Environ. Sci. Technol. 2003, 37, 2889-2897
Airborne Haloacetic Acids J O N A T H A N W . M A R T I N , * ,†,‡ SCOTT A. MABURY,§ C H A R L E S S . W O N G , §,| FRANCIS NOVENTA,† KEITH R. SOLOMON,† MEHRAN ALAEE,⊥ AND DEREK C. G. MUIR⊥ Department of Environmental Biology, University of Guelph, Bovey Building, Guelph, Ontario, Canada N1G 2W1, Department of Chemistry, University of Toronto, 80 St. George Street, Toronto, Ontario, Canada M5S 3H6, and National Water Research Institute, Environment Canada, 867 Lakeshore Road, Burlington, Ontario, Canada L7R 4A6
Haloacetic acid (HAA) concentrations were measured in air samples from a semi-rural and a highly urbanized site in southern Ontario throughout 2000 to investigate their sources and gas-particle partitioning behavior. Denuders were efficient for collection of gaseous HAAs, and the particle phase was collected on a downstream quartz filter with negligible breakthrough. Total HAA concentrations (i.e., gas + particles) ranged between 0.98). Quantitation was performed by point calibration with the spiked water sample producing the closest response to the sample and by blank subtraction. Recovery of HAAs from spiked water and from spiked quartz filters was quantitative for all HAAs (>80%) as previously reported (3, 42). A blank response was nearly always present for TFA and MCA (∼95% of analyses) and to a lesser extent DCA and TCA (∼50% of analyses). Very low concentrations of MFA, DFA, and CDFA were detected in blanks in less than 5% of analyses. The method detection limits (MDLs) ranged from 1.2 to 100 pg m-3, based on the average blank concentration plus three standard deviations (Table 1). Additionally, because blank response varied from day to day and actually decreased throughout the year, environmental concentrations were not reported unless the response was twice as high as the associated blank response from the same day of analysis.
TABLE 1. Optimized Nonresonant Collision-Induced Dissociation (CID) Amplitude (V), and Excitation Storage Level (CID rf) for Analysis by MS/MS, Ion Storage Windows for Analysis by Selected Ion Storage (SIS), and Scan Range for Both Modes of Analysisa HAA
optimized MS parameter
MS/MS
parent/ molecular ion
CID amplitude (V)
CID rf ( m / z)
scan range (m/z)
instrumental detection limit (fg)b
method detection limit (pg m-3)
TFA DFA MFA CDFA PFPA
225 207 189 241 275
72 68 189 241 275
91 91 92 91 91
125-160 125-160 125-160 125-160 125-160
5.1 11 8 15
99 4.6 1.2 5.4
HAA
optimized MS parameter
SIS
molecular ion
ion storage window (m/z)
scan range (m/z)
instrumental detection limit (fg)b
method detection limit (pg m-3)
MCA DCA TCA
205 239 275
100-102, 128-130, 204-208 127-129, 155-157, 238-242 127-129, 155-157, 272-276
100-210 125-245 120-280
30 36 52
100 50 46
a The instrumental detection limits and method detection limits are also shown. b Based on the injected mass producing a peak having a S/N of 3.
TABLE 2. Results of Gaseous Recovery Experiments (n ) 3) Showing Percent Recovery ((1 RSD (%)) for HAAs on First and Second Denuder denuder 1 denuder 2 total
TFA
DFA
MFA
CDFA
MCA
DCA
91 ( 5.2 5.7 ( 13 97
90 ( 2.1 8.3 ( 20 99
94 ( 4.2 9.0 ( 18 103
95 ( 7.1 6.3 ( 9.7 101
82 ( 8.7 13 ( 25 95
98 ( 5.6 3.1 ( 12 101
Lower responses were reported as nondetectable (nd) and were arbitrarily assigned a value of half the MDL for comparative and trend analysis. Multiple Linear Regression Analysis. At the Guelph site, meteorological data was collected throughout the year from a weather station situated 200 m from the sampler. Hourly mean relative humidity, temperature, short-wave radiation, and precipitation data were averaged over a 96-h period that included the 48-h sampling interval as well as the 48-h period prior to sampling. Multiple linear regressions (The SAS System 8.01, Cary, NC) were used to investigate the influence of PM10 and meteorological variables on total, gas, and particle HAA concentrations. All independent variables were included in the initial model and were removed by backward stepwise elimination if their partial F statistic was not significant (R ) 0.05), and the model was re-run until each remaining variable met the criteria. Relative humidity was never a significant variable, and in several instances the model was reduced to a simple linear model. Gaseous Recovery Test. To determine the gaseous recovery efficiency of HAAs on alkalized annular denuders, recovery tests were performed in the field. The same sampling apparatus was used, except that three denuders were used in series rather than two and a glass elbow was installed upstream of the inlet to accommodate introduction of the test compounds. The sampling pump was turned on and 50 µL of acetone, containing approximately 500 ng of each HAA, was immediately delivered to the glass elbow. Acetone was used as a solvent because a previous attempt to deliver HAAs in water yielded no volatilization. The sample was drawn for 1 h, and the denuders and glass elbow were capped and stored until analysis. Each denuder was analyzed individually, and analysis of the glass elbow extract was controlled for unvolatilized HAAs. This procedure was repeated three times, and the mean recovery on each denuder section was
determined relative to the acetone spiked directly into 25 mL of water.
Results and Discussion GC-MS(/MS) Analysis. The absolute detection limits ranged from 5 to 52 fg of the acid per injection using the parameters outlined in Table 1. Chlorinated HAAs were less sensitive than the fluorinated HAAs because there was more noise under SIS conditions than in the MS/MS operation (Table 1). In SIS mode, sensitivity decreased with increasing degree of chlorination among chlorinated HAAs because of decreased molecular ion abundance. Sample anilide mass spectra are displayed in Scott and Alaee (42). Gaseous Recovery. The recovery of all HAAs exceeded 80% on the first denuder, indicating quantitative gaseous collection efficiency, and the low standard deviation indicated that method precision was adequate (Table 2). Trichloroacetic acid recovery is not reported because it reacted readily with acetone to produce a stable ester derivative, CCl3C(O)OCH2CH3. The second denuder revealed traces ( 0.5). This may be partially attributable to the long sampling time, the associated variation in temperature, relative humidity, aerosol properties (34), and the different rates at which gases equilibrate with particles on hot and cold days (49). Aerosol acidity can affect the partitioning behavior of HNO3, particularly in rural locations where acidity is highest (50), and as such this is likely an important confounding variable for HAA partitioning behavior that was not examined.
The temporal Kp trend and the increased φ with increasing degree of chlorination among chlorinated HAAs are both indications that HAA gas-particle partitioning is controlled in part by vapor pressure. A complication is that HAAs may be interacting with particles in more than one manner, but Kp and φ are only appropriate estimates of partitioning behavior if HAAs are adsorbing to particle surfaces as protonated acids, otherwise both parameters may be overpredicted. Alternatively, HAAs may be present on particles as anions, through salt formation with particle bases, or by uptake into the adsorbed water content of particles. Uptake VOL. 37, NO. 13, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2893
FIGURE 3. Mean gas-particle partition coefficient (Kp) for each month in 2000 for TFA, TCA, MCA, and DCA (n ) 47). into particle-associated water is quite possible given that particle nitrate concentrations increase with increased water absorption (51) and because the very low KH (12, 13) and high mass accommodation coefficients for HAAs (52) favor uptake by atmospheric water. Once the anion is formed in aqueous solution, it is not likely to revolatilize as the acid, as illustrated by our failed denuder recovery trials using HAAs spiked into water. In such a case, the particle-bound HAAs may be considered nonexchangeable, the influence of which is described by Pankow (53). If HAA partitioning was purely a function of volatility, then the φ for TFA would be expected to be much less than that for MCA; however, the opposite was observed, suggesting the greater acid strength of TFA affords additional particle adsorption. By the same argument, the increasing trend of φ with increasing degree of chlorination among MCA, DCA, and TCA may be a result of increased acid strength in addition to decreased vapor pressure. Overall, the data suggest that both acid strength and vapor pressure are important factors in determining HAA gas-particle partitioning behavior. Source Discussion. The observations of elevated total TFA, DFA, MFA, and TCA concentrations in the urban environment of Toronto (Figure 1) are highly suggestive of contributions from local urban sources. On average, the TFA concentrations in Toronto were more than twice as high as the concentrations observed in Guelph for the same sampling period, exceeding the Guelph concentration by 1.2 ng m-3. Urban enrichment of TFA in atmospheric deposition has previously been shown to be as high as 6-fold surrounding the San Francisco area (54). Ellis et al. (7) have attributed some of this enrichment to the thermolysis of poly(tetrafluoroethylene); however, additional alternative sources may also contribute. Interestingly, DFA and MFA were also identified as minor thermolysis products of poly(tetrafluoroethylene) (7, 9). The mere presence of MFA and DFA in the atmosphere indicates that fluoropolymer thermolysis may indeed be contributing to atmospheric HAAs, and this argument is further supported by the significantly higher concentrations in urban samples relative to the semi-rural site. These measurements represent the first quantitative measurements of MFA or DFA in atmospheric samples, and currently there are no other proposed atmospheric sources. Among several HAAs, strong associations (p < 0.0005) were observed between their total concentrations in indi2894
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 13, 2003
vidual samples at Guelph. Total TFA concentrations were associated with MCA, DCA, and DFA, while the strongest relationship was observed between MCA and DCA (r 2 ) 0.48). It followed that total DFA was also strongly associated with MCA and DCA concentrations (r 2 ) 0.39 and 0.40, respectively). Strengthening the argument for PTFE thermolysis as a source of fluorinated acids, the source of DFA and MFA may be similar based on their strong association (r 2 ) 0.37, p < 0.00001). A weak association was also observed between TFA and MFA (r 2 ) 0.11, p ) 0.026). For MCA and DCA and for TFA and DFA, associations suggest either that they are evolved from the same organohalogen precursor or that their formation in the atmosphere is controlled by the same atmospheric conditions. To gain further insight into the mechanism(s) of production and sources of individual HAAs, multiple linear regressions were applied to meteorological data and gas, particle, and total HAA concentrations in Guelph (Table 3). It was hypothesized that the amount of precipitation previous to and throughout the sampling event may cause a decrease in airborne HAA concentrations through purging of the gas and particle phases. However, precipitation was not a useful predictor of airborne HAAs in our study, nor was it a useful predictor of wet HAA deposition in a previous report (14). Furthermore, because of the high water solubility of HAAs, it was hypothesized that higher relative humidity, leading to higher water content of aerosols, would favor partitioning into the particle phase; however, no statistically significant associations between particle concentrations and relative humidity were revealed. Associations may be shrouded by the relatively long sampling time and the associated variation in temperature, relative humidity, and aerosol properties. Gas-phase TFA, DFA, MFA, and MCA concentrations were statistically associated with and directly proportional to PM10 (Table 3). Higher particle loading may be indicative of air masses originating from urban environments (i.e., primary aerosols). For example, average PM10 was 2.7 times greater in Toronto than in Guelph, indicating that this parameter may be a good tracer of air masses originating from urban environments. The possibility that particle-phase HAAs are lost to the denuder surface as they pass through is unlikely given the results of Peters et al. (55). Particle DCA concentrations decreased with increasing temperature, and the significant negative interaction with PM10 and temperature
TABLE 3. Parameter Estimates for Temperature, Short-Wave Radiation (SWR), Particle Mass (PM10), and Y Intercept for Simple or Multiple Linear Regression Models Applied to Gas, Particle, or Total HAA Concentrations in Guelph (n ) 47)a independent variable coefficient (p value)
dependent variable phase/HAA gas log TFA log DFA log MFA log CDFA log MCA log DCA particle log MCA log DCA total log TFA log MCA log DCA a
temp (°C)
SWR (kJ m-2 × 106)
0.0193 (0.0006)
regression parameters interaction term
240 (0.0082) 282 (0.01) 452 (0.001) 1120 (
