Environ. Sci. Technol. 2005, 39, 7036-7045
Secondary Organic Aerosol Production from Terpene Ozonolysis. 1. Effect of UV Radiation ALBERT A. PRESTO,‡ KARA E. HUFF HARTZ,‡ AND N E I L M . D O N A H U E * ,†,‡ Department of Chemistry and Department of Chemical Engineering, Doherty Hall, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
We report secondary organic aerosol (SOA) yields from the ozonolysis of R-pinene under both dark and UV-illuminated conditions. Exposure to UV light reduces SOA yield by 2040%, with a maximum reduction in yield coinciding with a minimum in the amount of terpene consumed (15-30 ppb). The data are consistent with a constant absolute reduction in the yield of ∼0.03. Gas chromatography mass spectrometry analysis of filter samples indicates that the major products found in R-pinene SOA include organic acids (e.g., pinic acid), keto acids (e.g., pinonic acid), and hydroxy keto acids (e.g., 10-hydroxypinonic acid). Analysis of filter-based results suggests that yield reduction is a result of the formation of a more volatile product distribution when experiments are conducted in the presence of UV light. These results imply that previous “dark bag” experiments may overestimate SOA generation from monoterpenes and also that SOA generation in the atmosphere may depend significantly on actinic flux.
1. Introduction Emissions of biogenic volatile organic compounds (VOCs) are approximately an order of magnitude larger than emissions of anthropogenic VOCs on a global scale (1). A majority of biogenic VOCs contain double bonds (1) and are therefore susceptible to oxidation by OH, NO3, or O3 (2). The monoterpenes make up a significant fraction of biogenic VOC emissions (1). Previous research has shown that the oxidation of monoterpenes can lead to the formation of semivolatile species that partition between the gas and the aerosol phases, forming secondary organic aerosol (SOA) (3-12). Biogenic SOA appears to derive primarily from the terpenes, including monoterpenes and sesquiterpenes (C15H24). Oxidation of isoprene may also lead to significant SOA production in lowNOx environments (13). Previous experimental studies of SOA formation from monoterpene oxidation have focused heavily on the terpeneozone reaction (5-7, 9, 10, 12). The reason for this emphasis on ozone is twofold. Reaction with ozone is a significant atmospheric sink for terpenes (2), and ozonolysis of endocyclic alkenes (many terpenoid species have an endocyclic double bond) generates products that are multiply oxidized (2, 14). In fact, experimental results show that ozonolysis may be the most potent source for biogenic SOA (6). * Corresponding author phone: (412)268-4415; e-mail: nmd@ andrew.cmu.edu; fax: (412)268-7139. † Department of Chemistry. ‡ Department of Chemical Engineering. 7036
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005
In most experimental studies, SOA is generated in an environmental (or “smog”) chamber (3-12, 15-19). Ozonolysis is typically studied under dark conditions (5, 6) to limit interference from VOC oxidation via reaction with photochemically generated OH radicals (20). The ozonolysis reaction is assumed to be independent of photochemical factors, so observing SOA formation in dark terpene-ozone systems is a widely accepted practice. However, because the SOA yields from monoterpenes are typically on the order of 25%, there is reason to expect the oxidation products to be sensitive to ambient conditions, even from ozonolysis. There is evidence that any carbonyl oxides (Criegee intermediates) produced in the ozonolysis of monoterpenes decompose rapidly before the excess energy released from the cycloaddition reaction is lost to the atmospheric bath gas. This includes the large observed yields of OH radical (2, 21) and recent master equation modeling of the reaction dynamics (22). Carbonyl oxide decomposition will result in organic radical fragments that should eventually follow a reaction mechanism similar to other organic compounds (2). Indeed, measured aerosol yields are known to depend on the scavenger used to remove hydroxyl radicals generated during ozonolysis (23, 24), which clearly indicates that chemistry after the ozonolysis reaction influences SOA formation. These mechanisms characteristically depend strongly on NOx levels, and products are typically multifunctional oxidized organic compounds containing carbonyl, hydroperoxide, organic acid, and alcohol moieties. These moieties are responsible for the lowered vapor pressures of the products, but they also typically lead to elevated absorption cross sections in the ultraviolet (25) and thus vulnerability to photodecomposition. There is thus every reason to expect the product distribution from ozone-monoterpene reactions to depend on both NOx levels and UV exposure. As a consequence, we also expect the aerosol yields to depend on both NOx and UV light. In this study, we present an investigation of SOA production from the ozonolysis of R-pinene in the presence and absence of UV light. R-Pinene is both the most abundant (1) and the most studied of the terpene species (4-12). Ozonolysis of R-pinene generates a significant amount of aerosol, with SOA yields of up to 20% at 37 °C (6). In a companion paper, we shall examine the dependence of SOA yields from R-pinene ozonolysis on NOx levels (26). We have conducted experiments under varying conditions of initial terpene concentration, initial seed concentration, and lighting (dark or UV-irradiated). Our results show that SOA formation is dependent upon whether the chamber is held in darkness and refute the assumption that terpene ozonolysis is independent of photochemical branching.
2. Experimental Section 2.1. Chamber Experiments. SOA yield experiments are conducted in a 10-m3 Teflon chamber (Welch Fluorocarbon) (27). The chamber is suspended inside of a temperaturecontrolled room (15-40 °C). Ozone and NOx concentrations are measured by gas-phase analyzers for each species (Dasibi 1008-PC and API 200A). Particle concentrations are monitored by a scanning mobility particle sizer (SMPS, TSI 3936). For most experiments, a gas chromatograph with a flame ionization detector (GC-FID, Perkin-Elmer AutoSystem XL; J&W Scientific DB-624 capillary column, 30 m × 0.530 mm) coupled to a preconcentrator (Entech 7100A) is available for measuring gas-phase concentrations of organic species. The chamber is also equipped with three banks of UV lights (General Electric model 10526 black lights). 10.1021/es050174m CCC: $30.25
2005 American Chemical Society Published on Web 08/16/2005
FIGURE 1. Detail of smog-chamber setup. The SMPS is held inside the temperature-controlled room to minimize temperature differences between the instrument and the chamber. Prior to an experiment, the chamber is flushed with purified air for 12-48 h. Air is purified by passage through HEPA and carbon filters to remove particles and gas-phase organics and silica gel to reduce relative humidity to 90% of the OH radicals formed by the ozonolysis reaction. The scavenger is added to the chamber by bubbling a stream of clean air through a known volume of 2-butanol. The primary product of the 2-butanol-OH reaction is 2-butanone (2), which is not known to contribute to atmospheric aerosol even at high (ppm) concentrations. However, recent work has shown that the experimental aerosol yield is dependent on the choice of OH scavenger (23, 24). R-Pinene is introduced to the chamber via a septum injector located inside of the temperature-controlled room. An alkane tracer, either n-pentane or cyclopentane, is injected along with the R-pinene. Due to the high concentration of 2-butanol and the very slow reaction between ozone and alkanes (2), the tracer species is not consumed during the experiment. Thus, we can use the tracer concentration, as measured by GC-FID, to accurately determine the initial R-pinene concentration in experiments where the reaction is initiated by injection of the terpene into the chamber. This study considers SOA formation under two basic operating conditions: light and dark. Dark experiments are conducted with all chamber lights off and all windows covered. In light experiments, the chamber is irradiated by three banks of UV lights (Figure 2). Table 1 outlines the experimental conditions for the experiments conducted for this work. Figure 3 shows the time series of a typical experiment where the reaction was initiated by injection of R-pinene into the chamber (experiment 8/10/04). Prior to R-pinene injection, 2-butanol and ozone were added to the chamber, and the temperature was allowed to equilibrate (22 °C for all experiments) for approximately 1 h. In experiments where
FIGURE 2. Emission spectrum of the UV lamps used in this study. The lamps are classified as UVA and have a maximum output near 350 nm. In addition to emissions at ultraviolet wavelengths, some light is emitted in the purple region of the visible spectrum. the chamber is exposed to UV light, ozone is used as the initiating reactant. In these experiments, R-pinene is introduced to the chamber concurrently with the 2-butanol. We conducted several tests of R-pinene versus ozone initiation and found that under otherwise identical conditions the SOA yield is independent of the initiating reactant. In both the R-pinene- and the ozone-initiated experiments, particles begin to nucleate within 15 min of the start of the reaction. We determine aerosol mass based on the measured mobility diameter, assuming spherical particles with a density of 1.0 g cm-3. The aerosol increases in modal diameter and mass through coagulation and condensation as the R-pinene is consumed by ozone. Aerosol mass concentration peaks shortly (15-30 min) after complete consumption of the R-pinene. The aerosol mass concentration then begins to decrease as particles are lost to the chamber wall. In many experiments, the oxidation products were allowed to self-nucleate, with new particle formation occurring at the same rate as R-pinene removal; however, in some experiments conducted with UV lights on, there did appear to be a delay between terpene loss and particle formation. Consequently, we conducted several experiments where the chamber was seeded with SOA (34). In these experiments, a small amount of R-pinene was oxidized to generate a background of organic particles. After the aerosol seed mass peaked, a second, much larger injection of R-pinene was introduced to the chamber. Figure 4 shows the particle mass and number concentrations for a typical seed experiment (experiment 10/14/04). The first injection of R-pinene produces a maximum particle number concentration of 1.2 × 104 cm-3 and a maximum mass concentration of 27 µg m-3. These concentrations decrease to 9.2 × 103 cm-3 and 24.5 µg m-3 by the time of the second R-pinene injection. Waiting for the mass concentration to peak before adding the second injection of R-pinene ensures that each dose of terpene is consumed and thus produces SOA separately (though the oxidation products from the first injection are presumed to remain stable in the system throughout the experiment). The second injection of R-pinene leads to a large increase in aerosol mass but only a small ( 3500 s), aerosol mass concentration decreases because of wall losses.
FIGURE 4. Mass and number concentrations for a seed experiment (experiment 10/14/04). The first injection of r-pinene, ∼35 ppb at t ) 0 s, generates a large number concentration (∼104 cm-3) but a relatively small mass concentration (∼30 µg m-3). The second injection of r-pinene at ∼5000 s generates a large increase in mass concentration but a k0. While we cannot verify the exact nature of k1, as noted above, k2 is critical to the yield determination, and k1 simply provides a more accurate functional form to eq 2. We validated the aerosol yields determined by eqs 2 and 3 by comparing the results to yields determined by a separate method. For each experiment presented here and in a companion study (26), aerosol yield was calculated by first determining the first-order wall-loss rate exhibited during the wall-loss regime (time > 2 h in Figure 5). This wall-loss rate was then applied over the course of the entire experiment to determine the amount of organic mass on the wall. ∆M0 is thus the sum of the observed aerosol mass and the mass lost to the wall, and the yield is calculated from eq 1. For a sample of over 20 experiments, the two yield determination methods are consistent to within 5%. Previous chamber studies (5, 6, 8, 17, 42-45) have considered aerosol yield in the context of the partitioning theory of Pankow (17, 46, 47) and have characterized the yield using a two-product model. Though we will present aerosol yield in the familiar yield versus M0 plot (here M0 ) COC0), the focus of this study is to investigate the effect of UV radiation on aerosol formation, and we will not offer new two-product model parameters in this article. In a companion study (26), we consider our results in the context of the Pankow partitioning model.
4. Results 4.1. Aerosol Yield. We conducted experiments with three different initial concentrations of R-pinene: unseeded experiments using 15-30 and 180-210 ppb R-pinene and seeded experiments using 150-170 ppb R-pinene. Both dark and UV-irradiated experiments were conducted for each initial condition. Results are shown in Figure 6. In all cases, aerosol yield is observed to decrease in the presence of UV light. Yield reduction is more pronounced at lower initial R-pinene concentrations (lower M0), with reductions of ∼40% for 15-30 ppb R-pinene consumed and ∼20% for consumption of 180-210 ppb R-pinene. In fact, the data are consistent 7040
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005
FIGURE 7. Aerosol products identified by GC-MS. with a constant absolute reduction in the yield of approximately 0.03. Calculated k2 values for light and dark experiments are similar, typically ∼10-4 s-1. We thus assume that the UV lights only affect the contents of the chamber and not the chamber itself. For example, the similarity in k2 values between light and dark experiments strongly suggests that wall-loss rates in the chamber are not affected by the presence or absence of UV radiation. Aerosol yield increases as the amount of R-pinene oxidized increases. This effect is consistent with increasing supersaturation of condensable species as more VOC is oxidized, and the increase in yield with increasing VOC consumption is described by the partitioning theory of Pankow (47). Iinuma et al. (12) and Czoschke et al. (48) observed enhanced aerosol production in the presence of acidic seed particles and attributed the increase in yield to acid-catalyzed oligomerization reactions (49). Previous studies have identified organic acids in SOA (11, 38); the presence of these acids may make the SOA seed acidic enough to catalyze oligomer formation. However, because yield reduction is observed in both the seeded and the nucleation experiments, we surmise that any differences in chemistry between seeded and nucleation experiments do not alter the impact of UV exposure on SOA formation. 4.2. Filter Analysis. In most experiments, filter samples were collected for analysis by GC-MS. Figure 7 shows the SOA species identified in this study. Major identified products include pinonaldehyde, pinonic acid, and pinic acid. The least volatile species identified in all of the samples is 10hydroxypinonic acid (50). The species identified in this study
and seeded experiments, ratios for UV-irradiated experiments are smaller than the ratios for corresponding dark experiments with similar M0 values. The changes seen here are therefore evidence for changes in the ratio of total concentrations due to changes in reaction chemistry. Figure 8 does not include the filter data for the high VOC (180-210 ppb R-pinene) experiments. The ratios for these experiments vary greatly and display almost no repeatability. Several possibilities exist to explain this variability. Filters collected during the high VOC experiments contain greater SOA mass than filters collected during low VOC or seeded experiments. Increasing the mass on the filter may exacerbate any extraction inefficiencies that may exist, thereby changing the observed product ratios. Increasing the mass of SOA on the filter (and consequently the concentration in the filter extract) may also impact product derivatization. Both pinic acid and 10-hydroxypinonic acid form TMS diesters in the derivatization reaction, while pinonic acid is singly derivatized. In some of the high-mass filters, the ratios are lower than the ratios for the low VOC experiments. This decrease could be explained by incomplete derivatization of pinic acid and 10-hydroxypinonic acid. The extent of oligomer formation may also impact the measured product ratios. Gao et al. (52) showed that oligomers comprise a significant fraction of the SOA generated by R-pinene ozonolysis. The high VOC experiments typically exhibit larger SOA yields than the low VOC experiments. If this increase in SOA yield is related to an increase in the extent of oligomerization, then it is quite possible that the pinic/pinonic and 10-hydroxypinonic/pinonic ratios could decrease if either of the heavier species (pinic acid and 10-hydroxypinonic acid) participate in oligomer formation. Because of these uncertainties, discussion of filter-based results will focus on filters collected during low VOC and seeded experiments. FIGURE 8. Normalized product ratios from GC-MS data. Each ratio measures the relative abundances of a less volatile product (pinic acid or hydroxy pinonic acid) to a more volatile product (pinonic acid). Error bars show the precision (1σ) for the calculated ratios. Ratios increase with increasing M0 and are consistent with increasing aerosol yield as M0 increases. Ratios for UV-irradiated experiments (open symbols) are lower than ratios for corresponding dark experiments and suggest that aerosol yield reduction is a result of an increase in product volatility. agree well with past observations of condensed-phase R-pinene oxidation products (9, 11, 38). Because each filter contains a different mass of SOA, it is difficult to compare filters directly. Instead, we consider the ratios of identified species: pinic acid to pinonic acid and 10-hydroxypinonic acid to pinonic acid. We chose pinonic acid as a reference compound because it is formed directly from the rearrangement of the Criegee intermediate (51) and therefore its production should not be affected by UV light. In each case, the ratio compares the relative amounts of a less volatile species (lower saturation vapor pressure) to a more volatile species (higher saturation vapor pressure). Thus, the larger the ratio, the less volatile the aerosol. Figure 8 shows the normalized ratios for the experiments considered in this study. Both the pinic/pinonic acid and 10-hydroxypinonic/ pinonic acid ratios increase with increasing M0. This observation is consistent with increasing aerosol yield, and therefore a less volatile product distribution, at higher M0. According to gas-particle partitioning theory (47), the concentration of each compound in the particles will increase with M0, but the ratio of the two compounds will remain constant unless the ratio of total concentrations (gas + particle) changes. In both the low VOC (15-30 ppb R-pinene)
5. Discussion 5.1. Comparison to Other Work. 5.1.1. Aerosol Yield. We measured aerosol yields of up to 0.29 for dark experiments and 0.23 for UV-irradiated experiments. These yields are substantially larger than the maximum yield observed (0.18) in dark chamber experiments conducted by Griffin et al. (6) and Hoffmann et al. (5). The major difference is the higher maximum ∆VOC explored here, though temperature may also have an influence. All of the experiments reported here were conducted at 22 °C. Yields reported by Hoffmann et al. (5) and Griffin et al. (6) were obtained in experiments conducted at higher temperatures, typically 30-40 °C. Recent results (27) indicate that aerosol yields from R-pinene ozonolysis change only slightly with temperature; raising the chamber temperature by 10 °C should produce a change in aerosol yield of ∼10%. Our results echo this finding; there is good agreement between the results of Griffin et al. (6) and the data presented here and in a companion study (26) where the two data sets overlap. Takekawa et al. (44) measured SOA yields for R-pinene photooxidation of 0.23 at 10 °C and 0.10 at 30 °C with ∆M0 ≈ 90 µg m-3. This result bounds our measurements; our results (Figure 6) suggest an aerosol yield of about 0.2 at 22 °C for UV-irradiated experiments with ∆M0 ≈ 90 µg m-3. 5.1.2. GC-MS Product Identification. The most abundant gas-phase product of R-pinene ozonolysis is pinonaldehyde (53). Pinonaldehyde can partition into the aerosol phase and was observed by both Yu et al. (38) and Jaoui et al. (11) in chamber-generated aerosol. A significant number of aerosol products from R-pinene ozonolysis, including pinonic, norpinonic, and pinic acids (9, 11, 38), contain the organic acid moeity. Glausius et al. (9) identified pinonic acid and pinic acid as the most abundant organic acids in R-pinene aerosol. Our results are consistent with this finding. VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7041
10-Hydroxypinonic acid is the least volatile SOA species identified in this work (50). This is consistent with Jaoui et al. (11), who also identified pinalic-4-acid, 10-hydroxypinonaldehyde, and oxopinonic acid (both 4- and 10-). The C9H16O2 species with molecular weight 156 was identified by Yu et al. (38). Previous studies (11,38) identified norpinonaldehyde (C9H14O2) in experimentally generated aerosol. This species was not detected in any of our filter samples. Jaoui et al. (11) also identified several species, such as 1-hydroxypinonaldehyde, 4-oxopinonaldehyde, 3-hydroxypinaketone, and 2-hydroxy-3-pinanone, that were not observed here. We feel that these discrepancies are of minor importance, because we have identified the species previously noted as significant semivolatile products of R-pinene oxidation. Recent work has focused on the presence of oligomers in experimentally generated SOA (49, 52). Gao et al. (52) estimate that oligomeric products comprise more than 50% of the SOA mass generated from the ozonolysis of R-pinene. These oligomers have molecular weights up to 1600 g mol-1 and are difficult to detect using the GC-MS methods employed in this study. Our GC-MS results agree well with the findings of Gao et al. (52) that the low molecular weight (
