Environ. Sci. Technol. 2004, 38, 2802-2812
Full-Scale Chamber Investigation and Simulation of Air Freshener Emissions in the Presence of Ozone X I A O Y U L I U , * ,† M A R K M A S O N , * ,‡ KENNETH KREBS,‡ AND LESLIE SPARKS‡ ARCADIS, P.O. Box 13109, Research Triangle Park, North Carolina 27713, and National Risk Management Research Laboratory, Office of Research and Development, United States Environmental Protection Agency, Research Triangle Park, North Carolina 27711
Volatile organic compound (VOC) emissions from one electrical plug-in type of pine-scented air freshener and their reactions with O3 were investigated in the U.S. Environmental Protection Agency indoor air research large chamber facility. Ozone was generated from a device marketed as an ozone generator air cleaner. Ozone and oxides of nitrogen concentrations and chamber conditions such as temperature, relative humidity, pressure, and air exchange rate were controlled and/or monitored. VOC emissions and some of the reaction products were identified and quantified. Source emission models were developed to predict the time/concentration profiles of the major VOCs (limonene, R-pinene, β-pinene, 3-carene, camphene, benzyl propionate, benzyl alcohol, bornyl acetate, isobornyl acetate, and benzaldehyde) emitted by the air freshener. Gasphase reactions of VOCs from the air freshener with O3 were simulated by a photochemical kinetics simulation system using VOC reaction mechanisms and rate constants adopted from the literature. The concentration-time predictions were in good agreement with the data for O3 and VOCs emitted from the air freshener and with some of the primary reaction products. Systematic differences between the predictions and the experimental results were found for some species. Poor understanding of secondary reactions and heterogeneous chemistry in the chamber is the likely cause of these differences. The method has the potential to provide data to predict the impact of O3/VOC interactions on indoor air quality.
Introduction Fragrance chemicals are often used as air fresheners. Several methods such as evaporation from wicks, aerosol sprays, and heated pots have been used to introduce the fragrance chemicals into the indoor air. Plug-in air fresheners using small electric heaters to volatilize the fragrance chemicals have recently become a dominant product in the air freshener market. There is little information about the volatile and semivolatile compounds (VOCs and SVOCs) emitted from these products and on their impact on indoor air quality (IAQ). Because many fragrance chemicals are unsaturated VOCs, they may react with indoor O3 to produce aldehydes, ketones, * Corresponding authors (X.L.) e-mail:
[email protected]; (M.M.) e-mail:
[email protected]. † ARCADIS. ‡ United States Environmental Protection Agency. 2802
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
organic acids, particles (1), and free radicals (2-5) in a manner similar to the reactions between O3 and VOCs emitted from carpet (6, 7), paint (8), and other materials (9, 10). Heterogeneous processes have been shown to play an important role in these ozone-initiated reactions (11, 12). The ozone necessary for these reactions can come from the outdoors or from numerous indoor products that generate ozone, for example, photocopiers, laser printers, and ozone-emitting devices sold as air cleaners. These latter devices are of special concern because they can produce indoor ozone concentrations greater than the EPA’s health based ambient air quality standards (13). Prolonged exposure to ozone above certain levels is believed to cause lung and tissue damage (14-16). The combination of reaction products, VOCs and SVOCs, and O3 may have a greater impact on IAQ than VOC, SVOC, and O3 alone. The reactants and products of O3 and VOCs may be associated with mucous membrane irritation, unpleasant odor, and sick building syndrome (17) and have adverse health effects on building occupants. Epidemiological studies have shown a significant correlation between exposure to ambient fine particles and cardiac and respiratory mortality and levels of pulmonary lung function in children and adults with obstructive airways disease (18-20). Knowledge of VOC-O3 reactions is necessary to fully understand and model the impact of various consumer products and materials on IAQ. In principle, the chemistry of outdoor pollutants can be extended to indoor pollutants in indoor air models. Indoor air chemistry has its own peculiarities, such as lower photolysis rates, ventilation, filtration, penetration, and relatively large adsorption surface areas. Nazaroff et al. (21, 22), Weschler et al. (23), Sarwar et al. (24), and others (2, 4, 21, 24-28) have developed models by incorporating explicit chemical kinetic mechanisms to allow analysis of the fate of chemically reactive compounds and particles in the indoor environment. These models provide useful starting points but have not been verified with a large body of experimental data. The present study was undertaken to develop experimental methods for studying the impact of the emissions from plug-in air fresheners and their reactions with indoor ozone on IAQ and to investigate the suitability of existing models for describing these impacts. The experimental study involved the characterization of the VOC emissions from the air freshener, the characterization of the reaction products of the air freshener chemicals with ozone, the measurement of the size distribution and mass of fine particles produced by the reactions, and the modeling of the gas-phase VOCozone system.
Experimental Procedures Large Chamber Experiments. Experiments were performed in a 29.8 m3 (57 m2 surface area) insulated environmental chamber equipped with clean air generation, conditioning, and distribution systems as well as gas and particle sampling and analysis systems. The internal chamber surfaces are polished stainless steel. The clean air supply system is equipped with coarse and high efficiency particulate air (HEPA) filtration and VOC removal subsystems. Environmental conditions of the chamber, such as temperature, pressure, relative humidity (RH), air exchange rate (ACH), and concentrations of O3, NOx, and VOCs, are controlled and/or monitored by online instruments. Light is provided in the chamber by fluorescent light fixtures mounted outside the chamber and separated from the chamber by sealed glass 10.1021/es030544b CCC: $27.50
2004 American Chemical Society Published on Web 04/09/2004
windows. Details of the chamber and instrumentations are described elsewhere (29). Here, we report the results of three experiments that were conducted with pine-scented air fresheners. Experiments 1 and 3 investigated the impact of ozone on air freshener emissions, and experiment 2 was a source emission test without ozone input. Ozone was produced using an ozone generator air cleaner that had been well-characterized in previous studies (13). In each experiment, four pine-scented refills were randomly selected among six weighed refills and placed in the electrical plug-in devices mounted on a rack in the corner of the chamber under a mixing fan. In experiments 1 and 3, the plug-in air fresheners were first turned on for about 24 h, and then the ozone generator, which was set to generate ozone at a rate of about 4.5 mg/h (76 ppb/h), was remotely activated for a period of 24 h. At the elapsed time of 72 or 120 h, the pine-scented refill sources were removed from the chamber and reweighed. In experiment 1, after the sources were removed, VOCs in the chamber were monitored for an additional 24 h. Sulfur hexafluoride (SF6) was injected in all the tests to monitor the chamber air exchange rate. The chamber environmental control system was set to maintain test conditions of 23 ( 0.5 °C and 50 ( 5% RH with about 20 Pa positive pressure relative to the laboratory housing the chamber and a fresh air supply rate of 0.5ACH. Sampling and Analysis Methods. Concentration of the tracer gas, SF6, was determined as previously described (13). The following methods were employed to determine concentrations of reactants and products in the chamber air. O3 and NOx concentrations were continuously monitored by methods ASTM D5156-95 (30) and ASTM D3824-95 (31), respectively, with a Thermo Environmental Instruments Inc. (TEI) Model 49 UV O3 photometer and a TEI Model 42 chemiluminescence NOx monitor. VOC concentrations were determined by (a) cryogenic concentration (Entech Model 7100) with thermal desorption to a Hewlett-Packard (HP 5890) gas chromatograph (GC) equipped with a flame ionization detector (FID) and (b) by collection of VOCs on sorbent tubes (Tenax TA) followed by thermal desorption to the GC/FID or to a HP 6890 GC equipped with an HP 5970 mass selective detector (MSD) and automated thermal desorber (Perkin-Elmer ATD 400). An N2 purge method was employed to remove ozone from Tenax sorbent tubes. Immediately after sample collection, each sorbent tube was purged with a 100 mL of N2, sealed in a labeled Teflon bag, and stored at -4 °C. This method was evaluated by collecting gas-phase limonene, R-pinene, and β-pinene over a concentration range of 50-1000 µg/m3 on Tenax tubes in a flowing air stream containing up to 200 ppb of ozone. No significant recovery differences were observed between 0 and 100 ppb of ozone for limonene and among 0, 100, and 200 ppb of ozone for R-pinene and β-pinene. Carbonyl compounds were determined by EPA Method IP-6A, with collection of analytes on dinitrophenylhydrazine (DNPH) coated silica gel cartridges (Waters, Sep-Pak) and analysis by HP1090 high performance liquid chromatograph (HPLC) equipped with photodiode array detector. When O3 was present, O3 scrubbers (Waters, Sep-Pak) were used in front of DNPH cartridges. Polar and multi-functional carbonyl compounds were determined by the (O-[2,3,4,5,6]-pentafluorobenzyl)-hydroxylamine (PFBHA) derivatization method (32), using midget impingers to collect samples and analysis of derivitized compounds with a Varian GC/Saturn III ion trap mass spectrometer (MS). Semivolatile and particulate organic compounds were collected with a sampling system consisting of a glass annular denuder coated with XAD-4 (University Research Glassware, Chapel Hill, NC) followed in parallel by a 47 mm Teflon
membrane filter (Pall Gelman) and a 47 mm quartz fiber filter (Pall Gelman, 2500 QAOT-UP). Quartz and Teflon filter samples were collected at 16 L/min each for 4.5 (experiment 1) or 24 (experiment 3) hours, and denuder samples were simultaneously collected with a rate of 32 L/min. Denuders were extracted with solvent mixture of dichloromethane, acetone, and hexane in 2:3:5 (v/v) in triplicate, and the extract volume was reduced by N2 blowdown. A small portion of the quartz filter was taken for organic carbon/elemental carbon (OC/EC) measurement by NIOSH Method 5040. The rest of the filter was extracted with five successive 10 min sonication steps. The first two extractions were performed with hexane followed by three consecutive extractions with a 2:1 mixture of benzene and isopropyl alcohol solvent. The extract was then concentrated and stored in a vial with a Teflon-lined cap in a freezer until derivatization and analysis. The concentrated extracts from the denuders and quartz filters were each split into two fractions; one fraction remained neutral, and the other was methylated with fresh diazomethane for organic acids analyses. The neutral and methylated extracts were analyzed on the Varian GC/Saturn III ion trap MS with a DB-5 column and HP 6890 GC/5973 MSD with a DB-1 column, then derivatized by PFBHA and reanalyzed by a Varian GC/Saturn III ion trap MS. Each Teflon filter was weighed in accordance with the guidelines described in the Code of Federal Regulations (33) and the EPA Quality Assurance Guidance Document 2.12 (34), placed into a 7 mL vial, sonicated for 20 min with 5 mL of HPLC water, and analyzed on a Dionox DX-120 ion chromatograph (IC) for water-extractable ions. Particle size distributions were monitored with a laser aerosol spectrometer (LasX, Particle Measuring Systems Inc.) and scanning mobility particle spectrometers (SMPS 3934, SMPS 3936, TSI Inc.). A Model 3934 SMPS consisting of a Model 3071A Electrostatic Classifier (EC) and a Model 3010 Condensation Particle Counter (CPC) were used in experiment 1 to classify particles into 24 physical diameter bins between 0.019 and 0.523 µm. A Model 3936 SMPS equipped with a Model 3080 EC, Model 3025A CPC, and Model 3081 long Differential Mobility Analyzer (DMA) was used in experiment 3, where particles were classified in 106 physical diameter bins between 0.016 and 0.661 µm. The LasX, an optical spectrometer, was used to measure size distributions for particles with optical diameters between 0.1 and 7.5 µm. Quality assurance and quality control procedures were implemented by following approved quality assurance plans. Analytical instruments were calibrated by standards before tests. Data quality was evaluated based on instruments and method performance, which was measured by analyses of quality control samples that consisted of background samples collected prior to the test, field blanks, spiked field controls, duplicates, and daily calibration check samples. Modeling. SCIENTIST, developed by MicroMath Scientific Software (Salt Lake City, UT) was employed to model the air freshener source emissions, and the photochemical kinetics simulation system (PKSS) developed by Jeffries (35) was utilized to model the chemical reactions in the chamber system. The SCIENTIST program provides model equations for the least-squares fit of experimental data (36). It has the capability of solving systems of models that include simple, nonlinear algebraic, and differential equations, as well as Laplace transforms. PKSS has been widely used for solving certain kinds of mathematical models with differential equations from photochemical kinetic reaction mechanisms of typical indoor and outdoor chamber experiments (37, 38). The empirical VOC emissions models for the air freshener were obtained by fitting GC measurements with the SCIENTIST program. Reaction mechanism and rate constants used for the PKSS kinetic simulations in this study were adopted from the atmospheric chemistry literature. Data VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2803
TABLE 1. Summary of Experiment Conditions
a
experiment ID
1
2
3
experiment date ACH (h-1) (SD T (°C) ( SD pressure (Pa) ( SD %RH ( SD O3 emission rate (mg/h) highest O3 concentration (ppb) O3 generation duration (h) NO2 emission rate (mg/h) highest NO2 concentration (ppb) source emission duration (h) total weight loss (g) of four cartridges avg. TVOC emission rate during test (mg/h)
5/8/01 to 5/15/01 0.55 ( 0.01 23.45 ( 0.77 24.88 ( 9.95 48.56 ( 2.99 4.05 74 24.17 0.28 13 120.23 3.868 32.17
6/27/01 to 7/5/01 0.55 ( 0.00 22.94 ( 0.14 24.88 ( 2.49 48.59 ( 1.55 0 NAa 0 0 NAa 193.00 5.687 29.47
7/31/01 to 8/3/01 0.55 ( 0.06 24.10 ( 0.51 24.88 ( 2.49 49.37 ( 3.14 3.80 62 24.00 0.28 11 71.23 2.288 32.12
NA stands for ozone not available due to no injection.
FIGURE 1. VOCs detected in the air freshener large chamber experiment by HP 6890 GC/5973 MSD with ATD400. summarized by Calogirou (39) were used to represent the gas-phase reaction mechanism of limonene, R-pinene, β-pinene, camphene, and 3-carene. The ozone-terpene reaction system produces a significant amount of OH and other radicals as well as various primary reaction products. There is considerable uncertainty due to the limitation of current analytical techniques regarding such factors as the fate of biradical intermediates, ultimate products, their reaction rate constants, particle formation pathways, etc. Hence, the reaction mechanism employed in this study included only terpene reactions with O3, OH, and NO3 in the gas phase, plus subsequent reactions of some major products. This mechanism mainly focuses on predictions of ozone generation from ozone generator air cleaner and consumption of VOCs from air freshener in reactions with ozone. Chemistry of OH with other VOCs from the air freshener, such as benzaldehyde, bornyl acetate, and isobornyl actetate, were also considered in the mechanism. When an ozone generator air cleaner is used to generate ozone, inorganic species such as NO, NO2, NO3, N2O, and N2O5 will be produced via silent discharge when air is used as the oxygen source. If trace water is present, HNO3 can also be formed (40, 41). The inorganic chemistry and formaldehyde chemistry from the Carbon Bond-IV (CB IV) mechanism 2804
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
(42) were employed in the present PKSS simulation. No photolysis reactions were considered in the simulation since the photolytic reactions were found to proceed at very small or even negligible rates indoors (5). The chamber surface exhibited a sink effect in removing reactants and products. After adsorption, the chemicals can permanently remain on the chamber surface, evaporate back into chamber air, or chemically degrade on the chamber surface by heterogeneous reactions (11, 12, 43). However, little is reported in the literature about these potentially very important surface reactions; thus, no explicit description of these processes was incorporated in the present mechanism. Instead, deposition velocities of O3, NO2, and limonene were measured in the chamber and have been accounted for by comparing their decay at specific air exchange rates with an inert tracer gas SF6 (13).
Results and Discussion Chamber Test Measurement Data. Three large chamber experiments were conducted. The test conditions such as temperature; percent RH; chamber pressure; and ACH, O3, and NO2 concentrations are summarized in Table 1. Figure 1 shows the VOCs emitted into the chamber from the air freshener that were detected by the HP 6890 GC/5973
FIGURE 2. Carbonyl compounds detected in the air freshener large chamber experiment by DNPH-HPLC method. MSD with ATD400. The internal standard (IS) was d8-toluene. Among the compounds, R-pinene, camphene, β-pinene, benzaldehyde, 3-carene, limonene, benzyl alcohol, benzyl acetate, benzyl propionate, bornyl acetate, and isobornyl acetate were identified by matching the quality of the NIST98 Mass Spectra Library search results and confirmed by matching their retention times and mass spectra with standards. Dipropylene glycol methyl ether, β-myrcene, terpinolene, thujone, R-methyl benzyl acetate, 4-carene, R-cedrene, and lilial were tentatively identified by matching the quality of NIST98 Mass Spectra Library search results with matching quality g90%. The PFBHA derivative method did not provide much information about carbonyl identification in the source emissions and subsequent reactions with ozone due to the low concentrations. Some unidentified carbonyl compounds were detected by the DNPH-HPLC method as shown in Figure 2. Formaldehyde, benzaldehyde, and acetaldehyde derivatives were positively identified by matching retention times with their standards. Acetaldehyde was a compound that existed in the field blank. Figure 2 indicates that there were carbonyl compounds in the original air freshener sources, and new carbonyl compounds were formed in O3VOC reactions (e.g., peaks at 7.5, 13, and 19 min). It has been reported (ref 44 and references therein) that terpenes react with ozone to produce pinonaldehyde (R-pinene-O3 reaction), nopinone (β-pinene-O3 reaction), caronaldehyde (3carene-O3 reaction), 4-acetyl-1-methyl-cyclohexene (4-AMC, limonene-O3 reaction), camphenilone (camphene-O3 reaction), acetone, formaldehyde, organic acids, organic nitrates, and other compounds. More research is needed to further characterize these compounds. Particle sampling by denuder and filters and subsequent analyses by GC/MS with a derivatization method and IC were expected to shed light on the chemical composition of the particles. However, because of low concentrations of VOCs and their reaction products in the tests, no reliable data were available. In experiment 3, denuder and filters samples were collected for 24 h with ozone present. Unfortunately, it was found at the end of the test that the Teflon filter was broken. Extract from the 1/4 Teflon filter was analyzed by IC. A concentration of 0.5 mg/L of nitrate was detected in the 5
mL extraction solution with two additional unidentified species. The nitrate concentration was below the practical quantification limit, which was 0.7-0.8 mg/L. Source Emissions. The emissions from the electrical plugin type air freshener fell into two patterns. The chamber concentrations of VOCs such as benzaldehyde, benzyl alcohol, benzyl proprionate, bornyl acetate, and isobornyl actetate increased to a high initial value with minimal change over several hours. The chamber concentrations of limonene, R-pinene, β-pinene, camphene, and 3-carene continuously decreased from peak concentrations. Emission rates of major VOCs from the air freshener in the large chamber were determined based on the GC/MS data of VOCs in the absence of ozone in experiments 1 and 3 using the SCIENTIST program. In experiment 2, only R-pinene, β-pinene, and limonene were quantified; thus, experiment 2 data were not used for source emissions modeling and are not presented here. However, as shown in Table 1, the average weight loss rates of total VOCs were consistent among experiments 1-3, demonstrating that ozone generation had no apparent impact on the overall VOC emission rate. No sink effect was taken into account due to lack of data. Two models were developed for the two distinctive patterns of source emissions. Model 1 is for VOCs such as benzaldehyde, benzyl alcohol, benzyl proprionate, bornyl acetate, and isobornyl actetate that contain one or more oxygen atoms in their structures
R(t) ) R10 + R20(1 - e-kt)
(1)
where R(t) stands for the emission rate of a VOC (µg/h) at time t, R10 is the initial emission rate (µg/h), R20 is the time variable component of the emission rate (µg/h), k is the emission rate decay constant (h-1), and t is the elapsed time (h), 0 < t e 120. Three parameters, R10, R20, and k, are included in the model and are determined by least-squares fit of the experimental data. The resultant model parameters are tabulated in Table 2. Model 2 is a first-order decay model that was fit to the data for terpenes, such as limonene, R-pinene, β-pinene, VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2805
TABLE 2. Estimated Model Parameters for VOCs in the Absence of Ozone in Experiments experiment ID
VOCs
R10
1
benzaldehyde benzyl alcohol benzyl proprionate bornyl acetate isobornyl actetate benzaldehyde benzyl alcohol benzyl proprionate bornyl acetate isobornyl acetate
119.98 1691.00 537.05 64.52 5417.60 240.85 1792.50 557.43 75.79 6227.50
3
R20
k
77.98 0.0726 1695.60 0.0549 722.05 0.0395 58.83 0.0476 3370.90 0.4137 95.36 -0.0034 2164.00 0.0321 847.44 0.0210 158.05 0.0067 3567.30 0.1004
TABLE 3. Estimated Model Parameters for Terpenes in the Absence of Ozone in Experiments experiment ID
VOCs
R0
k
1
limonene R-pinene β-pinene camphene 3-carene limonene R-pinene β-pinene camphene 3-carene
3801.9 2945.3 2498.7 997.93 555.54 3555.1 2712.1 2529.7 908.53 547.12
0.0114 0.0097 0.0074 0.0056 0.0108 0.0113 0.0103 0.0071 0.0043 0.0095
3
compounds were detected by the DNPH-HPLC method. Among these compounds, Cmpd-1 (Figure 2, retention time 18.42 min) varied in a manner similar to limonene and other VOCs that can quickly react with ozone, implying that it existed in the original air freshener source and was also an unsaturated carbonyl that could react readily with ozone and thus consumed more ozone. The fact that another 10 unidentified compounds had time profiles similar to the profiles of nopinone and 4-AMC suggests that they were products of ozone reactions with VOCs. Figure 3 also shows the experimental and simulation profiles of NO2, where one finds that simulation results predicted only half as much as measured in the experiment, especially during the time when the ozone generator was activated. Underestimation of the NO2 concentration might be due to the uncertainties of the fate of NO2 in the system. In the presence of NO2, peroxyacetyl nitrate (PAN) analogue compounds with the general structure of R-C(O)O2N will be formed by the oxidation of VOCs. It is known that the radical initiated oxidation of pinonaldehyde, an R-pinene oxidation product, is expected to produce R-pinonyl peroxynitrate (RPPAN) (45). The differences seen in Figure 3 between observed and predicted concentrations of NO2 may be due to interference from RONO2 and ROONO2 compounds in the chemiluminescence NOx analyzer. Although above the detection limit, NO2 concentrations were below the practical quantification limit, which was about 26 µg/m3. Furthermore, the ozone and NO2 emission rates from the ozone generator air cleaner were calculated by the equations obtained from a previous experimental study (13):
camphene, and 3-carene
R(t) ) R0e-kt
(2)
where R(t) denotes the emission rate of a VOC (µg/h) at time t, R0 is the initial emission rate (µg/h), k is the emission rate decay constant (h-1), and t is the elapsed time (h), 0 < t e 120. The two parameters R0 and k fitted from chamber data are listed in Table 3. In experiments 1 and 3, when the ozone generator was turned on for the period of 24-48 h of the elapsed time, limonene, R-pinene, β-pinene, and 3-carene were found to decay rapidly because of their quick reactions with ozone. Camphene decayed at a lower rate consistent with its slower reaction rate with ozone. At the end of experiment 1, the air freshener sources were removed from the chamber, and VOCs were monitored for an additional 24 h. A rough calculation using Tenax GC/MS data indicated that the decay rates for VOCs emitted by the air freshener were slower (range 0.32-0.52 h-1) than the air exchange rate for the large chamber (0.55 h-1), indicating moderate reversible wall adsorption. Kinetic Simulations. Kinetic simulations with the PKSS package were performed, and results were compared to the experimental measurements. In general, modeling results of reactions of terpenes with ozone agreed well with the experimental data for both experiments 1 and 3. Figures 3-6 present part of the results from experiment 1 as examples. In Figure 3, comparisons of concentration-time profiles were made between the experimental and the simulated O3 and NO2 concentrations. The simulation reproduced the general tendency and curve shape of the ozone time series; however, the model prediction of O3 concentration was a little higher than the experimental data. In this complicated reaction system, only major VOC reactions with known kinetic data from the literature were included in the model. Minor components and secondary products can consume ozone as well, leading to a lower measured concentration of ozone in the reaction system. For example, 11 unidentified carbonyl 2806
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
O3 emission rate (mg/h) ) -0.0785 RH (%) + 7.8341 NO2 emission rate (mg/h) ) -0.0017 RH (%) + 0.3646 where the units of coefficients are in milligrams per hour. Thus, the variation of chamber test conditions is another possible explanation for the differences between measurement and model predictions of O3 and NO2 concentrations. All these uncertainties of the fate of NO2 in the system have to be considered in future tests. Figures 4a-6c show the measured and simulated concentration-time profiles of R-pinene, β-pinene, limonene, camphene, 3-carene, benzyl alcohol, benzyl propionate, benzaldehyde, bornyl acetate, isobornyl acetate, formaldehyde, 4-AMC, and nopinone, respectively, among which nopinone, 4-AMC, and formaldehyde were products. The simulated results for R-pinene, β-pinene, limonene, 3-carene, and nopinone agreed well with the experiment data, indicating that their reactions with ozone dominated their oxidation and degradation. The agreement between experimental data and the simulation of nopinone, a major product from the β-pinene-ozone reaction, in Figure 6c suggests that good predictions of yields of primary products are possible even when reaction mechanisms in the model are incomplete. The differences between model and experimental measurement for the reaction products, formaldehyde and 4-AMC, are likely due to lack of complete formation and decomposition mechanisms. Also, note that the rate constant of the 4-AMC-ozone reaction is 1.5E-16 cm3 molecule-1 s-1 (39), about 30000 times faster than that of nopinone. There are systematic discrepancies, however, between the simulation and the measurement for camphene, benzyl alcohol, benzyl propionate, benzaldehyde, bornyl acetate, and isobornyl acetate, especially during the time period when ozone was generated in the chamber. Although many factors, such as VOC loss to the chamber walls, contribute to these discrepancies, one of the most likely factors is the incomplete
FIGURE 3. O3 and NO2 measurements and simulations in experiment 1. Experimental data are represented by symbols, and model simulations are represented by solid lines. understanding of the complex reactions that occur in the presence of ozone. The benzaldehyde concentrations determined in the presence of ozone resulted from air freshener source emissions, VOCs-ozone reactions, and benzaldehyde-OH reaction removal. Noticeable differences between observed and simulated concentrations occurred where ozone was in the system (∼24 to ∼48 h), confirming again that resolution of uncertainties of reaction mechanisms is crucial to improved indoor air chemistry modeling. Sampling bias must also be considered. Differences between benzaldehyde concentrations determined by DNPH-HPLC and Tenax/thermal desorption/GC methods during the period where ozone was present in the chamber indicate that further investigation is needed to evaluate ozone and NO2 interference with Tenax sampling (refs 46 and 47 and references therein). The benzaldehyde data shown in Figure 5c were measured by the DNPH-HPLC method.
Particle Formation. It is well-known that many terpeneozone reactions produce particles. It has also been reported that reactions of VOCs, such as bornyl acetate, with OH radicals leads to organic aerosols (48). Particle formation was observed when the ozone generator air cleaner was turned on (Figure 7). For clarity of presentation, SMPS diameter bins were condensed into six composite bins with physical diameter range (µm) of 0.02-0.05, 0.05-0.1, 0.10.2, 0.2-0.3, 0.3-0.4, and 0.4-0.5. The total particle mass concentration was estimated from the particle number concentrations and diameters measured by the SMPS. The particles were assumed to be spherical and have a density of 1 g/cm3 (49, 50). Figure 7a unambiguously clarifies the carbon sources of aerosol particles, when the ultrafine and fine particles were formed, and how much aerosol was produced. Ultrafine and fine particles were formed when the ozone generator was on. Particle formation is part of the VOC, O3, and NO2 reaction process. It is possible that other VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2807
FIGURE 4. Terpenes measurements and simulations in experiment 1. Experimental data are represented by symbols, and model simulations are represented by solid lines. oxidant species were also involved in the particle formation process. Figure 7b shows the changes in particle number concentrations and particle size distributions for particles in the physical diameter range of approximately 0.02-0.5 µm. Ultrafine particles (physical diameter less than 0.1 µm) were formed with high number concentrations when ozone was initially introduced into the system. During the course of the experiment, the particle diameters increased, but their number concentration decreased. This phenomenon implies that the underlying mechanism of particle growth was through condensation in the early reaction process and subsequent gas-to-particle partitioning of the reaction products (24). 2808
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
The average particle mass concentration determined gravimetrically from a 4 h collection on a Teflon filter when ozone was in the system was 48.4 µg/m3 in experiment 1. With a defective Teflon filter in experiment 3, particle mass concentrations were estimated from the sum of the organic and inorganic mass determined from extraction of the quartz filter. From the OC/EC data, it was estimated that there was about 0.9 mg of organic carbon and 0.004 mg of elemental carbon collected on the quartz filter; thus, the minimum average concentration of total PM was estimated to be approximately 33.1 µg/m3 with the sample volume of 23.7 m3 over the 24 h sampling period. This is consistent with the SMPS measurements during the time when filter samples
FIGURE 5. VOCs measurements and simulations in experiment 1. Experimental data are represented by symbols, and model simulations are represented by solid lines. were collected, given that the average concentration was 43.2 µg/m3 in experiment 1 and 50.2 µg/m3 in experiment 3. The results of this study show clear potential for particle formation in indoor environments as a result of reactions of VOCs from the air freshener with ozone. Mass Balance. To assess the overall test procedure, mass balance was evaluated by comparing the total weight loss of air freshener refills with the sum of integrated mass of each identified and quantified VOC emitted to the chamber. The VOCs quantified by Tenax GC/MS include R-pinene, camphene, β-pinene, 3-carene, limonene, benzyl alcohol, benzyl propionate, bornyl acetate, and isobornyl acetate. Benzaldehyde was measured by Tenax GC/MS and DNPH-HPLC. The HPLC data were used for this calculation. The recovery was calculated by integration of the measured VOC concentrations over time multiplied by airflow and then divided
by the total weight loss of the air freshener refills. The mass recovery obtained was 64.3 and 59.5% with reactions taken into account as compared with 66.5 and 64.1% for experiments 1 and 3, respectively, if concentrations were integrated across the period of ozone generation without regard for concentration decreases due to reactions. Low recoveries are due to incomplete characterization of source emissions, chamber sink effects, and particle deposition. The recoveries for experiments 1 and 3, which were performed under very similar conditions, were comparable, and the relative percent difference (%RPD ) difference/average) was found to be 7.6% with reactions taken into account and 3.7% without, demonstrating good precision of the overall mass balance for these two tests. Implications. In this study, emissions and chemical degradation of VOCs from one electrical plug-in type air VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2809
FIGURE 6. Products measurements and simulations in experiment 1. Experimental data are represented by symbols, and model simulations are represented by solid lines. 2810
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
FIGURE 7. (a) Particle formation in experiment 1. (b) Particle number concentrations in the chamber under different size in experiment 1. freshener with pine-scented refills were investigated in the presence of ozone from an ozone generator air cleaner in the U.S. EPA large chamber facility. Models that employed SCIENTIST and PKSS software were utilized to simulate VOC emissions from the air freshener and their interactions with ozone under simulated indoor conditions. Particle size distributions and mass concentrations were also determined. The model developed here provides a framework to understand indoor air experiments performed in the EPA large chamber. Despite numerous uncertainties and the need to optimize some aspects of the mechanism to fit the chamber data, the indoor chemical mechanism employed in the PKSS package served as a reasonably good starting point. In some cases, the results were quite encouraging, indicating that the mechanism may have reasonably good predictive capability, at least for consumption of reactants. The method described herein has the potential to provide data to demonstrate the impact on IAQ of appliances that are marketed as VOC air
cleaners and provides the type of data that is needed to advance our understanding of indoor air chemistry and IAQ modeling. The present study indicates that full-scale chamber experiments can provide inputs to sophisticated indoor air chemistry models. However, better understanding of source emissions, reaction products, pathways, particle composition, gas-particle partitions, and the yields is also needed. The lack of detailed pathways for subsequent oxidation processes is a significant limitation since this precludes prediction of products that partition between gas and particle phases. More work is also needed to characterize particle deposition velocities and heterogeneous reactions on indoor surfaces so that explicit descriptions of important surface reactions other than unimolecular decomposition and irreversible adsorption may be incorporated into the model. Concentrations of many species are significantly perturbed by chemical reactions. Determination of concentrations of OH, NO2, VOL. 38, NO. 10, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2811
HONO, H2O2, and other reactive species in the test chamber are needed to advance indoor air chemistry modeling. Prediction, identification, and quantification of a variety of aldehydes from the test systems show the importance of the present study because of potential adverse health concerns of aldehydes. The highest concentration of formaldehyde measured in experiment 1 was 28.2 µg/m3 at elapsed time 29.13 h in the presence of ozone, and the total measured aldehyde concentration (formaldehyde, benzaldehyde, 4-AMC, and nopionone) at this time was about 50 µg/m3. Clearly, more data and improved models are needed to predict particle formation in the real world. The particle concentrations in the test systems that resulted from the interactions of ozone and air freshener emissions were close to the 24 h national ambient air quality standard of 65 µg/m3 for fine particles less than 2.5 µm in diameter, demonstrating that indoor air chemical reactions have the potential to significantly increase human exposure to fine particles. Source exposure could be reduced, however, by avoiding indoor sources of O3 (e.g., from ozone air cleaners) and by reducing the use of consumer products that contain or emit biogenic VOCs that react readily with ozone.
Acknowledgments The authors thank Scott Moore, Ivan Dolgov, and Richard Perry from the U.S. EPA for operating the large chamber control systems and online instruments and Nancy Roache from ARCADIS for collecting Tenax and DNPH samples. We also thank Dr. Yuanji Dong from ARCADIS for performing denuder and filters extraction and methylation and Dr. Kara Linna from the U.S. EPA for ion analysis using ion chromatography.
Literature Cited (1) Wainman, T.; Zhang, J.; Weschler, C. J.; Lioy, P. J. Environ. Health Perspect. 2000, 108, 1139. (2) Sarwar, M.; Corsi, R. L.; Allen, D. T.; Weschler, C. J. Production and levels of selected indoor radicals: a modeling assessment, In Indoor Air 2002: Proceedings of the 9th International Conference on Indoor Air Quality and Climate, Santa Cruz, CA, 2002. (3) Weschler, C. J.; Shields, H. C. Environ. Sci. Technol. 1997, 31, 3719. (4) Weschler, C. J.; Shields, H. C. Environ. Sci. Technol. 1996, 30, 3250. (5) Sarwar, G.; Corsi, R.; Kimura, Y.; Allen, D.; Weschler, C. J. Atmos. Environ. 2002, 36, 3973. (6) Morrison, G. C.; Nazaroff, W. W. Environ. Sci. Technol. 2002, 36, 2185. (7) Weschler, C. J.; Hodgson, A. T.; Wooley, J. D. Environ. Sci. Technol. 1992, 26, 2371. (8) Reiss, R.; Ryan, P. B.; Koutrakis, P.; Tibbetts, S. J. Environ. Sci. Technol 1995, 29, 1906. (9) Reiss, R.; Ryan, P. B.; Tibbetts, S. J.; Koutrakis, P. J. Air Waste Manage. Assoc. 1995, 45, 811. (10) Lee, D. S.; Holland, M. R.; Falla, N. Atmos. Environ. 1996, 30, 1053. (11) Fick, J.; Pommer, L.; Andersson, B.; Nilsson, C. Atmos. Environ. 2002, 36, 3299. (12) Fick, J.; Pommer, L.; Andersson, B.; Nilsson, C. Atmos. Environ. 2003, 37, 4087. (13) Mason, M. A.; Sparks, L. E.; Moore, S. A.; Dolgov, I.; Perry, R. B. Characterization of ozone emissions from air cleaners equipped with ozone generators and sensor and feedback control circuitry, In Engineering Solutions to Indoor Air Quality Programs Symposium, Research Triangle Park, NC, 2000. (14) Yang, Q.; Chen, Y.; Shi, Y.; Burnett, R. T.; McGrail, K. M.; Krewski, D. Inhalation Toxicol. 2003, 15, 1297. (15) Moolgavkar, S. H. Inhalation Toxicol. 2003, 15, 877. (16) Nikasinovic, L.; Momas, I.; Seta, N. J. Toxicol. Environ. Health 2003, 6, 521.
2812
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 10, 2004
(17) Wolkoff, P.; Clausen, P. A.; Jensen, B.; Nielsen, G. D.; Wilkins, C. K. Indoor Air 1997, 7, 92. (18) Brunekreef, B.; Holgate, S. T. Lancet 2002, 360, 1233. (19) El-Fadel, M.; Massoud, M. Sci. Total Environ. 2000, 257, 133. (20) Oberdo¨rster, G. Int. Arch. Occup. Environ. Health 2000, 74, 1. (21) Nazaroff, W. W.; Salmon, L. G.; Cass, G. R. Environ. Sci. Technol. 1990, 24, 66. (22) Nazaroff, W. W.; Cass, G. R. Environ. Sci. Technol. 1989, 23, 157. (23) Weschler, C. J.; Shields, H. C.; Shah, B. M. J. Air Waste Manage. Assoc. 1996, 46, 291. (24) Sarwar, M.; Corsi, R. L.; Kimura, Y.; Allen, D. T.; Weschler, C. J. Atmos. Environ. 2003, 37, 1365. (25) Nazaroff, W. W.; Cass, G. R. Environ. Sci. Technol. 1986, 20, 924. (26) Weschler, C. J.; Shields, H. C. Indoor Air 2000, 10, 92. (27) Sorensen, D. N.; Weschler, C. J. Atmos. Environ. 2002, 36, 9. (28) Sinclair, J. D.; Psota-Kelty, L. A.; Weschler, C. J.; Shields, H. C. Atmos. Environ. 1990, 24A, 627. (29) Howard, E. M.; Mason, M. A.; Zhang, J.; Brown, S. A comparison of design specifications for three large environmental chambers, In the Proceedings of an International Symposium Sponsored by the US EPA & Air Waste Management Association, Research Triangle Park, NC, 1995. (30) ASTM Standard Test Methods for Continuous Measurement of Ozone in Ambient, Workplace, and Indoor Atmospheres (Ultraviolet Absorption), ASTM Designation D 5156-95, Annual book of ASTM standards; American Society for Testing and Materials: Philadelphia, PA, 1995; Vol. 11.03. (31) ASTM Standard Test Methods for Continuous Measurement of Oxides of Nitrogen in the Ambient or Workplace Atmosphere by the Chemiluminescent Method, ASTM Designation D 3824-95, Annual book of ASTM standards; American Society for Testing and Materials: Philadelphia, PA, 1995; Vol. 11.03. (32) Liu, X.; Jeffries, H. E.; Sexton, K. G. Atmos. Environ. 1999, 33, 3005. (33) Ambient air monitoring reference and equivalent methods. Code of Federal Regulations, Part 53, Title 40, 2002. (34) Quality assurance guidance document 2.12, monitoring PM2.5 in ambient air using designated references or class 1 equivalent methods; U.S. Environmental Protection Agency, Office of Research and Development, National Exposure Research Laboratory: Research Triangle Park, NC, 1998. (35) Jeffries, H. E. PC-Photochemical Kinetics Simulation System (PC-PKSS), software Version 3.0; Department of Environmental Sciences and Engineering, School of Public Health, University of North Carolina: Chapel Hill, NC, 1991. (36) Guo, Z.; Roache, N. F. Ann. Occup. Hyg. 2003, 47, 279. (37) Kamens, R. M.; Myoseon, J.; Chien, C.; Leach, K. Environ. Sci. Technol. 1999, 33, 1430. (38) Kamens, R. M.; Jaoui, M. Environ. Sci. Technol. 2001, 35, 1394. (39) Calogirou, A. Atmos. Environ. 1999, 33, 1423. (40) Braun, D.; Kuchler, U.; Pietsch, G. Pure Appl. Chem. 1988, 60, 741. (41) Kaiga, N.; Takase, O.; Todo, Y.; Yamanashi, I. Ozone 1997, 19, 169. (42) Adelman, Z. CARBON BOND-IV•99. Master’s Thesis. Department of Environmental Sciences and Engineering, School of Public Health, University of North Carolina, Chapel Hill, NC, 1999. (43) Febo, A.; Perrino, C. Atmos. Environ. 1991, 25A, 1055. (44) Atkinson, R.; Arey, J. Atmos. Environ. 2003, 37 Suppl. 2, S197. (45) Noziere, B.; Barnes, I. J. Geophys. Res. 1998, 103, 25587. (46) Fick, J.; Pommer, L.; Andersson, B.; Nilsson, C. Environ. Sci. Technol. 2001, 35, 1458. (47) Pommer, L.; Fick, J.; Andersson, B.; Nilsson, C. Atmos. Environ. 2002, 36, 1443. (48) Coeur, C.; Jacob, V.; Foster, P. Atmos. Environ. 1999, 33, 1615. (49) Sioutas, C.; Abt, E.; Wolfson, J. M.; Koutrakis, P. Aerosol Sci. Technol. 1999, 30, 84. (50) Rohr, A. C.; Weschler, C.; Koutrakis, P.; Spengler, J. D. Aerosol Sci. Technol. 2003, 37, 65.
Received for review July 10, 2003. Revised manuscript received February 26, 2004. Accepted March 1, 2004. ES030544B
