Optimization of a Mist Chamber (Cofer Scrubber) for Sampling Water

Ambient Air Measurement of Acrolein and Other Carbonyls at the Oakland-San Francisco Bay Bridge Toll Plaza. Hugo Destaillats, Reggie S. Spaulding, and...
3 downloads 0 Views 126KB Size
Download PDF
Environ. Sci. Technol. 2002, 36, 1798-1808

Optimization of a Mist Chamber (Cofer Scrubber) for Sampling Water-Soluble Organics in Air

power of the mist chamber in concert with PFBHA derivatization and mass spectrometry to measure pptv concentrations of water-soluble organics with a sampling time of 10 min.

Introduction REGGIE S. SPAULDING, ROBERT W. TALBOT,† AND M. JUDITH CHARLES* Department of Environmental Toxicology, University of California, Davis, Davis, California 95616

While the atmospheric fate and transport of biogenic and anthropogenic hydrocarbons has been extensively studied, little is known about the behavior of first-, second-, and third-generation photo-oxidation products that arise from OH radical oxidation of the parent species. The results of chamber experiments establish that •OH oxidation of biogenic and anthropogenic hydrocarbons yields carbonyls, dicarbonyls, hydroxycarbonyls, and keto-acids. However, little is known about the generation and fate of these products in the ambient atmospheric environment. This is changing because of the advent of methods that rely on O-(2,3,4,5,6pentafluorobenzyl)hydroxylamine (PFBHA) derivatization of carbonyls in concert with gas chromatography/ion trap mass spectrometry. Such methods provide the means to identify and quantify water-soluble organics, which historically have been difficult to measure. A limitation of existing sampling methods, however, is the use of devices that require low flow rates (0.5-1 L min-1). Accordingly, long sampling times (3-4 h) are needed to obtain pptv-ppbv detection limits. The mist chamber is an attractive device because of the high flow rates (25-70 L min-1) compatible with its use. Herein, we evaluate a mist chamber using a flow rate of 25-30 L min-1 to provide short (10 min) sampling times and pptv limits of detection. The results establish a relationship between the Henry’s law constant (KH) and the collection efficiency and demonstrate the suitability of the method to measure analytes with KH g 103 M atm-1. Adjusting the pH, adding quaternary ammonium salts, or decreasing the temperature of the collecting solution in the mist chamber did not significantly affect the collection efficiency. We tested the method by sampling photooxidation products of isoprene (glyoxal, methylglyoxal, hydroxyacetone, and glycolaldehyde) in the Blodgett Forest, CA. This is the first report of a study the employs the mist chamber to sample hydroxycarbonyls. The accuracy and the reproducibility of the method were evaluated by the analysis of duplicate samples and field spikes. The mean recovery of field spikes was g80%, and the relative standard deviation was e22% between duplicate measurements. The detection limits were 48, 15, 7.7, and 2.7 pptv for glycolaldehyde, hydroxyacetone, methylglyoxal, and glyoxal, respectively. This work demonstrates the * Corresponding author phone: 530-754-8757; fax: 530-752-3394; e-mail: [email protected]. † Current address: Institute for the Study of Earth, Oceans, and Space, University of New Hampshire, Durham, NH. 1798

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 8, 2002

Volatile organic compounds (VOCs) emitted from biogenic and anthropogenic sources play a key role in the generation of tropospheric ozone, secondary aerosols, and atmospheric radical species (1-3). The formulation of regulatory strategies to control tropospheric ozone and fine particles, components of secondary organic aerosols, requires a thorough understanding of atmospheric transformation reactions and atmospheric conditions that influence the generation and fate of the reaction products. Chamber studies establish that OH radical oxidation of biogenic and anthropogenic VOCs yields first-, second-, and third-generation products. Many of these products are carbonyls, hydroxycarbonyls, dicarbonyls, and keto-acids (2, 4). Several studies document the formation and concentrations of carbonyls in ambient air (5-11). However, few studies measure a broad range of carbonyls, including dicarbonyls, hydroxycarbonyls, and keto-acids in air, and those that do exist often focus on only one class of compound (12-14). Absence of such ambient air data on mulitfunctional carbonyls is primarily due to the lack of suitable analytical methods. Methods exist that employ derivatization with 2,4dinitrophenylhydrazine (DNPH) along with HPLC and UVvisible absorption (12, 15, 16) and derivatization with O-(2,3,4,5,6-pentafluorobenzyl)hydroxylamine (PFBHA) in concert with gas chromatography/mass spectrometry (GC/ MS) or GC/electron capture detection (17, 18) to measure carbonyls. Hydroxyl and carboxyl groups on the PFBHA derivatives can be further derivatized with bis-(trimethylsilyl)trifluoroacetamide (BSTFA) to improve the chromatography and sensitivity (19, 20). An advantage of using PFBHA as the derivatization reagent, in contrast to DNPH, is that R-dicarbonyls can be distinguished from hydroxycarbonyls. In addition, the method affords unambiguous molecular weight determinations by interpreting the electron ionization (EI), methane chemical ionization (methane CI), and pentafluorobenzyl alcohol chemical ionization (PFBOH CI) ion trap mass spectra (20-22). (In previous work, we established that ion-molecule reactions are critical to achieve unambiguous identification of the derivatives; refs 20-22.) Sampling devices used in previous studies are impingers, a coil scrubber, a mist chamber containing an aqueous solution of the derivatization reagent (12, 16, 23, 24), cartridges coated with the derivatizion reagent (24), and cooled solid sorbents (5, 6, 9) to sample carbonyls. The low flow rates (0.5-1 L min-1) compatible with impingers and coated cartridges necessitate long sampling times (3-4 h) (15, 20). Full automation and short sampling times (e.g., 10 min) are achieved by using cryogenic collection, or cryogenic collection with solid sorbents, along with thermal desorption and GC with flame ionization (FID) or MS detection (5, 8). These methods, however, do not afford the measurement of trace levels of the more polar multifunctional carbonyls. The mist chamber was previously utilized to sample formaldehyde, pyruvic acid, glyoxal, and methylglyoxal, each in a different study (12, 25, 26), but no information exists regarding the collection of hydroxycarbonyls. Because the mist chamber can operate at flow rates from 25 to 70 L min-1, it is a promising technique for the collection of trace levels of watersoluble organics with short sampling times. 10.1021/es011189x CCC: $22.00

 2002 American Chemical Society Published on Web 03/12/2002

Herein, we evaluate a modified mist chamber that operates at flow rates of 25-30 L min-1 and 10-min sampling periods for collection of water-soluble organics. We explore operational parameters that may affect the collection efficiencies of carbonyls, dicarbonyls, hydroxycarbonyls, and keto-acids that are first-, second-, and third-generation products from photooxidation reactions of isoprene, 2-methyl-3-buten-2ol (MBO), and alkylbenzenes. These parent hydrocarbons originate from biogenic and anthropogenic sources and are important sources of VOCs (27-29). Herein, we provide the first measurements of a range of carbonyls by using the mist chamber. Most notably, it is the first study that reports the collection of hydroxycarbonyls. The sampling technique was used in conjunction with PFBHA, PFBHA/BSTFA derivatization of carbonyls and dicarbonyls, and carbonyls containing hydroxy and carboxy moieties, respectively, followed by gas chromatography/ion trap mass spectrometry (GC/ITMS). We tested and proved the power of the method by measuring isoprene photooxidation products in the Blodgett Forest, CA.

Experimental Section Preparation of Reagents and Glassware. HPLC grade water (Fisher Scientific, Pittsburgh, PA) was purified by distillation in glass in the presence of KMnO4 to oxidize any organic contaminants. Other solvents were purchased in the highest purity available (g99.9%) and further purified by distillation in glass. O-(2,3,4,5,6-Pentafluorobenzyl)hydroxylamine hydrochloride (PFBHA) was recrystallized two times in glassdistilled 2-propanol. Carbonyl standards were purchased from Aldrich Chemical Co. (Milwaukee, WI) at the highest purity available and used without further purification. All glassware was soaked overnight in 15% (v/v) dichlorodimethylsilane in toluene. The glassware was then rinsed with toluene, methanol, and dichloromethane (three times each) and dried at 150 °C. PFBHA and BSTFA Derivatization. Carbonyls were allowed to react with PFBHA in a 1 mM (ambient air studies) or 2 mM (laboratory studies) aqueous PFBHA solution for 24 h at room temperature. After derivatization, the PFBHA derivatives were extracted two times with 4 mL of CH2Cl2. Five milliliters of H2O containing 2 drops of concentrated H2SO4 was added to each vial to remove excess PFBHA. The organic fraction was filtered through a 70 mm × 7 mm o.d. column of anhydrous Na2SO4(s) to remove water. The eluate was collected and evaporated under a stream of N2(g) to a final volume of 4.0 mL in laboratory studies and 200 µL in ambient air studies. Each time a vial was opened, the headspace was evacuated with N2(g) before it was resealed. For the derivatization of hydroxyl and carboxyl groups, BSTFA, containing 10% trimethylchlorosilane (TMCS) was employed. This solution was prepared daily by adding 100 µL of TMCS to a 1-mL ampule of BSTFA. A 100-µL aliquot of the PFBHA derivatives in CH2Cl2 was transferred to a 500-µL silanized glass autosampler vial insert. Twenty microliters of BSTFA/TMCS (90:10, v/v) was added to the vial, the headspace in the vial was evacuated with N2(g), and the vial was sealed with a Teflon-coated cap. The solution was allowed to react for 12 h at 42 °C, after which the derivatives were analyzed by GC/ITMS. GC/ITMS. In previous studies, we and others established the power of ITMS to identify analytes for which authentic standards did not exist and to identify analytes in the presence of coeluting interferences (19-22). Ion-molecule reactions that facilitate the identification of molecules do not occur when using a quadrupole or double-focusing magnetic sector mass spectrometer (30). A Varian Star 3400 CX gas chromatograph, with a temperature-programmable injector port interfaced to a Saturn 2000 ion trap mass spectrometer (Varian Analytical Instruments, Walnut Creek, CA) was

employed. Gas chromatographic separation of the derivatives was accomplished by using a DB-XLB column (30 m × 0.25 mm i.d., 0.25 µm film thickness; Agilent Technologies, Inc., Wilmington, DE). A 5-m integrated guard column was used prior to the analytical column. The injector temperature was held at 280 °C for 1 min, increased to 310 °C at 50 °C min-1, and held at 310 °C for 25 min. The oven of the gas chromatograph was held at 70 °C for 1 min. The temperature was increased to 100 °C at 5 °C min-1, increased to 280 °C at 10 °C min-1, increased to 310 °C at 30 °C min-1, then held at 310 °C for 10 min. The sample injection volume was 2 µL. EI, methane CI, and PFBOH CI mass spectra were acquired as previously described (20-22). The power of these techniques, and their complementary nature, was established in previous work (20-22). The masses scanned were from m/z: 65-650 under EI conditions, from m/z: 150-550 under methane CI conditions, and from a m/z: 230-650 under PFBOH CI conditions. We limited the masses scanned under CI and PFBOH conditions to increase the sensitivity of the techniques. The addition of a pentafluorobenzyl group adds 185 amu to the native species. The ion trap mass spectrometer in this study has an upper mass limit of 650 amu. Assuming that the monocarbonyls are derivatized with one pentafluorobenzyl group, the dicarbonyls are derivatized with one carbonyl group, and the tricarbonyls are derivatized with three pentaflurobenzyl groups, it is unlikely that we would identify tricarbonyl species because the molecular weight of the derivatives exceed 650 amu. Identification and Quantification. Identification of analytes for which authentic standards exist was accomplished by interpreting the methane CI mass spectra and by matching the relative retention time and mass spectra of an authentic standard to the analyte in the sample extract. Compounds for which authentic standards do not exsit were tentatively identified by interpreting their EI, methane CI, and PFBOH CI mass spectra (20-22). All stock solutions were stored at -80 °C in the laboratory and kept on dry ice during field sampling. A working standard solution contained formaldehyde, acetone, methacrolein, methyl vinyl ketone, glycolaldehyde, hydroxyacetone, glyoxal, methylglyoxal, and pyruvic acid and was prepared in methanol. This solution was utilized in the laboratory experiments. The working standard solution for the experiments conducted in the ambient environment was also prepared in methanol and contained formaldehyde, glyoxal, methylglyoxal, glycolaldehyde, hydroxyacetone, and pyruvic acid. Methacrolein and methyl vinyl ketone were not components of this solution because the results of the laboratory experiments established low collection efficiencies for these compounds. The working standard solution was used to produce solutions for the generation of calibration curves and to prepare solutions enriched with the analytes (field spikes). The internal standard stock solution was prepared in methanol and contained 13C3-acetone (used to quantify aldehydes and ketones), 4-fluorobenzaldehyde (used to quantify dicarbonyls), and 4-hydroxybenz-13C6-aldehyde (used to quantify hydroxycarbonyls). Calibration solutions were prepared by adding known quantities of the analytes and internal standards to 10 mL of a 2 mM aqueous PFBHA solution (laboratory studies) or 20 mL of a 1 mM aqueous PFBHA solution (ambient air studies). The resulting extracts, comprised of the PFBHA and PFBHA/BSTFA derivatives of authentic standards, ranged in concentration from 62.5 to 6250 pg µL-1 of the analyte and 250 pg µL-1 of the internal standards (laboratory studies) or from 40 to 4000 pg µL-1 of the analyte and 400 pg µL-1 of the internal standards (ambient air studies). Standard concentrations greater than 1000 pg µL-1 were used for calibration curves only when sample concentrations exceeded this concentration. VOL. 36, NO. 8, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1799

Standard solutions were analyzed before and after the analysis of sample extracts. Quantification was accomplished by internal standardization, using a regression equation obtained by plotting the (peak areaanalyte(M+H)+ ion)/(peak areainternal standard(M+H)+ ion) generated by methane CI versus the concentration of the analyte. A “mid-point” standard solution was analyzed after every five samples to evaluate the stability of the calibration curve. Typically, during the analytical period, 15 samples and 2 midpoint standard solutions were analyzed. The calculated concentration of the midpoint standard was within (20% of the actual value. The linearity of the calibration curves was evaluated by (1) examining the coefficient of determination (r2) of the linear regression equations and (2) comparing the relative response factors [(peak areaanalyte)/(peak areainternal standard) × (concentrationinternal standard)/(concentrationanalyte)] of the analyte over the range of concentrations. The r2 values were typically g0.98 for acetone, methacrolein, and methyl vinyl ketone; g0.95 for formaldehyde, glycolaldehyde, and hydroxyacetone; and g0.90 for glyoxal and methylglyoxal. The relative standard deviation among the relative response factors was e25%. Blanks, Matrix Spikes, and Limits of Detection. For both laboratory and ambient air studies, ultrahigh purity (UHP) zero-grade air was collected in the scrubber as a blank and treated identically to the samples. The concentration measured in the blank extracts was subtracted from the concentration measured in the sample extracts. For ambient air studies, the limit of detection (LOD) was calculated as 3σblank, where σblank is the standard deviation of the response factor for the analyte in three or more reagent blanks. In experiments in which the analytes were extracted from filters or a stainless steel annular denuder, the matrix spike was prepared by enriching the filter or stainless steel denuder with a solution of the analytes. Preparation of Gas-Phase Standards. A 200-L Tedlar bag was partially filled with UHP zero-grade air. A 25-µL aliquot of a solution of 10 ng µL-1 of analytes in methanol was placed in a silylanized 1/4 in. o.d. glass U-tube. The U-tube was connected to the zero-grade air tank with 1/4 in. stainless steel tubing at one end and to the Tedlar bag with 1/4 in. PTFE tubing at the other end. All tubing was heated with heating tape. UHP zero-grade air was passed through the U-tube to the Tedlar bag at 1 L min-1. The U-tube was held at room temperature for 1 min and then heated to 200 °C at approximately 5 °C min-1. The bag was filled to a final volume of 200 L of air. Mist Chamber. Operating principles of the mist chamber, or Cofer scrubber, have been described previously (23, 25). Briefly, the scrubber consists of a glass mist chamber with an air inlet, a nebulizing nozzle, and a port for addition and removal of solution. The collecting solution is drawn from the reservoir into the nozzle, where impaction with the incoming air forms a fine aqueous mist. The fine aqueous drops provide a large surface area for efficient extraction of water-soluble compounds from air. The mist is drawn upward through the mist chamber until it reaches a hydrophobic membrane, where aqueous droplets containing scrubbed atmospheric gases coalesce and are refluxed into the reservoir. In this work, air was drawn through the scrubber with a GAST 1023, 50 Hz, 3/4 horsepower vacuum pump (TECO Pneumatic, Inc., Pleasanton, CA) connected to the outlet. The flow rate was measured with a Hastings mass flow controller (HFC203; Williams & Associates, Commercial City, CO) and ranged from 28 to 30 L min-1 (laboratory studies) and from 23 to 25 L min-1 (ambient air studies). The reservoir was filled with a collecting solution that consisted of 10 mL of 2 mM (laboratory studies) or 20 mL of 1 mM (ambient air studies) aqueous PFBHA. A 1-µm diameter pore size 1800

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 8, 2002

Zefluor filter (Pall-Gelman Laboratory, Ann Arbor, MI) fitted to the top of the glass chamber prevented the loss of liquid water. Collection Efficiency (CE). If Henry’s law equilibrium is established for carbonyl compounds during sample collection in the mist chamber, the CE for these compounds can be estimated using a gas/aqueous-phase distribution calculation, where CE represents the fraction of analyte that is present in the aqueous phase at equilibrium (32). The theoretical CE was calculated using eq 1

CE )

KHRTwL 1 + KHRTwL

(1)

where KH is the Henry’s law equilibrium constant (M atm-1), R is the ideal gas law constant, T is the temperature, and wL is the liquid water content (volume water/volume air). For all of the carbonyls, except methacrolein and methyl vinyl ketone, we used the “effective” KH, determined from the total amount of carbonyl that dissolves in water (including the gem-diol) (33). For methacrolein and methyl vinyl ketone, we used the “physical” KH determined as the amount of carbonyl dissolved in its native form (without inclusion of the gem-diol), because to our knowledge, effective KH are not available (34). However, we expect, as with other aldehydes and ketones, that minimal gem-diol formation would occur for methacrolein and methyl vinyl ketone (35, 36). For example, if we assume that the hydration equilibrium constants for methacrolein and methyl vinyl ketone are within the hydration constants for acetone (Khyd ) 2 × 10-3) and butyraldende (Khyd ) 0.83), the KH would increase by a factor less than 2. For hydroxyacetone and pyruvic acid, we estimated the KH from solubility and vapor pressure because experimental values were not available (37-39). Another reaction that may affect the CE is PFBHA derivatization of the analyte during sample collection. We assume that at pH 3, the reaction between the carbonyl and PFBHA is too slow to affect the Henry’s law equilibrium. Two lines of evidence exist to support this assumption. First, data published by LeLacheur et al. in 1993 (30) and unpublished data from experiments conducted in our laboratory that investigate the effect of the reaction time on the yield of PFBHA derivatives indicate that greater than 10 min is required for complete derivatization of carbonyls. Second, the reaction of the analytes with PFBHA would presumably drive the reaction to effect greater dissolution of the analytes. Therefore, if the reaction with PFBHA were significant, the empirical CE would be greater than the theoretical CE. As presented in Table 2, good agreement was obtained between the theoretical and empirical efficiency for most of the compounds. Thus, existing data do not indicate that the reaction of the analyte with PFBHA affects the CE during sample collection. To explain empirical observations and to support the choice of operational parameters that may increase the CE by altering the KH, we evaluated the potential for Henry’s law equilibrium to be established during sample collection. We want to emphasize that the focus of this work is the development and application of an analytical method. We acknowledge that our analysis is not precise, but the analysis conducted is sufficient to suggest the potential for equilibrium to be established. A more rigorous analysis is beyond the scope of this work. When collecting air at 30 L min-1 in a 190-mL mist chamber containing 10 mL of aqueous solution, the contact time between the air and the water is 0.36 s. Therefore, for equilibrium conditions to apply, equilibrium must be established in e0.36 s. Establishment of the effective Henry’s law equilibrium between air containing an analyte and an aqueous solution

that does not contain the analyte involves (1) gas-phase diffusion of the analyte to the surface of the mist droplet, (2) transport of the analyte across the interface of the droplet, (3) aqueous-phase diffusion of the analyte through the droplet, and (4) hydration of the carbonyl to form a gemdiol. A rough approximation for the characteristic times for each process was calculated (equations, constants, and references used can be found in the Supporting Information). The calculations, which are not meant to predict precise times, indicate that the time for each transport process will be at least an order of magnitude less than 0.36 s for carbonyls, hydroxycarbonyls, dicarbonyls, and keto-acids. However, the time for hydration can range from 0.06 to 10 s. Therefore, establishment of the effective Henry’s law equilibrium should not be limited by mass transport, but may be limited by the hydration reaction for certain compounds. The rate and extent of hydration can vary considerably for different carbonyl compounds, and constants are not available for most of the analytes of interest. However, for compounds that become hydrated rapidly or for which little hydration occurs, Henry’s law equilibrium may dictate the CE of the mist chamber, and eq 1 should provide a reasonable approximation of the CE. For two samplers in series, such as in our experiments, the CE is determined by eq 2

CE ) 1 - (C2/C1)

(2)

where C2 and C1 are the concentration of analyte measured in the second and first sampler, respectively. A discussion of the reason for using this equation is presented in the Supporting Information. The efficiency of the mist chamber to collect gas-phase carbonyls, hydroxycarbonyls, dicarbonyls, and keto-acids was investigated in the laboratory by sampling gas-phase standards from a Tedlar bag into two scrubbers in series. The theoretical CE (eq 1) was compared to the empirical CE (eq 2) to evaluate whether Henry’s law was controlling CE for the analytes. This approach was also used by other researchers to calculate the empirical and theoretical CE (16). Experiments were also conducted to evaluate the effect of operational parameters on the CE for the mono- and multifunctional carbonyls. The conditions examined were modifications of the pH and temperature and the addition of tetraalkylammonium salts to the collecting solution. Effect of pH on CE. A Tedlar bag experiment (n ) 2) and an experiment in the ambient environment (n ) 5) were conducted to investigate the effect of the pH of the collecting solution on the CE. The pH was 3 for the PFBHA solution, without the addition of buffer. The pH of the PFBHA solution was altered to pH ) 1, 5, or 7 by the addition of H3PO4 and NaH2PO4 to a final PO43- concentration of 2 mM (lab experiments) or 5 mM (ambient air experiments) and by adjusting the pH with NaOH. NaCl was added as needed to make the conductivity of each sample the same. The pH of the solution was measured after collection of samples and was within (0.5 pH units of the initial pH. Effect of Temperature on CE. A Tedlar bag experiment (n ) 3) was conducted to determine the effect of temperature of the PFBHA solution on the CE. The temperature of the collecting solution was decreased with a plastic bag containing ice and KCl that was wrapped around the mist chamber. Effect of Addition of Salts on CE. A Tedlar bag experiment was conducted to determine the effect of tetraalkylammonium salt additions on the CE. Air was collected into mist chambers containing (1) aqueous PFBHA (control), (2) aqueous PFBHA containing 0.25 molal tetramethylammonium bromide, or (3) aqueous PFBHA containing 0.25 molal tetrapropylammonium bromide (n ) 2).

Interferences from Ozone. We evaluated potential positive and negative interferences from ozone first by theoretical calculations and then by sampling in the ambient environment in the presence and absence of an ozoneremoving device. Theoretical Calculations of Negative Interferences. We estimated the rate of aqueous reaction between analytes and ozone using 100 ppbv as the maximum mixing ratio of ozone in the Blodgett Forest (Schade, G., personal communication). A reaction rate of 100 M-1 s-1 for reaction of a carbonyl with ozone in water was employed (40). The Henry’s law constant for ozone is 0.0113 M atm-1. We used actual aqueous concentrations of carbonyls measured after collection of a 250-L air sample in the Blodgett Forest and a reaction time of 24 h (reaction time of analytes with PFBHA in aqueous solution). Theoretical Calculations of Positive Interferences. Positive interferences were calculated on the basis of the assumption that the analytes of interest (formaldehyde, glycolaldehyde, hydroxyacetone, glyoxal, and methylglyoxal) can be formed by aqueous ozonolysis of isoprene, MBO, methacrolein, and methyl vinyl ketone and that ozonolysis of the precursors can form any one of the analytes at 100% yield. We used maximum mixing ratios for isoprene (10 ppbv), MBO (10 ppbv), methacrolein (5 ppbv), and methyl vinyl ketone (10 ppbv) measured in the Blodgett Forest (9, 41). We calculated the aqueous-phase concentrations of these compounds in the mist chamber by using their Henry’s law constants (34, 42, 43). Because the compounds are unsaturated, they react with ozone more rapidly than analytes discussed previously (104-105 M-1 s-1 vs 102 M-1 s-1) and will quickly deplete the aqueous ozone in the sample (44, 45). Therefore, the total amount of each parent compound that could react with ozone was calculated as the sum of the amount that would react during the 10-min sampling period (assuming Henry’s law equilibrium for ozone) and the amount that would react with the remaining ozone that is present at the end of the sampling period. Air Sampling in the Absence and Presence of a KI-Coated Stainless Steel Annular Denuder. Denuders were constructed from a 12 in. × 1/8 in. o.d. stainless steel tube inside a 12 in. × 1/4 in. o.d. stainless steel tube. The denuders were coated with KI by passing a saturated aqueous KI solution through the denuder and drying it under a stream of N2(g). At a flow rate of 30 L min-1, we established that the denuder removes 91% ( 2% of air containing 120 ppbv of O3. Six duplicate air samples were collected at the Blodgett Forest, CA, on July 29, 2000, at 11:02, 11:59, 12:01, 13:04, 13:59, and 14:45 local standard time (LST) in the absence and presence of a KI-coated stainless steel annual denuder. At 08:55 and 16:49, duplicate samples were collected without denuders to determine precision. Ozone mixing ratios during the sampling period ranged from 49 to 95 ppbv, which are typical summer levels. Sampling Site. The Blodgett Forest field site is a ponderosa pine plantation, owned and operated by Sierra Pacific Industries. The plantation is located (38°53′42.9′N, 120°37′57.9′W, 1315 m) adjacent to the Blodgett Forest Research Station, a research forest of the University of California, Berkeley. Detailed descriptions of the site can be found elsewhere (9, 41). Statistical Analyses. All statistical analyses were accomplished by using MINITAB statistical software for Windows, release 13 (Minitab Inc., State College, PA). A student t-test was used to compare means for experiments in which two parameters were varied. When the experiment contained more than two variables, analysis of variance (ANOVA) was used to test for differences between means. When a difference between means was detected by ANOVA, Tukey’s pairwise comparison was used to indicate which means differed. We VOL. 36, NO. 8, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1801

TABLE 1. Distribution of Analytes between the Zefluor Filter and the Aqueous PFBHA Solutionin the Mist Chamber

compound formaldehyde glycolaldehyde hydroxyacetone glyoxal methylglyoxal pyruvic acid

a

sample number 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3

concentration (pg/µL) aqueous filter solution

percent of total aqueous filter solution

0 35 0 0 165 153 5 0 12 17 104 32 86 224 55 5 0 10

0 6a 0 0 11 22 1 0 20 2 28 4 4 19 4 1 0 2

2447 1602 6013 509 1304 542 668 2149 49 832 273 816 2329 973 1504 600 477 654

100 94a 100 100 89 78 99 100 80 98 72 96 96 81 96 99 100 98

percent of total in solution mean ( SD (% RSD) 99 ( 1 (1) 89 ( 11 (12) 93 ( 12 (12) 87 ( 14 (17) 91 ( 9 (10) 99 ( 1 (1)

For formaldehyde, the percent collected on the filter was divided by 0.34 to correct for the 34% recovery obtained from enriched filters.

TABLE 2. Comparison of Theoretical and Empirical Collection Efficiencies for Sampling Carbonyls in a Mist Chamber (n ) 4) collection efficiency compound methacrolein acetone methyl vinyl ketone hydroxyacetone formaldehyde pyruvic acid glycolaldehyde glyoxal methylglyoxal d

“effective” KH at 298 K (M atm-1) 4.3a 26b 46a 2927c,d 2970b 6730c,e 41400b >300000b 371000b

theoretical

empirical mean ( SD

difference

0.00 0.02 0.04 0.70 0.71 0.85 0.97 1.00 1.00

0.08 ( 0.36 -1.02 ( 2.05 0.06 ( 0.36 0.84 ( 0.14 0.84 ( 0.12 0.94 ( 0.12 0.93 ( 0.06 0.78 ( 0.20 0.85 ( 0.04

+0.08 -1.00 +0.02 +0.14 +0.13 +0.12 -0.04 -0.22 -0.15

a “physical” K from Allen et al. (34). b “effective” K from Betterton (50). c K estimated from solubility from Yalkowsky and Dannenfelser (37). H H H KH estimated from solubility and vapor pressure from Neely and Blau (39). e KH estimated from vapor pressure from Perry and Green (38).

performed the t-test, ANOVA, and Tukey’s pairwise comparison using 90% confidence intervals (p ) probability of a type I error ) 0.10).

Results and Discussion Sorption of Analytes to the Zefluor (Teflon) Membrane Filter. Because collection efficiencies and analyte concentrations measured in the mist chamber could be affected by sorption of the analytes to the Zefluor filter during sample collection, we evaluated whether sorption to the filter was significant. After the collection of gas-phase standards, the concentrations of model carbonyl compounds sorbed to the Zefluor filter were compared to the concentration measured in the scrubber by using an extraction method described by Rao et al. (22). To determine the efficiency of the extraction method, we also extracted carbonyls from Zefluor filters that were enriched with a solution containing the compounds. The average recoveries from the enriched filters were 91% for glycolaldehyde (n ) 1), 77% for hydroxyacetone (n ) 1), 100% for pyruvic acid (n ) 1), 134% ( 42% for glyoxal, and 80% ( 25% for methylglyoxal (n ) 2), indicating that these compounds can be extracted efficiently from the Zefluor membrane filter. The average recovery of formaldehyde was 34% ( 6% (n ) 2). Formaldehyde concentrations measured on Zefluor filters, after sample collection, were corrected to account for the low recoveries. The recoveries for acetone, methacrolein, and methyl vinyl ketone were e1%. Therefore, 1802

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 8, 2002

sorption of these compounds onto the filters was not measured. The distribution of carbonyls between the aqueous collecting solution and the Zefluor filters after collection of gas-phase samples in the mist chamber is presented in Table 1. The fraction in the aqueous phase ranged from 87% to 99% of the total collected. We consider (20% to be acceptable error for the method; thus, the sorption of formaldehyde, hydroxycarbonyls, dicarbonyls, and keto-acids on the Zefluor membrane filter during sample collection is essentially zero. Relationship Between CE and Henry’s Law. The results of experiments conducted to evaluate the relationship between the CE and KH are presented in Table 2, as the theoretical and empirical collection efficiencies for model carbonyl compounds in the mist chamber. For all compounds except acetone, good agreement (e22% difference) was obtained between the theoretical and empirical collection efficiencies. The difference between theoretical and empirical collection efficiencies for acetone (0.02 vs -1.02) is likely a reasult of analytical errors that arise from measuring acetone near the LOD. The data indicate that Henry’s law equilibrium is established for carbonyls, hydroxycarbonyls, dicarbonyls, and ketoacids during sample collection in the mist chamber and that the CE of a given carbonyl compound can be predicted using eq 1 when effective Henry’s law constants are available. The

FIGURE 1. Two-step reaction of PFBHA with a carbonyl to form a pentafluorobenzyl oxime.

FIGURE 2. Collection efficiencies observed for analytes collected from a Tedlar bag with alteration of the pH of the aqueous PFBHA solution in the mist chamber (n ) 2). data also suggest that the method is suitable for sampling water-soluble species with KH g 103. Evaluation of Operational Parameters. Given the correlation between the collection efficiency and the Henry’s law constant, we hypothesized that we could increase the CE by altering operational parameters that would affect the KH. Accordingly, we investigated the effect of pH, temperature, and addition of organic salts on the CE. We did not investigate the effect of relative humidity on the CE because previous studies establish that accurate measurements of watersoluble species can be obtained by using the mist chamber, without adjusting for the water volume lost due to the relative humidity (51, 52). This assumption was proved valid by examining the collection efficiencies for samples collected in July and August, 2000, under variable relative humidity conditions (12-19%) (Goldstein, A., personal communication). For a total of 108 samples, the mean and standard deviation of the CE was 0.92 ( 0.07 for hydroxyacetone, 0.93 ( 0.07 for glycolaldehyde, 0.90 ( 0.21 for glyoxal, and 0.83 ( 0.11 for methylglyoxal. If relative humidity affected the CE, we would observe a greater variability in the CE and lower collection efficiencies because of the evaporation of the analytes with the water vapor on days with low relative humidities. Under low relative humidity conditions, the water and dissolved analytes in the water may evaporate from the mist chamber and condense into the aqueous phase in the second mist chamber. If relative humidity did affect the CE, we expected to observe a positive correlation due to a lower CE and lower volume of water in the first chamber due to evaporation of the water and the analyte. We thus investigated whether a correlation exists between the CE and the water volume in the first scrubber. The correlation of determination (r2) between the analytes and the water volume in the first chamber was e0.07 (p e 0.098). Accordingly, the experimental data obtained in this study and later work does not indicate that relative humidity affects the CE.

Effect of pH. In theory, reaction of the dissolved carbonyl with PFBHA allows more of the gas-phase carbonyl to dissolve in order to satisfy Henry’s law equilibrium. We thus reasoned that if the reaction of the analytes with PFBHA occurs within the contact time of air and water (0.36 s), then the reaction with PFBHA could increase the CE by increasing dissolution of the analytes. To our knowledge, kinetic data do not exist regarding the reaction of carbonyls with PFBHA. Limited data do exist, however, for the reaction of carbonyls with hydroxylamine (46). These data demonstrate that the reaction is a two-step process, illustrated for PFHBA in Figure 1. In the first step of the reaction, the hydroxylamine reacts with the carbonyl through nucleophilic attack to form an amino alcohol. This reaction is faster at a pH above the pKa of hydroxylamine (pH ) 6 for hydroxylamine; pH ) 3.2 for PFBHA), because hydroxylamine is not protonated. The second step is dehydration of the amino alcohol, which leads to formation of an oxime. The dehydration reaction is faster at low pH, where the amino alcohol is easily protonated (46). For the reaction of hydroxylamine with acetone, the overall reaction rate produces a bell-shaped curve, with the maximum rate occurring between pH 4 and 5. We expect the kinetics for PFBHA to be similar, and hypothesized that, by changing the pH, we could increase the rate of reaction of PFBHA with the carbonyls to a rate that would cause an increase in the CE. We therefore determined the CE for collection of carbonyls with the collecting solution at pH ) 1, 3, 5, and 7. The data are presented in Figure 2 as a plot of CE versus compound. Although analysis of variance (ANOVA) indicated no significant difference in collection efficiencies at the different pH (p ) 0.10), visual inspection indicates an optimum pH of 5 for methyl vinyl ketone, acetone, formaldehyde, and methylglyoxal; an optimum pH of 5 or 7 for hydroxyacetone; and a pH of 3 or 5 for glyoxal. Further, the data appear to follow a bell-shape curve, in which there is VOL. 36, NO. 8, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1803

TABLE 3. Comparison of Theoretical and Empirical Collection Efficiencies at 16 and 11 °C collection efficiency at 16 °C (298 K) mean ( SD compound methacroleina acetoneb methyl vinyl ketonea hydroxyacetone formaldehydeb glycolaldehydeb pyruvic acid glyoxal methylglyoxalb a

∆KH from Allen et al. (34).

theoretical

empirical

difference

theoretical

empirical

difference

0.01 0.03 0.07

0.02 ( 0.03 -0.28 ( 0.50 0.04 ( 0.10 0.83 ( 0.06 0.78 ( 0.04 0.86 ( 0.07 0.99 ( 0.01 0.51 ( 0.08 0.73 ( 0.03

+0.01 -0.31 -0.03

0.01 0.04 0.11

-0.01 +0.23 +0.08

-0.04 -0.12

0.87 0.99

-0.27

1.00

0.00 ( 0.00 0.27 ( 0.07 0.19 ( 0.03 0.78 ( 0.09 0.99 ( 0.02 0.90 ( 0.06 0.98 ( 0.09 0.49 ( 0.05 0.66 ( 0.06

0.82 0.98 1.00 b

KH(T2) 1 R 1 ) ln T2 T1 ∆KH KH(T1)

(3)

We attempted to increase KH and, thus, the CE, by wrapping the chamber in a KCl ice bath (-8 °C) to decrease the temperature of the collecting solution. The temperature of the collecting solution, measured after sample collection, was 11 °C in the chamber wrapped with the KCl ice bath and 16 °C in the control chamber. The mean collection efficiencies (n ) 3) for each analyte at 16 and 11 °C are presented in Table 3. For methacrolein, hydroxyacetone, pyruvic acid, glycolaldehyde, glyoxal, and methylglyoxal, the collection efficiencies do not appear to be affected by the temperature of the collecting solution. For methyl vinyl ketone, acetone, and formaldehyde, the collection efficiencies appear to increase at the lower temperature. However, significant differences (p ) 0.10) were only observed for formaldehyde. To explain these results, we calculated the theoretical KH at 16 and 11 °C, for compounds for which ∆KH was available, according to eq 3. Using values obtained for KH at 16 °C (289 K) and 11 °C (284 K), we calculated the theoretical collection efficiencies at the two temperatures using eq 1 (see Table 3). Overall, good agreement was obtained between 9

+0.12 -0.09 -0.34

∆KH from Betterton (50).

an increase in the CE from pH ) 1 to 5, and a decrease from pH ) 5 to 7, representing the balance in the rates of nucleophilic attack and dehydration. To resolve differences between conclusions reached by visual inspection and statistical analysis of the data, an additional experiment was conducted in the ambient environment. In this experiment, we compared collection efficiencies at pH ) 5 (the pH was adjusted with PO43- buffer) to collection efficiencies at pH ) 3 (no pH adjustment). Collection efficiencies at pH ) 3 and 5 were 0.98 ( 0.05 and 0.99 ( 0.03 for glycolaldehyde, 0.96 ( 0.08 and 1.00 ( 0.00 for hydroxyacetone, 1.00 ( 0.00 and 1.00 ( 0.00 for glyoxal, and 0.68 ( 0.17 and 0.79 ( 0.19 for methylglyoxal. Collection efficiencies for methylglyoxal appeared to be higher at pH ) 5, but again, a two-sample t-test showed no significant differences (p ) 0.10). Overall, the results of experiments conducted in the laboratory and in the ambient environment indicate that derivatization of the analytes does not occur at a fast enough rate to significantly increase dissolution of the analytes, and thus, the CE, at a pH between 1 and 7. Accordingly, we chose to collect subsequent samples at pH ) 3. Effect of Temperature. For most organic compounds, decreasing the temperature will increase the Henry’s law constant (M atm-1), as demonstrated by eq 3, where T1 and T2 are the original and modified collecting solution temperatures, respectively, and ∆KH is the reaction enthalpy, which is negative for most compounds.

1804

collection efficiency at 11 °C (284 K) mean ( SD

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 8, 2002

the theoretical and empirical collection efficiencies. According to the theoretical calculations, the greatest difference in CE as a result of a decrease in temperature should be for formaldehyde (+0.05), followed by methyl vinyl ketone (+0.04). The CE for formaldehyde did, in fact, increase significantly with a decrease in temperature. The CE for methyl vinyl ketone also appeared to increase with a decrease in temperature, but the increase was not statistically significant because of poor precision. While a decrease in the temperature of the collecting solution indicated differences in the CE, the differences were small, and submersing the chambers in an ice bath in the field is cumbersome. We thus chose to collect field samples at ambient temperature. Effect of Addition of Tetraalkylammonium Salts. Dissolution of an organic compound in water involves (1) the breaking of interactions between organic molecules, (2) the breaking of interactions between water molecules, and (3) the formation of interactions between the organic molecule and water molecules. Because breaking the organic-organic and water-water interactions generally requires more energy than the energy released by formation of organic-water interactions, organic compounds are often only slightly soluble in water. The addition of a large organic salt, such as a tetraalkylammonium salt, to water increases the structure of the water molecules and, thus, the strength of the waterwater interactions. However, the large alkyl portion of the salt will interact favorably with the organic compound. This process is exothermic and can offset the endothermic processes of breaking organic-organic and water-water interactions, causing the organic compound to become more soluble (47, 48). The phenomenon described is known as “salting in” and has been demonstrated for alkanes and alcohols in the presence of tetraalkylammonium salts (49). To our knowledge, data on the effect of tetraalkylammonium salts on the solubility of carbonyl, dicarbonyl, hydroxycarbonyl, and keto-acid compounds is not available. We hypothesized that tetraalkylammonium salts would increase the solubility of carbonyls and, thus, investigated the effect of adding tetraalkylammonium salts to the collecting solution in the mist chamber on CE. We measured the CE after the addition of 0.25 molal tetramethylammonium bromide, 0.25 molal tetrapropylammonium bromide, or no salt (control) to the collecting solution. The results are presented in Figure 3 (n ) 2) as a plot of the CE versus each analyte for the three conditions employed. No significant differences (p ) 0.10) in CE were detected for methacrolein, methyl vinyl ketone, acetone, hydroxyacetone, formaldehyde, pyruvic acid, and glycolaldehyde by analysis of variance (ANOVA). For glyoxal and methylglyoxal, the collection efficiencies in tetrapropylammonium bromide were significantly lower than in tetramethylammonium bromide, which, in turn, were significantly lower than in the

FIGURE 3. Collection efficiencies for analytes collected from a Tedlar bag with quaternary ammonium salts added to the aqueous PFBHA solution in the mist chamber (n ) 2). control (p ) 0.10). The reason for the latter result is unclear. However, glyoxal and methylglyoxal owe their high watersolubilities to the formation of gem-diols (50). Interaction of glyoxal and methylglyoxal with the alkyl portion of the tetraalkylammonium salts may inhibit formation of the gemdiol and, thus, decrease the solubility. In addition, glyoxal and methylglyoxal would be expected to interact more strongly with the larger propyl group of tetrapropylammonium bromide, which would inhibit gem-diol formation to a greater extent than with tetramethylammonium bromide. In summary, the collection efficiencies for the carbonyls, dicarbonyls, hydroxycarbonyls, and keto-acids in the mist chamber were measured by altering the pH and temperature and by adding tetraalkylammonium salts to the collecting solution. Differences in collection efficiencies were observed for certain compounds, but these differences were small and do not warrant altering the pH or temperature or adding salts. Interferences from Ozone. Previous work indicates that atmospheric ozone causes interferences in the measurement of carbonyl compounds (20). Two types of interference are (1) depletion of the carbonyls or carbonyl derivatives (negative interference) and (2) formation of carbonyl derivatives (positive interference) (53). Depletion of carbonyl derivatives occurs through aqueous ozonolysis of the derivatives or native carbonyls in the collecting solution. Positive interferences result from the reaction of ozone with other unsaturated compounds, such as terpenes, isoprene, and unsaturated carbonyls and alcohols in the collecting solution. We addressed these interferences by conducting theoretical calculations and experiments in the ambient environment. Theoretical Calculations. Using the assumptions and calculations discussed in the Experimental section, theoretical calculations of negative interferences, we calculated that at the detection limit of the analyte, aqueous ozonolysis can destroy e1% of the analyte. Aqueous ozonolysis of the analytes is, thus, not expected to cause significant depletion of the analytes of interest at the Blodgett Forest. Using assumptions and calculations explained in the Experimental section “theoretical calculations of positive interference”, we calculated that 86 pptv (back-calculatedto-mixing ratio in air) of the analyte could be formed as a

result of aqueous ozonolysis of the parent hydrocarbons (isoprene, methacrolein, methyl vinyl ketone, and MBO) in the Blodgett Forest. Because the effect of an interferent will be greatest at low concentrations, we compared the potential concentration of the analyte formed by aqueous ozonolysis to the detection limit of each analyte. Eighty-six parts per trillion by volume represents approximately 0.1x, 2x, 5x, 30x, and 10x of the mixing ratios of formaldehyde, glycolaldehyde, hydroxyacetone, glyoxal, and methylglyoxal, respectively, at their detection limits. The potential thus exists that a positive interference can arise from aqueous ozonolysis of precursor compounds. Air Sampling in the Absence and Presence of a KI-Coated Stainless Steel Annular Denuder. To determine whether artifacts are formed during sample collection in the Blodgett Forest, we compared the concentration of analytes measured when sampling air in the absence or presence of a KI-coated annular denuder. The results are presented in Table 4. Glycolaldehyde, hydroxyacetone, glyoxal, and methylglyoxal were detected in all samples. Formaldehyde and pyruvic acid were not quantified because of high levels of contamination in the field blanks. Data analysis using a paired-sample t-test found no significant differences (p ) 0.10) between the concentrations of glycolaldehyde and hydroxyacetone in the presence and absence of the KI-coated annular denuder. However, methylglyoxal and glyoxal concentrations were 36% and 65% lower in samples collected in the presence of the KI-coated annular denuder versus the concentrations collected in the absence of the denuder. These lower concentrations could be due to sorption of the compounds onto the denuder or to formation of glyoxal and methylglyoxal through aqueous ozonolysis of precursors in the absence of the KI-coated annular denuder. Sorption of Analytes to the Stainless Steel Annular Denuder. To determine whether sorption to the stainless steel annular denuder occurs, we collected samples from a Tedlar bag into a mist chamber fitted with a stainless steel annular denuder at the inlet. We then measured carbonyl concentrations in the collecting solution and sorbed to the denuder. We also measured recoveries from denuders enriched with carbonyls (spiked denuders). We obtained recoveries of 114% ( 5%, 147% ( 11%, 128% ( 11%, 91% ( 3%, 89% ( 1%, and VOL. 36, NO. 8, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1805

TABLE 4. Concentration of Analytes Collected in the Absence and Presence of a KI-Coated Stainless Steel Annular Denuder mixing ratio (pptv)

analyte glycolaldehyde

hydroxyacetone

glyoxal

methylglyoxal

KI-coated KI-coated annular annular % denuder denuder relative absent present difference mean ( SD 1062 1397 1930 1552 1264 1277 498 945 1481 1048 879 879 32 59 24 20 24 167 232 297 265 247 390

1166 804 1757 1487 1700 1278 697 672 1450 1021 1142 890 4 16 17 5 9 75 173 240 165 194 182

+10 -42 -9 -4 +34 0 +40 -29 -2 -3 +30 +1 -88 -73 -29 -75 -63 -55 -25 -19 -38 -21 -53

-2 ( 25

+6 ( 25

-66 ( 22

-35 ( 16

139% ( 5% for formaldehyde, glycolaldehyde, hydroxyacetone, glyoxal, methylglyoxal, and pyruvic acid, respectively (n ) 2), from the denuders, thereby demonstrating our ability to quantify analytes sorbed on the denuder. In the sample extracts, we measured 4.3% ( 1.7%, 1.6% ( 0.4%, 55% ( 13%, 32% ( 6%, and 17% ( 5% of the total concentration (∑ of the concentration on the denuder and in the collecting solution in the mist chamber) of hydroxyacetone, glycolaldehyde, pyruvic acid, glyoxal, and methylglyoxal, respectively, on the stainless steel annular denuder (n ) 2). These data demonstrate that glyoxal, methylglyoxal, and pyruvic acid sorb onto the stainless steel annular denuder during sample collection. Therefore, the differences between the ambient air concentrations measured with and without a denuder, discussed previously, were likely due to sorption of the compounds onto the denuder rather that to aqueous ozonolysis in the absence of the denuder. We thus chose to collect subsequent air samples without ozone removal. Application of the Method to Measurement of MultiFunctional Carbonyls in the Blodgett Forest. Glycolaldehyde, hydroxyacetone, glyoxal, and methylglyoxal were identified and quantified in air samples collected in the

FIGURE 4. Mixing ratios of isoprene and its photooxidation products, glycolaldehyde, hydroxyacetone, methyglyoxal, and glyoxal at the Blodgett Forest Research Station, CA. Blodgett Forest. Interpretation of the EI, methane CI, and PFBOH CI ion trap mass spectra led additionally to tentative identification of a C3-dicarbonyl (not methylglyoxal), a saturated C3-dioxo-acid, a saturated C4-hydroxycarbonyl, an unsaturated C4-hydroxycarbonyl, a saturated C5-dicarbonyl, an unsaturated C5-dicarbonyl, and a saturated C5-dihydroxycarbonyl. The identification of these compounds was not confirmed. The accuracy and precision of the measurements as indicated by the recovery of the compounds in the field spikes, the CE, and a comparison of the values obtained in this study to other studies are presented in Table 5. Overall, the accuracy and precision were good, as demonstrated by recoveries of field spikes g80% (n ) 8) and deviation between duplicate measurements e22% (n ) 2). The mean collection efficiencies were similar to, if not somewhat higher than, the mean collection efficiencies for all control samples collected in laboratory studies (0.82 ( 0.20, 0.80 ( 0.13, 0.88 ( 0.15, and 0.73 ( 0.14 for glycolaldehyde, hydroxyacetone, glyoxal, and methylglyoxal in laboratory studies, respectively (n ) 11)). The detection limits for glycolaldehyde, glyoxal, and methylglyoxal were similar to detection limits obtained from methods utilizing C18-cartridges or a coil scrubber for collection, with DNPH derivatization of carbonyls (12, 15, 16), and to detection limits obtained utilizing 400-mL impingers, with PFBHA derivatization of carbonyls (20). However, the detection limit for hydroxyacetone obtained here is more than an order of magnitude lower than that obtained by Klotz et al. utilizing a coil scrubber and NaHSO3adduct formation (14). Most importantly, the sample collection time used in this research is much shorter than those used for impinger or cartridge methods (15, 20). The range of mixing ratios measured in this work fall within the ranges measured in previous studies (12-14, 16, 20). The sample data are plotted, along with isoprene mixing ratios (Schade, G., personal communication), in Figure 4 as

TABLE 5. Statistics for Samples Collected at the Blodgett Forest and Comparison of the Limits of Detection and Mixing Ratio Ranges to Literature Values

analyte

field spike recoverya

collection efficiency (mean ( SD)b

deviation between duplicates

glycolaldehyde

86% ( 17%

0.94 ( 0.05

22%

0.048 [10]

hydroxyacetone

98% ( 16%

0.86 ( 0.05

8%

0.015 [10]

glyoxal

80% ( 9%

0.86 ( 0.20

15%

0.0027 [10]

methylglyoxal

92% ( 10%

0.83 ( 0.08

8%

0.0077 [10]

a

LOD (ppbv) [sampling time (min)] this work literature values 0.020 [20]c 0.067 [180-240]d 0.20 [180-240]d 0.40 [15]e 0.020 [20]c 0.020 [60-120]f 0.020 [20]c 0.050 [60-120]f

range of mixing ratios measured (ppbv) this work literature values 0.52-1.9 0.27-1.5 0.020-0.054 0.069-0.39