The significance of secondary organic aerosol formation and growth in buildings: experimental and computational evidence

Atmospheric Environment 37 (2003) 1365–1381

The significance of secondary organic aerosol formation and growth in buildings: experimental and computational evidence Golam Sarwara, Richard Corsia,*, David Allena, Charles Weschlerb a

Center for Energy and Environmental Resources (R7100), The University of Texas at Austin, 10100 Burnet Road, Austin, TX 78758, USA b Environmental and Occupational Health Sciences Institute, UMDNJ/Robert Wood Johnson Medical School and Rutgers University, Piscataway, NJ 08854, USA Received 16 July 2002; accepted 27 September 2002

Abstract Experiments were conducted in an 11 m3 environmental chamber to investigate secondary particles resulting from homogeneous reactions between ozone and a-pinene. Experimental results indicate that rapid fine particle growth occurs due to homogeneous reactions between ozone and a-pinene, and subsequent gas-to-particle partitioning of the products. A new indoor air quality model was used to predict dynamic particle mass concentrations based on detailed homogeneous chemical mechanisms and partitioning of semi-volatile products to particles. Chamber particle mass concentrations were estimated from measured particle size distributions and were in reasonable agreement with results predicted from the model. Both experimental and model results indicate that secondary particle mass concentrations increase substantially with lower air exchange rates. This is an interesting result, given a continuing trend toward more energy-efficient buildings. Secondary particle mass concentrations are also predicted to increase with lower indoor temperatures, higher outdoor ozone concentrations, higher outdoor particle concentrations, and higher indoor a-pinene emissions rates. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: Indoor chemistry; Ozone; Secondary particles; Terpenes; a-Pinene

1. Introduction Recent epidemiological and related studies have linked fine particles to human health and welfare effects including premature mortality (Pope and Dockery, 1996; USEPA, 1997). These findings have prompted the United States Environmental Protection Agency (USEPA) to promulgate a new national ambient air quality standard for fine particles. While the aforementioned regulations focus entirely on outdoor air quality, the average American spends 18 h indoors for every hour spent outdoors (Robinson et al., 1991). Thus, *Corresponding author. Tel.: +1-512-232-3611; fax: +1512-471-1720. E-mail address: [email protected] (R. Corsi).

human exposure to fine particles in indoor environments can be a major fraction of total exposure to fine particles. Additionally, based on the results of bioassay experiments involving rats, Long et al. (2001) recently concluded that particles generated indoors may be more bioactive than particles generated outdoors. Santannam et al. (1990) reported the results of indoor and outdoor sampling completed at 280 residences in Ohio and Wisconsin. The average ‘‘unexplained’’ percentage of indoor fine particles, i.e., those that could not be attributed to specific sources, ranged from 19% to 42%. Similar results (average of 25% ‘‘unexplained’’ fine particles) were observed during the USEPA’s PTEAM study (Wallace, 1996). Wallace concluded that the unexplained fraction of fine particles should be the focus of future studies. Wainman (1999) further

1352-2310/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S1352-2310(02)01013-0

1366

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

reported that the ‘‘unknown’’ indoor fine particle concentration in the PTEAM study was 8.75 mg/m3 during the daytime, and 6.75 mg/m3 during nighttime. Recent studies suggest that indoor air chemistry, particularly as related to reactions between terpenes that originate from indoor sources and ozone (O3) that originates outdoors, can be an important source of indoor fine particles (Weschler and Shields, 1999; Wainman et al., 2000; Long et al., 2000; Rohr et al., 2003; Weschler and Shields, 2002). Weschler and Shields (1999) selectively introduced limonene, a-terpinene and a terpene-based cleaner into two adjacent unoccupied and identical offices. In one of the offices, an O3 generator raised the indoor O3 levels to 200–300 ppb; the adjacent office did not have an O3 generator. The authors reported a significant increase in particle number concentrations in the sub-micron range in the office with elevated O3. The authors attributed the formation and growth of indoor particles to O3/terpene reactions and condensation of the subsequent byproducts. In subsequent experiments conducted in these same adjacent offices, rather than using an ozone generator, the authors depended on outdoor-to-indoor transport as the source of ozone—typically between 2 and 40 ppb. In these experiments, one of the offices contained a limonene source; the other did not. The office with the limonene source had significantly higher levels of sub-micron particles when indoor ozone levels were elevated. The difference in particle levels between the two offices tracked the indoor ozone level and was roughly 20 mg/m3 when indoor ozone was in the range of 25–30 ppb. Long et al. (2000) reported evidence that chemical reactions are a potentially important source of ultra-fine particles in indoor environments. The authors conducted six sampling events in Boston homes during which a pine-oil-based cleaner was used for mopping floors and toilet cleaning. Ozone was not deliberately added to these homes; however, O3 was present indoors due to outdoor-to-indoor transport. Five of the six sampling events demonstrated significant fine particle formation. Particle number concentrations increased by 7–100 times relative to the original particle number concentrations. More than 50% of the particles (by volume) generated during these five events were ultrafine in nature. A nested dynamic system consisting of an inner and an outer chamber was used by Wainman et al. (2000) to investigate indoor particle formation and growth due to O3/limonene reactions. Air exchange could occur between the inner and outer chambers. Limonene and O3 were injected incrementally into the inner chamber. Increases in particle numbers in the 0.1–0.2 mm size range were observed shortly after the first O3 injection, but formation of particles in the 0.2–0.3 mm size range did not occur until a second ozone injection. Combining

the results of this study with the earlier study of Weschler and Shields (1999), the investigators concluded that reactions of terpenes that originate from indoor sources and O3 that originates outdoors can increase indoor particle mass concentration by more than 20 mg/m3. The authors noted that indoor ozone/terpene reactions provide a mechanism whereby indoor secondary particle mass concentrations can vary coincidentally with outdoor fine particles during periods of elevated outdoor ozone. Rohr et al. (2003) recently conducted experiments with a-pinene and O3 in a plexiglas chamber as part of a mouse bioassay study. Ozone concentrations during these experiments were in the range of 100–400 ppb and a-pinene concentrations were about 100 ppm, both much higher concentrations than typically encountered in indoor environments. Significant increases in ultrafine particle number concentrations were reported as a result of a-pinene and O3 reactions. Weschler and Shields (2002) also conducted experiments to determine the effects of air exchange rates on indoor secondary particle size distribution and mass concentrations. They introduced limonene into two adjacent unoccupied and identical offices. In one of the offices, an O3 generator was operated to raise the indoor O3 levels to 50–425 ppb. The other office did not have an O3 generator. The authors reported that the lower air exchange rates led to increased particle mass concentrations and shifted the resulting secondary particle size distribution toward larger sizes. Such findings, albeit sparse and somewhat preliminary in nature, might provide an explanation for the ‘‘unexplained’’ indoor fine particles reported in previous studies. In this study, the effects of reactions between O3 and a-pinene on indoor particles were investigated using chamber experiments and a new indoor air quality model.

2. Methodology 2.1. Experimental methodology Experiments were conducted in a stainless steel chamber (11 m3) located at the Center for Energy and Environmental Resources (CEER) at the University of Texas at Austin. The chamber has dimensions of 2.4 m
1.8 m
2.5 m (L
W
H), yielding a surfaceto-volume ratio of 2.7 m1, a value similar to that found in many commercial and residential buildings (Nazaroff et al., 1993). A dedicated blower was used to draw air through the chamber (once through system). Inlet air was provided from within the general occupied space of the building. The inlet air was not conditioned for the experiments described herein. Thus, it contained some amount of common indoor pollutants, including

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

terpenes, O3, and particles. A diagram of the environmental chamber and associated instrumentation is shown in Fig. 1. A commercial corona discharge ozone generator (Living Air, BORA-IV) was placed in the middle of the chamber and operated for 10–20 h to achieve steadystate O3 concentrations prior to the introduction of a-pinene. A 4 ml vial containing pure reagent-grade a-pinene was then placed on the floor of the chamber, allowing for evaporation of a-pinene. The difference in vial mass before and after each experiment was combined with the duration of the experiment to obtain the average a-pinene emission rate. This rate was also confirmed gravimetrically with separate experiments under a laboratory fume hood. Two fans (SMCs, Model TR12 and Coll-Breezet, Model EB24925) were operated near the floor of the chamber to enhance mixing. A space heater was used to maintain elevated temperatures in the chamber during two of the ten experiments. Chamber particle number concentrations were continuously measured using two particle counters (Particle Measuring Systems Inc., LASAIRs—model 1002; and TSIt, P-TRAKt—model 8525). The P-TRAKt was used to measure number concentrations of particles between 0.02 and 1.0 mm in diameter. The sample averaging time of the P-TRAKt was set to 60 s. The LASAIRs measures total particle numbers in eight different channels corresponding to diameters of (all in mm) 0.1–0.2, 0.2–0.3, 0.3–0.4, 0.4–0.5, 0.5–0.7, 0.7–1.0, 1.0–2.0, and >2.0. The sampling interval of the instrument was set at 60 s. Total particle number in each channel was divided by the total sampling volume to determine the average particle number concentration during the sampling interval. Particle number concentrations from the first six channels of the LASAIRs were added to obtain the total particle number concentrations for the size range of 0.1–1.0 mm. This total particle number concentration was then subtracted from the particle number concentration detected by the P-TRAKt to obtain the particle number concentration in the range of 0.02–0.1 mm. Thus, the two particle

Blower

O 3 Analyzer

Particle Counters

O 3 Generator

SF6 Analyzer

Terpene Source

HVAC

Hygrometer

Fans (2)

Fig. 1. Schematic diagram of the environmental chamber.

1367

analyzers collectively provided particle number concentrations in nine different diameter ranges. This subtraction technique introduced some level of uncertainty for concentrations in the 0.02–0.1 mm diameter range. Nevertheless, the results provided useful information in the ultra-fine range of particles. Mass concentration was estimated from particle number concentrations and particle density. Two different approaches were employed as described below. Method 1: A geometric mean diameter was estimated from the minimum and maximum particle diameters in each size range. The average volume of particles in each size range was determined by assuming that each particle was a sphere with a diameter equal to the geometric mean diameter. The total particle volume in each of the size ranges was estimated by multiplying the average volume of the particle by the particle number concentration in that size range. Particle volumes in each of the size ranges were summed to obtain the total particle volume. Total particle volume was then multiplied by particle material density to obtain particle mass concentration. Turpin and Lim (2001) analyzed atmospheric particle density data and suggested a material density of 1.2 g/cm3 for atmospheric organic aerosols, a value adopted for this study. This method has been shown to produce reasonable particle mass concentrations in previous studies. For example, Fan (2002) concurrently measured particle number and mass concentrations during experiments involving reactions of O3 with organic compounds. Particle mass concentrations were measured using a filter-based technique. Particle mass concentrations were also derived using measured particle number concentrations and the geometric mean diameters as described above. The derived particle mass concentrations were 5% lower than the results obtained using the filter-based technique. For this study, the experimental mass concentrations obtained using Method 1 contained initial oscillations due to growth wave kinetics and the coarseness of particle size channels measured by the analyzers. Therefore, this method was discarded. Method 2 (functional fit to data): The particle size ranges and number concentrations measured by the LASAIRs were used to develop an exponential best-fit curve through the data. As discussed earlier, particle number concentrations in the 0.02–0.1 mm had potentially high uncertainties and were not included in the curve fitting. For purposes of curve fitting, each size range was represented by the geometric mean diameter estimated from the minimum and maximum particle diameters in each size range. For example, particle number concentrations in the 0.1–0.2 mm size range were represented by a geometric mean diameter of 0.14 mm. The resulting curve was used to estimate particle number concentrations in small size increments, starting from 0.1 mm and ending at 0.7 mm with increments of 0.01 mm.

1368

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

The resulting particle number concentrations were used to estimate particle mass concentrations using the procedure described for Method 1. For each experiment, this procedure was repeated for each time step. The LASAIRs and P-TRAKt particle counters were located outside the chamber during each experiment. Sample tubing inlets were located in the middle of the chamber approximately 1.5 m above the floor. The LASAIRs sample tube inlet was placed within about 20 cm of the P-TRAKt sample tube inlet. The LASAIRs drew sample air from within the chamber through a 3.5 m long (3 mm OD) polyurethane tube. The P-TRAKt pulled air through a 5 m long (6 mm OD) tygon tube. Particle losses in tubing were determined by running the analyzers with and without tubing. The average loss of particles was 25% in the 5 m tube used for the P-TRAKt and 14% in the 3.5 m tube used for the LASAIRs. Particle numbers measured by these analyzers were adjusted to account for losses in tubing. Chamber O3 (inside and outside) was measured using an ozone analyzer (Dasibi Environmental Corporation, model 1003-AH). A solenoid valve was used to switch the intake of the O3 analyzer between the interior and exterior of the chamber. The solenoid valve was programmed to maintain the intake of the ozone analyzer inside the chamber for 50 min before switching to outside the chamber for a period of 10 min for each hour of the experiment. The analyzer was placed outside the chamber and drew sample air through approximately 3 m (6 mm OD) of teflon tubing. The chamber air exchange rate was measured by injecting sulfur hexafluoride (SF6) and monitoring its decay using a GC/ECD optimized for analysis of SF6 (Lagus Applied Technology, Inc., model 101, Autotrac). Chamber temperature and relative humidity were measured during each experiment using a portable digital hygrometer (VWR, model 35519–041). Temperature and relative humidity were recorded every 30–60 min during each experiment. Each experiment began with the measurement of particle number concentrations outside the chamber for 10–30 min followed by measurements inside the chamber for another 10–30 min before the introduction of a-pinene into the chamber. Experiments continued for a period of 10–12 h after source introduction into the chamber. Particle number concentrations outside the chamber were measured for 10–30 min at the end of each experiment. The LASAIRs and P-TRAKt particle counters were calibrated by the manufacturers as per recommendation. Periodic zero checks were conducted on the LASAIRs and on the P-TRAKt during these experiments. The ozone analyzer was calibrated using primary standards 2 weeks prior to and at the end of all experiments. The analyzer responses were within 1% of primary standards.

2.2. Modeling methodology The indoor chemistry and exposure model (ICEM) was used in this study. Details of the ICEM have been described elsewhere (Sarwar et al., 2002). The building interior was modeled as a single well-mixed environment with homogeneous chemistry. In its simplest mathematical form, the ICEM can be represented by n X dCi Ei ¼ ki lCoi  lCi þ  Vdi Ci a þ Rij ; dt V j¼1

ð1Þ

where Ci is the indoor concentration of pollutant i (ppm), Coi is the outdoor concentration of pollutant i (ppm), ki is the outdoor-to-indoor penetration factor for pollutant i (0–1; unitless), l is the fresh air exchange rate (min1), Ei is the whole-building emission rate (all sources) for pollutant i (ppm m3/min), Vdi is the deposition velocity for pollutant i to indoor materials (m/min), a is the surface area to volume ratio for the indoor environment (1/m), Rij is the reaction rate between pollutant i and pollutant j (ppm/min), V is the volume of the indoor environment (m3), and t is the time (min). The first term in Eq. (1) corresponds to the time rate of change of pollutant concentration (Ci ) within a building environment. The second term represents the transport of pollutants from outdoors to indoors. The third term represents the transport of pollutants from indoors to outdoors. The fourth term corresponds to the whole-building emission rate for pollutant i: The fifth term corresponds to the removal of pollutant i by deposition on indoor materials. The sixth term corresponds to the summed results of all reactions involving pollutants i and j: The ICEM employs the detailed homogeneous chemistry of SAPRC-99 (Carter, 2000). Particle chemistry for a-pinene, developed by Kamens et al. (1999), was also incorporated into ICEM and was combined with the homogeneous reactions of SAPRC99. Outdoor photon energy was replaced by typical indoor photon energy. The current ICEM does not include adsorption/desorption phenomena that may occur in indoor environments, but as shown in Eq. (1), does include a term for irreversible deposition of particles, ozone, free radicals, and semi-volatile organic compounds. 2.3. Particle chemistry for a-pinene Kamens et al. (1999) conducted experiments in a 190 m3 outdoor Teflon film chamber under dark conditions and used the results to develop the particle formation chemistry for a-pinene. An overview of the particle chemistry of a-pinene is provided here; details can be found in Kamens et al. (1999). Particle chemistry for a-pinene can be divided into two parts: (1) gas-phase

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1369

chemistry leading to semi-volatile products and (2) particle growth by partitioning of semi-volatile products onto existing particles. The particle chemistry of apinene consists of six semi-volatile model compounds: pinald, pinacid, diacid, oxy-pinald, p3gas, and oxypinacid as shown in Fig. 2. The model compound pinald represents pinonaldehyde and nor-pinonaldehyde, pinacid represents pinonic and nor-pinonic acids, diacid represents pinic and nor-pinic acids, oxy-pinald represents hydroxy and aldehyde-substituted pinonaldehydes, p3gas represents a pinic acid precursor (2,2-dimethylcyclobutane-3-acetylcarboxylic acid—C9H14O3), and oxypinacid represents hydroxy and aldehyde-substituted pinonic acids.

The decomposition of CRIEGEE2 leads to oxypinald, CRGPROD2 (a generalized intermediate product), a stabilized criegee 2 radical, HCHO, OH  and HO2 . Stabilized Criegee 2 radicals can react with H2O to produce p3gas and methanol (CH3OH). The compound p3gas reacts with OH  producing a peroxyacyl radical (PREDI-OO), which in turn produces diacid via oxidation with HO2 . The compound CRGPROD2 reacts with HO2 to form diacid. a-Pinene also reacts with OH  to produce pinald, oxy-pinald, and volatile oxygenated products called FRAG.

2.4. Gas-phase chemistry

The six semi-volatile products described above can partition on to ‘‘seed’’ particles that originate largely from outdoor-to-indoor penetration. Some of these products have very low vapor pressures and can selfnucleate to form new particles and are designated as ‘‘seed1’’ in the model. Partitioning was treated as sorption to, and desorption of semi-volatile products from, particles. The ratio of the sorption rate (kon;i ) and desorption rate (koff;i ) constants was taken as the equilibrium constant (Kp;i ) for gas–particle equilibrium:

a-Pinene reacts with O3 to produce an ozonide. However, the ozonide is highly unstable and rapidly decomposes into two Criegee biradicals (CRIEGEE1 and CRIEGEE2). Several reaction pathways are available for the decomposition of these Criegee biradicals. The decomposition of CRIEGEE1 leads to pinacid, stabilized criegee 1 radical, pinald, oxy-pinald, OH  , HO2 , and CO. Stabilized Criegee 1 biradicals react with H2O to form pinacid. In the presence of O2, pinald reacts with OH  to form a peroxyacyl radical (PINALD-OO), which is subsequently oxidized via HO2 to produce pinacid. Oxy-pinald is oxidized to oxy-pinacid via OH  and HO2 .

+ O3

2.5. Partitioning of semi-volatile products on to ‘‘seed’’ particles

Kp;i ¼ kon;i =koff;i :

Here, Kp;i represents the ratio of the sorbed-phase concentration of species i (ng/mg) to the gas-phase concentration of i (ng/m3) and has a unit of m3/mg. It

(ozonide)

CRIEGEE1

STABILIZED CRIEGEE 1

ð2Þ

OH.

CRIEGEE2

HO2.

CO

OXY-PINALD

CRGPROD2

STABILIZED CRIEGEE 2

HCHO OH. HO2.

PINACID

PINALD OH.

H2O

HO2.

OH.

HO2.

PINALD-OO

H2O

OXY-PREPINACID DIACID HO2.

P3GAS

CH3OH

OH.

OXY-PINACID

HO2.

PREDI-OO

+ OH. FRAG

PINALD

OXY-PINALD

Fig. 2. Reaction between a-pinene and O3 (semi-volatile organic compounds are shown in bold italics).

1370

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

was estimated from knowledge of the vapor pressure and molecular weight of a product according to Kp;i ¼ 7:501RT=109 MWom i p0L ;

ð3Þ 3

where Kp;i is the equilibrium constant (m /mg) for compound i; R is the universal gas constant (8.31 J/ K mol), T is the temperature (K), MWom is the average molecular weight of the ‘‘liquid’’ that constitutes the organic particles (g/mol), i p0L is the vapor pressure of the product (Torr), and 7.51 is a conversion factor (Kamens et al., 1999). Absorption of semi-volatile products on to particles is a weak function of temperature; thus it was assumed that sorption rates (kon;i ) are independent of temperature. Desorption rate constants (koff;i ) of semi-volatile products from particles are strongly dependent on temperature and were estimated using koff;i ¼ b expðEa =RTÞ;

ð4Þ

where b is a pre-exponential factor, R is the universal gas constant (8.31 J/K mol), T is the temperature (K), and Ea is the activation energy (J/mol). A general equation for evaporation from a liquid surface was used to estimate values of b and Ea (Kamens et al., 1999). Sorption rates (kon;i ) for semi-volatile products were estimated as the product of equilibrium constant, Kp;i ; and the desorption rate constant (koff ) for any given compound. Secondary particle mass associated with a-pinene reactions can be estimated by keeping track of all seven products (six semi-volatile products and ‘‘seed1’’, the self-nucleated product) in the particle phase. Based on experiments by Kamens et al. (1999), an average molecular weight of 120 g/mol was used to convert secondary particle concentrations from volumetric units (ppm) to mass units (mg/m3). ‘‘Seed’’ particles needed for the model were estimated based on outdoor-to-indoor transport. Based on compositional analysis of outdoor particles measured in Houston (Tropp et al., 1998), an average molecular weight of 90 g/mol was used to convert ‘‘seed’’ particle concentrations from volumetric units (ppm) to mass units (mg/m3). Kamens et al. (1999) also determined losses of semivolatile reaction products and particles to Teflon film chamber walls. Several reactions were included in the particle formation mechanism of a-pinene to account for these wall losses. The ICEM estimates surface removal rates using prescribed deposition velocities, the surface area to volume ratio, and pollutant concentrations in indoor environments. Thus, reactions that account for chamber wall losses in the a-pinene particle formation mechanism were switched off, allowing the ICEM to estimate surface removal rates of these compounds based on prescribed deposition velocities. Deposition velocity data for semi-volatile products in actual building environments are not known. In the

absence of any measured data in indoor environments, deposition velocities for semi-volatile organic products of a-pinene were assigned twice the deposition velocity of fine particles following Kamens et al. (1999). The authors acknowledge that this assumption is a source of uncertainty in the model and will require future refinement. Deposition velocities for O3 and particles were previously measured in the chamber by comparing decay rates at specific air exchange rates with an inert tracer gas (SF6) and were used in the ICEM. The O3 deposition rate in the stainless steel chamber used for this study was 0.48 h1, much lower than the reported average O3 deposition rate of 3.6 h1 for indoor environments (Nazaroff and Cass, 1986). The surface area to volume ratio of the environmental chamber used in this study was 2.7 m1. With this surface area to volume ratio, the deposition velocity of O3 in the chamber was 0.005 cm/s. The average deposition rate of fine particles in the chamber was 0.34 h1. The measured particle deposition rate compares favorably with the fine particle deposition rate of 0.33 h1 reported by Wainman (1999) and 0.39 h1 reported by Wallace (1996). The average deposition velocity of particles in the chamber was 0.0035 cm/s. 2.6. Model input parameters Measured a-pinene emission rates, O3 emission rates, chamber outside O3 concentrations, chamber outside particle mass concentrations, chamber average air exchange rates, chamber temperature, relative humidity, and chamber surface area and volume were provided as input data to the model to estimate chamber particle mass concentrations as a function of time.

3. Results and discussions A summary of the ten experiments involving a-pinene is presented in Table 1. Number concentrations represent particles in the 0.02–0.7 mm diameter range and mass concentrations represent particles in the 0.1–0.7 mm diameter range. The ozone generator was not operated during experiment 1; so for this experiment the O3 levels were in the range of 10–15 ppb and were derived from outdoor-to-indoor penetration. A typical evolution of particle number concentrations is shown in Fig. 3, in this case for experiment 3. A rapid ‘‘burst’’ of small particles in the 0.02–0.1 mm size range was detected shortly after the introduction of a-pinene into the chamber. Particle numbers in the 0.02–0.1 mm size range then decreased before attaining a steady concentration of approximately four times the initial concentration. Particle numbers in the next size range then increased and subsequently decreased before attaining a steady concentration. This process continued

Note: l=air exchange rate, E=a-pinene emission rate, T=Time at which the maximum particle number concentrations occurred in the chamber, Mi =initial particle mass concentration in the chamber, Mf;e =final experimental particle mass concentration in the chamber, Mf;p =final predicted particle mass concentration in the chamber, SOA=secondary organic aerosol=Mf;e  Mi ; difference=ðMf;e  Mf;p Þ=Mf;e :

16 44 26 22 3 35 57 20 35 146 5.0 55 27 20 30 66 52 21 8.8 181 0.7 35 19 12 20 43 22 10 3.2 71 4.3 39 21 16 29 49 33 17 6.5 74 3.6 3.1 2.2 4.3 9.2 5.4 11 6.9 3.3 2.7 1997 10004 7227 6774 7324 9511 6327 35975 25332 9180 108 17 34 39 31 48 26 48 48 32 3423 96989 34662 11852 18529 14532 30114 31062 37729 25131 2760 7047 1383 9202 4138 10391 2836 8472 12529 4438 25 24 26 24 25 24 25 32 33 25 15 132 134 111 136 170 142 157 144 159 1.25 0.55 0.68 1.06 0.70 0.70 0.75 1.34 1.50 0.71 1 2 3 4 5 6 7 8 9 10

196 75 81 78 70 91 98 89 102 185

T (1C) Initial O3 (ppb) E (mg/ min) l (h1) Expt no.

Table 1 Summary of chamber experimental and modeling results

Initial particle (#/cm3)

Max particle (#/cm3)

Time (min)

Final particle (#/cm3)

Mi (mg/ m3)

Mf;e (mg/ m3)

SOA (mg/ m3)

Mf;p (mg/ m3)

Difference (%)

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1371

for particles with diameters up to 0.5–0.7 mm, after which no appreciable increase in particle numbers was observed. The initial ‘‘burst’’ of small particles followed by a decrease in particle numbers in a given size range and increase in particle numbers in subsequent size ranges created an effective particle growth ‘‘wave’’ that was observed for all experiments. Ultimately, particle number concentrations reached steadystate conditions controlled primarily by the reaction rate of O3 with a-pinene and the prevailing air exchange rate. The particle size distributions during experiment 3 at the time of a-pinene introduction into the chamber (time=0) and at steady state (time=700 min) are shown in Fig. 4. Particle number concentrations in all size ranges at steady state were consistently higher than the particle number concentrations at time 0. Similar results were also observed for other experiments. The increase in particle number concentrations occurred via the O3/apinene reactions and partitioning of subsequent reaction by-products. Particle surface area and volume at time 0 for experiment 3 are shown in Fig. 5. While the initial particle number concentration was dominated by particles with diameter with o0.1 mm, the maximum particle surface area and particle volume both occurred in the 0.2–0.3 mm size range. The bulk of the particle surface area was contained among the 0.1–0.2, 0.2–0.3, and 0.3–0.4 mm size ranges. While there was some initial growth of larger size particles, the initial ‘‘burst’’ of particles was observed in the 0.02–0.1 mm size range. This suggests that the initial ‘‘burst’’ of small particles may not be solely the result of condensation/absorption of semi-volatile particles onto smaller seed particles, in which case an initial ‘‘burst’’ would be expected in the larger size ranges. The initial ‘‘burst’’ of small particles may be the result of the nucleation of very low vapor pressure products of the O3/a-pinene reactions. It is, however, also possible that initial particles in the chamber during experiment 3 were bimodally distributed and particles in the smaller size range (lower than the detectable range of 0.02 mm) may have contained higher surface area or volume. Details of the reaction products that can form new particles via nucleation are not currently known. However, Kamens et al. (1999) suggested two possible mechanisms for the nucleation process to occur. Stabilized Criegee biradicals may react with ‘‘the carbonyl portion of product compounds to produce an extremely low-pressure secondary ozonide product’’. It is also possible that Criegee biradicals can react with dicarboxylic acids to form low volatility anhydrides. These products were suggested to have high molecular weights (B350) and very low vapor pressures (o1015 Torr) with significant potential for self-nucleation (Kamens et al., 1999).

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1372

100, 000. 0 0.02- 0.1

3

PARTICLES [#/cm ]

10, 000. 0

0.2-0.3

0.1-0.2

1,000. 0

0.3-0.4

100. 0

0.4-0.5

10. 0

0.5-0.7

1.0

α-pinene introduced

0.1 -50

50

150

250

350 450 TIME [min]

550

650

750

Fig. 3. Particle number concentrations during experiment 3.

Fig. 4. Particle size distributions at times 0 and 700 min for experiment 3.

The initial particle number concentration during experiments 4 and 6 were relatively high compared to experiment 3. Particle surface area and volume at time 0 for experiment 4 are shown in Fig. 6. The maximum particle surface area and particle volume occurred in the 0.1–0.2 mm size range and the bulk of the particle surface area and volume was contained among the 0.02–0.3 mm size range. The evolution of particle number concentrations for experiment 4 is shown in Fig. 7. The initial ‘‘burst’’ of particles in the 0.02–0.1 mm size range was negligible compared to experiment 3. The initial increases in particle number concentrations occurred in the 0.1–0.2, 0.2–0.3, and 0.3–0.4 mm size ranges and reflects particle growth from the immediate lower size ranges containing the bulk of the particle surface area

and volume. This suggests that the initial particles in the chamber were not bimodally distributed for experiment 4, and the initial particle growth occurred via condensation and/or absorption of semi-volatile particles. Similar results were observed for experiment 6. Thus, it is conceivable that initial particles during experiment 3 were not bimodally distributed and the initial ‘‘burst’’ of small particles may have occurred via nucleation of very low vapor pressure products of O3/a-pinene reactions, and via the growth of smaller particles into the detectable size range. Unfortunately, the limited number of particle size ranges associated with the instruments used for this study lend enough uncertainty to surface area calculations to preclude a definitive statement regarding the relative importance of particle formation

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

0.8 15

surface area 0.6 10

volume

0.4

5 0.2

0

PARTICLE VOLUME [µm3/cm3]

1.0

20

PARTICLE SURFACE AREA [µm2/cm3]

1373

0.0 0.02-0.1

0.1-0.2

0.2-0.3

0.3-0.4

0.4-0.5

0.5-0.7

PARTICLE SIZE [µm]

Fig. 5. Particle surface area and volume as a function of particle size at time 0 during experiment 3.

2.0 surface area 1.5

50 40

1.0 30

volume

20

0.5

10

3

3

60

PARTICLE VOLUME [µm /cm ]

PARTICLE SURFACE AREA [µm2/cm3]

70

0.0

0 0.02-0.1

0.1-0.2

0.2-0.3

0.3-0.4

0.4-0.5

0.5-0.7

PARTICLE SIZE [ µ m]

Fig. 6. Particle surface area and volume as a function of particle size at time 0 during experiment 4.

versus growth. The issue of particle nucleation/growth during the initial phase of the experiment should be further explored using particle analyzers that can measure much finer size distributions. With the exception of experiments 8 and 9, there was a time lag of between 2 and 15 min after the introduction of a-pinene in the chamber prior to a rapid increase in particle number concentration. Experiments 8 and 9 were performed at 8–91C higher temperature than other experiments. The burst of particles during these two experiments was detected by the P-TRAKTM 10–15 min after the introduction of a-pinene in the chamber. These observations were not due to conveyance times between sample inlet and particle instruments. The estimated travel time for particles from the chamber to the PTRAKTM was only about 0.5 min. The travel time for

particles from the chamber to the LASAIRs was even shorter. It takes a finite amount of time for the a-pinene to evaporate and to mix into the chamber. The effective time to evaporate into, and mix within, the chamber should have been fairly similar for all experiments, and likely contributed partly to the delay in the initial burst of particles in the chamber. However, the primary reason for the delay in the initial burst of particles was likely due to the time needed for the reaction between O3 and a-pinene to raise concentrations of low vapor pressure semi-volatile products to sufficient levels for nucleation to occur, or for condensation on existing smaller particles. The elevated temperatures during experiments 8 and 9 caused the vapor pressures of these products to be higher than that of other experiments.

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1374

100,000.0 0.02-0.1

3

PARTICLES [#/cm ]

10,000.0

0.1-0.2

1,000.0

0.2-0.3

100.0

0.3-0.4 0.4-0.5

10.0 0.5-0.7

1.0

α -pinene introduced 0.1 -50

50

150

250

350 450 TIME [min]

550

650

750

Fig. 7. Evolution of particle number concentrations during experiment 4.

50

150 120

predicted ozone experimental ozone

30

90

predicted particles

20

60 experimental particles

10

OZONE [ppb]

3

PM [µg/m ]

40

30

0

0 0

150

300

450

600

750

TIME [min]

Fig. 8. Experimental and predicted particle mass and O3 concentrations during experiment 3.

Thus, it should have taken longer to raise the concentrations of these products to sufficient levels for nucleation, a plausible reason why the initial burst of particles was further delayed during each of these two experiments. The evolutions of experimental and predicted particle mass and O3 concentrations for experiment 3 are presented in Fig. 8. Particle mass concentration in the chamber was initially 2.2 mg/m3 and increased after the introduction of a-pinene into the chamber, eventually reaching a steady-state concentration of 21 mg/m3; the secondary organic aerosol (SOA) concentration was approximately 19 mg/m3. The predicted and measured particle mass concentrations for experiment 3 were in reasonable agreement; predicted mass concentrations were within 26% of experimental results at steady-state

conditions and at worst a factor of two higher between minutes 150 and 250. Predicted (using ICEM) constituents of SOA are shown in Fig. 9 for experiment 3. Model compounds diacid, oxy-pinald, and oxy-pinacid were the three highest contributors and represented almost 90% of the predicted SOA mass concentrations. Self-nucleated product (‘‘seed1’’) accounted for approximately 5% of predicted SOA mass concentrations. Model simulations suggest that O3- and OH  -initiated reaction pathways produced approximately 92% and 8% of the predicted secondary particle mass concentrations, respectively. Similar results were predicted for other experiments. The evolution of particle number concentrations for experiment 2 is shown in Fig. 10. Similar to experiment 3, a rapid ‘‘burst’’ of small particles in the 0.02–0.1 mm

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1375

12 diacid

9

3

PM [µg/m ]

oxy-pinald

6 oxy-pinacid pinacid

pinald

3

p3gas seed seed1

0 0

100

200

30 0

400

500

600

70 0

800

TIME [min]

Fig. 9. Constituents of the predicted particle mass concentrations for experiment 3.

1,000,000.0 0.02 -0.1

100,000.0

3

PARTICLES [#/cm ]

10,000.0 0.1-0.2

1,000.0

0.2-0.3 0.3-0.4 0.4-0.5

100.0 10.0

0.5-0.7

1.0 α-pinene introduced

0.1 -50

50

150

250

350

450

550

650

750

TIME [min]

Fig. 10. Evolution of particle number concentrations during experiment 2.

size range was detected shortly after the introduction of a-pinene into the chamber. A particle growth ‘‘wave’’ similar to experiment 3 was also observed. An analysis of relative time scales suggests that coagulation phenomena played an insignificant role in defining the particle growth wave. Hinds (1982) provided an equation to relate particle number concentrations for coagulation: NðtÞ ¼ N0 =ð1 þ N0 KtÞ;

ð5Þ

where K is the coagulation coefficient, NðtÞ is the particle number concentration at time t; and N 0 is the initial particle concentration. The coagulation coeffi-

cient, K; for polydisperse particles was estimated using the following equation (Hinds, 1982; Lee and Chen, 1984): K¼

2kT f1 þ expðln2 sg Þ 3Z 2:49l þ ½expð0:5 ln2 sg Þ þ expð2:5 ln2 sg Þ g; CMD

ð6Þ

where k is the Boltzmann constant (1.38
1016 dyne cm/K), T is the temperature (298 K), Z is the viscosity of air (1.90
104 dyne s/cm2), sg is the geometric standard deviation of the particles, and CMD is the count median diameter. The count median

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1376

diameter (CMD) and the geometric standard deviation (sg ) were estimated by the following equations (Hinds, 1982): P ni ln di ; ð7Þ lnðCMDÞ ¼ N P ln sg ¼

1 ni ðln di  ln CMDÞ2 2 ; N 1

ð8Þ

where ni is the particle number concentration in size range i; di is the geometric mean particle diameter for size range i; and N is the total particle number concentration. For purposes of illustration, Eqs. (7) and (8) were used to estimate particle CMD and sg for experiment 2. Eq. (6) was used to estimate K for polydisperse particles. Eq. (5) was used to estimate the time (t1 ) needed for the reduction in peak number concentration of smaller particles to peak number concentration of larger particles via coagulation. The estimated (t1 ) and observed times (t2 ) are presented in Table 2 for experiment 2, and are consistent in magnitude with other experiments. The diameter ‘‘d’’ in Table 2 represents the geometric mean diameter of each size range. Estimated times for coagulation were greater than the observed times by 1–4 orders of magnitude. Thus, coagulation was not the dominant process via which the number concentrations of smaller particles decreased after the initial burst of 0.02–0.1 mm particles. More importantly, chamber particle mass concentrations continued to increase until a stable condition was attained, reinforcing that coagulation was not the dominant growth process in the system; coagulation would lead to changes in particle number concentrations but would result in a relatively constant mass concentration. Rather than coagulation, we propose that the particle growth wave occurs via partitioning of semivolatile reaction products to smaller particles causing them to grow in size. As the particle size increases, more mass is required to effect an incremental growth in particle diameter. Therefore, the time interval between successive peaks increases. The evolutions of experimental and predicted particle mass and O3 concentrations during experiment 2 are presented in Fig. 11. The initial ozone concentration was

132 ppb. It decreased after the introduction of a-pinene into the chamber, and ultimately attained a relatively steady-state concentration. The predicted O3 concentrations were within 5% of experimental results. Particle mass concentration was initially 3.1 mg/m3 and increased after the introduction of a-pinene into the chamber, eventually reaching a steady-state concentration of 39 mg/m3. The SOA concentration was therefore 35 mg/m3. The predicted particle mass concentration was 44% higher than the measured concentration at steadystate conditions. The air exchange rate for experiment 2 was lower than all other experiments and the resulting SOA mass concentration was also higher than all experiments except for experiments 6 and 10. However, the initial particle number concentrations, the initial O3 concentrations, and a-pinene emission rate during experiment 6 were also higher than experiment 2, all conditions that could have led to higher SOA concentrations during experiment 6. Experiment 10 was conducted at a much higher a-pinene emission rate than experiment 2, likely the reason for a higher SOA concentration for experiment 10. Predicted particle mass concentrations were within 3–57% higher than measured concentrations for all experiments except experiment 10 (Table 1). The current ICEM does not include adsorption/desorption phenomena that may occur in indoor environments, and as described previously, all surface interactions for semivolatile organic compounds were modeled using an irreversible deposition term that has not been determined experimentally. As such, it is conceivable that the consistent over-prediction of particle mass concentration was due to an underestimation of semi-volatile organic compound removal to chamber surfaces. Through model sensitivity analyses, we have confirmed that predicted particle mass concentrations are highly sensitive to changes in prescribed deposition velocities for semi-volatile organic compounds. As such, additional knowledge of surface phenomena involving semivolatile organic compounds is required to improve model accuracy. Over all experiments, the reasonably close agreements between predicted and experimental particle mass concentrations are encouraging. The model appears to

Table 2 Estimated (t1 ) and observed time (t2 ) for the reduction in peak particle number concentrations during experiment 2 Range (mm)

d (mm)

K (cm3/s)

N0 (#/cm3)

NðtÞ (#/cm3)

t1 (min)

t2 (min)

0.02–0.1 0.1–0.2 0.2–0.3 0.3–0.4 0.4–0.5

0.045 0.141 0.245 0.346 0.447

1.3E09 1.4E09 1.8E09 1.6E09 1.7E09

92,870 14,816 4129 926 244

14,816 4129 926 244 5

725 2059 7723 3.1
104 1.9
106

98 140 136 118 106

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1377

10 0

15 0 experimental ozone

80

12 0

60

90

predicted particles

40

60

OZONE [ppb]

3

PM [µg/m ]

predicted ozone

experimental particles

20

30

0

0 0

150

300

450

600

750

TIME [min]

Fig. 11. Experimental and predicted particle mass and O3 concentrations during experiment 2.

capture processes associated with particle formation and/or growth. However, additional research will be needed to improve model performance. 3.1. Effects of selected parameters on indoor SOA levels Model simulations were performed to investigate the effects of the following parameters on indoor SOA mass concentrations: air exchange rate, outdoor fine particle concentration, outdoor O3 concentration, indoor apinene emission rate, and indoor temperature. A basecase scenario was defined as follows: indoor a-pinene emission rate=2.3 mg/min; outdoor O3 concentration=100 ppb and, outdoor fine particle concentration (PM2.5)= 15 mg/m3. An indoor volume of 500 m3 was used for this analysis. The prescribed indoor emission rate would produce an indoor a-pinene concentration of about 100 ppb at an air exchange rate of 0.5 h1 in the absence of O3. For simplicity, it was also assumed that no pollutants other than O3, a-pinene, associated gaseous by-products, and fine particles existed in the indoor environment. All model simulations were performed with an assumed indoor relative humidity of 50%, indoor temperature of 297 K (except when the effects of temperature were investigated), and an air exchange rate of 0.5 h1 (except when the effects of air exchange rate were investigated). An average deposition velocity of 0.036 cm/s for O3 was used in the model (Nazaroff and Cass, 1986). An average deposition velocity of 0.07 cm/s was used for both hydroxyl and hydroperoxy radicals in the model (Nazaroff and Cass, 1986). An average deposition velocity of 0.004 cm/s was used for fine particles (Wallace, 1996). The deposition velocities of the semivolatile organic products of a-pinene were assigned twice

the average deposition velocity of fine particles following Kamens et al. (1999). It was also assumed that no other sources of fine particles were present in the building, and that transport of outdoor fine particles provided ‘‘seed’’ particles in the model. The predicted indoor O3 concentration for the basecase scenario was 11 ppb. The total indoor fine particle concentration was predicted to be 12.1 mg/m3. The predicted indoor SOA concentration resulting from reactions between O3 and a-pinene for the base-case scenario was 3.7 mg/m3. The predicted indoor ‘‘seed’’ particle concentration resulting from the transport of outdoor fine particles was 8.4 mg/m3. 3.2. Effects of air exchange rate on indoor SOA concentrations The air exchange rate was varied from 0.1 to 3.0 h1. Resulting steady-state indoor SOA concentrations are shown in Fig. 12. Indoor seed and outdoor particle concentrations are also shown in Fig. 12. Predicted indoor SOA concentrations increased with lower air exchange rates (above 0.4 h1). Lower air exchange rates were associated with lower indoor O3 and ‘‘seed’’ particle levels. However, indoor a-pinene concentrations also increased with lower air exchange rates. The O3/apinene reaction rate and time available for reactions increased with the reduction in the air exchange rate (above 0.4 h1). The combination of higher reaction rates and longer reaction times resulted in increased indoor SOA concentrations. The predicted indoor SOA concentrations decreased with further reductions in air exchange rate below 0.4 h1, primarily due to a decrease in the O3/a-pinene reaction rate below an air exchange rate of 0.4 h1. Interestingly, at air exchange rates of 0.3–1.0 h1, i.e., typical of many residential dwellings, SOA was predicted to account for approximately

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1378

20

outdoor particles

3

PM [µg/m ]

16

12

seed particles

8

4 secondary particles

0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

-1

AIR EXCHANGE RATE [hr ] Fig. 12. Indoor fine particle concentration as a function of air exchange rate.

20–35% of the total indoor fine particle mass concentrations.

3.3. Effects of outdoor fine particle concentration on indoor SOA concentrations Outdoor fine particle concentrations were varied from 0 to 100 mg/m3. Predicted indoor SOA mass concentrations increased with higher outdoor fine particle mass concentrations. Higher outdoor fine particle concentrations caused higher indoor ‘‘seed’’ particle concentrations, increased partitioning of semi-volatile products, and increased indoor SOA concentrations. The predicted indoor SOA concentration was only 0.4 mg/m3 when the outdoor particle level was zero and resulted from self-nucleation of low vapor pressure reaction products. However, the predicted indoor SOA concentration increased to almost 12 mg/m3 at an outdoor fine particle concentration of 100 mg/m3, a very high concentration in outdoor environments.

3.4. Effects of outdoor ozone concentration on indoor SOA concentrations The outdoor O3 concentration was varied from 0 to 200 ppb. The higher outdoor O3 concentrations caused an increase in indoor O3 concentrations, which in turn increased the O3/a-pinene reaction rate. Increased amounts of reaction products were generated by the higher O3/a-pinene reaction rate, and resulted in increased indoor SOA mass concentrations. The predicted indoor SOA mass concentration at an outdoor O3 concentration of 200 ppb was almost 50% of the total indoor fine particle concentration.

3.5. Effects of indoor a-pinene emission rate on indoor SOA concentrations The indoor a-pinene emission rate was varied from 0 to 3 times the base-case emission rate. Increased indoor a-pinene emission rates increased the O3/a-pinene reaction rate and resulted in elevated indoor SOA concentrations. The contribution of SOA to indoor fine particle mass equaled that of outdoor-to-indoor penetration for an a-pinene emission rate of 4.6 mg/min (twice the base case). 3.6. Effects of indoor temperature on indoor SOA concentrations The indoor temperature was varied from 280 to 315 K (45–1081F). The resulting steady-state indoor SOA concentrations are shown in Fig. 13. Indoor seed and outdoor particle concentrations are also shown in Fig. 13. The indoor SOA concentrations increased dramatically as indoor temperature was reduced. Lower temperatures reduced the rates of homogeneous chemical reactions. However, this effect was more than compensated for by an increase in gas-to-particle partitioning of reaction products at lower temperatures. Predicted indoor SOA concentrations increased by more than a factor of two for every 101C decrease in indoor temperature. Experiments 8 and 9 were conducted at elevated chamber temperatures. Temperatures during these two experiments were 32–331C, about 8–91C higher than other experiments. The lower particle growth observed during these experiments cannot be entirely attributed to elevated temperature, since the air exchange rates during these experiments were also higher than for other experiments; operation of the space heater tended to increase the air exchange rate in the chamber.

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

1379

20

3

PM [µg/m ]

16

outdoor particles

12 indoor seed particles

8

4 indoor secondary particles

0 40

50

60

70

80

90

100

110

INDOOR TEMPERATURE [F] Fig. 13. Indoor fine particle concentration as a function of indoor temperature.

The combination of elevated outdoor O3 concentrations or the presence of indoor O3 sources, low air exchange rates, high indoor terpene emission rates, and low indoor temperatures are predicted to produce the highest secondary particle concentrations. Thus, the building designer/operator should avoid this combination to reduce occupant exposure to fine particles. For example, a control device for O3 can be installed in buildings to reduce indoor O3 concentration, which can also prevent the formation of elevated SOA concentrations. During the urban summertime ozone season, especially midday, sources of indoor terpenes should be avoided to limit the formation of indoor secondary particles. Buildings can be operated with higher air exchange rates in combination with enhanced particle filters. The use of high-efficiency particle filters can reduce the transport of outdoor fine particles in commercial buildings, and higher air exchange rates can reduce SOA levels. Outdoor-to-indoor transport is the main source of indoor ozone for most buildings. However, sources of ozone can also be present in buildings. For example, photocopy machines, laser printers, electrostatic air filters and electrostatic precipitators have been shown to produce ozone (Weschler, 2000). Typical indoor ozone concentrations range between 20% and 70% of outdoor ozone concentrations due primarily to heterogeneous reactions between ozone and indoor surfaces, e.g., carpet (Weschler, 2000). Indoor ozone concentrations are expected to be somewhat lower than ozone concentrations used during these experiments. For example, indoor ozone concentrations in some urban areas have been observed to be less than a factor of two lower than used in this study (Weschler et al., 1989; Gold et al., 1999). Thus, particle mass concentrations resulting from reactions between ozone transported from outdoors and a-pinene emitted in actual furnished

houses are expected to be somewhat lower than particle mass concentrations presented in this paper. Our research team is currently conducting field experiments to evaluate the importance of indoor ozone/a-pinene reactions in actual furnished houses. However, it should be noted that Long et al. (2000) reported a peak particle mass concentration of 38 mg/m3 during six sampling events in Boston area homes in which a pine-oil-based cleaner was used. Weschler and Shields (1999) also reported an increase of approximately 20 mg/m3 in indoor fine particle mass concentration when a limonene source was present in an office in New Jersey. Indoor ozone generators were not used in any of these experiments; indoor ozone was present due to the transport of outdoor ozone.

4. Conclusions The research described herein included experimental as well as computational investigations. Ten chamber experiments were completed to assess the effects of ozone/a-pinene reactions on indoor secondary organic particles. A significant contribution of this research was the prediction of dynamic particle mass concentrations based on detailed homogeneous chemical mechanisms and partitioning of semi-volatile products to particles using a new indoor chemistry and exposure model (ICEM). The ICEM allows for the simulation of airexchange processes, indoor emissions, chemical reactions, deposition, and variations in outdoor air quality. Predicted indoor secondary particle mass concentrations were in reasonable agreement with experimental results, which suggests that the model is capturing the essence of particle formation/growth processes. Nevertheless, many uncertainties remain to be resolved. For example, deposition velocities of the semi-volatile products of

1380

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381

ozone/a-pinene reactions were assigned twice the deposition velocity of fine particles, which warrants further exploration. Experimental results indicate that significant particle formation/growth can occur indoors due to reactions between O3 and a-pinene. Particle growth occurs through an initial ‘‘burst’’ of small particles followed by a decrease in particle numbers in a given size range and increase in particle numbers in subsequent size ranges. This process leads to an effective particle growth ‘‘wave’’. If the initial indoor particle levels are low, the initial ‘‘burst’’ of particles is pronounced and may be the result of nucleation of O3/a-pinene reaction products and the growth of smaller particles into larger detectable sizes. If the initial indoor particle levels are high, the initial ‘‘burst’’ of particles is smaller and particle growth appears to occur via partitioning of O3/a-pinene reaction products. Additional research is recommended to resolve the issue of particle formation/growth during the initial phase of the experiment using particle analyzers that can measure smaller particles and finer size distributions. Indoor secondary organic particles may account for some of the previously reported ‘‘unexplained’’ particle mass in indoor environments. A significant fraction of indoor secondary particle numbers are in the ultra-fine range (o0.1 mm). Most of the particle number and mass associated with ozone/terpene reactions is associated with particles o0.7 mm in diameter. Lower air exchange rates increase indoor secondary particle mass concentrations, a phenomenon attributed to longer indoor residence times. The combination of more time to generate reaction products and longer particle residence times increases particle mass concentrations and distributes the resulting mass into larger size ranges. Lower indoor temperatures promote increased gas-to-particle partitioning of low vapor pressure by-products of ozone/terpene reactions, and increased particle mass concentrations. Higher outdoor fine particle mass concentrations cause higher indoor ‘‘seed’’ particle concentrations, which in turn increase the partitioning of semi-volatile products, resulting in increased indoor secondary particle mass concentrations. Ozone generators that are marketed as indoor air ‘‘purifiers’’ should be avoided since, in the presence of terpenes, they can lead to extremely high levels of by-products including formaldehyde, higher molecular weight aldehydes, organic acids, and fine particles.

Acknowledgements The authors acknowledge Professor Richard Kamens of the University of North Carolina for providing particle chemistry for a-pinene. The Texas Air Research Center (TARC) provided all necessary funding for this

study. The authors thank Dr. Yosuke Kimura for assistance with computational analysis. References Carter, W., 2000. Documentation of the SAPRC-99 chemical mechanism for VOC reactivity assessment. Report to California Air Resources Board, Contract 92-329 and 95-308, 8 May. Fan, Z., 2002. Environmental and Occupational Health Sciences Institute, UMDNJ-Robert Wood Johnson Medical School, New Jersey, personal communication. Gold, D.R., Allen, G., Damokosh, A., Serrano, P., Hayes, C., Castillejos, M., 1999. Comparison of outdoor and classroom ozone exposures for school children in Mexico City. Journal of the Air and Waste Management Association 46, 335–342. Hinds, W.C., 1982. Aerosol Technology: Properties, Behavior, and Measurements of Aerosol Particles. Wiley-Interscience Publication, New York. Kamens, R., Jang, M., Chien, C.J., Leach, .K., 1999. Aerosol formation from the reaction of a-pinene and ozone using a gas-phase kinetics-aerosol partitioning model. Environmental Science and Technology 33, 1430–1438. Lee, K.W., Chen, H., 1984. Coagulation rate for polydisperse particles. Aerosol Science and Technology 3, 327–334. Long, M.C., Suh, H.H., Koutrakis, P., 2000. Characterization of indoor particle sources using continuous mass and size monitors. Journal of the Air and Waste Management Association 50, 1236–1250. Long, C.M., Suh, H.H., Kobzik, L., Catalano, P.J., Ning, Y.Y., Koutrakis, P., 2001. A pilot investigation of the relative toxicity of indoor and outdoor fine particles: in vitro effects of endotoxin and other particulate properties. Environmental Health Perspectives 109, 1019–1026. Nazaroff, W.W., Cass, G.R., 1986. Mathematical modeling of chemically reactive pollutants in indoor air. Environmental Science and Technology 20, 924–934. Nazaroff, W.W., Gadgil, A.J., Weschler, C.J., 1993. In: Nadga, N.L., (Ed.), Modeling of Indoor Air Quality and Exposure, Vol. 1205. ASTM Special Technical Publication, Philadelphia, pp. 81–104. Pope, A., Dockery, D., 1996. In: Wilson, R., Spengler, J.D. (Eds.), Epidemiology of Chronic Health Effects: CrossSectional Studies. Particles in our Air: Concentrations and Health Effects. Harvard University Press, Cambridge, MA, pp. 149–167. Robinson, J.P., Thomas, J., Behar, J.V., 1991. Report to the United States Environmental Protection Agency under Contract No. 69-01-7324, Delivery Order 12, Exposure Assessment Research Division, National Exposure Research Center, USEPA, Las Vegas, NV. Rohr, A., Weschler, C.J., Koutrakis, P., Spengler, J.D., 2003. Generation and quantification of ultrafine particles through terpene/ozone reactions in a chamber setting. Aerosol Science and Technology 37, 65–78. Santannam, S., Spengler, J.D., Ryan, P.B., 1990. Particulate matter exposures estimated from an indoor–outdoor source apportionment study. Proceedings of the Fifth International Conference on Indoor Air Quality and Climate, Toronto, 29 July–3 August, pp. 583–588.

G. Sarwar et al. / Atmospheric Environment 37 (2003) 1365–1381 Sarwar, G., Corsi, R.L., Kimura, Y., Allen, D., Weschler, C., 2002. Hydroxyl radicals in indoor environments. Atmospheric Environment 36, 3973–3988. Tropp, R.J., Kohl, S.D., Chow, J.C., Frazier, C.A., 1998. Final Report for the Texas PM2.5 Sampling and Analysis Study, Document No. 6570-685-7770.1F, Prepared for the Bureau of Air Quality, City of Houston, 15 December. Turpin, B.J., Lim, H.J., 2001. Species contribution to PM2.5 mass concentrations: revisiting common assumptions for estimating organic mass. Aerosol Science and Technology 35, 602–610. United States Environmental Protection Agency. Federal Register, Vol. 62, No. 138, Friday, 18 July 1997. Wainman, T., 1999. Use of a two tiered dynamic chamber to investigate indoor air chemistry. Ph.D. Dissertation. The State University of New Jersey, Rutgers. Wainman, T., Zhang, J., Weschler, C.J., Lioy, P.J., 2000. Ozone and limonene in indoor air: a source of submicron particle

1381

exposure. Environmental Health Perspectives 108, 1139–1145. Wallace, L., 1996. Indoor particles: a review. Journal of the Air and Waste Management Association 46, 98–126. Weschler, C.J., 2000. Ozone in indoor environments: concentrations and chemistry. Indoor Air 10, 69–288. Weschler, C.J., Shields, H.C., Naik, D.V., 1989. Indoor ozone exposures. Journal of the Air Pollution Control Association 39, 1562–1568. Weschler, C.J., Shields, H.C., 1999. Indoor ozone/terpene reactions as a source of indoor particles. Atmospheric Environment 33, 2301–2312. Weschler, C.J., Shields, H.C., 2002. Experiments probing the influence of air exchange rates on particles generated by indoor chemistry. In: Levin, H. (Ed.), Indoor Air 2002: Proceedings of the Ninth International Conference on Indoor Air Quality and Climate, Vol. 3. Indoor Air 2002, Santa Cruz, CA, 2002, pp. 500–505.