Environ. Sci. Technol. 2004, 38, 5064-5072
Formation of Secondary Organic Aerosol by Reactive Condensation of Furandiones, Aldehydes, and Water Vapor onto Inorganic Aerosol Seed Particles CHARLES A. KOEHLER, JEREMIAH D. FILLO, KYLE A. RIES, J O S EÄ T . S A N C H E Z , A N D DAVID O. DE HAAN* Department of Chemistry, University of San Diego, 5998 Alcala Park, San Diego, California 92110
Volatile furandiones and aldehydes are significant atmospheric oxidation products of aromatic compounds. The mechanism of secondary organic aerosol formation by these compounds was probed using particle chamber observations and macroscale simulations of condensed phases. Growth of inorganic seed aerosol was monitored in the presence of humidity and high concentrations of 2,5furandione (maleic anhydride), 3-methyl-2,5-furandione (citraconic anhydride), benzaldehyde, and trans-cinnamaldehyde. Particle growth commenced when the gasphase saturation level of each organic compound and water vapor (relative to its pure liquid), when summed together, reached a threshold near one, implying the formation of a nearly ideal mixed organic/aqueous phase. However, these organics are immiscible with water at the high mole fractions that would be expected in such a phase. Highly acidic dicarboxylic acids produced by the reactions between furandiones and water were shown to rapidly acidify an aqueous phase, resulting in greatly increased benzaldehyde solubility. Thus, the uptake of these organics onto particles in the presence of humidity appears to be reaction-dependent. Finally, it is shown that dicarboxylic acids produced in these reactions recyclize back to furandiones when subjected to normal GC injector temperatures, which could cause large artifacts in gas/particle phase distribution measurements.
Introduction Secondary organic aerosol (SOA) is recognized as a significant component of urban haze (1-4) and as a major source of uncertainty in global climate change through the indirect aerosol effect on cloud formation (5). A correlation between fine aerosol (PM2.5) levels and excess mortality rates has also been widely observed (6-8), although a general mechanism of particulate lung damage that could explain the correlation has yet not been identified (9, 10). If fine aerosol particles are indeed adversely affecting human health, then SOA, as an important type of fine aerosol, is also implicated. A significant research effort is being directed at understanding and predicting the partitioning of secondary organic * Corresponding author phone: (619)260-6882; fax: (619)260-2211; e-mail:
[email protected]. 5064
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004
compounds into the particulate phase. Studies on SOA formation (11-18) have used absorptive partitioning theory (19), where individual compounds equilibrate between the gas phase and a liquid organic matter (om) phase, with reasonable success. In both smog chamber and field studies, good agreement between theoretical and measured partitioning of identified compounds has been found for compounds with vapor pressures below 10-3 Torr (20, 21). For low-polarity compounds such as substituted phenols and PAHs, the good agreement extends at least up to compounds with vapor pressures of 1 Torr (13, 22, 23). However, for more polar organic compounds with acid, ketone, and/or aldehyde functional groups, an apparent theoretical underprediction of condensed phase concentrations increases with vapor pressure and reaches a factor of 105 for compounds with vapor pressures near 1 Torr (20, 21). While denuder sampling artifacts have been invoked to explain this disagreement between theory and measurement (20, 24), such artifacts should depend primarily on vapor pressure and not so strongly on compound polarity or functional groups present. It is likely that misidentification of condensed organic material, due to reactions that occur in the particle phase or during the analytical process, is contributing to the apparent discrepancy. If an organic compound converts to a form with lower vapor pressure when exposed to aerosol or if a compound in the aerosol phase reacts to form a volatile product during extraction and chemical analysis, partitioning theory will be unable to describe the observed particle-phase concentrations for that compound until its condensed-phase form is correctly identified. As an example of this type of mechanism, it has recently been proposed that carbonyl-containing compounds undergo acid-catalyzed particle-phase reactions to form hydrates and polymers in the presence of water and to form hemiacetals and acetals in the presence of alcohols (24-27). It has also been shown that aldehydes and hydroperoxides can form peroxyhemiacetals via gas-phase reactions (28). Such reaction products would have reduced vapor pressures relative to the reactants, and the reactions are reversed during GC analysis (28). These reactions would rationalize the apparent theoretical underprediction of om-phase concentrations of volatile carbonyl-containing compounds if such compounds are actually present in the condensed phase as hydrate, hemiacetal, acetal, or peroxyhemiacetal forms. Another class of secondary organic compounds that could be involved in seemingly anomalous SOA formation is furandiones. These compounds form from butenedial (and derivatives), which are themselves ring-cleavage products formed during atmospheric oxidation of benzene, toluene, and xylenes (29). Because of their stability, furandiones have been observed in the gas phase (at molar yields of 4%) during smog chamber experiments on benzene and toluene (29) and at even higher yields (up to 30%) during the hightemperature surface-catalyzed oxidation of isoprene (30). Using six different simple aromatic compounds as reactants, Forstner et al (31) reported that 37-67% of identified condensed-phase material (or up to 20% of the total GCelutable particulate mass) was attributed to five specific furandiones: 2,5-furandione (maleic anhydride, MA), 3,4dimethylfurandione, 3-ethyl-2,5-furandione, 3-methyl-2,5furandione (citraconic anhydride, CA), and dihydro-2,5furandione (succinic anhydride). These compounds were extracted from particles using supercritical fluids and analyzed by GC-MS. Later studies of aromatic oxidation products using functional group derivatization followed by 10.1021/es034672b CCC: $27.50
2004 American Chemical Society Published on Web 09/02/2004
TABLE 1. Sample Experimental Conditions in Aerosol Chamber Experimentsa gas-phase concn (mg/m3) (ppmv) (Torr)
expt
time (s)
chamber vol (L)
% RH
seed type
organics present
vol added (µL)
mass added (mg)
1
3 000b
122
40
NaCl
BZ CA MA
167 0 na
174.5 0 7.9
796.0 0.0 35.6
311.3 0.0 15.1
2
5 400b
117
40
AS
BZ CA MA
271 0 na
283.2 0 8.9
1297.2 0.0 41.0
3
19 000b
168
29
NaCl
BZ CA Cinn
288 32.2 4
301.0 40.2 4.2
4
8 000c
100
34
none
CA Cinn
64.8 12.2
5
max.d
53
18 MΩ). The humidified gas was filtered with a 0.2 µm membrane filter to remove any particles produced from the bubbler. Since the entire experimental system was at room temperature, relative humidity in the chamber is equal (within (10%) to the ratio of gas volume added through the bubbler and the total volume of the chamber (36), as verified by hygrometer measurements (Fisherbrand, NIST-traceable calibration). Condensed-phase chemistry was examined on a macroscopic scale by attenuated total reflectance (ATR)-FTIR with diamond crystals (JASCO model 480 FTIR, SpecAc Golden Gate single reflection ATR, SensIR double reflection ATR). Each FTIR experiment began with a 3-mm-thick water layer held inside a small O-ring fitting on the ATR crystal. Organic liquids (benzaldehyde, citraconic anhydride) and solids (maleic anhydride) were added in milligram quantities on VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5065
top of the water layer as data acquisition began. In some experiments, a small pH microprobe (IQ Scientific Instruments) was inserted into the aqueous layer. Organics in the aqueous phase were quantified by scaled, full spectral subtraction of aqueous phase or neat standards until absorbance features were minimized. Furandiones and their corresponding dicarboxylic acidss cis-butenedioic acid (maleic acid, Matheson) and cis-methylbutenedioic acid (citraconic acid, Acros)swere analyzed by 1 H NMR (Varian, 350 MHz) and by GC-MS (Hewlett-Packard 5890/5823) in order to determine purity and chemical reactivity during analysis. The reactivity of related dicarboxylic acidss1,4-butanedioic acid (succinic acid), 1,5-pentanedioic acid (glutaric acid), 1,6-hexanedioic acid (adipic acid), (all obtained from Sigma-Aldrich), and trans-butenedioic acid (fumaric acid, J. T. Baker)swas also determined by GC analysis. Chromatographic separation was achieved by oven temperature programming from 35 to 250 °C (Alltech EC-5 column, 30 m × 0.25 mm i.d., 25 µm film, detector temperature ) 280 °C).
Results and Discussion Aerosol Chamber Observations of Particle Growth. Experimental conditions for five aerosol chamber experiments are summarized in Table 1. Because chamber volume and gas-phase concentrations vary constantly during each experiment due to gas additions and sampling, conditions given are valid only at the specific times listed. These times are selected to correspond with significant events in the chamber, such as the onset of particle growth or a particle nucleation event. The experiments are analyzed individually below. Data from experiment 1, where a humidified, monodisperse population of sodium chloride seed particles was exposed to benzaldehyde, maleic anhydride, and citraconic anhydride, is shown in Figure 1. The simultaneous increases in particle volume and geometric mean diameter that begin at approximately t ) 3000 s are a clear indicator of particle growth due to gas uptake. Previous work on monodisperse particle populations (34) showed that geometric mean particle diameters are stable to within a few nanometers in the chamber for a period of hours. Evidently, coagulation is limited by the low number densities that result from SMPS particle size selection on the way into the chamber. Furthermore, since only a narrow size range of particles is present, all particles are lost to the chamber walls at rates similar to the total decline in number density observed in Figure 1. Thus, wall losses and coagulation do not significantly affect measures of average particle size under these conditions, and the increase in geometric mean diameter in Figure 1 must be attributed to particle growth by condensation of material from the gas phase. Similarly, total particle volumes normally decline continuously due to wall losses, at rates similar to declines in number density, and the observed increase in Figure 1 is only attributable to particle growth by condensation. Partitioning theory predicts particle growth by condensation when the condition is met that
Si
∑ζ g 1 i
(2)
i
where the summation is performed over each organic compound present (12, 34, 37). If the condensed phase behaves as an ideal mixture, all activity coefficients ζi ) 1 and the condensation threshold simplifies to (34)
Sorg )
∑S g 1 i
(3)
i
5066
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004
FIGURE 1. Experiment 1: total particle volumes (scaled up by a factor of 100, in µm3/cm3), geometric mean diameters (in nm), and number densities (reduced by a factor of 10, in cm-3) of monodisperse sodium chloride seed particles in the presence of water vapor during the addition of 350 µL of benzaldehyde (BZ), 19.4 mg of maleic anhydride (MA), and 90 µL of citraconic anhydride (CA). Addition times are indicated by horizontal arrows and vertical dotted lines. Calculated total organic saturation levels, Sorg (open circles connected by a line), were corrected for dilution. The calculation assumed constant organic evaporation rates and resulted in total uncertainties of (10%. Relative humidity levels (heavy solid line) were calculated based on total volumes of wet and dry nitrogen added to particle chamber, corrected for dilution, and are accurate to within (10%. where the total organic saturation (Sorg) is the sum of the gas-phase saturation level (relative to the pure liquid) of each organic compound i present. If, on the other hand, activity coefficients in the condensed phase average less than one, eq 2 could be satisfied even though eq 3 is not, and condensation may be observed before Sorg reaches 1. Prior work at low humidities has shown that benzaldehyde and furandiones condense only when Sorg reaches 1.0, which implies that the organics mix in a nearly ideal fashion (34). In Figure 1, the particle growth observed beginning at approximately 3000 s clearly precedes the point where Sorg reaches 1 (at t ) 11 000 s). Thus, the condensed phase must contain more than the three organic compounds present in the chamber. If the organics and water vapor condensed together to form a mixed organic/aqueous phase, the relative humidity (Swater) could be meaningfully added to the sum Sorg to generate a combined organic/water saturation level (Stot):
Stot ) Sorg + Swater
(4)
Formation of an ideal mixed organic/aqueous phase would then be predicted to occur when Stot reaches 1. It is seen in Figure 1 that Stot reaches 1 (within experimental error) at approximately t ) 5000 s, shortly after the onset of particle growth. Thus, the formation of a mixed organic/aqueous particulate phase is the likely explanation for the observed particle growth. The observation that particle growth begins before Stot reaches 1 suggests that the activity coefficients (including ζwater) in the mixed phase may average slightly less than one. The interactions of monodisperse ammonium sulfate seed particles with humidity, benzaldehyde, maleic anhydride, and citraconic anhydride during experiment 2 are shown in Figure 2. Growth is observed by a simultaneous increase in
FIGURE 2. Experiment 2: average volumes per particle (scaled up by a factor of 104, in µm3), geometric mean diameters (in nm), and number densities (reduced by a factor of 20, in cm-3) of monodisperse ammonium sulfate seed particles in the presence of water vapor during the addition of 355 µL of benzaldehyde (BZ), 11.7 mg of maleic anhydride (MA), and 90 µL of citraconic anhydride (CA). Sorg and relative humidity level calculations include dilution corrections as described in Figure 1.
FIGURE 3. Experiment 3: average volume per particle (in µm3) and mode particle diameter (in nm) of monodisperse sodium chloride seed particles in the presence of humidity as dry nitrogen gas, 288 µL of benzaldehyde (BZ), 52 µL of citraconic anhydride (CA), and 10 µL of cinnamaldehyde (Cinn) were added to the chamber. Sorg and relative humidity level calculations include dilution corrections as described in Figure 1. the average volume per particle and the geometric mean particle size that begins as Stot reaches 1 at t ) 5400 s, As in Figure 1, growth clearly precedes the point where Sorg reaches 1. Data from experiment 3, where humidified monodisperse sodium chloride seed particles were exposed to additions of dry air, benzaldehyde, trans-cinnamaldehyde, and citraconic anhydride is shown in Figure 3. The particles were observed for 3 h before any organic gases were added. During this period, average particle volumes and mode diameters declined by approximately 25% and 7%, respectively, during the first 18 000 s of the experiment. This relative stability in particle size in the face of a decline in relative humidity from 64 to 30% indicates that the aerosol population consists of solid NaCl particles rather than deliquesced liquid droplets.
FIGURE 4. Experiment 4: total particle volumes (scaled up by a factor of 200, in µm3/cm3), geometric mean diameters (in nm), and number densities (reduced by a factor of 10, in cm-3) of particles formed by nucleation in the presence of gas-phase water vapor and during the addition of 85 µL of citraconic anhydride (CA) and 15 µL of trans-cinnamaldehyde (Cinn). Geometric mean diameters before 8000 s are omitted since no particle population was present. Sorg and relative humidity level calculations include dilution corrections as described in Figure 1. Solid NaCl particles carry water as a thin liquid-like surface layer (38, 39), and the loss of most of this water as humidity declines would have only a minimal effect on particle size, as observed. Particle growth, detected as a simultaneous increase in mode diameter (+14 nm) and in average volume per particle (+1.3 × 10-4 µm3), can be observed beginning at approximately 19 000 s during the simultaneous addition of citraconic anhydride and trans-cinnamaldehyde. As in Figure 2, this growth begins when Stot reached 1, within experimental error. Results from experiment 4, where particle nucleation occurred in the presence of water, citraconic anhydride, and cinnamaldehyde, are displayed in Figure 4. The particle-free chamber was subjected to slowly decreasing humidity levels (due to dilution) as the organic compounds were added. The small initial fluctuations in particle volume are due to periodic random laser counts triggered by condensation in the particle counter. (Due to these random counts and the absence of a population of particles in the chamber, before t ) 8000 s the geometric mean diameters were randomly distributed between 25 and 600 nm. This “white noise” is omitted for clarity.) Particle nucleation occurred after t ) 8000 s, as indicated by the simultaneous appearance of continuous, matching trends in particulate volume, number density, and geometric mean diameters. At this point Sorg had reached 1.2, and since relative humidity had declined to 34%, Stot ) 1.5. Under these conditions, the nucleation of organic (waterfree) particles is theoretically possible because Sorg exceeded 1. However, in experiments 5 and 6 (shown in Table 1), where benzaldehyde, citraconic anhydride, maleic anhydride, and trans-cinnamaldehyde were added to particle chambers containing only dry air, no particle nucleation was observed when Sorg reached as high as 1.2. This suggests that water was involved in the nucleation process observed in experiment 4 and that mixed organic/aqueous particles were formed. To summarize, mixed organic/aqueous particulate phases likely formed in experiments 1-4 since the amounts of organic gases present at the onset of growth in each case were not enough to establish an organic phase in the absence of humidity. Furthermore, the condensation threshold near the point where Stot ) 1 in experiments 1-3 implies that the VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5067
TABLE 2. Mole Fractions of Mixtures Simulating Organic/ Aqueous Particulate Phasesa benzcinnam- citraconic maleic liquid expt water aldehyde aldehyde anhydride anhydride phases 1, 2 3 4
0.43 0.27 0.23
0.50 0.36
0.07 0.16 0.44
0.21 0.33
2 2 2
a Mixtures prepared with mole fractions selected by taking approximate organic saturation levels and relative humidity levels from each experiment at the point where Stot ≈ 1 (experiments 1-3) or where nucleation was observed (Figure 4). These values were scaled slightly to add to 1. Mixtures were then observed for phase separation. Maleic anhydride was dissolved in benzaldehyde before water addition.
condensed phase behaves as a nearly ideal mixture, with activity coefficients in the mixed phase near 1. To examine the plausibility of water and the organics involved in this study existing as a nearly ideal mixture, the condensed organic/aqueous phase expected to have formed in each experiment was simulated on a macroscopic scale. Mixtures were made where the mole fraction Χi ) Si for all gases present at the point at which particle growth was observed in each experiment (normalized so that ∑Χi ) 1), as shown in Table 2. Mixtures prepared with these mole fractions were observed for phase separation. In each case, the separation of the mixture into two phases indicates that a mixed aqueous/organic phase is thermodynamically unstable and nonideal. The nonreactive condensation of these organic gases and water vapor into a mixed phase can therefore be ruled out as a plausible explanation for these observations of particle growth. Instead, the condensation threshold may be explained by condensed-phase chemical reactions, as shown below. Experimental Simulations of Particle-Phase Chemistry. One condensed-phase reaction that is known to occur between the compounds present in the chamber is a reaction between water and the furandiones, producing dicarboxylic acids. For example:
This type of reaction was investigated by FTIR-ATR spectroscopy and pH monitoring. Figure 5 displays IR spectral intensity data for maleic acid and citraconic acid produced in a 3-mm-thick aqueous layer, recorded as maleic anhydride or citraconic anhydride was added to the top of the layer. Aqueous activities of H+ ions measured by pH microprobe during maleic anhydride or citraconic anhydride additions are shown in Figure 6. As both figures show, the appearance of maleic acid and citraconic acid can be successfully monitored by the rapid acidification of the aqueous layer or by the appearance of IR spectral features. By either measure the reactions are reasonably fast on this time scale, with the production of maleic acid essentially complete by the first measurement at 1 min. Citraconic acid, the weaker acid of the two, is produced more slowly, which could be due to the limited contact area between the two liquid phases. (Maleic acid is added to the aqueous layer as a powder.) If reaction 5 occurred in a 50 nm diameter particle that initially had 5068
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004
FIGURE 5. Attenuated total reflectance FTIR measurements of maleic acid and citraconic acid produced in 50 µL aqueous layers held inside a stainless steel O-ring fitting with a 4.8 mm i.d. during contact with the corresponding furandione. Open diamonds: spectral intensity of citraconic acid features that appeared as 50 µL of citraconic anhydride was pipetted into fitting, Citraconic acid signals were converted to an aqueous-phase molality scale by comparison to aqueous citraconic acid standards. Closed triangles: spectral intensity of maleic acid features that appeared as 100 mg of maleic anhydride powder was added. Maleic acid signals were also converted to an aqueous-phase molality scale by comparison to aqueous maleic acid standards.
FIGURE 6. H+ activities (in M) measured by pH microprobe in 70 µL water layers held inside O-ring fittings as furandiones were added. Open diamonds: acid produced as citraconic anhydride was pipetted into fitting. Simultaneous FTIR-ATR measurements indicated the organic compound stayed on top of the water layer for the period of measurement shown. Closed triangles: acid produced as maleic anhydride powder was added to fitting. All quantities added were similar to those described in Figure 5. separate organic and aqueous phases, the ratio of interphase contact area to phase volume would be larger than in this macroscopic-scale experiment by a factor of almost 105. Assuming that the reaction occurs at the phase boundary and the processing rate is proportional to this ratio, anhydride/acid conversion within such a particle would be completed within a few milliseconds. The dicarboxylic acids formed by this process have unusually strong acidity. First and second acid dissociation constants, expressed as pKa values, are shown in Table 3. It is seen that the cis-dialkenoic acids (citraconic acid and maleic acid) have lower pKa1 values and higher pKa2 values than other dicarboxylic acids. The strong tendency of these two
TABLE 3. Dicarboxylic Acid Dissociation Constantsa dicarboxylic acid
pKa1
pKa2
adipic acid (1,5-pentanedioic acid) (40) citraconic acid (cis-methylbutenedioic acid) (41) fumaric acid (trans-butenedioic acid) (40) glutaric acid (1,6-hexanedioic acid) (42) maleic acid (cis-butenedioic acid) (40) succinic acid (butanedioic acid) (40)
4.42 2.29 3.05 4.35 1.91 4.21
5.42 6.15 4.49 5.42 6.33 5.64
a First and second acid dissociation constants of relevant dicarboxylic acids, expressed as pKa values, taken from the literature. All values are given at 25 °C.
FIGURE 8. GC-MS peak heights of maleic anhydride, observed in the mass spectrometer at m/z ) 54 beginning at the usual retention time of maleic anhydride produced during injections of maleic acid as the gas chromatograph injection port temperature was varied. No maleic acid was observed to elute. Each point represents a single run. Maleic anhydride standards showed a 50% run-to-run variability, so similar uncertainties may be assumed for these data.
FIGURE 7. Attenuated total reflectance FTIR measurements of benzaldehyde present in 54 µL aqueous layers held in O-ring fittings. Spectral intensity of benzaldehyde features are shown as percentages of the spectral intensity of a neat standard of benzaldehyde. Diamonds: spectral intensity of benzaldehyde features that appeared as 54 µL of benzaldehyde was added to fitting containing the aqueous layer. Triangles: spectral intensity of benzaldehyde features that appeared as 27 µL of benzaldehyde and 27 µL of citraconic anhydride were simultaneously added to the fitting. acids to lose their first proton can be explained by the stabilization of the deprotonated carboxyl group by the hydrogen of the protonated carboxyl group and the steric effect of the cis-double bond holding the two groups in constant proximity. Experiments 4-6, performed in the absence of seed particles, showed that no particle nucleation occurred unless Stot, the sum of relative humidity and organic saturation, exceeded 1.5. This indicates that furandione/water reactions (e.g., reaction 5) are slow enough in the gas phase that the amounts of dicarboxylic acids produced, even after a few hours of reaction time, are insufficient to substantially lower the barrier to nucleation. However, once a mixed organic/ aqueous phase is established on preexisting inorganic seed particles when Stot reaches 1, furandione/water reactions occur rapidly in the condensed phase, stabilizing it against evaporation and phase separation as dicarboxylic acids form and are ionized. Data from two FTIR-ATR experiments with benzaldehyde are summarized in Figure 7. When a sample of benzaldehyde was pipetted on top of an aqueous layer, very little dissolved into the water. However, when benzaldehyde and citraconic anhydride were pipetted simultaneously onto the water layer, the intensity of the benzaldehyde IR spectrum in the water layer increased by an order of magnitude. This result suggests that the mixing of benzaldehyde into a mixed organic/ aqueous aerosol phase is also a reaction-dependent process.
The large increase in aqueous-phase benzaldehyde can be rationalized by a mechanism that begins with the production of citraconic acid by reaction 5, which rapidly acidifies the aqueous layer. The increase in acidity then catalyzes the conversion of benzaldehyde into its hydrate (24-27), which would have a higher solubility in water. More benzaldehyde hydrate, in rapid equilibrium with benzaldehyde, can enter the acidified aqueous phase. The cause of the slow decline in benzaldehyde IR spectral intensity that is observed after 2 min of reaction time has not been identified but may be caused by further acid-catalyzed hydrate reactions such as the formation of hemiacetals or polymerization (25-27). Furandione Formation During Chemical Analysis. Reaction 5 is known to be reversed for maleic anhydride at or above temperatures of 230 °C (43). The interconversion of citraconic acid and citraconic anhydride during GC analysis has also been reported (44). Thus, maleic acid, citraconic acid, and perhaps other dicarboxylic acids produced in the condensed phase may be misidentified as the corresponding anhydrides when analyzed by GC without prior derivatization. The effect was analyzed in a series of GC experiments where the injector port temperature was varied while dicarboxylic acids were injected, as shown in Figures 8-10. Single ion monitoring was used to detect the acids and the appearance of the corresponding anhydrides. Neither maleic acid nor citraconic acid eluted from the GC. Instead, Figure 8 shows that maleic anhydride, detected by its major fragment (loss of CO2) at m/z ) 54, was produced when the injector temperature was 230 °C or higher, matching the reported temperature threshold (43). The reaction was not instantaneous, as evidenced by chromatographic peaks that began at the retention time of maleic anhydride but extended for several minutes. The formation of citraconic anhydride during citraconic acid injections (shown in Figure 9) is much more facile, occurring at temperatures as low as 120 °C. Furthermore, chromatographic peak widths for citraconic anhydride produced in the injection port were as narrow as when citraconic anhydride was injected directly. (For this reason, peak heights in Figure 9 are much larger than in Figure 8.) Other dicarboxylic acids were analyzed to find if they also recyclize into anhydride forms during GC analysis at high temperatures. Other anhydride conversions were much less favorable, because in every case the injected acid was the dominant peak that eluted. The size of the chromatographic VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5069
FIGURE 9. GC-MS peak heights of citraconic anhydride, observed in the mass spectrometer at m/z ) 39, 53, and 68, produced during injections of citraconic acid as the gas chromatograph injection port temperature was varied. No citraconic acid was observed to elute. Each point is an average of 2-3 runs. Error bars are (1σ.
FIGURE 10. GC-MS peak area ratios indicating ease of anhydride formation during injections of various dicarboxylic acids as a function of GC injection port temperature. Each point represents an anhydride peak area divided by the corresponding acid peak area, observed by the mass spectrometer in scanning mode. In every run the dicarboxylic acid that was injected was observed to elute. Each point represents 1-2 runs. peaks due to corresponding anhydrides produced in the injector, expressed as peak area ratios against the acid peaks, are plotted against temperature in Figure 10. The observed trend in anhydride formation (succinic . glutaric . adipic) is consistent with Bruice and Pandit (45), who generalized that each free rotation in the carbon backbone reduced the speed of anhydride formation by a factor of 230 due to the effect of rotamer distributions on the proximity of the two carboxylic acid groups that must react with each other. Fumaric acid, the trans isomer of maleic acid, has no free rotations due to its double bond, and its carboxyl groups are fixed in a nonreactive position. However, the cis and trans isomers interconvert at 300 °C, resulting in the formation of maleic acid. Since the interconversion temperature is above 230 °C, maleic anhydride is then produced. It should also be noted that longer-chain dicarboxylic acids, such as adipic acid, would form anhydrides rings with seven atoms, and thus product stability would be further lowered by ring strain. 5070
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004
It is probable that the high temperatures of supercritical fluid extraction would cause the same reactions to occur, which would result in large negative artifacts for particlephase citraconic acid, maleic acid, and succinic acid and large positive artifacts for the corresponding furandiones in the particle phase. This may explain the identification of high levels of furandiones in the particulate phase in the study that used this extraction method (31) but not in others (24, 32, 33). Potential Atmospheric Significance. It is important to note that the gas-phase benzaldehyde concentrations used in this study are approximately 6 orders of magnitude higher than the average urban atmospheric concentration of benzaldehyde and about 4 orders of magnitude higher than typical total aldehyde concentrations (46, 47). Similar statements can be made about the concentrations of the other three organic gases used in this study. Thus, the process of condensation observed in this study will certainly differ in many regards from atmospheric SOA formation. For example, experiments in this study using seed particles began with desiccated inorganic seed particles, and particles took up organics and water simultaneously only after a sharply defined condensation threshold was exceeded. Atmospheric particles, in contrast, are not expected to have well-defined condensation thresholds because they are believed to contain at least a small quantity of organic material and, because they are mixtures, typically do not have sharp deliquescence points such as pure salts. Thus the condensation thresholds observed in this work are used to identify components and processes in the condensed phase but are not expected to govern furandione condensation in the atmosphere. We can safely conclude that the formation of organic aerosol from very high concentrations of volatile furandiones and aldehydes in the presence of humidity is greatly enhanced by particle-phase reactions. Our measurements indicate that a still unidentified mechanism dependent on acidity increases aqueous-phase benzaldehyde solubility. This suggests the interesting possibility that gas/particle partitioning of aldehydes involves an interaction between inorganic/aqueous and om particulate phases, an area of current interest in modeling SOA formation (48). When volatile furandiones form in the atmosphere during the oxidation of aromatic compounds, their primary contribution to SOA mass will be unlikely to occur through partitioning directly into the particulate om phase but rather by reaction with particle-phase water to form low-volatility dicarboxylic acids. To our knowledge, there are no published measurements of ambient gas-phase furandione levels except in smog chamber experiments. If we assume that the 4% molar yield of maleic anhydride during the oxidation of simple aromatics observed by Bandow et al. (29) is a reasonable estimate of total atmospheric furandione production from these precursors, we can then use emission rates of benzene (49) to estimate an atmospheric furandione production rate of approximately 1 t/day in the Los Angeles air basin. While the atmospheric rates of furandione/water reactions cannot be estimated from this work for reasons described above, we note that the reaction products (dicarboxylic acids) are widely detected in tropospheric aerosol (50-54), appear to be secondary (53, 54) and anthropogenic (51, 52, 55) in origin, and have been observed to be an important component in the particulate phase in urban air masses (55, 56). Although there are certainly other atmospheric reaction pathways to dicarboxylic acids, these facts are at least consistent with the possibility that furandione/water reactions could be a significant source of SOA formation. Any significance of these reactions would be expected to depend on the water content of local aerosol, which is a function of relative humidity and the hygroscopic nature of the inorganic and organic components of the particle. Two
groups have reported that organic aerosol yields in smog chamber experiments using aromatic precursors are not affected by water vapor or by the presence or absence of wet or dry (NH4)2SO4 seed particles (33, 57). This may indicate that the influence of hydrophobic and hydrophilic (or waterreactive) organic oxidation products present in those experiments was approximately equal, causing the net effect of condensed-phase water to be negligible. However, water uptake measurements (57) showed that om phase polarity increases with time, suggesting that the balance tips toward more hydrophilic organic compounds as oxidation proceeds. We predict that if the establishment of organic phases on seed particles occurs predominantly by reactive pathways, then preferential growth of certain types of particles into SOA may occur. Differences in SOA yields that depended on seed particle type have been observed for deliquesced seed particles in the R-pinene/ozone system (17). SOA yields in the isoprene/O3 and acrolein/O3 systems have also been shown to increase in the presence of acidified particles (26). The effect of seed particle type, however, has not been examined for aromatic oxidation systems. Placing emission limits on types of particles that are preferentially activated into SOA is a potential PM2.5 mitigation strategy that should be addressed by further research. (The strategy would have to be considered in addition to existing strategies of limiting volatile organic carbon (VOC) or oxidant levels, since high levels of both make SOA formation inevitable.) Several major questions remain. What fraction of VOC will condense only if a suitable, reactive particle surface is present? What fraction of reactive particles are anthropogenic and potentially controllable? Answers to these questions depend on better characterization of chemical reactions occurring at particle surfaces.
Acknowledgments This research was supported by an award from Research Corporation. C.A.K. participated as a Merck/AAAS Scholar. We thank William P. Hastings for making confirmatory humidity measurements.
Literature Cited (1) Hildemann, L. M.; Cass, G. R.; Mazurek, M. A.; Simoneit, B. R. Environ. Sci. Technol. 1993, 27, 2045-2055. (2) Barthelmie, R. J.; Pryor, S. C. Sci. Total Environ. 1997, 205, 167178. (3) Chazette, P.; Liousse, C. Atmos. Environ. 2001, 35, 2497-2506. (4) Mallet, M.; Roger, J. C.; Despiau, S.; Dubovik, O.; Putaud, J. P. Atmos. Res. 2003, 69, 73-97. (5) Rivera-Carpio, C. A.; Corrigan, C. E.; Novakov, T.; Penner, J. E.; Rogers, C. F.; Chow, J. C. J. Geophys. Res. 1996, 101, 1948319493. (6) Schwartz, J. Environ. Res. 1994, 64, 36-52. (7) Schwartz, J. Environ. Res. 1994, 64, 26-35. (8) Dockery, D. W.; Pope, C. A.; Xu, X.; Spengler, J. D.; Ware, J. H.; Fay, M. E.; Ferris, B. G.; Speizer, F. E. N. Engl. J. Med. 1993, 329, 1753-1759. (9) Harrison, R. M.; Yin, J. Sci. Total Environ. 2000, 249, 85-101. (10) Kao, A. S.; Friedlander, S. K. Inhalation Toxicol. 1995, 7, 149156. (11) Odum, J. R.; Hoffmann, T.; Bowman, F.; Collins, D.; Flagan, R. C.; Seinfeld, J. H. Environ. Sci. Technol. 1996, 30, 2580-2585. (12) Bowman, F. M.; Odum, J. R.; Seinfeld, J. H.; Pandis, S. N. Atmos. Environ. 1997, 31, 3921-3931. (13) Jang, M.; Kamens, R. M.; Leach, K. B.; Strommen, M. R. Environ. Sci. Technol. 1997, 31, 2805-2811. (14) Odum, J. R.; Jungkamp, T. P. W.; Griffin, R. J.; Forstner, H. J. L.; Flagan, R. C.; Seinfeld, J. H. Environ. Sci. Technol. 1997, 31, 1890-1897. (15) Odum, J. R.; Jungkamp, T. P. W.; Griffin, R. J.; Flagan, R. C.; Seinfeld, J. H. Science 1997, 276, 96-99. (16) Kamens, R. M.; Jang, M.; Chien, C.-J.; Leach, K. B. Environ. Sci. Technol. 1999, 33, 1430-1438. (17) Cocker, D. R., III; Clegg, S. L.; Flagan, R. C.; Seinfeld, J. H. Atmos. Environ. 2001, 35, 6049-6072.
(18) Kavouras, I. G.; Stephanou, E. G. Environ. Sci. Technol. 2002, 36, 5083-5091. (19) Pankow, J. F. Atmos. Environ. 1994, 28, 185-188. (20) Pankow, J. F.; Seinfeld, J. H.; Asher, W. E.; Erdakos, G. B. Environ. Sci. Technol. 2001, 35, 1164-1172. (21) Seinfeld, J. H.; Erdakos, G. B.; Asher, W. E.; Pankow, J. F. Environ. Sci. Technol. 2001, 35, 1806-1817. (22) Liang, C.; Pankow, J. F.; Odum, J. R.; Seinfeld, J. H. Environ. Sci. Technol. 1997, 31, 3086-3092. (23) Chandramouli, B.; Jang, M.; Kamens, R. M. Atmos. Environ. 2003, 37, 853-864. (24) Jang, M.; Kamens, R. M. Environ. Sci. Technol. 2001, 35, 36263639. (25) Jang, M.; Kamens, R. M. Environ. Sci. Technol. 2001, 35, 47584766. (26) Jang, M.; Czoschke, N. M.; Lee, S.; Kamens, R. M. Science 2002, 298, 814-817. (27) Jang, M.; Lee, S.; Kamens, R. M. Atmos. Environ. 2003, 37, 21252138. (28) Tobias, H. J.; Ziemann, P. J. Environ. Sci. Technol. 2000, 34, 2105-2115. (29) Bandow, H.; Washida, N.; Akimoto, H. Bull. Chem. Soc. Jpn. 1985, 58, 2531-2540. (30) Philipps, R.; Walter, S.; Liauw, M. A. Chem. Eng. J. 2002, 90, 107-116, doi: 110.1016/S1385-8947(1002)00072-00074. (31) Forstner, H. J. L.; Flagan, R. C.; Seinfeld, J. H. Environ. Sci. Technol. 1997, 31, 1345-1358. (32) Kleindienst, T. E.; Smith, D. F.; Li, W.; Edney, E. O.; Driscoll, D. J.; Speer, R. E.; Weathers, W. S. Atmos. Environ. 1999, 33, 36693681. (33) Edney, E. O.; Driscoll, D. J.; Speer, R. E.; Weathers, W. S.; Kleindienst, T. E.; Li, W.; Smith, D. F. Atmos. Environ. 2000, 34, 3907-3919. (34) Fillo, J. D.; Koehler, C. A.; Nguyen, T. P.; De Haan, D. O.; Gilbert, B. A.; Flinn, K. P. Environ. Sci. Technol. 2003, 37, 46724677. (35) National Institute of Standards and Technology. Chemistry WebBook; NIST Standard Reference Database No. 69, March 2003 release; http://webbook.nist.gov/chemistry. (36) De Haan, D. O.; Brauers, T.; Oum, K.; Stutz, J.; Nordmeyer, T.; Finlayson-Pitts, B. J. Int. Rev. Phys. Chem. 1999, 18, 343-385. (37) Griffin, R. J.; Cocker, D. R., III; Seinfeld, J. H. Environ. Sci. Technol. 1999, 33, 2403-2408. (38) Allen, H. C.; Laux, J. M.; Vogt, R.; Finlayson-Pitts, B. J.; Hemminger, J. C. J. Phys. Chem. 1996, 100, 6371-6375. (39) Vogt, R.; Finlayson-Pitts, B. J. J. Phys. Chem. 1994, 98, 37473755. (40) Harvey, D. Modern Analytical Chemistry; McGraw-Hill: Boston, 2000. (41) Brown, H. C.; et al. In Determination of Organic Structures by Physical Methods; Braude, E. A., Nachod, F. C., Eds.; Academic Press: New York, 1955. (42) Weast, R. C., Ed. CRC Handbook of Chemistry and Physics; The Chemical Rubber Publishing Co.: Cleveland, OH, 1972. (43) O’Neal, M. J., Ed. The Merck Index. An Encyclopedia of Chemicals, Drugs, and Biologicals, 13th ed.; Merck and Co.: Whitehouse Station, NJ, 2001. (44) Di Lorenzo, A. J. Chromatogr. 1971, 55, 303-308. (45) Bruice, T. C.; Pandit, U. K. J. Am. Chem. Soc. 1960, 82, 58585865. (46) Grosjean, E.; Grosjean, D.; Fraser, M. P.; Cass, G. R. Environ. Sci. Technol. 1996, 30, 2687-2703. (47) Grossmann, D.; Moortgat, G. K.; Kibler, M.; Schlomski, S.; Schmann, K. B.; Alicke, B.; Geyer, A.; Platt, U.; Hammer, M.-U.; Vogel, B.; Mihelcic, D.; Hofzumahaus, A.; Holland, F.; Volz-Thomas, A. J. Geophys. Res. 2003, 108, 8250-8269, doi: 8210.1029/2001JD001096. (48) Griffin, R. J.; Nguyen, K.; Dabdub, D.; Seinfeld, J. H. J. Atmos. Chem. 2003, 44, 171-190. (49) California Air Resources Board. 1996 Air Toxics Inventory; http://www.arb.ca.gov/toxics/cti/cti.htm. (50) Silva, P. J.; Prather, K. A. Anal. Chem. 2000, 72, 3553-3562. (51) Kawamura, K.; Umemoto, N.; Mochida, M.; Bertram, T.; Howell, S.; Huebert, B. J. J. Geophys. Res. 2003, 108, 8639-8645, doi: 8610.1029/2002JD003256. (52) Mochida, M.; Kawabata, A.; Kawamura, K.; Hatsushika, H.; Yamazaki, K. J. Geophys. Res. 2003, 108, 4193-4203, doi: 4110.1029/2002JD002355. (53) Mochida, M.; Umemoto, N.; Kawamura, K.; Uematsu, M. Geophys. Res. Lett. 2003, 30, 1672, doi: 1610.1029/2003GL017451. VOL. 38, NO. 19, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5071
(54) Narukawa, M.; Kawamura, K.; Anlauf, K. G.; Barrie, L. A. J. Geophys. Res. 2003, 108, 4575-4583, doi: 4510.1029/ 2003JD003646. (55) Brown, S. G.; Herckes, P.; Ashbaugh, L.; Hannigan, M. P.; Kreidenweis, S. M.; Collett, J. L. J. Atmos. Environ 2002, 36, 5807-5818. (56) Fraser, M. P.; Cass, G. R.; Simoneit, B. R. T. Environ. Sci. Technol. 2003, 37, 446-453.
5072
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 19, 2004
(57) Cocker, D. R., III; Mader, B. T.; Kalberer, M.; Flagan, R. C.; Seinfeld, J. H. Atmos. Environ. 2001, 35, 6073-6085.
Received for review June 27, 2003. Revised manuscript received June 28, 2004. Accepted July 15, 2004. ES034672B
