.
compd
2-methyl-2-butene 21 ethane neohexaneb 22 acetylene 3-methyl-1-pentene 23 ethene 4-methyl-1-pentene 24 1,l-difluoroethaneb 4 cyclopentane 25 propene 5 2-methylpentane 26 propane 6 3-methylpentane 27 isobutane 7 n-hexane 28 isobutene 8 cis-3-hexene 29 1-butene 9 methylcyclopentane 30 n-butane 10 2,4-dimethylpentane 31 trans-2-butene 11 l,l,l-trichloroethane 32 cis-2-butene 12 benzene 33 3-methyl-1-butene 13 cyclohexane 34 isopentane 14 2,3-dimethylbenzene 35 1-pentene 15 trichloroethylene 36 ’ 2-methyl-1-butene 16 methylcyclohexane 37 17 n-pentane 2,3,4-triethylpentane 38 isoprene 18 39 toluene trans-2-pentene 19 40 3-fluorotolueneb cis-2-pentene 20 Southern Oxidant Research Program on Ozone Non-Attainment. Added internal standard. gas standard; presence in ambient samples indicated by GUMS. 1
2 3
Quantitation of the concentration of identified peaks was undertaken on the basis of the response of the FID detector of the GC as a carbon detector, using the NISTcertified n-butane mass standard. The average area response factor to a range of hydrocarbon concentrations, corresponding to those expected in ambient air, was established for both the n-butane and a secondary benzene mass standard. The butane standard contained 10.34 f 0.30 ppb butane (expressed as mol % in the mixture) and, hence, 41.36 f 1.20 ppb C. Samples of this mixture were injected as E O - , 300-, and 600-mL quantities through the concentrator GC system, i.e., in exactly the same manner as an ambient air sample. A linear best-fit relationship between the FID peak area response and the injected volume was thus obtained. By interpolating this best fit line to a 300-mL volume, an area response factor for the 41.36 ppb C (butane) standard was calculated. The stability of this peak area response factor with time was ensured by running, on a daily basis, 300-mL samples of the NIST mass standard and plotting these on a control chart (see below). Individual hydrocarbon species in ambient air samples were quantified using the n-butane response factor; consequently, concentrations reported here are in units of ppb C. The effective carbon number response factor of the FID to hydrocarbons has been experimentally demonstrated to be unity (21-23)in general for compounds to CS. In the absence of a definitive measurement of the response factor of the higher molecular weight hydrocarbons, we have also taken this to be unity per atom of carbon, based on the response of the instrument to a known mass of n-butane. No attempt was made to determine response factors for individual target hydrocarbons. It was felt that whatever potential gains in accuracy such an approach may yield would be far outweighed by the inconvenience, the cost, and the corresponding loss of precision associated with handling so many individual standards. This procedure also facilitated an estimation of the concentration of unidentified hydrocarbons. Given the variability and uncertainty of the FID response factors of multiply halogenated hydrocarbons, semiquantitative information only is suggested for these compounds. Several of these peaks are useful confirmatory markers for the chromatograms.
no. 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57
58 59 60
compd n-octane perchloroethylene ethylbenzene p-xylene m-xylene styrene o-xylene isopropylbenzene a-pinene n-propylbenzene 1-ethyl-3-methylbenzene 1-ethyl-4-methylbenzene 1,3,5-trimethylbenzene 1-ethyl-2-methylbenzene 8-pinene 1,2,4-trimethylbenzene n-decaneb limonenec 1,3-diethylbenzene n-butylbenzene
Isomerization product of unstable 8-pinene
The target quantitation limit of compounds for the study was set to be 0.1 ppb C. Peak areas integrating to less than this amount were excluded from the data analysis. Under optimum chromatographic conditions (Le., observing a peak well separated from any others, on a stable baseline) such a detection limit was possible with the 300mL sample volume employed during the study. The practical quantitation limit attained at the Lost Mountain site was more likely to be of the order of 1ppb C, reflecting the signal to noise of ambient samples of varying humidity and composition. This value was estimated from the consideration of repeated analyses of some commonly analyzed intercomparison samples made available to the study (see below). A complete validation of the data set currently being undertaken aims to set confidence limits for the concentrations of each of the components of the target analyte list. For several reasons these are not necessarily the same throughout the chromatographic run and may also show some variance among the three sites of the study. At low concentrations (