the Creative Commons Attribution 3.0 License. Atmospheric Chemistry and Physics

Measurements of particle formation following the gas phase oxidation of volatile organic compounds (VOCs) emitted by Scots pine (Pinus sylvestris L.) seedlings are reported. Particle formation and condensational growth both from ozone (O 3) and hydroxyl radical (OH) initiated oxidation of pine emissions (about 20-120 ppb) were investigated in a smog chamber. During experiments, tetramethylethy-lene (TME) and 2-butanol were added to control the concentrations of O 3 and OH. Particle formation and condensa-tional growth rates were interpreted with a chemical kinetic model. Scots pine emissions mainly included α-pinene, β-pinene, 3-carene, limonene, myrcene and β-phellandrene, composing more than 95% of total emissions. Modeled OH concentrations in the O 3-and OH-induced experiments were on the order of ∼10 6 molecules cm −3. Our results demonstrate that OH-initiated oxidation of VOCs plays an important role in the nucleation process during the initial new particle formation stage. The highest average particle formation rate of 360 cm −3 s −1 was observed for the OH-dominated nucleation events and the lowest formation rate of less than 0.5 cm −3 s −1 was observed for the case with only O 3 present as an oxidant. In contrast to the particle formation process, ozonolysis of monoterpenes appears to be much more efficient to the aerosol growth process following nucleation. Higher contributions of more oxygenated products to the SOA mass loadings from OH-dominated oxidation systems were found as compared to the ozonolysis systems. Comparison of mass and volume distributions from the aerosol mass spectrometer and differential mobility analyzer yields estimated SOA effective densities of 1.34±0.06 g cm −3 for the OH+O 3 oxidation systems and 1.38±0.03 g cm −3 for the O 3 dominated chemistry.


Introduction
On a worldwide basis, the annual biogenic volatile organic compounds (VOCs) budget is estimated to be in the range of 491∼1150 Tg C (Guenther et al., 1995;Griffin et al., 1999).Biogenic VOCs react with O 3 , OH and nitrate radical (NO 3 ), leading to the formation of secondary organic aerosols (SOA).Globally, the annual SOA production from these biogenic precursors is estimated to range from 2.5 to 44.5 Tg C (Tsigaridis and Kanakidou, 2003), comprising about 60% of the organic aerosol mass on the global scale and an even higher fraction regionally (Kanakidou et al., 2005).SOA can directly affect the Earth's radiation budget by scattering and absorbing incoming solar radiation (Kanakidou et al., 2005).Additionally, SOA can indirectly affect the radiation budget by modifying cloud properties through its potential contribution to the population of cloud Published by Copernicus Publications on behalf of the European Geosciences Union.condensation nuclei (CCN).SOA may also have adverse effects on human health through their inhalation.Understanding these effects requires a greater knowledge of the mechanisms inducing aerosol formation in the atmosphere.
Monoterpenes and isoprene are important classes of globally emitted organic biogenic species.Their oxidation can lead to low volatility products (low saturation vapor pressure) which are believed to be able to nucleate once their saturation ratio increased sufficiently.Compounds with higher saturation vapor pressure, known as semi-volatile organic compounds, will partition into previously formed particles, contributing to the growth and mass loadings of atmospheric aerosols.Thus, the SOA formation from oxidation of monoterpenes (Hoffmann et al., 1997;Jang and Kamens, 1999;Gao et al., 2004;Shilling et al., 2008;Presto and Donahue, 2005;Odum et al., 1996;Yu et al., 1999) and isoprene (Ng et al., 2007;Kroll et al., 2006;Claeys et al, 2004) by OH, O 3 and NO 3 has been intensively studied in laboratory and field campaigns (e.g.Kulmala et al., 2004;Laaksonen et al., 2008).These studies have demonstrated that the atmospheric oxidation of monoterpenes and isoprene represents a significant source of SOA.Despite the important advances in this area, SOA formation in the atmosphere is still poorly understood.For example, the relative roles OH and O 3 play in the SOA formation is still largely unconstrained and results from previous studies are somewhat contradictory.Bonn and Moortgat (2002) and Koch et al. (2000) concluded that OH oxidation of monoterpenes in the atmosphere is not an important source of new SOA, whereas Burkholder et al. (2007) reported that OH-initiated oxidation allows for new particle formation.Moreover, in most chamber studies, only single organic precursors are traditionally employed as model compounds to simulate SOA formation.Since the mechanisms of SOA formation from oxidation of biogenic hydrocarbons are complex, it is uncertain as to what degree the results from the single monoterpene chamber studies can be extrapolated to the conditions prevailing in the real atmosphere (VanReken et al., 2006).Recent studies have started to address these types of questions.In these studies, aerosol formation events have been simulated in more realistic conditions by using the direct biogenic emissions from macroalgae (McFiggans et al., 2004), white cabbage (Joutsensaari et al., 2005;Pinto et al., 2007), oak and loblolly pine (VanReken et al., 2006), Scots pine (Vaattovaara et al., 2009), birch, pine and spruce (Mentel et al., 2009).In some of these studies, new particle formation efficiencies from OH-or O 3 -induced oxidation were also investigated.
Here we presented results from chamber studies of new particle formation using direct VOC emissions of living Scots pine seedlings.Scots pine was selected because it is one of the dominant tree species in the European boreal forest (Räisänen et al., 2008).Both O 3 and OH are used as oxidants in the controlled conditions of these experiments.Our motivation is to evaluate the roles that these oxidants play in the formation of new particles and in the subsequent condensational growth.We also carry out measurements of aerosol chemical composition and densities.By using a mixture of VOCs emitted from plants, rather than a single SOA precursor, we are able to get more realistic and important information on which processes are more important in atmospheric aerosol formation.

Experimental description
The major components of the experimental system consisted of a biogenic emission enclosure, reaction chamber and both gas-phase and aerosol particle sampling systems.A block diagram of this system is shown in Fig. 1 and the overall system components and analysis methods are described in detail below.

Plant materials
Scots pine seedlings (1 year-old) were obtained from a forest nursery (Finnish Forest Research Institute, Suonenjoki Research Unit, Suonenjoki, Finland) in May and planted in 7.5 l plastic pots in 2:1 (v/v) quartz sand (Ø0.5-1.2 mm, SP Minerals Partek, Finland) and fertilized sphagnum peat (Kekkilä PP6, Finland).After planting, seedlings were grown in an open field site (Research Garden, University of Kuopio) until the end of July.Then the seedlings were taken into a greenhouse to avoid onset of winter dormancy.In October, prior to the experiments, the seedlings were kept in growth chambers (see Vuorinen et al., 2004 for more detail) at a temperature of +22 • C. To simulate herbivore attack on tree bark and to activate the chemical defense of seedlings, 3 cm long and 2 mm deep cuts were made on the base of main stem by a knife.The purpose of pretreatment on the seedlings is to increase the total VOC emissions.After these preparations, two seedlings were selected as the trial plants and placed carefully in two individual transparent Teflon TM FEP bags.These bags served as manifolds through which clean air passed to transport the pine emissions to the reaction chamber.Six lamps (Lival Shuttle Plus, Lival Oy, Sipoo, Finland, 24 W, PAR ca.350 µmol m −2 s −1 ) were utilized to ensure the optimal photosynthesis activity and VOC emissions.

Reaction chamber experiments
The reaction chamber for this study is a 6 m 3 rectangular chamber made of Teflon TM FEP film, which is attached to a stainless steel frame.The height of the chamber is 2.5 m with base dimensions of 1.0 m×2.4 m.Five experiments were carried out in the present study during 12-26 October 2007.The initial conditions of these experiments are summarized in Table 1.Prior to each experiment, the chamber was continuously flushed overnight with laboratory compressed clean air.The compressed air was processed through activated charcoal, Purafil TM select and absolute HEPA filter, to remove water, non-methane hydrocarbons, NO x and particles.After chamber purification, the air flow was rehumidified by bubbling through distilled water and introduced into the chamber to adjust the chamber relative humidity (RH).Then another additional clean air stream transported pine emissions from the plant enclosure to the chamber through Teflon tubing (approximately 1h prior to the start of the experiment).In addition to the plant emissions, in some experiments (e.g.E1a-c), tetramethylethylene (TME; 99+%, Aldrich) was also added.TME was delivered to the chamber by sending air through a heated bulb containing a known volume of TME.TME was used as a source of hydroxyl radicals (OH).It was chosen because OH production by TME ozonolysis requires neither NO x nor UV photolysis, and OH yields are near unity (Lambe et al., 2007;Paulson, et al., 1997).For experiments E2 and E3, no TME was applied.In E3, 2-butanol, used as OH scavenger, was introduced into the chamber in a similar manner as TME.The concentration of 2-butanol injected was approximately 600 times higher than that of the terpenes so that the reaction rate of OH with the scavenger exceeded that of OH with parent hydrocarbons by a factor of 100 (Keywood, et al., 2004;Docherty et al., 2003;Ziemann et al., 2003).Major products of the OH+2-butanol reaction are 2-butanone (with yield of 69% at room temperature) and acetone (with yield range from 12-29%) (Aschmann et al., 2002;Jiménez et al., 2005), which have high vapor pressures suggesting that they are too volatile to partition into the particle phase.After the terpenes, TME and 2-butanol were added and given time for mixing, O 3 were introduced to initiate oxidation.Ozone was generated by a UV lamp O 3 generator.Ozone enriched air (60-800 ppb) was introduced into the chamber (at 40 L/min).In three experiments, two ozone injections were made in order to investigate the roles of oxidants in the new particle growth process, as well as to obtain larger particle sizes and higher aerosol mass loading.The beginning of the first ozone injection marks the start of each experiment.In all experiments, temperature was controlled in the range of 25±2 • C and RH in 35±5%.No seed aerosols were added and the chamber was kept dark with a black polyethylene covering.
Three types of experiments were conducted.In experiments E1a-E1c, TME was added to investigate the role of OH-initiated oxidation in new particle formation.In experiment E2, No TME was added in order to simulate lower OH conditions.Experiment E3 was performed in the presence of 2-butanol (the OH scavenger) to investigate the role of ozonolysis alone in new particle formation.

Analytical methods and instrumentation
Nitric oxide (NO), total NO x and SO 2 concentration inside the chamber were measured with an AC 30M NO x analyzer and AF21M SO 2 analyzer (Environment s.a.), respectively.O 3 concentration was monitored at the inlet and inside the chamber by two DASIBI 1008-RS O 3 analyzers (Dasibi Environmental Corporation, Glendale, CA, USA).Temperature and RH were measured continuously using a RH and ).In all the experiments, SO 2 concentration was lower than the detection limit of analyzer (∼1 ppb).Measurements of VOC emissions from pine seedlings were made using two instruments, a gas chromatographymass spectrometer (GC-MS, Hewlett-Packard GC model 6890, MSD 5973) and a proton transfer reaction mass spectrometer (PTR-MS, Ionicon Analytik GmbH).For off-line analysis, VOCs were sampled onto about 150 mg Tenax TA adsorbent (Supelco, mesh 60/80) for 15 or 30 min.The collected VOCs were then analyzed with GC-MS.Identification of the compounds was based on the retention time and comparison of mass spectra with those of pure standards.The analysis method used here has been described in detail elsewhere (Pinto et al., 2007;Joutsensaari et al., 2005;Vuorinen et al., 2004).PTR-MS was used for the continuous on-line measurements of concentrations of selected parent hydrocarbons and various gas-phase intermediates and products.A detailed description of PTR-MS technique is given by Lindinger et al. (1998) and de Gouw and Warneke (2007).The calibration procedure and mixing ratio calculations were described in Taipale et al. (2008).During every experiment full scan mode was used to see if there were important compounds other than the selected ones that were measured continuously.Selected ions monitored were those with mass-to-charge (m/z) consistent with formaldehyde (m/z 30+1), methanol (m/z 32+1), acetone (m/z 58+1), isoprene (m/z 68+1), TME (m/z 84+1), monoterpenes (m/z 136+1) and sesquiterpenes (m/z 204+1).
Aerosol chemical composition and mass size distributions were measured by an Aerodyne Aerosol Mass Spectrometer (AMS).Detailed descriptions of the AMS measurements, calibration, modes of operation and data processing can be found in other publications (Jayne et al., 2000;Allan et al., 2003Allan et al., , 2004;;Jimenez et al., 2003aJimenez et al., , 2003b;;Zhang et al., 2004).Briefly, the AMS uses an aerodynamic lens to focus particles into a tight beam that is introduced into a high vacuum detection region (10 −8 Torr).Particles are detected by flash-vaporization on a porous tungsten surface heated to 600 • C, followed by ionization by a 70 eV electron source.The resulting positive ion fragments are detected by a Quadrupole Mass Spectrometer (QMA 430, Balzers Instruments, Balzers, Liechtenstein).Particle size is calculated from the particle time of flight.Mass spectra are achieved by the difference between sample and background and ion signals.
Aerosol particle size distributions over a range of 5.6-560 nm were measured using a fast mobility particle sizer spectrometer (FMPS, TSI model 3091).The FMPS was operated with aerosol and sheath flows of 10 and 40 Lmin −1 , respectively.The high measurement resolution (sampling in 1 s resolution and averaged in 5 min resolution in this study) enables the visualization of particle size distributions during new particle formation events.

Calculating particle formation and growth rates
Particles in the sub-25 nm size range are considered newly formed particles from precursor vapors (Dal Maso et al., 2005).Thus, in this work the particle formation rate was derived from the rate of change of the number concentration of aerosols with diameters between 5.6 nm (the lower size limit of the FMPS) and 20 nm.At the onset of new particle formation this formation rate is equal to the instantaneous particle formation rate at ∼6 nm (J 6 ), which can be related to the nucleation rate using various modeling approaches (Weber et al., 1997;Kerminen and Kulmala, 2002;Lehtinen et al., 2007).One means of assessing the relationship between the nucleation rate and J 6 is to calculate the factor η (Kerminen and Kulmala, 2002): where γ is a proportionality factor that is approximately equal to 0.2 nm 2 m 2 hr −1 , CS α=1 is the flux of non-volatile vapors to the pre-existing particle population in m −2 , and GR is the observed growth rate in nm hr −1 .η is conceptually similar to the "L-parameter" of McMurry et al. (2005), since the numerator represents the rate at which the critical cluster coagulates with the pre-existing particle population and the denominator the rate at which it grows by condensation.
In the limit that η 1 then growth rates of clusters greatly exceed loss of the clusters to coagulation, and thus the observed particle formation rate is an adequate representation of the nucleation rate.Using the measured average growth rates shown in Fig. 5, and adding the assumption that wall losses are negligible over the timescales of the growth of the nuclei from the critical cluster (ca. 1 nm in diameter) to 6 nm in diameter, we estimate this ratio to lie in the range of 0.1 to 0.001, where 0.1 corresponds to slower growth rates observed with E1c.Thus for the higher growth rates such as those observed during E1a, we expect the particle formation rates plotted in Fig. 5 to be approximately equal to the nucleation rate.For the growth rate calculations presented below, the geometric mean diameters (GMD) from the FMPS measurements were adopted to estimate the condensational growth rates of new particles.

Effective density
AMS measures the vacuum aerodynamic diameter (D va ), defined as the diameter of a standard density (ρ 0 ) sphere with the same terminal velocity as the particle of interest in a free molecular regime.The FMPS, on the other hand, provides a measurement of the mobility diameter (D m ), the diameter of a sphere with the identical migration velocity of particle in a constant electric field at atmospheric pressure.The effective density of particle, ρ eff., is defined as ratio of the vacuum aerodynamic and mobility diameter multiplied by unit   (1996) density (Jimenez et al., 2003b;DeCarlo et al., 2004;Bahreini et al., 2005).
where ρ p is the material density, ρ o is the unit density (1 g cm −3 ), D ve is the diameter of a spherical particle having the same volume as the particle under consideration, and X ve is the dynamic shape factor in the free molecular regime.
For spherical particles, the effective density is equivalent to the particle density, while the effective density can be significantly different for a particle with an irregular shape.Note that various definitions of effective density are used in the literature.These definitions do not yield the same numerical values for irregular particles (DeCarlo et al., 2004).In this study, a series of mobility diameters measured by FMPS was compared to their corresponding vacuum aerodynamic diameters measured by AMS in parallel in order to determine the densities of SOA.

Modeling of VOC oxidation
Since new particle formation is initiated by the gas-phase oxidation of VOCs, a kinetic model of VOC oxidation by O 3 and OH was built to interpret the new particle formation and growth events from the perspective of gas-phase chemistry.In this model, the oxidation of VOCs was calculated using the initial measured concentrations of VOCs and known reaction rates.O 3 and OH initiate the oxidation reactions of VOCs and TME and secondary loss process of OH and O 3 are taken into account.Briefly, O 3 -and OH-initiated oxidations of these VOCs produce condensing and nucleating species, OH radicals and other volatile compounds: If TME is present, its ozonolysis leads to the production of OH radicals: Thus, from rates of Eqs. ( 4) and ( 5), the formation rate of OH radical (f OH ) can be calculated using the following equation: where MT represents monoterpenes as listed in Table 2, ω i and ω i denote the yields of hydroxyl radicals from the O 3 reactions, and k represents rate constants.
Due to the high reaction rate constant of ozone with TME, there is competition of O 3 between TME and monoterpenes during the initial stage of reaction.The average consumption rate of O 3 can expressed as: where γ and γ are the correction factors to take into account the secondary loss of ozone with products.These correction factors were estimated according to the VOC molecular structures and the complexities of their reactions with O 3 and OH and further confirmed by a model sensitivity test.For the O 3 -induced oxidation, the correction factors were estimated to be 1.0 (Ng et al., 2006), 1.1 and 1.2 for the monoterpenes with one, two and three double C=C bonds, respectively.Similarly, for the OH-initiated oxidation, due to the high reaction activity of OH with hydrocarbon compounds, the correction factors are estimated to be higher than the O 3 oxidation case, with 1.1, 1.2 and 1.3 for those compounds with one, two and three double C=C bonds, respectively.The added TME also resulted in OH loss due to the high reaction potential of OH +TME: Thus reactions (2-5, 7) give the time-dependent OH concentration as: where γ and γ are the correction factors to take into account the secondary loss of OH by reaction with products.
Values of all parameters used in the model are listed in Table 2. Products from the O 3 -and OH-initiated TME oxidation are formaldehyde, methanol and acetone as well as some minor compounds (Niki et al., 1987;Tuazon et al., 1997Tuazon et al., , 1998)).The vapor pressures of these species are too high to take part into aerosol nucleation and growth processes and so are omitted from the model.Berndt et al. (2004) have also shown that ozonolysis of TME in the absence of SO 2 leads to no observable new particle formation.Thus, in this paper, O 3 and OH oxidation rates (excluding the contributions from TME) with VOCs are defined as (hereafter referred to as O 3 and OH rates): New particle formation rate and growth rate can be constrained directly by the oxidation rates of VOCs: Similarly, for the loss of other compounds and formation of products, the kinetics are described using the following rate equations: where β i , β i are the yields of acetone (AT) from reactions of O 3 and OH with VOCs and TME.Similarly, α i and α i are the formaldehyde (FAH) yields.VOC i includes all the monoterpenes, isoprene and TME listed in Table 2.

Results and discussion
3.1 Gas-phase chemistry

Scots pine emissions
The observed VOCs emitted from the Scots pine seedlings included monoterpenes, sum of isoprene and 2-methyl-3buten-2-ol (MBO), and minor amounts of other small molecular compounds, e.g.methanol, with monoterpenes dominating the total emissions in this work.Measured monoterpenes included α-pinene, β-pinene, 3 -carene, β-phellandrene, limonene and myrcene, making up more than 95% of the total VOC emissions as tabulated in Table 3.In E1c and E3, there was no detectable 3 -carene, with E1c exhibiting a very high contribution of β-pinene to total VOC emissions.The variations in relative emissions of species, specially α-pinene, β-phellandrene, limonene, and 3 -carene, may be due to emission rates that were sensitive to ambient environmental conditions (Räisänen et al., 2008, Tarvainen et al., 2005) and the frequency of 3 -carene-rich genotype (Manninen et al., 1998).Field results in Finnish forests have demonstrated the dependence of monoterpene emission rates on environmental temperature and light intensity (Räisänen et al., 2008;Tarvainen et al., 2005).In all cases, sesquiterpenes in the emitted compounds were lower than 0.05 ppb as determined with PTR-MS measurements.This is consistent with observations in Hyytiällä forest in Finland that sesquiterpene emissions from Scots pine are only 2-5% of the total monoterpene emission rates, with emissions ceasing in September (Tarvainen et al., 2005).
According to the measurements made in two locations in Finland, the major emitted compounds from Scots pine are α-pinene, β-pinene and 3 -carene, with approximate contributions of 60-85% to the total observed monoterpene emission rates, although late in autumn β-phellandrene can make 20% of total monoterpene emission (Räisänen et al., 2008).Lindfors and Laurila (2000) estimated that total biogenic VOC emissions from Finnish forest are dominated by monoterpenes, which contribute approximately 45%, while in deciduous trees the percentage can be as low as 10-15% (Mäntylä et al., 2008), and isoprene emissions are only 7% of the annual total.In the USA the main biogenic monoterpene emissions are α-pinene, β-pinene, and limonene (Sakulyanontvittaya et al., 2008).Thus the emission spectrum in this study is very consistent with natural Scots pine and the monoterpene distributions in Europe and USA, making this work more representative of the atmospheric reaction conditions than studies involving single VOCs or mixtures of a few VOCs commonly used in chamber experiments.

Reaction of emitted VOCs
Figure 2 shows the decay profiles of the ten VOCs (panel A) tabulated in  parent precusors are very consistent with the modeled results.Measured and modeled acetone concentrations also agree extremely very well, though the modeled formaldehyde concentrations are 10-40% higher than the measurements, which suggests that formaldehyde formation yields might be somewhat too high.Good agreement between modeled and measured VOC and product concentrations provides further confirmation to the validity of the model used in this study.

New particle formation from OH-and ozoneinduced chemistry
As the emissions of pine seedlings were oxidized by O 3 and OH, various products with differing volatilities were formed.Homogeneous nucleation can occur when the non-volatile products exceed their saturation vapor pressure.Later on, the semi-volatile species begin condensing on the surface of the pre-formed particles, causing an increase in particle size and aerosol volume concentrations.Contour plots of aerosol particle number size distributions and graphs of total number and volume concentrations as well as gas-phase oxidant concentrations are shown in Fig. 3.In these systems, OH radicals are produced not only by the ozonolysis of TME but also by the ozonolysis of monoterpenes.Both the OHand O 3 -initiated oxidation of monoterpenes might contribute to new particle formation.At high concentration of TME, ozonolysis of TME was expected to produce large amounts of OH radicals based on Eq. ( 5).It can be seen that a rapid increases in the particle number concentration occurred and obvious new particle formation events took place after the first addition of ozone in all cases.However, the maximum number concentration was 27 700 particle cm −3 in the highest TME experiment (E1c), which was much lower than the 128 000 particle cm −3 in the lowest TME case (E1a).This is caused by the higher OH concentration and 3 times faster OH oxidation rate in E1a than in E1c as shown in Table 1 and Fig. 5 and will be discussed further in Sect.3.2.3.In experiments with two ozone additions, the second ozone injection induced a very weak increase of aerosol number concentrations, but a large increase in aerosol volume concentrations.

New particle formation from ozonolysis
In experiment E3, 2-butanol was added to the chamber to remove the OH.2-butanol is known to scavenge the OH radicals efficiently thus ensuring that only ozonolysis reactions occur.Figure 4 shows the particle profiles of the ozonolysis experiment.Only a very weak nucleation event was observed in this experiment.The maximum number concentration was only 8120 particle cm −3 , much lower than that was found during OH+O 3 initiated cases.On the other hand, a relatively high rate of increase in aerosol volume was observed.

Particle formation and growth rates
The experiments described here emphasize the variability in new particle formation when different oxidants initiate reaction.The roles of OH-and O 3 -induced oxidation in new particle formation can be interpreted using the results from our kinetic model.The particle formation rates (see Sect. 2.4) and condensational diameter growth rates along with the oxidation reaction rates of monoterpenes with OH and O 3 are shown in Fig. 5 and Table 4. Intensive particle formation usually ended at the initial stages of reaction in this study (less than 60 min after the beginning of experiment) and growth continued as long as VOCs and O 3 were available.In E1a, very fast aerosol formation, with rate as high as 360 cm −3 s −1 , was observed,

E2 E1c
E1b E1a  consistent with the rapid OH oxidation rate (as defined in Sect.2.6) of 140×10 6 molecules cm −3 s −1 calculated with the model.In the pure ozonolysis experiment (E3, with OH scavenger present), the formation rate was quite low (<0.5 cm −3 s −1 ).In this case, OH oxidation rate is also low (0.72×10 6 molecules cm −3 s −1 ), though the O 3 oxidation rate was quite high (49.4×10 6molecules cm −3 s −1 ).These results indicate that OH-initiated oxidation reactions play a very important role in nucleation and the early stages of particle growth, whereas the ozonolysis of monoterpenes is less effectively involved in these processes under the current experimental conditions.Once the new particles are formed, OH oxidation reactions appear to have a weak influence on aerosol formation during the second ozone addition period   In these panels, formation and growth rates were determined from the FMPS size distributions data as discussed in Sect.2.4.OH and O 3 oxidation rates were calculated using the model described in Sect.2.6.

Fig. 5.
Effects of ozone-and OH-initiated oxidation rates with emitted VOCs on new particle formation and growth rates.In these panels, formation and growth rates were determined from the FMPS size distributions data as discussed in Sect.2.4.OH and O 3 oxidation rates were calculated using the model described in Sect.2.6.Data missed here Fig. 6.Mass spectra of the aerosol products of the oxidation of pine VOCs from the first two hours of each experiment.All mass spectra are normalized to the sum total of all organic mass fragments, providing a quantitative fractional contribution of each mass fragment to the total measured signal.The left and right panels correspond to the mass spectra of the first and the second hours of reactions, respectively.In E1a, AMS scanned in the range of 1-100 amu, whereas the other four experiments scanned a wider range (1-200amu), leading to small difference in percentage contributions in E1a.No AMS measurements were made in the first hour of experiment of E1a.39 Fig. 6.Mass spectra of the aerosol products of the oxidation of pine VOCs from the first two hours of each experiment.All mass spectra are normalized to the sum total of all organic mass fragments, providing a quantitative fractional contribution of each mass fragment to the total measured signal.The left and right panels correspond to the mass spectra of the first and the second hours of reactions, respectively.In E1a, AMS scanned in the range of 1-100 amu, whereas the other four experiments scanned a wider range (1-200 amu), leading to small difference in percentage contributions in E1a.No AMS measurements were made in the first hour of experiment of E1a.
(second OH concentration peak, shown for E1a and E1c in Fig. 5), instead participating in the condensational growth process.This can be explained by the importance of reaching a high supersaturation in order for strong nucleation and early particle growth to occur.Such conditions are possible in the beginning of the experiment when there are no particles present in the chamber, i.e. there is no surface onto which vapor may condense.At the second OH peak, total particulate surface area in the chamber is high, and the reaction products will start condensing to existing aerosol as soon as they reached supersaturation.High supersaturations are therefore not possible leading to only weak (or no) new particle formation.
The modeled OH and O 3 reaction rates also show correlations with the condensational growth rates in the experiments.The peak growth rate in E1a (126 nm h −1 ) is 1.7 times higher than that of E3 even though its O 3 reaction rate (25.5×10 6 molecules cm −3 s −1 ) is two-times lower.Similarly, an O 3 reaction rate of 5.5×10 6 molecules cm −3 s −1 was observed in E1c, which is 9 times lower than in E3.However, a growth rate of 25.0 nm h −1 in E1c was observed, which is only 3 times lower than in E3.These results indicate that not only OH but also O 3 play roles in the aerosol condensational growth process.In E3, the OH reaction rate is at least 25 times lower (0.72×10 6 molecules cm −3 s −1 ) than the other four cases (>18.3×10 6molecules cm −3 s −1 ).However, a high growth rate of 73.8 nm h −1 was achieved, suggesting that the ozonolysis of VOCs is more efficient that OH oxidation for aerosol growth.
Comparison between experiments E1a and E1b (Table 4) further verifies the above conclusions.These experiments were selected because their initial emission distributions were roughly the same.The initial emitted VOCs distributions were found to have an influence on the nucleation and growth rates as discussed below.The particle formation rate in E1a was 4.5 times higher than that in E1b, which is consistent with 3.8 times faster of OH oxidation rates in E1a than in E1b.Similarly, a 126.0 nm h −1 diameter growth rate in E1a was observed, 6.6 times higher than that in E1b, which corresponds well to the O 3 oxidation rates in E1a of 6.3 times faster than in E1b.
Aerosol formation and growth rates were also impacted by the amounts and molecular structures of emitted VOCs.Oxidation of compounds with different positions of unsaturated bonds and number of carbon atoms in the molecules can result in different aerosol growth rates and SOA mass yields (Griffin et al., 1999;Ng et al., 2006).Since 3 -carene compound has one endo double bond (internal to ring structure) and is more likely to form condensable products than compounds with external double bonds (Griffin et al., 1999), the absence of 3 -carene compound in the E1c and E3 might contribute to the lower nucleation rates than other cases.Furthermore, the high contributions of β-pinene (with an exo double bond) to the total VOC mix in E1c might lead to the lower nucleation rate.The conclusion that β-pinene may have inhibited particle formation in the Holm oak (Quercus ilex) experiments was also reached in previous work ( Van-Reken et al. 2006).Nucleation and growth from the oxidation of monoterpenes by OH and O 3 has been studied over the past ten years (Bonn and Moortgat, 2002;Burkholder et al., 2007;Hoppel et al., 2001;Griffin et al., 1999;Mentel et al., 2009).Most studies were performed using one monoterpene such as α-pinene and β-pinene as representative compounds to assess the OH-and O 3 -initiated oxidation on new particle formation efficiency.Bonn and Moortgat used a scanning mobility particle sizer (SMPS) to study the new particle formation during the oxidation of α-and β-pinene by O 3 and OH.By comparing the aerosol number and volume concentrations in different cases, they reported that ozonolysis of monoterpenes was much more efficient for particle formation than what was the observed for OH-initiated case.This contradicts our results in which ozonolysis is less effective in nucleation process.Our results are consistent with the results of Burkholder et al. (2007), which find that monoterpene ozonolysis alone (with no OH radical reaction) cannot yield significant particle formation.However, we find that ozonolysis does contribute to particle growth.The results of this study are in strong agreement with recent work showing that for the oxidation of pine emissions, ozonolysis does not induce particle formation and OH radicals are essential for new particle formation (Mentel et al., 2009).

Chemical signatures
Chemical signatures of the newly formed aerosols during the above experiments were investigated utilizing the AMS. Figure 6 displays the average mass spectra of SOA for the first hour and second hour for each experiment.All spectra were normalized to total mass signal.We can see that most of the spectra were characterized by some intense mass fragments at m/z 27,29,41,43,44,51,55,65,77,91 and 141. Mass Fragments 27,29,41,43 and 55 represent the ion series of C n H + 2n−1 and C n H + 2n+1 and separated by 14 mass units due to loss of CH 2 , which can indicate the presence of heavily saturated and non-oxidized hydrocarbon-like organic species (Alfarra et al., 2006;Allan et al., 2006).Another potential contribution to mass 43 is an oxygen-containing fragment, C 2 H 3 O + .Fragment 44 corresponds to CO + 2 , which represents highly oxidized classes of compounds containing carbonyl and carboxylic acid functional groups (Alfarra et al., 2006).The ion series 77 (C 6 H 5 ) and 91(C 7 H 7 ) are usually considered to be associated with aromatic fragments (Zhang et al., 2005;Bahreini et al., 2005) but can be observed in this study.Mass fragments of 51 and 141 might be derived from the ozonolysis reaction of VOCs but their identifications are unclear at present.
In the TME experiments (E1b and E1c), mass spectra were dominated by mass fragment 44, contributing to as high as 20.5% and 13.8% to the total first hour mass signal, respectively.As a comparison, during the OH-scavenged experiment (left panel E3 in Fig. 6), no fractional contribution by m/z 44 signals was found in the first hour.This indicates OH plays a key role in forming species which fragment to give m/z 44 species.The importance of OH reaction for each VOC (as a percentage of total reaction) during new particle formation is presented in where r i,OH(t) and r i,O 3 (t) were the reaction rates of VOC i with OH and O 3 at time t (min), respectively.In all the OH-dominated experiments (E1a-1c), all VOCs but terpinolene reacted primarily (>75%) with OH rather than O 3 .For E2, both O 3 -and OH-initiated oxidation were important, with the fraction of OH reaction varying from 8.4% (terpinolene) to 97.8% (myrcene).And in the ozonolysis experiment (E3), ozonolysis dominated for all VOC oxidation except camphene.The relative importance of OH reactions in all experiments corresponds well to the mass fraction of m/z 44 ions, strongly suggesting that OH-initiated oxidation leads to the formation of highly oxidized aerosol species.On the other hand, the very high contribution of m/z 43 signal to the mass loadings in E3 suggests that ozonolysis reaction might contribute more to the somewhat less oxidized products.
Averaged mass spectra appeared to have highly similar fragmentation patterns during the second hour, implying the products formed have broadly similar chemical functionality in these experiments.These similarities in the mass spectra were further confirmed in longer reaction time (not shown in Fig. 6), suggesting that once particles were formed, the chemical signature detected by AMS did not change significantly under these experimental conditions.However, there are still significant differences in the minor mass fragments, suggesting that the aged SOA might have different chemical composition.Contribution of m/z 44 to the mass spectrum in the second hour of oxidation decreased compared to the first hour in E1b (from 20.4% to 5.3%) and E1c (13.8% to 4.9%).This might be due to the fact that terpenes react with OH and O 3 in the first hour, leading to formation of the gas-phase nucleating and condensing species.The most highly oxidized material generally has the lowest vapor pressure, and therefore, condenses earlier.Once new particles are formed, less oxidized (more volatile) species may also partition into the aerosol phase, leading to an increase in aerosol volume and a decrease in oxygen content of the aerosol.
The mass spectra measured in the present studies also share some common features with those measured in the real atmosphere.The mass spectrum from particles following a nucleation event at Hyytiälä forest in Finland and Mexico city exhibit peaks at m/z 27, 29, 41, 43, 44, 51, 55, 57, 77, 91 and 141 (Alfarra et al., 2006;Allan et al., 2006;Salcedo et al., 2006).Comparison of the measurements made in this study with the average ambient mass spectrum from one nucleation event from 09:30 a.m. to 11:30 a.m. on 1 June 2007 at Hyytiälä is shown in Fig. 7.The similarities (R 2 =0.89) in these two types of mass spectra indicate that the experimental results are able to reproduce the product distributions of SOA formed in the real atmosphere.

Effective density (ED)
Effective density of SOA was calculated following the approach outlined in Sect.2.5.In this method, the mobility and vacuum aerodynamic diameters were identified from the peak values by using a log-normal fitting to volume/mass size distributions.The determined effective density from the OH and O 3 oxidation induced SOA was 1.34±0.06g cm −3 , nearly identical to the value of 1.38±0.03g cm −3 of particles from the ozonolysis reaction systems.The results show that the effective density for both types of SOA is not substantially impacted by the initiating oxidants.The density is also found to be independent of particle size within the accumulation mode.However the effective density showed a dependence on the particle size with Aitken mode in this work and these results need further confirmation.Our reported aerosol densities are in agreement with measured densities of 1.25±0.15g cm −3 of SOAs from real pine emissions (Mentel et al., 2009).These results are also consistent with density measurements of 1.2-1.7 g cm −3 for ambient aerosol and laboratory-generated monoterpene SOA (Shilling et al., 2009;Saathoff et al., 2008;Bahreini et al., 2005;Alfarra et al., 2006;Kostenidou et al., 2007;Kannosto et al., 2008).
Accurate measurements of density are very important to convert the aerosol volume concentration to mass loadings.The effective density is heavily affected by the actual material density and particle shape factor (Bahreini et al., 2005;DeCarlo et al., 2004).Alfarra et al. (2006) reported shape of SOA formed from the photooxidation of α-pinene to be spherical.In their experiments, effective densities were equivalent to the material densities.However, SOA are reported to be in solid or waxy phase with densities larger than 1.25 g cm −3 from the ozonolysis of biogenic emissions (Kostenidou et al., 2007).Irregular shapes of the solid or waxy aerosol make the effective density lower than the real material density.In this study, our aerosol might be nonspherical and our ongoing biogenic SOA experiments are investigating this possibility.This will be further discussed in future work.

Conclusions
We have studied freshly formed aerosols following the gas phase oxidation of the direct emissions of Scots pine (Pinus sylvestris L.) seedlings in a smog chamber.New particle formation events following ozonolysis and O 3 plus OH initiated oxidation were characterized.2-butanol and TME were used to control the ratios and the amount of O 3 and OH.Emitted gaseous species, aerosol formation, aerosol condensational growth, aerosol chemical composition and aerosol density are measured and investigated in this study using GC-MS, PTR-MS, FMPS, AMS and a chemical kinetics model.
Major emitted VOCs from Scots pine are observed to be monoterpenes.Ozone-and OH-initiated oxidation of these VOCs produced large amounts of new particles.OH-initiated oxidation plays a critical role during the initial stages of nucleation and new particle formation.Ozonolysis seems to be more important for the particle condensational growth, but it does not contribute substantially to particle formation.These results demonstrate that the OH oxidation of VOCs could be a source of new organic particles in the atmosphere.
Mass spectral results show that highly oxidized compounds (m/z 44 signals) contribute more to SOA formed from OH-initiated oxidation of emitted VOCs during the initial stage of nucleation events.Ozonolysis of emitted VOCs produces SOA which is less oxidized, with the mass spectrum dominated by the m/z 43 ions.Overall, the mass spectral patterns in different experiments do not differ significantly, indicating the SOA produced have broadly similar chemical functionality.
From AMS and FMPS data it was determined that the aerosol effective densities are almost identical in these two oxidation systems, implying that the products formed by the different oxidants have roughly similar densities.
The experiments carried out here were designed to closely match atmospheric conditions as much as possible, so that the present results would be applicable to new particle formation and growth in the atmosphere, particularly in pristine forested environments.The oxidants (O 3 and OH) and organics (monoterpenes) studied are among the most important in the troposphere; concentrations of OH and monoterpenes approached tropospheric levels; and the organics included an atmospherically relevant mixture of monoterpenes (direct vegetation emissions).

35Fig. 2 .
Fig. 2. Comparisions of measured and modeled results from experiment E1a: (a) Changes in emitted VOC concentration upon reaction (from GC-MS measurements) (b) TME decay profiles (from GC-MS measurements) and gas-phase product concentration of acetone and formaldehyde (from PTR-MS measurements).Results from other four experiments are similar and so are not shown here.

Fig. 3 .
Fig. 3. New particle formation from OH-and O 3 -initiated oxidations of pine emissions.The frames contain, from top to bottom: (A) particles size distribution intensity plot, (B) measured aerosol number concentration (red curve) and volume concentration (black curve), (C) modeled OH concentration inside the chamber (blue curve), and (D) measured ozone concentrations at the inlet (black dashed curve) and inside the chamber (pink curve) and modeled ozone concentration inside the chamber (green dashed curve).

Fig. 3 .
Fig. 3. New particle formation from OH-and O 3 -initiated oxidations of pine emissions.The frames contain, from top to bottom: (a) particles size distribution intensity plot, (b) measured aerosol number concentration (red curve) and volume concentration (black curve), (c) modeled OH concentration inside the chamber (blue curve), and (d) measured ozone concentrations at the inlet (black dashed curve) and inside the chamber (pink curve) and modeled ozone concentration inside the chamber (green dashed curve).

Fig. 4 .
Fig. 4. New particle formation from ozonolysis of pine emissions in the presence of 2-butanol OH scavenger.The panel legends are the same as in Fig.3.Modeled OH concentration was quite low and is not shown.

37Fig. 4 .
Fig. 4. New particle formation from ozonolysis of pine emissions in the presence of 2-butanol OH scavenger.The panel legends are the same as in Fig. 3. Modeled OH concentration was quite low and is not shown.

Fig. 5 .
Fig.5.Effects of ozone-and OH-initiated oxidation rates with emitted VOCs on new particle formation and growth rates.In these panels, formation and growth rates were determined from the FMPS size distributions data as discussed in Sect.2.4.OH and O 3 oxidation rates were calculated using the model described in Sect.2.6.

Table 1 .
Initial conditions of chamber experiments performed and analysis results.

Table 2 .
Gas phase kinetic constants, OH yield (Y OH ), acetone (AT) yield (Y AT ), formaldehyde (FAH) yield (Y FAH ) and correction factors used in the study.

Table 3 .
Percentage molar contributions of different chemical compounds to pine emissions.
Table3and the formation curves of gasphase acetone and formaldehyde as functions of time (panel B) shown for experiment E1a.The modeled concentrations were calculated based on the OH-and O Comparisions of measured and modeled results from experiment E1a: (A) Changes in emitted VOC concentration upon reaction (from GC-MS measurements) (B) TME decay profiles (from GC-MS measurements) and gas-phase product concentration of acetone and formaldehyde (from PTR-MS measurements).Results from other four experiments are similar and so are not shown here.

Table 4 .
Comparisons of particle formation and growth rates vs. OH and O 3 oxidation rates.

Table 5 .
Fraction by which each VOC reacts with OH during the first hour of new particle formation.

Table 5 .
Values were calculated rison of the AMS mass spectra of aerosol products in the chamber experiments from this work with those erosol at Hyytiälä forest station.The plotted data are the fractional contribution of each mass peak to the hown in Fig.6.The color scale is a function of m/z number and the black dotted line is made by a linear fit.trum in this work (y-axis) is the average from the first hour in E2 and ambient measurement is from twomass spectra during one nucleation event from 9:30AM to 11:30AM on June 1 st , 2007, in Hyytiälä forest in Comparison of the AMS mass of aerosol products in the chamber experiments from this work with those from ambient aerosol at Hyytiälä forest station.The plotted data are the fractional contribution of each mass peak to the total signal as shown in Fig. 6.The color scale is a function of m/z number and the black dotted line is made by a linear fit.The mass spectrum in this work (y-axis) is the average from the first hour in E2 and ambient measurement is from two-hour averaged mass spectra during one nucleation event from 09:30 a.m. to 11:30 a.m. on 1 June 2007, in Hyytiälä forest in Finland.

Table 6 .
Effective density for different selected sizes of OH and O 3 induced SOA.