Measurements of particle nucleation following the gas phase oxidation of α-pinene and β-pinene are reported. Particle nucleation following ozonolysis and O3 plus OH initiated monoterpene oxidation was measured in a 70-L Teflon bag reactor over the temperature range 278 to 320 K. Particle concentration temporal profiles were measured for a range of initial monoterpene and ozone concentrations using ultrafine condensation particle counters. Profiles were interpreted using a coupled gas phase chemistry and kinetic multicomponent nucleation model to determine the molar yield of the nucleating species in the ozonolysis experiments to be 1 × 10−5 and 0.009, for α-pinene and β-pinene, respectively. OH initiated oxidation was found to increase the nucleator yield for α-pinene, approximately a factor of three, but not for β-pinene. The molar yield of condensable reaction products (vapor pressure < 20 ppt at 296 K) was determined to be 0.06 for both α-pinene and β-pinene and was independent of the oxidation source. Particle growth, which is determined by the condensable reaction products, was nearly temperature independent. Atmospheric box model calculations of nucleation and particle growth for α-pinene and β-pinene oxidation under typical tropospheric conditions are presented and showed that (1) nucleation can be significant under favorable conditions, (2) nucleation is dominated by OH initiated oxidation, and (3) the partitioning of the condensable monoterpene oxidation products made a significant contribution to the growth of atmospheric aerosol and was capable of explaining the observed particle growth rates commonly observed in remote forests.
 Atmospheric aerosols have important implications to human health, processing of trace species in the atmosphere, and Earth’s climate through both direct and indirect radiative effects [IPCC, 2001]. The global significance of aerosol and its role in these environmental issues is an area of active research. It is generally accepted that semivolatile organic compounds, produced in the atmospheric oxidation of anthropogenic and biogenic compounds, contribute to the growth and mass of atmospheric aerosol. However, at present, the occurrence of “new” particle formation (nucleation) in the atmosphere originating from biogenic compounds is uncertain.
 Rapid nucleation events in the atmosphere have been observed in a variety of locations, in different seasons, and over a range of ambient conditions. Kulmala et al.  have compiled a review of field studies that report such observations. The mechanisms responsible for the rapid new particle nucleation observed differ depending on atmospheric conditions and the precursor concentrations. Mechanisms proposed to explain the observed nucleation events include ion-induced nucleation, homogeneous nucleation of iodine oxides, biogenic organic compounds, and H2SO4/H2O mixtures, as well as nucleation of ternary mixtures of H2SO4/H2O/NH3. Of particular interest to the present study are observations of significant nucleation events in clean remote forested regions. These events are best suited for observing biogenic organic aerosol formation. For example, campaigns at the SMEAR station in Finland [Kulmala et al., 2001] report regional nucleation bursts which yield 103 to 104 ultrafine particles cm−3 (<50 nm dia.) of predominately organic composition [O'Dowd et al., 2004] being formed over a period of several hours.
 Monoterpenes (C10H16) is an important class of globally emitted organic biogenic compounds that under laboratory conditions are known to lead to nucleation. Global emissions of monoterpenes account for a significant fraction, ∼11% of the total organic biogenic flux [Guenther et al., 1995; Lee et al., 2005] with the most abundant species being α-pinene and β-pinene. α-Pinene and β-pinene are bicyclic hydrocarbons with a single double bond that is endocyclic for α-pinene and exocyclic for β-pinene. The differences in molecular structure lead to different atmospheric degradation products being formed following reaction with OH, NO3, or O3 and therefore possibly influence the efficiency of nucleation. The oxidation of α-pinene and β-pinene and subsequent nucleation is commonly taken to be representative of the oxidation of other endocyclic and exocyclic monoterpenes and alkenes. A general review of the atmospheric oxidation mechanisms and gas phase reaction product yields for α-pinene and β-pinene is given by Atkinson and Arey  and references cited within.
 In this paper, laboratory measurements of particle nucleation following the gas phase oxidation of α-pinene and β-pinene at temperatures in the range 278 to 320 K are presented. Monoterpene oxidation was initiated either by ozonolysis or a combination of ozonolysis and reaction with the OH radical. The aim of this work was to evaluate several key factors in the nucleation process. These include: (1) the nucleation efficiency of monoterpenes (taking α-pinene and β-pinene as representative examples) by evaluating the dependence of nucleation on molecular structure (endocyclic versus exocyclic double bonds), oxidant (O3 or OH), and temperature (i.e., the thermodynamic stability of “small” clusters); and (2) the yield of condensable reaction products and its partitioning between the gas and particulate phases. Ultimately, the goal of our work is to assimilate this information in the development of an accurate representation of monoterpene nucleation in the atmosphere. As a first step toward this goal, simple atmospheric box model calculations of nucleation and particle growth for α-pinene and β-pinene under conditions representative of the tropospheric boundary layer are presented and the implications are discussed. A brief discussion of results reported in a few of the previous α-pinene and β-pinene aerosol studies are also presented.
2. Experimental Details
 Measurements of particle nucleation and growth following the oxidation of α-pinene and β-pinene were performed in a 70-L temperature regulated FEP-Teflon bag reactor. Experiments were performed over a range of experimental conditions that included variations in (1) the initial monoterpene concentration, (2) the source of gas phase oxidation of the monoterpene (ozonolysis or O3 plus OH radical reactions), (3) the rate of monoterpene oxidation (variations in the initial [O3]), and (4) temperature. The experimental apparatus and procedures used in this study are similar to those used in our previous nucleation study of iodine oxides [Burkholder et al., 2004]. A general outline of the experimental apparatus, methods, and some details particular to this work are described below. The experimental data were interpreted using a coupled gas phase chemical kinetics: kinetic particle nucleation model that is described separately below.
 The apparatus consisted of a vacuum manifold, a temperature regulated 70-L FEP-Teflon bag reactor, and two ultrafine condensation particle counters (UCPC, TSI models 3025a and 3022) (Certain commercial equipment, instruments, or materials are identified in this article in order to adequately specify the experimental procedure. Such identification does not imply recognition or endorsement by the National Oceanic and Atmospheric Administration, nor does it imply that the material or equipment identified is necessarily the best available for the purpose.) with 3 and 10 nm diameter particle detection limits. The particle counters were operated under high flow conditions, 1.5 L min−1. The particle concentration in the reactor was monitored continuously during a nucleation experiment with data recorded at 1 Hz. The duration of an experiment was ∼1000 s and was determined by the UCPC sampling rate and the collapse of the bag.
 The temperature of the reactor was varied over the range 278 to 335 K during this study. The reactor was enclosed in a thermally insulated box and its temperature regulated by heating or cooling the air inside the box (the particle counters were operated externally at room temperature). Cooling was accomplished by circulating a temperature-regulated fluid through heat exchanging coils inside the box. Heating was achieved by circulating heated air inside the box. The measured reactor temperature was uniform and constant to within 1 K over the duration of a nucleation experiment. The temperature of the gas within the bag was measured, in test measurements, and was identical to the outside temperature.
 A nucleation experiment was performed using the following procedure. The reactor was first thoroughly flushed using four cycles of filling the reactor with N2 and evacuating. The reactor was then filled with ozone/synthetic air at an ozone concentration comparable to that used in the nucleation experiments. Background nucleation was then measured. No nucleation, <1 particle cm−3, was observed when the reactor was adequately flushed. Following the background test, the reactor was flushed again before starting a nucleation experiment.
 Ozone was added to the reactor first, in the majority of measurements, and followed by the addition of the OH radical scavenger (cyclohexane), if used. In several experiments, SO2 was added to the reaction mixture at this time. The reactor was then filled to ∼75% total capacity with synthetic air. Finally, the monoterpene sample was rapidly flushed into the reactor and the bag filled to capacity (625 Torr). Particle sampling was started after all gases were added to the reactor. It is worth noting that no particle production was observed for the range of monoterpene and ozone concentrations used in our experiments when N2 was used as the buffer gas (only O2 from the ozone source was present, <0.5 Torr). The initial reactant concentrations in the reactor were calculated using the measured pressures of the dilute sample mixtures in the vacuum manifold and the calibrated volumes of the manifold and reactor [Burkholder et al., 2004]. All experiments were performed at zero relative humidity in the absence of seed particles and nitrogen oxides. The range of concentrations used is outlined in Table 1. In the majority of the measurements, ozone was in large excess over the monoterpene to enable gas phase reaction rates fast enough for nucleation to be observed on the timescale of our measurements.
Table 1. Summary of Experimental Parameters and Range of Values Used in the Particle Nucleation Experiments
Range of Values
Cyclohexane was used as an OH radical scavenger in some experiments.
 The α-pinene and β-pinene samples (>99% purity) were degassed using freeze-pump-thaw cycles before use. Dilute gas phase mixtures of α-pinene and β-pinene in synthetic air were prepared manometrically in 12 L Pyrex bulbs. The monoterpene mixing ratio in the bulb, ∼1 × 10−4, was periodically checked using infrared absorption and found to be stable to within 5%. O3 was prepared using a commercial ozonizer. Mixtures of O3 in O2 were prepared in a darkened 12 L Pyrex bulb at a total pressure of ∼800 Torr. The O3 mixing ratio, ∼5 × 10−5, was measured by UV absorption using a diode array spectrometer and was constant to within 3%. Synthetic air (UHP; 80% N2/ 20% O2) was passed through a SO2 scrubber, hydrosil HS-600 6 × 8 mesh, and a particle filter before entering the vacuum manifold. The SO2 impurity level in the scrubbed synthetic air was measured using cw KrCl (222 nm) excitation combined with SO2 fluorescence detection to be <20 ppt. Pressure measurements were made using 10, 100, and 1000 Torr capacitance manometers.
 Cyclohexane was used as the OH radical scavenger. The reaction of cyclohexane with OH forms mostly volatile products and therefore is considered to be a good OH radical scavenger for our nucleation experiments. We refer the reader to the work of Keywood et al. [2004a] and their evaluation of OH radical scavengers commonly used in aerosol studies. Cyclohexane (>99% purity) was degassed using freeze-pump-thaw cycles before use. Dilute mixtures of cyclohexane in He were prepared manometrically in a 12 L Pyrex bulb.
2.2. Coupled gas Phase Chemistry and Kinetic Nucleation Model
 A coupled gas phase chemistry and kinetic particle nucleation model was used to interpret the experimentally observed particle concentration and growth temporal profiles. The modeling analysis provides a quantification of the molar product yields of nucleating and condensable (low vapor pressure) species as well as thermodynamic parameters for nucleator clustering and the gas to particle partitioning of the condensable species.
 Our experiments do not elucidate the identity of the nucleating or condensing species. In our model, we describe the gas phase oxidation of α-pinene and β-pinene in terms of lumped classes of compounds, which we refer to as nucleating, condensing, and semivolatile and volatile species:
The semivolatile and volatile products, which account for the majority of the product mass, do not contribute to nucleation or growth under the conditions of our experiments. However, these volatile products may partition to the aerosol phase under atmospheric conditions. The nucleating species are responsible for particle formation and also contribute to particle growth. The condensing (low vapor pressure) species only contribute to particle growth through partitioning between the gas and particulate phases. In this system, the nucleating and condensing products, in all likelihood, represent a combination of product species. Although this mechanism represents an oversimplification, it is appropriate for the interpretation of our nucleation experiments considering the present level of understanding of the identity and number of the nucleating and condensing species. Table 2 gives a summary of the gas phase kinetic parameters used in the data analysis.
Table 2. Summary of Gas Phase Kinetic Parameters Used in the Present Studya
Kinetic parameters are taken from the study of Atkinson  unless noted; rate coefficients are in units of cubic centimeter per molecule per second.
OH yields were measured at room temperature. The temperature dependence of the OH yields in the O3 + monoterpene reactions is currently not available but expected to be small. Values taken from the study of Atkinson and Arey .
 The formation of OH radicals in reaction (1d) will be shown to play an important role in the nucleation process for α-pinene. The OH yields in reaction (1d) are significant for both α-pinene and β-pinene (Table 2). We have taken advantage of this “internal” OH radical source to evaluate nucleation and growth following OH initiated monoterpene oxidation. Nucleation experiments performed with an OH radical scavenger (cyclohexane) added to the gas mixture
were used to evaluate nucleation from ozonolysis alone. Experiments performed without an OH radical scavenger were used to evaluate nucleation from a combination of ozonolysis and OH radical initiated chemistry
In the nucleation model, the ozone and OH sources of nucleating and condensing species are accounted for separately. This approach assumes that the reaction products formed by O3 and OH initiated chemistry can be treated separately but behave similarly with respect to nucleation and particle growth. We do not account for secondary loss processes for OH in this model and assume reactions (3a), (3b), and (3c) as its only loss process. During the initial stages of particle nucleation this assumption is valid with >90% of the OH produced in reaction (1d) reacting via reactions reactions (3a), (3b), and (3c). We assume that the gas phase monoterpene reaction mechanism and reaction product yields to be independent of temperature over the range of temperatures used in our experiments.
2.2.2. Nucleation Model
 The theoretical approach and computational details of the kinetic particle nucleation model are described in our recent iodine oxide [Burkholder et al., 2004] and ion-induced nucleation [Lovejoy et al., 2004] studies. In this work, the model was modified to couple single component homogeneous nucleation (the nucleating species) [Burkholder et al., 2004] with multicomponent particle growth (by the nucleating and condensing species).
 The model combines a full kinetic treatment of the nucleation steps with a “sectional” aerosol model [Raes and Janssens, 1985] to treat the growth and coagulation of the larger nucleated particles. The nucleation portion of the model consists of a variable number of bins (typically 7) that increment by a single nucleating molecule starting with the nucleating monomer. The thermodynamics for growth and evaporation of the nucleation clusters are parameterized as described previously [Burkholder et al., 2004]. The nucleation portion of the model is coupled to a second set of bins in which the number of molecules in a bin is incremented geometrically, by a factor of 1.5. There are usually about 40 bins for the large clusters, giving a maximum particle diameter of about 1 μm.
 The kinetics of growth and evaporation of the nucleating clusters are treated explicitly. The nucleating species is also allowed to condense irreversibly onto the larger geometric clusters. The condensing (low vapor pressure) species only interacts with the geometric clusters. The thermodynamics for growth and evaporation of the geometric clusters is parameterized using the liquid drop model and specifying the surface tension, density, and bulk liquid thermodynamics. Evaporation of the geometric clusters is not treated explicitly. The first-order rate coefficient for growth of geometric clusters by condensation of the condensable species is given by
This gives an accurate treatment when growth dominates, as is the case in our experiments. We assume that the nucleator and condensor have the same density (1.0 g cm−3) to simplify the model. The results are most sensitive to the density of the condensable species since it overwhelmingly controls particle growth.
 As will be shown in section 3, the modeling approach described above, which includes both nucleating and condensing species, reproduces the experimentally observed particle concentrations and growth over a broad range of initial concentrations and temperatures. An alternative model approach, one that includes only a single component responsible for both nucleation and growth (single component homogeneous nucleation), would do an equally good job of reproducing the experimental data at a given temperature (however the nucleator yields and thermochemistry for cluster formation would be significantly different). However, the single component homogeneous nucleation model does a poor job of reproducing the observed temperature dependence of nucleation and significantly overpredicts the nucleation at low temperatures. Therefore we have chosen to use the multicomponent (nucleator and condensor species) model in our analysis.
3. Results and Discussion
 Experimental data highlighting the results from the ozonolysis, ozonolysis plus OH, and temperature dependence are presented in separate sections below. The loss of the monoterpene in the gas phase and buildup of reaction products including possible nucleating and condensing species depends on the initial ozone concentration, temperature, and whether an OH radical scavenger was used. These differences are quantitatively accounted for in our modeling analysis. The model interpretation of the experimental results follows the presentation of the experimental data.
 Representative particle temporal profiles measured during the ozonolysis of α-pinene and β-pinene at room temperature are shown in Figures 1 and 2. A range of initial monoterpene and ozone concentrations were used in our measurements, outlined in Table 2, and include 60 α-pinene and 25 β-pinene separate nucleation experiments. All experimental data were used in the nucleation modeling analysis. Cyclohexane was added in large excess, 1000 × [monoterpene], and scavenged >99% of the OH radicals generated during the oxidation of α-pinene and β-pinene. Several experiments performed with half this cyclohexane amount yielded nearly identical nucleation confirming that OH was efficiently scavenged in these experiments.
 Measured particle concentrations ranged from ∼20 particle cm−3 up to 100,000 particle cm−3 (UCPC upper limit). Measurements were reproducible to within a factor of two for particle concentrations > 1000 particle cm−3. The variability observed for low particle formation experiments was typically ±50 particle cm−3 (see data figures). The time delay to the detection of particles (>1 particle cm−3) was a key observable for the evaluation of the nucleation efficiency. The measured delays were reproducible to within 20 s for rapid nucleation, delay times of ∼100 s or less, while for conditions with slower nucleation rates the variability was ∼40 s. The variability in the initial mixing and filling of the reactor may account for some of the observed variability. However, the measured delays in nucleation were independent of the order that the reactants were added to the reactor. The particle temporal profiles, i.e., the rate of rise in the particle concentration and the growth of the particles, were also reproducible (see data figures).
 The majority of our experiments were performed using high initial ozone concentrations. High concentrations were used to obtain monoterpene loss rates and therefore particle formation rates that were faster than the diffusive loss of gases and particles to the walls of the reactor. Although the temporal profiles for nucleation were dependent on the rate of monoterpene loss, the results showed no systematic dependence on the initial ozone concentration. In other words, for experiments in which [O3] × [Monoterpene] was constant the particle formation was the same.
Figures 1 and 2 clearly illustrate the nonlinear dependence of particle production on the loss of monoterpene (i.e., the rate of gas phase reaction product formation given by k1[Monoterpene][O3]). In Figure 1, a 104 increase in particle concentration is observed for a 10-fold increase in α-pinene reaction product formation rate. Accompanied with the increase in particle concentration is a decrease in the delay time. Particle formation was not observed for initial α-pinene concentrations less than 0.55 ppb and for β-pinene concentrations below 2.6 ppb with up to ∼1 ppm [O3]. A quantitative interpretation of this data is included in the modeling section below.
3.2. Ozonolysis + OH Radical Initiated Chemistry
 Particle formation following OH initiated monoterpene oxidation was evaluated using experimental methods identical to those used in the ozonolysis measurements but without the addition of an OH radical scavenger. In this system, OH radicals are produced as a reaction product in the ozonolysis of the monoterpenes. The OH yields in the ozonolysis of α-pinene and β-pinene have been measured to be 0.8 and 0.3, respectively [Atkinson, 1997]. At early times in the reaction, during the nucleation stage of the measurement, OH radicals react primarily with the excess monoterpene and account for a significant fraction of the monoterpene loss.
 In this study, 40 α-pinene and 10 β-pinene OH oxidation measurements were performed using a range of initial concentrations and temperatures. Figures 3 and 4show representative back-to-back experiments performed under identical conditions, with and without an OH radical scavenger, for α-pinene and β-pinene, respectively. It is clear, at least qualitatively, that OH produces a dramatic increase in nucleation for α-pinene. The effect is much less evident for β-pinene. The particle growth rate also increased as the rate of particle production increased. Measurements made at other temperatures also showed increases in particle production for α-pinene due to OH. A quantitative evaluation of these observations is obtained from the nucleation modeling given below.
3.3. Temperature Dependence
 Nucleation experiments were performed at temperatures between 278 and 320 K, a range relevant to particle formation in the lower troposphere. Measurements were made at 278 and 320 K for α-pinene and 278, 284, 310, and 320 K for β-pinene. At each temperature, measurements were performed with a range of monoterpene concentrations with some variation in O3 concentration (9 α-pinene and 17 β-pinene measurements). Measurements were also made with and without the addition of an OH radical scavenger. Figures 5–8 illustrate the temperature dependence of particle production for α-pinene and β-pinene.
Figure 5 shows particle profiles measured at reduced temperatures, 278 and 296 K, for initial α-pinene mixing ratios of 2.67 and 8.9 ppb with 500 ppb O3 (without scavenger). These concentrations cover low and high particle production conditions. The profiles recorded at room temperature were measured prior to cooling the reactor. The profiles show that particle production is weakly enhanced for an 18 K reduction in temperature. For β-pinene, the nucleation temperature dependence is noticeably stronger with a factor of five increases in particle concentration at 278 K for the data shown in Figure 6. It must be kept in mind that changes in temperature also change the monoterpene loss rate (see Table 2). Therefore the observed increase in particle formation with decreasing temperature is masked by the decrease in the production rate of nucleating and low volatility reaction products.
Figures 7 and 8 show nucleation measurements for α-pinene and β-pinene, respectively, at elevated temperatures. As expected, particle production is suppressed for both α-pinene and β-pinene. The particle profiles also show longer delay times and slower rates of particle production. The effect is significantly greater for β-pinene. In fact, it was necessary to perform experiments with higher β-pinene loss to observe appreciable nucleation. The particle growth rates, however, for both α-pinene and β-pinene do not show a strong dependence on temperature.
 A consideration for nucleation experiments performed at reduced or elevated temperatures is the possibility of particle evaporation (sampling temperature > reactor temperature) or nucleation (sampling temperature < reactor temperature) outside the reactor, either in the sampling line or UCPC. Particle evaporation was judged to be insignificant in our experiments based on an estimate of the evaporation rate. For a 2 ppb loss of monoterpene and a condensable yield of 0.06, the uptake rate of condensable products would be <1 × 1013 molecule cm−2 s−1 and the evaporation rate would be <5 molecule s−1 for a 3 nm diameter particle. Therefore, evaporation would be insignificant for our sampling time and particle evaporation does not impact the interpretation of our observed nucleation at reduced temperatures. Nucleation outside of the reactor was clearly observed in experiments performed at 334 K with 5.0 and 10.0 ppb β-pinene (560 ppb O3). In the 10.0 ppb β-pinene experiment, ∼1000 particle cm−3 were detected in less than 10 s of reaction time and no increase in particle concentration was observed with increasing reaction time. Measurements made under these conditions with an OH scavenger added showed no particle formation at all. Rapid particle production was also observed but in the 5.0 ppb β-pinene experiment but the particle concentration was less, ∼100 particle cm−3. These observations are consistent with nucleation outside of the reactor following the formation of gas phase nucleator and condensable products in the reactor. Measurements made at 320 K did not show these systematic discrepancies and were judged to be free of interference from particle production outside of the reactor and were included in our data analysis.
3.4. Influence of SO2 on Organic Nucleation
 Particle sources other than monoterpene oxidation could influence the interpretation of our nucleation data. The formation of H2SO4/H2O particles following the reaction of OH with SO2 (a possible system impurity) in the presence of H2O is a possible source of nonorganic particle production. Hoppel et al.  and Gao et al.  observed that particle formation following the ozonolysis of α-pinene was sensitive to the addition of SO2. In the Hoppel et al. study, measurements were performed with SO2 added (0.5 and 6 ppb) to a α-pinene/O3/H2O mixture (30 to 50% relative humidity) and an increase in the rate of particle formation and particle concentration was observed for the highest SO2 concentration used. The enhancement in particle formation was presumed to result from the formation of sulfuric acid particles.
 In our work, the SO2 impurity level was determined to be <20 ppt. However, due to the current uncertainties associated with predicting the influence of SO2 on particle production, we have performed standard SO2 addition experiments to directly evaluate the impact of SO2 on organic nucleation. The conditions of our measurements differ from the previous studies primarily in that our relative humidity was very low, [H2O] < 5 ppm. The low relative humidity will reduce the nucleation rate of sulfuric acid particles [Ball et al., 1999; Berndt et al., 2005]. In our worst-case (lowest monoterpene mixing ratio, 0.55 ppb α-pinene), less than 0.03% of the OH produced via reaction (1d) will react with SO2. In experiments with the OH scavenger added, the fraction of OH reacting with SO2 is ∼100 times smaller. Results from our SO2 addition experiments are shown in Figure 9. Measurements were performed at 296 K with 3.3 and 5.5 ppb α-pinene (500 ppb O3) with 0.10 and 0.53 ppb SO2 added to the reactor. In the 3.3 ppb α-pinene experiment, the addition of SO2 had no observable impact on particle production. In the 5.5 ppb α-pinene experiment, the addition of 0.10 ppb SO2 did not significantly alter particle production. However, the addition of 0.53 ppb SO2, 50 times our estimated SO2 impurity level, showed a small decrease in delay time and increase in particle concentration. This is consistent with H2SO4 nucleation. However, the differences are near our level of experimental reproducibility. For the high SO2 conditions, we estimate [H2SO4]gas, at 200 s, to be ∼3 × 107 molecule cm−3. Particle profiles measured under these conditions with an OH radical scavenger added were found to be independent of SO2 addition. Berndt et al. [2004, 2005] have recently reported that [H2SO4]gas of ∼107 molecule cm−3 leads to efficient nucleation of H2SO4 particles. In their experiments, H2SO4 was produced using the ozonolysis of α-pinene to produce OH in the presence of SO2. The nucleation observed in their experiments is greatly enhanced from that reported in the H2SO4/H2O nucleation study of Ball et al. . The discrepancies between these nucleation studies are presently not well understood.
 We conclude, based on our direct experimental measurements, that SO2 did not influence our organic nucleation experiments through either sulfuric acid particle formation or by altering the gas phase monoterpene chemistry (scavenging of OH or reaction with organic intermediate species).
3.5. Nucleation Model Analysis
 The multicomponent homogeneous nucleation and growth model includes reaction products that are responsible for nucleation and low volatility products that condense on newly formed stable clusters resulting in particle growth. The nucleating species also condense on the stable clusters while the low volatility (condensable) species do not form particles but are involved in particle growth. To date, the identification of multicomponent nucleation following monoterpene oxidation has been based on the compositional analysis of aerosols [Hoppel et al., 2001; Iinuma et al., 2004; Kalberer et al., 2004; Yu et al., 1999; Ziemann, 2002]. It is difficult to identify a multicomponent nucleation/condensation system based on observations of particle formation at a single temperature. However, organic nucleation measurements made over a range of temperatures aid the identification of the mechanism (single versus multicomponent).
 Our experimental nucleation data were most consistent with a multicomponent homogeneous nucleation mechanism. A multicomponent analysis reproduces the experimental data at all temperatures reasonably well (see Figures 1–9). The yield of condensable products is well constrained by the observed particle growth rates, and the nucleation rate is well defined by the particle number density. In contrast, a single component analysis leads to a systematic overestimate (underestimate) of particle production at low (high) temperature with differences of a factor of 10 or more possible. In order to explain the data with a single component responsible for nucleation and growth, a significant nucleation barrier is required, since condensable concentrations are relatively high and observed particle concentrations are relatively small. Nucleation processes with high barriers are very nonlinear with respect to concentration and will have strong temperature dependencies. In order to reproduce the relatively temperature independence and low particle production coupled with efficient temperature independent growth, the production of a small yield of very low vapor pressure nucleator plus a significant yield of condensable material that has a low vapor pressure (but not low enough to nucleate itself) are required.
 A summary of the nucleation model parameters and their optimized values is given in Table 3. Simulations of the experimental data using the optimized model parameters are included in Figures 1–9. The agreement between the model and experimental data is good considering the overall reproducibility of the experimental data and the nonlinearity of the nucleation process. This level of agreement justifies the assumptions and simplifications used in our interpretation of the monoterpene gas phase chemistry and in the nucleation model. A quantitative evaluation of the uncertainties in the derived model parameters is however difficult. Changes in the parameter values may result in better fits for an individual experiment but at this time are not warranted. Refinements in the modeling would be greatly aided with a better understanding of the nucleating species and its physical-chemical properties. However, the present results do enable an evaluation of monoterpene nucleation under realistic atmospheric conditions (see section 3.7).
Table 3. Nucleation Model Parameter Definitions and Optimized Values
Values of the fixed parameters are set to physically reasonable values. For a discussion of the sensitivity of the model to the fixed parameter values see the work of Burkholder et al. .
A small increase in ΔH0Dimer for β-pinene with increasing temperature was required to obtain reasonable agreement with the experimental particle profiles data, see Figure 8. No change in ΔH0Dimer for α-pinene with temperature was used in the data analysis.
Molar yield of nucleating species: Ozonolysis
Molar yield of condensing species: Ozonolysis
Molar yield of nucleating species: Ozonolysis + OH
Molar yield of condensing species: Ozonolysis + OH
 The modeling analysis yields several interesting results. First, for α-pinene, the yield of the nucleating species is extremely small, 5 × 10−5 (ozonolysis) and 1.5 × 10−4 (ozonolysis plus OH). The low yield will make efforts to identify the nucleating species via currently available analytical techniques a challenge. For example, in our experiments, the gas phase concentration of the nucleating species would be <106 molecule cm−3 at the onset of particle detection. Second, the formation of a dimer of the nucleating species, the first step in the nucleation mechanism, is thermodynamically favorable with a bond energy of 23.5 kcal mol−1. The stability of the dimer reflects the weak temperature dependence observed in particle production. In the case of β-pinene, the nucleating species yield is significantly larger, 0.009, and was independent of the oxidation source. However, the dimer bond energy of the nucleating species, 18 kcal mol−1, is considerably smaller than for α-pinene resulting in a larger barrier to nucleation. Overall, β-pinene is more efficient in nucleation than α-pinene at room temperature. This is consistent with the general observations reported in previous studies such as that of Koch et al.  (see their Figure 1).
 As mentioned earlier, it has been proposed that carboxylic acids, such as pinic, norpinic, and pinonic acid, play a role in nucleation [Iinuma et al., 2004; Jenkin et al., 2000; Kavouras et al., 1998, 1999a]. However, the currently accepted vapor pressures of the dicarboxylic acids [Bilde and Pandis, 2001; Koch et al., 2000; Tao and McMurry, 1989] are on the order of 100 ppt at room temperature and these compounds would not be supersaturated in our experiments. The present study indicates that the nucleating species has a low vapor pressure (<1 ppt) and is formed with a small molar yield. Our results are consistent with that of Hoppel et al.  who estimated the vapor pressure of the condensable species to be <10 ppt.
 In our work, experiments were performed using dual UCPCs to simultaneously measure particle concentration profiles and the particle growth without loss of time resolution or sensitivity for small particles. The time dependence of the ratio [>10 nm particle]/[>3 nm particle] was used to determine the yield of condensable products, the volatility of the condensable products, and ultrafine particle growth rate. The yields of condensable products for α-pinene and β-pinene were the same, 0.06 ± 0.01, independent of the oxidation source, and independent of temperature over the range 278 to 320 K. The partitioning of the condensable products between the gas and particulate phases as calculated in the model is included in Figures 1–8. The condensable products are not in equilibrium with the particulate phase during our measurement and it is only possible to determine upper limits for the vapor pressure of the condensable species from our experiments, <20 ppt at 296 K and <200 ppt at 320 K. The similarity of the yields for condensable products for α-pinene and β-pinene implies that the physical properties of the minor reaction products (the products can be different) that account for the particle growth do not differ significantly.
 The yield of condensable products for α-pinene is in excellent agreement with the values reported by Hoppel et al. , Nozière et al. , and Presto and Donahue . On the other hand, previous studies of SOA yields for α-pinene and β-pinene are somewhat lower in the limit of low aerosol mass [Griffin et al., 1999a; Zhang et al., 1992]. These studies also report appreciable differences in the yields for α-pinene and β-pinene with α-pinene > β-pinene in ozonolysis experiments and β-pinene > α-pinene in photooxidation experiments. It is also worth noting that the yields determined in our work compare well with the recent acid catalyzed uptake experiments reported by Iinuma et al. .
 The model analysis also shows that the ultrafine particle growth is not very dependent on temperature. The condensable products either have low vapor pressure or the condensation process involves irreversible reactive uptake. Our experiments cannot distinguish between these two processes. It is possible that reaction products such as dicarboxylic acids, for example pinic acid, may contribute to the observed condensable yield. The lack of temperature dependence in particle growth over the range 278 to 320 K has possible implications for the growth of atmospheric aerosol as discussed in the section 3.7 below. Further studies to determine the vapor pressures of monoterpene oxidation products would aid this interpretation [Barsanti and Pankow, 2004, 2005; Chattopadhyay and Ziemann, 2005; Offenberg et al., 2006].
 A brief summary of our experimental results includes: (1) the experimental data are most consistent with a multicomponent nucleation and growth mechanism, (2) β-pinene (exocyclic) nucleation is more efficient at producing new particles than α-pinene (endocyclic) at room temperature, (3) OH oxidation significantly enhances nucleation for α-pinene but not for β-pinene, (4) the temperature dependence of particle nucleation for α-pinene and β-pinene differ but are both relatively weak, (5) the condensable yields for α-pinene and β-pinene are basically the same, and (6) the uptake of condensable reaction products is nearly temperature independent.
Hoppel et al.  performed α-pinene ozonolysis experiments in a large volume reaction chamber at room temperature. Experiments were made with 15 to 20 ppb α-pinene, ∼100 ppb ozone and no OH radical scavenger. Several experiments were also performed with the addition of known amounts of SO2 as discussed earlier. Results from our model for their conditions are in good agreement with their total particle concentration and its temporal profile, see their plate 1. It is worth noting that the Hoppel et al. experiments were performed at 30 and 50% relative humidity while our experiments were performed under dry conditions. The good agreement between model and experiments for these two studies implies that relative humidity does not significantly influence nucleation. However, this limited comparison may not be sufficient to draw reliable conclusions. It is known that water vapor does not affect the formation of OH radicals following monoterpene ozonolysis [Atkinson and Arey, 2003]; however, H2O could influence the yields of the low vapor pressure nucleating and condensing species. At present, there are contradictory results in the literature for the influence of water vapor on nucleation and growth following monoterpene ozonolysis (see Jonsson et al.  and references cited within for a thorough discussion). Hoppel et al.  concluded that (1) particle nucleation is rate limited in the formation of the dimer or trimer clusters of the nucleating species and (2) that particle growth can be explained by a single condensable species with a room temperature vapor pressure of 10 ppt or less. Our results are consistent with their conclusions.
Bonn and Moortgat , Bonn et al. , and Koch et al.  have reported studies of particle formation following the oxidation of α-pinene and β-pinene. Both chamber and flow reactors were used to examine OH initiated chemistry, using photochemical generation of OH radicals, and ozonolysis with an OH radical scavenger present. We have reproduced one of the ozonolysis experimental conditions reported by Bonn and Moortgat  and Bonn et al.  studies, 50 ppb α-pinene with 108 ppb O3 and 50 ppm cyclohexane, for direct comparison purposes. Our experimental data are shown in Figure 3. The total particle concentration at 800 s, ∼8000 particle cm−3, is in reasonable agreement with their values, ∼2000 and ∼5000 particle cm−3 (see their Figure 5). However, there is a significant difference in the observed delay time, 200 s in our study compared to ∼1000 s in theirs. The longer delay may reflect their lower sensitivity to small particles (and therefore longer delay while the particles grow to detectable size) due to the use of a scanning mobility particle sizer to measure particle size distributions. The particle growth rates reported by Moorgat and coworkers are also consistent with the condensable product yields obtained in this work. Therefore our ozonolysis results and model analysis are in general agreement with the data reported by Moorgat and coworkers.
Bonn and Moortgat  have reported that the OH initiated oxidation of α-pinene and β-pinene leads to much lower particle production than observed for ozonolysis. Also, particle growth is more rapid in the OH-initiated experiments (see their Figures 5 and 6). This contradicts our results in which enhanced nucleation due to OH for α-pinene was observed. In our work, particle growth rates obtained for the different oxidation sources were nearly indistinguishable. We have no clear explanation for these discrepancies although it may lie in the methods used to generate OH radicals. We utilized OH radicals produced in the monoterpene secondary chemistry (no scavenger) while Bonn and Moortgat produced OH radicals by UV photolysis of CH3ONO in the presence of a high concentration of NO. The secondary chemistry of peroxy radicals formed in the oxidation of the monoterpenes with NO may influence the yields of the nucleating and condensing species in their experiments. On the basis of their results, Bonn and Moortgat  and Koch et al.  conclude that OH oxidation of monoterpenes in the atmosphere is an unimportant source of new organic particles. In contrast, our results allow for nucleation under favorable atmospheric conditions as described in the next section.
3.7. Atmospheric Implications
 Ultrafine particle formation in remote forests is a well-documented phenomenon. The BIOFOR field campaign carried out at the SMEAR II research station in southern Finland represents one of the most extensive studies of particle nucleation bursts in remote forests (Tellus Special Issue 53B, 2001). In a typical nucleation burst, detection of ultrafine particle formation occurs several hours after sunrise with maximum particle concentrations of 1000 to 10,000 particles cm−3 several hours later [Kulmala et al., 2001]. The source(s) of particle nucleation and growth are however still unclear. Therefore a quantitative evaluation of possible biogenic VOC nucleation using atmospheric model calculations is worthwhile.
 In this section, we present results from a simple box model to evaluate the possible significance of biogenic VOC nucleation under conditions representative of the tropospheric boundary layer. The kinetic nucleation model and parameters derived in the present work for α-pinene and β-pinene are used for this evaluation. Other monoterpenes or other classes of biogenic compounds are not considered in these calculations. Therefore the results do not represent the total impact of biogenic compounds or the possible combined effects of these compounds and their oxidation products.
 Model parameters that have a significant influence on nucleation and growth include the O3, OH radical, and monoterpene concentrations, temperature, and background aerosol surface area. A typical O3 concentration of 50 ppb was used in all the calculations. Calculations started at sunrise using an OH radical temporal profile that peaked 4 hours after sunrise. Calculations were performed using peak OH concentrations of 1 × 106 molecule cm−3 and 1 × 107 molecule cm−3 and T = 278 K. The monoterpene mixing ratio was varied between 0.05 and 1 ppb and was constant during a model run. Background aerosol surface area has a large impact on nucleation. In the absence of experimental data, we have assumed unit uptake of the nucleating and condensable species on the background aerosol (i.e., the maximum possible particle growth rate) in our calculations. Calculations were made with initial surface areas of 0, 2, 5, and 10 μm2 cm−3 with a peak in the log normal particle size distribution at 60 nm diameter. A zero background aerosol surface area is not atmospherically realistic but represents an upper limit to nucleation in the model. In Figures 10 and 11, the particle concentration and growth rate after 4 hours of model integration are shown. In these calculations, only particles with diameters >3 nm (currently the limit for ultrafine condensation particle counters used for field measurements) are counted as new particles.
 We found, for both α-pinene and β-pinene, that ozonolysis alone (excluding the OH radical reaction) does not yield significant particle formation. This highlights the importance of OH radical initiated monoterpene gas phase chemistry in new particle formation. Ozonolysis of the monoterpenes does, however, contribute to particle growth.
 The atmospheric oxidation of both α-pinene and β-pinene produces significant particle concentrations, up to 10,000 particle cm−3, for high peak OH concentrations and monoterpene concentrations greater than ∼100 ppt. However, the concentration of new particles strongly depends on the background aerosol surface area as shown in Figure 10. For background aerosol surface area >10 μm2 cm−3 the loss of nucleating species via condensation on background aerosol dominates over nucleation for α-pinene and strongly diminishes the nucleation for β-pinene. The background aerosol surface area in a remote forest environment is strongly influenced by the local meteorology that is not accounted for in our box model calculations. Nucleation bursts at night are possible but are less likely to occur due to the normally higher background aerosol surface areas, >10 μm2 cm−3.
O'Dowd et al.  demonstrated that organic compounds of biogenic origin play an important role in the growth of these ultrafine particles. Tunved et al.  have recently correlated monoterpene emissions in a boreal region to particle formation. Particle growth rates observed during a remote forest nucleation burst fall into the range 2 to 20 nm h−1 [Kulmala et al., 2001]. Particle growth rates (calculated as the change in peak particle diameter of the log normal distribution after 4 hours of model integration) obtained in our calculations for a peak OH concentration of 1 × 106 molecule cm−3 are shown in Figure 11. The calculated growth rates fall in the range observed during atmospheric nucleation burst events. The yield of condensable products obtained from our laboratory work for both α-pinene and β-pinene are rather small, ∼0.06, but sufficient to account for observed growth rates under realistic atmospheric conditions consistent with the study of Tunved et al. .
 Our model calculations are based on results from our laboratory measurements that were not performed under actual atmospheric conditions. Notable differences in the conditions used being the relative humidity and the absence of NOx (= NO + NO2). Jonsson et al.  have observed an increase in particle and SOA formation with increasing relative humidity in the ozonolysis of limonene, Δ3-carene, and α-pinene. Previous studies of the relative humidity dependence in monoterpene ozonolysis are in poor agreement (see summary in Jonsson et al. ). Further work is needed to understand the discrepancies in previous laboratory studies and the relative humidity dependence, if any, appropriate for atmospheric model calculations. Nucleation in the OH initiated oxidation of monoterpenes has been observed to be independent of relative humidity [Bonn and Moortgat, 2002]. The presence of NOx will undoubtedly influence monoterpene secondary gas phase chemistry and product yields. There are numerous chamber studies of SOA formation in the presence of NOx in the literature. However, it is not yet clear how the presence of NOx will influence nucleation. In a remote forest environment, the NOx concentration is expected to be low and may have only a weak influence on nucleation. The possible role of biogenic VOC oxidation by NO3 also warrants further consideration [Bonn and Moortgat, 2002; Hoffmann et al., 1997].
α-Pinene and β-pinene are commonly considered to be representative monoterpene compounds. However, the impact of a biogenic compound on atmospheric aerosol depends on the combination of its atmospheric abundance and its efficiency for nucleation and SOA formation. This work has demonstrated that the higher nucleation efficiency of β-pinene offsets its lower atmospheric abundance. A quantitative evaluation of particle nucleation for biogenic compounds of low abundance is therefore desired for a more complete evaluation of biogenic nucleation and SOA formation.
 The presence of inhomogeneous monoterpene sources, that is, localized areas of high concentration, would enhance biogenic nucleation. The presence of low vapor pressure biogenic oxidation products from other sources may also influence nucleation and growth. More detailed atmospheric model calculations that include detailed meteorology are needed to quantify the role of monoterpenes as an atmospheric source of new particles.
 Laboratory measurements of nucleation and particle growth following the gas phase oxidation of α-pinene and β-pinene, monoterpenes (C10H16) commonly found in the troposphere, by ozonolysis and ozonolysis combined with OH-initiated chemistry are reported. Experimental measurements were made for a range of initial monoterpene and ozone concentrations at temperatures in the range 278 to 320 K. A coupled gas phase chemistry and kinetic particle nucleation model was used to interpret the experimental data and obtain a physical description of the nucleation process and secondary organic aerosol formation. The nucleation model analysis yielded absolute gas phase product yields for the nucleating and condensing species, as well as values for the thermodynamic stability of small clusters and the condensation process.
 A summary of the conclusions from this study include: (1) particle formation is most consistent with a multicomponent nucleation/growth mechanism, (2) β-pinene (exocyclic C=C bond) nucleation is more efficient than α-pinene (endocyclic C=C bond), (3) OH initiated oxidation significantly enhances particle production for α-pinene but not for β-pinene, (4) the temperature dependence of nucleation for α-pinene and β-pinene differ but are relatively weak (i.e., small molecular clusters are thermodynamic stable for the conditions of this study), and (5) the uptake of condensable products is nearly temperature independent under our experimental conditions and can be explained by either low vapor pressure species, <20 ppt at 296 K, or irreversible reactive uptake.
 Atmospheric box model calculations, under conditions typical of a remote forest, demonstrate that under favorable conditions monoterpene oxidation can lead to new particle formation. In addition, the uptake of the condensable products, formed in the oxidation of monoterpenes, on ultrafine particles can account for most, if not all, of the particle growth observed during remote forest nucleation bursts.
 We thank J. Holloway for the SO2 purity measurements and J. Kazil for helpful discussions. This work was supported in part by NOAA's Air Quality Program.