Atmospheric Chemistry and Physics Particle mass yield from β-caryophyllene ozonolysis

The influence of second-generation products on the particle mass yield of -caryophyllene ozonolysis was systematically tested and quantified. The approach was to vary the relative concentrations of first- and second- generation products by adjusting the concentration of ozone while observing changes in particle mass yield. For all wall- loss corrected organic particle mass concentrations Morg of this study (0.5 10 µg m 3 the par- ticle mass yield increased to as high as 70 % for the ulti- mate yield corresponding to the greatest ozone exposures. These differing dependencies on ozone exposure under dif- ferent regimes of Morg are explained by a combination of the ozonolysis lifetimes of the first-generation products and the volatility distribution of the resulting second-generation products. First-generation products that have short lifetimes produce low-volatility second-generation products whereas first-generation products that have long lifetimes produce high-volatility second-generation products. The ultimate particle mass yield was defined by mass-based stoichio- metric yields i of 0 = 0.17± 0.05, 1 = 0.11± 0.17, and 2 = 1.03± 0.30 for corresponding saturation concentrations of 1, 10, and 100 µg m 3 . Terms 0 and 1 had low sensitiv- ity to the investigated range of ozone exposure whereas term 2 increased from 0.32± 0.13 to 1.03± 0.30 as the ozone exposure was increased. These findings potentially allow for


Introduction
Sesquiterpene emissions have been estimated as 10-30 % of those of monoterpenes (Helmig et al., 2007;Sakulyanontvittaya et al., 2008a).Because of their fast ozonolysis reactivity and their tendency to form low-volatility products, sesquiterpenes (C 15 H 24 ) are important precursor molecules to secondary organic material (SOM) (Hoffmann et al., 1997;Jaoui and Kamens, 2003;Jaoui et al., 2003).For instance, βcaryophyllinic acid, which is an ozonolysis product of the sesquiterpene β-caryophyllene, has been measured at substantial concentrations for ambient particles in environments as diverse as the tropics and the arctic (Jaoui et al., 2007;Hu et al., 2008;Fu et al., 2009).For typical concentrations of tropospheric oxidants, the lifetimes of sesquiterpenes are about 2 min for the reaction with ozone, 2-3 min for the reaction with nitrate radicals at night, and 30-40 min for the reaction with hydroxyl radicals in the day (Shu and Atkinson, 1995).Sesquiterpene emissions usually occur during the day (Helmig et al., 2007).Reaction with ozone is therefore regarded as the dominant degradation pathway of many sesquiterpenes in the atmosphere (Atkinson and Arey, 2003).
As one of the most atmospherically prevalent sesquiterpenes, β-caryophyllene has been studied extensively in the laboratory.The reported particle mass yields associated with its oxidation range from 6-62 % for dark ozonolysis and from 37-125 % for photooxidation (Grosjean et al., 1993; Published by Copernicus Publications on behalf of the European Geosciences Union.Q. Chen et al.: Particle mass yield from β-caryophyllene ozonolysis Hoffmann et al., 1997;Griffin et al., 1999;Jaoui et al., 2003;Lee et al., 2006a, b;Winterhalter et al., 2009) (Table 1).Particle mass yield Y is defined as the organic particle mass concentration M org that is produced divided by the mass concentration of the precursor volatile organic compound VOC that is reacted (Odum et al., 1996): Y ≡ M org / VOC.Yield depends to a first approximation on the volatility of the reaction products.An implication is that the relative tendency of molecules to partition from the gas phase to the particle phase (i.e., M org ) depends on the extent of particle phase (i.e., M org ) that is present (Pankow, 1994a,b).Reaction conditions also influence yield by affecting the relative importance of competing kinetic pathways and hence the relative proportion of products of differing volatilities (Chan et al., 2007).Important experimental conditions include relative humidity, particle acidity, ozone concentration, and NO x concentration (Ng et al., 2007;Offenberg et al., 2009;Winterhalter et al., 2009).Winterhalter et al. (2009), for example, demonstrated that the addition of gas-phase formic acid or water scavenged reactive Criegee intermediates and thereby influenced the concentration profile of the reaction products, ultimately leading to a doubling of the particle mass yield by β-caryophyllene ozonolysis.Offenberg et al. (2009) showed that increased particle acidity led to greater particle mass yields from β-caryophyllene photooxidation.
Among the several factors influencing particle mass yield from β-caryophyllene ozonolysis, the stoichiometric ratio of ozone to β-caryophyllene is an especially important regulator.β-caryophyllene has two double bonds, and the endocyclic bond is more reactive to ozone by two orders of magnitude than is the exocyclic bond (Nguyen et al., 2009;Winterhalter et al., 2009).Second-generation products are produced by ozonolysis of the remaining double bond of the first-generation products.The second-generation products typically have lower vapor pressures and hence greater thermodynamic tendencies to condense to the particle phase (Li et al., 2011).As a result, particle mass yield increases significantly in the case that sufficient ozone is present for first-generation products to continue oxidation.For instance, when injecting excess ozone, Ng et al. (2006) observed continued and rapid particle growth even after βcaryophyllene was completely consumed.This observation was explained by the production and gas-to-particle condensation of second-generation products.
In the atmosphere, concentrations of biogenic volatile organic compounds (BVOCs) are ultimately limited by their surface emissions and are normally less than 10 ppbv in forested environments (Helmig et al., 1998).By comparison, ozone concentrations range from 10 to 30 ppbv at background sites (Fiore et al., 2003) to over 100 ppbv for urban locations (Solomon et al., 2000).Moreover, the atmospheric concentration of ozone is continuously renewed by atmospheric production.The implication is that under most circumstances the ozone exposure in the atmosphere is in excess of the possible consumption by BVOCs, espe-cially for sesquiterpenes.Few laboratory studies, however, have carried out experiments for conditions of excess ozone (i.e., ozone: β-caryophyllene > 2) (Table 1).As such, particle mass yield in the regime of second-generation dominance remains largely unknown, especially for conditions that overlap with atmospheric concentrations (M org < 10 µg m −3 ).
Herein, experimental observations for the dark ozonolysis of β-caryophyllene (0.8-46 ppbv) under various conditions of excess ozone are reported.The qualitative molecular identification of the particle-phase products for these experiments and an associated discussion of their production mechanisms were presented in Li et al. (2011).An analysis of oxygen-tocarbon (O:C) and hydrogen-to-carbon (H:C) elemental ratios was presented in Chen et al. (2011) in a comparative study of isoprene, α-pinene, and β-caryophyllene.In relation to these two earlier reports, the focus of the present report is on second-generation products and their effects on particle mass yield.

Experimental procedures and measurements
Experiments were carried out in the Harvard Environmental Chamber.Detailed descriptions of the chamber were previously published (Shilling et al., 2008;King et al., 2009).Compared to those reports, in this study a 0.13-mm (5-mil) thick PFA Teflon bag having a volume V of 4.7 m 3 was newly installed as a replacement for the previously described 0.05mm (2-mil) PFA bag.As in the earlier studies, the bag served as a continuously mixed flow reactor (CMFR) (Kleindienst et al., 1999;Seinfeld et al., 2003).A flow Q of 22 l min −1 was used, corresponding to a mean residence time τ of 3.6 h, as calculated by τ = V /Q.Temperature and relative humidity were held at 25 ± 1 • C and 40 ± 1 %, respectively.Ammonium sulfate particles were continuously present in the inflow of the CMFR.The surfaces of these particles served as a suspended substrate to accommodate the gas-to-particle condensation of the oxidation products of β-caryophyllene.An ammonium sulfate solution (Sigma-Aldrich, ≥ 99.0 %) was atomized (TSI, 3076, Liu and Lee, 1975) to produce a polydisperse particle population, which was then dried in a 160-cm silica gel diffusion dryer (RH < 10 %).A monodisperse electric-mobility fraction was selected from the population by a differential mobility analyzer (TSI, 3071 DMA; 85 Kr bipolar charger) (Knutson and Whitby, 1975).The DMA was operated with 10 l min −1 sheath and 2 l min −1 aerosol flows.Electricequivalent (+1 charge) mobility diameters d m,+1 of the seed particles for each experiment are summarized in Table 2.The corresponding surface area concentration inside the CMFR prior to initiating β-caryophyllene ozonolysis was (4.0 ± 2.0) × 10 −5 m 2 m −3 .For experiments at the highest initial β-caryophyllene concentration of 46.4 ppbv, polydis- perse seed particles were used instead of monodisperse ones so that the surface area concentration (2.0 × 10 −3 m 2 m −3 ) was sufficiently high to prevent new particle formation by homogeneous nucleation of the organic gases.
In a series of experiments, the initial concentration of βcaryophyllene inside the CMFR prior to ozonolysis was varied from 0.8 to 46.4 ppbv.A liquid solution (1:2500, v/v) of β-caryophyllene (Sigma-Aldrich, ≥ 98.5 %) in cyclohexane (Sigma-Aldrich, ≥ 99.9 %) was fed by a syringe pump into a round-bottom flask warmed to 70 • C. A continuous flow of air swept the evaporated molecules of β-caryophyllene and cyclohexane from the flask into the CMFR for the entire experimental period.The concentration of β-caryophyllene inside the CMFR was adjusted by using a variable rate of liquid injection from the syringe pump.Cyclohexane served as a scavenger of hydroxyl radicals produced by some ozonolysis pathways.A quadrupole proton-transfer-reaction mass spectrometer (Ionicon, PTR-MS) was used both to confirm the β-caryophyllene concentration in the CMFR prior to ozone injection and to track the extent of reaction after ozone injection (Shilling et al., 2008).Ozone concentrations of 50, 100, and 200 ppbv were used during the conducted experiments (Table 2).Within a single experiment, the ozone concentration was held constant by feedback control, meaning that sufficient ozone was added to react away nearly all the β-caryophyllene while maintaining a constant concentration of 50, 100, or 200 ppbv.For all experiments, the ozone concentrations were therefore in excess compared to the reacted concentrations of β-caryophyllene.
The outflow of the CMFR was continuously sampled by a scanning mobility particle sizer (TSI, 3936L22 SMPS, Wang and Flagan, 1989), by filter-based collection for off-line molecular analysis using an ultra-performance liquid chromatography coupled to an electrospray-ionization timeof-flight mass spectrometer (Waters, ACQUITY/LCT Premier XE UPLC-ESI-ToF-MS, Neue et al., 2010), and by an on-line particle-vaporization electron-impact high-resolution time-of-flight mass spectrometer (Aerodyne, HR-ToF-AMS, DeCarlo et al., 2006).The SMPS provided measurements of the number-diameter distribution of the particle population.The UPLC-ESI-ToF-MS molecular analysis of the material sampled onto the filters was previously presented in Li et al. (2011).The HR-ToF-AMS was used for the in situ collection of the mass spectrum of the particle population in the CMFR outflow.During a single steady-state experiment, no significant changes in the mass spectra were observed, and the AMS data sets under steady-state conditions were averaged for 4 to 12 h to increase the signal-to-noise ratio.The measured quantities at steady state typically fluctuated within 5 % during the course of an experiment.The particle population largely had diameters within the AMS acceptance window of 50 to 1000 nm vacuum aerodynamic diameter.The error in M org because of undetected particles is estimated as smaller than 1 % based on the SMPS number-diameter measurements.For experiment #27 (Table 2), a DMA (TSI, 3071; 85 Kr bipolar charger) was coupled to an aerosol particle mass analyzer (Kanomax, APM 3600, Ehara et al., 1996) to measure the particle effective density.
The CMFR was an experimental configuration that facilitated the study of ultimate yield by providing a mean residence time sufficient for multi-generational chemistry (Kleindienst et al., 1999).For example, the lifetime of β-caryophyllene was 20 s based on a rate constant of (1.2 ± 0.1) × 10 −14 molecule −1 cm 3 s −1 at 200 ppbv ozone.) outflow , M org,corr , and OS c , respectively represent the reacted concentration of the precursor VOC in an experiment, the mass concentration of particle-phase SOM measured by the AMS, the mass concentration of particle-phase SOM corrected by wall loss and used in the yield calculation (Eq.2), and the average oxidation state of carbon for particle-phase SOM, calculated as 2(O : C) -1(H : C) (Kroll et al., 2011).The O : C and H : C ratios were originally reported in Chen et al. (2011).Material density ρ org is reported only for experiments using monodisperse seed and having (M org ) outflow > 1 µg m −3 .Errors represent the measurement precision.For Exp. #10, a 210 Po charger was installed in the seed injection line, which neutralized the charged particles exiting the DMA and prior to entering the chamber.As a result of neutralization, electrostatic particle loss to the bag walls decreased (McMurry and Rader, 1985).

Measured quantities
Derived quantities at steady state Seed The average ozonolysis lifetime of the first-generation products of β-caryophyllene was about 0.5 h (1860 s) based on a rate constant of (1.1 ± 0.4) × 10 −16 molecule −1 cm 3 s −1 (Shu and Atkinson, 1995;Winterhalter et al., 2009).In comparison, the mean residence time of the Harvard Environmental Chamber for most experiments was 3.6 h (13 100 s) (Shilling et al., 2008;King et al., 2009).The CMFR operation also improved the precision and accuracy of yield measurements at low M org by providing sufficient observation time for significant signal averaging (Shilling et al., 2008).By establishing a steady state and thereby saturating the surface layers of the Teflon bag material (Matsunaga and Ziemann, 2010), CMFR operation also reduced the effects of wall interactions on reactive and non-reactive exchanges with gas-phase molecules, including the re-partitioning of condensable products (cf.further discussion in the Sect.A of the Supplement).

AMS data analysis
The spectra collected by the HR-ToF-AMS were used to calculate the mass concentrations and the elemental ratios of the particle-phase secondary organic material present in the outflow from the CMFR (Table 2).The spectra were analyzed using the software toolkits Sequential Igor Data Retrieval (SQUIRREL), Peak Integration by Key Analysis (PIKA), and Analytic Procedure for Elemental Separation (APES) (DeCarlo et al., 2006;Aiken et al., 2007).In the analysis, standard relative ionization efficiencies (RIE) were used, corresponding to 1.1 for nitrate, 1.2 for sulfate, 1.4 for organic molecules, 4.0 for ammonium, 1.3 for chloride, and 2.0 for water (Alfarra et al., 2004;Mensah et al., 2011).An AMS collection efficiency (a factor which potentially corrects for undetected particle mass concentration) of 1.0 was used, as supported by the agreement of the AMS-measured mass concentrations with the density-compensated volume concentrations measured by the SMPS (Sect.2.3).Determination of the air correction factors, several updates to the fragmentation table, and the contributions of organic material to CO + and H x O + signal intensities followed the method described in Chen et al. (2011).

Material density
Under an assumption of spherical particles, volume-diameter distributions measured by the SMPS can be used in conjunction with mass-diameter distributions measured by the AMS to estimate the particle effective density ρ eff (kg m −3 ).More specifically, ρ eff was calculated by dividing the mass-mode particle diameter measured by the AMS (i.e., vacuum aerodynamic diameter) by the volume-mode diameter measured by the SMPS (i.e., mobility diameter) (DeCarlo et al., 2004;Bahreini et al., 2005;Katrib et al., 2005).Figure S1 shows two examples of the mass-diameter distribution derived from the SMPS measurements compared to that measured by the AMS for particles in the CMFR outflow.The SMPS massdiameter distributions were derived by multiplying the SMPS volume-diameter distributions by the ρ eff value that was obtained using the mode diameters.The good agreement between the SMPS and AMS distributions suggests that within detection limits ρ eff does not vary with diameter.The integrated area under the distributions represents the total particle mass concentration.Figure S2 shows the scatter plot of the total mass concentrations obtained from the AMS measurements against the SMPS-derived concentrations.The slope of 0.99 ± 0.02 indicates that an AMS collection efficiency of unity can be assumed given the measurement uncertainties of the AMS and the SMPS (Matthew et al., 2008;Shilling et al., 2008).
For nonporous spherical particles, effective density is identical to material density (DeCarlo et al., 2004).Past work indicates the applicability of a nonporous spherical morphology for SOM-coated sulfate particles (King et al., 2007).Based on the ρ eff value obtained for each experiment, the material density ρ org of the SOM was calculated by the rule of volume additivity.As explained by Bahreini et al. (2005), in context of AMS data sets this rule states that ρ org = M org M particle ρ eff − M AS ρ AS , where ρ AS is the material density of ammonium sulfate (1770 kg m −3 ), M particle is the total particle mass concentration (µg m −3 ), and M AS is the ammonium sulfate mass concentration (µg m −3 ).The presented equation is valid provided that the chemical components either do not mix or alternatively have a numerically small excess volume of mixing.The ρ org values obtained by this analysis and their uncertainties are listed in Table 2.The uncertainties of 5 to 10 % were based on a Monte Carlo analysis.The parameters used in the analysis included the uncertainty in the SMPS and AMS mode diameters (i.e., as needed for calculating ρ eff ) as well as the standard deviations of the temporal variation of M particle , M org , and M AS at steady state.
For experiment #27 (Table 2), independent measurements of ρ eff and hence independent calculations of ρ org were also made by the DMA-APM methodology (Kuwata et al., 2012).In this case, ρ org was calculated from data of quasimonodisperse particles rather than for the entire particle population.The uncertainty of ρ org for the DMA-APM method was estimated as 2 % on the basis of calibrations using polystyrene latex particles.
Figure S3 shows that the ρ org values determined by the independent AMS-SMPS and DMA-APM methods were in agreement.The measured value of ρ org of 1320 ± 20 kg m −3 at high mass concentrations (M org > 50 µg m −3 ) was also consistent with the value reported previously by Bahreini et al. (2005) for β-caryophyllene ozonolysis for experiments at M org = 300 µg m −3 .Figure S3 also shows that there was a dependence of ρ org on organic particle mass concentration.A similar result was reported previously for α-pinene ozonolysis (Shilling et al., 2009).Particle mass yield Y is defined as Y ≡ M org / VOC, with the sign construed as positive.The term VOC is the difference in the β-caryophyllene concentration between the outflow and the inflow of the CMFR, meaning that negligible loss to any apparatus surface is assumed for unreacted β-caryophyllene at steady state.The accuracy of this assumption was supported by the agreement between the βcaryophyllene concentrations in the CMFR outflow prior to initiation of reaction (as measured by the PTR-MS) and the concentrations calculated using the liquid injection rate of the syringe pump.
For M org , a simplification is that the inflow concentration is zero but a complication is that significant mass is lost to the walls of the bag and therefore does not contribute to concentration measured in the outflow.Therefore, M org = (M org ) outflow + (M org ) wallloss -(M org ) inflow .We measure (M org ) outflow using the AMS, we know (M org ) inflow = 0, and we relate (M org ) wallloss = f ((M org ) outflow ) by calibration of wall-loss rates.More specifically, (M org ) wallloss = β τ (M org ) outflow for an assumed firstorder diameter-independent particle wall-loss coefficient β (s −1 ) and a mean residence time τ (s) in the CMFR (Seinfeld et al., 2003;Pierce et al., 2008).Wall-loss mechanisms include Brownian diffusion, electrostatic forces, and gravitational sedimentation (McMurry and Grosjean, 1985).The final form of the equation for particle mass yield for the conducted experiments is as follows: For further use, we also introduce here the definition M org,corr ≡ (1 + β τ )(M org ) outflow as the wall-loss corrected organic particle mass concentration.Derivation of this equation is provided in the Sect.A of the Supplement.The particle wall-loss coefficient β can be determined experimentally by measuring the change in number concentration of surrogate particles in the chamber (McMurry and Grosjean, 1985).For this purpose, experiments were conducted with populations of ammonium sulfate particles in the absence of organic material.On the basis of the particle number balance in the CMFR (Seinfeld et al., 2003), β = (2.9 ± 0.3) × 10 −4 s −1 was obtained, and no significant dependence of β on diameter was observed for the studied populations (Fig. 1).For numberdiameter distributions of SOM-coated sulfate particles of the actual experiments (Fig. S4), independent estimates of β = (2.8 ± 0.4) × 10 −4 s −1 were obtained by comparing the total particle number concentrations measured in the CMFR outflow to the total concentrations expected based on the seed particle concentrations in the CMFR inflow.
The measured wall-loss coefficients of the present study were greater than the measured values for the earlier experiments in the Harvard Environmental Chamber.showing the loss of β-caryophyllene, the appearance of particlephase secondary organic material following the introduction of ozone, and the increase of particle-phase organic material following the elevation of ozone concentration.In support of this explanation, the experiments described herein from April to June 2009 correspond to the reported β values, and no systematic temporal trend of wall losses was observed during this time period, suggesting that the effects of bag aging during the two months of experiments were minimal.In September 2010, with the by-then aged bag, β values in agreement with Shilling et al. (2008) were obtained, as represented by experiment #27 of Table 2.The additional aging of the bag surface significantly decreased the particle wall losses, perhaps because of surface alterations that influenced electrostatic charging (McMurry and Rader, 1985).In support of this explanation, re-neutralization of the seed particles prior to the injection decreased particle wall losses, as represented by experiment #10 of Table 2.
Experiments carried out at repeated conditions show differences no more than 30 %, supporting the reproducibility of the results (Table 2).The particle-phase organic mass concentrations were corrected by 68 to 370 % for the different β values of the experiments (Eq.1).The corrected values lay on a self-consistent trend line within instrument uncertainty, supporting the accuracy of the applied corrections (Fig. S5).

Results and discussion
Figure 2 shows a typical time series for the transient period prior to steady-state in the CMFR.The time traces show the loss of β-caryophyllene after the introduction of ozone and the appearance of particle-phase secondary organic material.Seed particles and β-caryophyllene were introduced into the CMFR and reached steady-state concentrations prior to the injection of ozone.Once ozone was introduced, the concentration of β-caryophyllene decreased rapidly.After a transient period, feedback control between the measured ozone concentration and the quantity of ozone injected into the bag maintained the ozone concentration at 50 ppbv.The steadystate concentrations of particle-phase SOM were reached after 8 h and were maintained until completion of the experiment after 56 h.
During the transient period of Fig. 2, the initial consumption of β-caryophyllene was primarily by the reaction of ozone with the endo-cyclic double bond.The ozonolysis rate constant of the endo-cyclic double bond of β-caryophyllene (k endo = 1.16 × 10 −14 molecule −1 cm 3 s −1 ) is 100× greater than that of the exo-cyclic double bond of the first-generation products (k exo = 1.1 × 10 −16 molecule −1 cm 3 s −1 ) (Shu and Atkinson, 1995;Winterhalter et al., 2009).The time decay of the β-caryophyllene concentration agreed well with that predicted for the CMFR (i.e., the modeled line in Fig. 2) using the rate constant k endo and the measured ozone concentrations.
Subsequent to the first transient period, Fig. 2 shows that a further increase in the ozone concentration to 200 ppbv caused an increase in the mass concentration of particlephase SOM.The remaining exo-cyclic double bond of the first-generation products reacted at the higher ozone concentration, leading to additional low-volatility products that partitioned to the particle phase (Kanawati et al., 2008;Li et al., 2011).The data of Fig. 3 demonstrate the generalization of this result for a single injected β-caryophyllene concentration to all experiments, showing an increase in yield for 200 compared to 50 ppbv ozone for all injected β-caryophyllene concentrations.
The representation in Fig. 2 (Li et al., 2011).(b) The percent change in mass concentration at 50 compared to 200 ppbv O 3 for the fifteen identified products.Experiments in this figure are for 13 ppbv of reacted β-caryophyllene.Error bars represent the one-sigma standard deviation of three replicate experiments.Labels x, y, and z represent a permutation of 1, 2, and 3. P254-1 is hashed because the data of the present study suggest that this product might have been assigned incorrectly in previous work as a first-generation product (Li et al., 2011); it can instead be assigned as a second-generation product (see main text).
of the remaining exo-cyclic double bond.This distribution in reactivity is assumed to be represented by a mean value τ exo and an associated variance σ 2 for the population of products.The calculations τ CMFR /τ exo (50) = 1.8 and τ CMFR /τ exo (200) = 6.6 suggest that for many products i the relationships 0.1 < τ CMFR /τ exo,i (50) < 10 and 0.1 < τ CMFR /τ exo,i (200) < 10 hold, implying that there are significant differences among the various first-generation products in the timescales for conversion from first-to second-generation products.These differences are sensitive to the ozone concentration in the CMFR.For instance, a kinetics calculation (cf.Sect.B of the Supplement) suggests a mean conversion of 63 % from first-to second-generation products for 50 ppbv ozone compared to a mean conversion of 87 % for 200 ppbv ozone.The actual overall conversion to second-generation products depends upon the unknown probability density function PDF(τ exo ,σ ) of reactant concentrations and reactivity.The experimental observations of Fig. 2 show that there is a 60 % increase in (M org ) outflow for 200 compared to 50 ppbv ozone.This increase is caused by the additional conversion of first-to second-generation products at 200 ppbv ozone.
Differences in particle-phase molecular products between 50 and 200 ppbv ozone exposure, as highlighted in Fig. 4, are consistent with an enhancement of second-generation products relative to first-generation ones.Figure 4a shows the ozone-dependent mass fractions of first-compared to second-generation products, as determined by UPLC-ESI-ToF-MS for 15 compounds (Li et al., 2011).As the ozone concentration increases, the second-generation products are increasingly dominant.
Figure 4b shows the percent change in mass concentration for exposure to 50 compared to 200 ppbv ozone for each of the 15 products.The structures of the products labeled in Fig. 4 are provided in Li et al. (2011) and are reproduced for convenience in Fig. S6. Figure 4b shows that the mass concentrations of the first-generation products decrease and the concentrations of the second-generation products increase at higher ozone concentration, as is consistent with the conversion of the former into the latter.The concentration of P270-1 represents the largest relative change, possibly suggesting that its first-generation precursor compound has a relatively low ozonolysis rate constant.There are, however, a few exceptions apparent to this trend.Unlike other second-generation products, the mass concentrations of the products P252-4 and P302 do not change, suggesting that at 50 ppbv ozone they are already produced to completion from their first-generation precursors.Another exception is P238, showing a negative relative change in mass concentration.This species is produced from the first-generation product P236 by a stabilized Criegee intermediate channel that competes with an RO 2 -assisted isomerization channel (Fig. S6).At higher ozone concentrations, the importance of the isomerization channel possibly increases because of higher RO 2 concentrations, providing one possible explanation for the observed decrease in P238 concentration.
Figure 5 presents the yield data for increasing organic particle mass concentration.Yield data are customarily parameterized with the objective of further upscale modeling (e.g., air quality models).In common usage, parameterizations are based on either a two-product mode (Odum et al., 1996) or a basis-set approach (Presto and Donahue, 2006).Our appli-cation employs the latter.The products formed by the oxidation of β-caryophyllene are binned into product classes of mass yields α i , and the volatilities of the product classes are prescribed in decadal units of 10 −i , where 10 −i is denoted as C * i .Particle mass yield is then written (Seinfeld and Pankow, 2003;Presto and Donahue, 2006): The yield data sets in panels a1-a3, which represent the three different ozone concentrations of this study, were each fit using Eq. ( 1) for i i = 0 and i f = 2.The fit treated C * i (µg m −3 ) as fixed quantities, Y and M org,corr as data, and α i as the quantities for optimization.The overall bar height in panels a1-a3 represents the cumulative potential yield (i.e., i f i=i i α i ).The coloring inside each bar represents the partitioning of the organic molecules between the gas (gray) and particle (green) phases (Donahue et al., 2006;Presto and Donahue, 2006).The different shades of green represent the volatility associated with a component of the particle phase.The optimized values for α i (Table S1) are plotted in the panels b1-b3 for each ozone concentration.

Q. Chen et al.: Particle mass yield from β-caryophyllene ozonolysis
An approximate upper limit of 150 % for the maximum potential particle mass yield from β-caryophyllene ozonolysis can be established by assuming that two ozone molecules add to the original molecule and that all product molecules partition to the particle phase.The representation in panel a1 of Fig. 5 shows that the cumulative yield approaches 60 % for 50 ppbv ozone.By comparison, panel a3 for 200 ppbv ozone shows that the cumulative yield approaches 130 %, suggesting that the overall reaction is nearly complete.The implication is that the particle-phase mass yield represented by each green bar in panel a3 can be taken as the approximate representation of the ultimate yield.
From 50 to 200 ppbv ozone, the optimized mass yields α i have similar values for products of low volatility (C * i = 1, 10 µg m −3 ) but a trend of increasing values for products of relatively high volatility (C * i = 100 µg m −3 ) (panels b1-b3 of Fig. 5).The similar values of mass yields suggest that the low-volatility products are mainly second-generation products that are formed nearly to completion from their firstgeneration precursors even at 50 ppbv, indicating fast formation pathways.In this case, the mass yields are not sensitive to the ozone concentration.Product P302 is the lowest volatility product among the 15 identified products and can therefore be supposed as the dominant species to condense to the particle phase at low M org (Li et al., 2011).This conclusion that P302 is fast-forming low-volatility product is supported by the molecular data; there is a negligible change in the mass concentration for an increase in ozone concentration (Fig. 4b).By comparison, the trend of increasing α i values with increasing ozone concentration for C * i = 100 µg m −3 suggests that the relatively high-volatility products are mainly second-generation products that are formed by relatively slow pathways.The increase of ozone concentration leads to greater conversion of these products from their first-generation precursors.Panels a1-a3 show the consequence of these processes: the increase in particle mass yield from 50 to 200 ppbv ozone is dominantly driven by products of relatively high volatility (cf.large light-green bar in panel a3 for C * i = 100 µg m −3 ).The findings for α i represented in panels b1-b3 provide a formal framework for the observation that differences in particle mass yield for 200 ppbv compared to 50 ppbv ozone are small for M org,corr < 10 µg m −3 but greater for higher M org,corr (panels a1-a3).For low M org,corr , the particlephase SOM is dominantly composed of products having low volatility.Therefore, the significantly increased production at 200 ppbv ozone of the high-volatility products of C * i = 100 µg m −3 causes little additional partitioning to the particle phase.By comparison, for 10 ≤ M org,corr ≤ 100 µg m −3 products of C * i = 100 µg m −3 increasingly partition to the particle phase, meaning that 10 % of these products partition to the particle phase for M org,corr = 10 µg m −3 and 50 % of them to the particle phase for M org,corr = 100 µg m −3 (Donahue et al., 2006).
There is a dependency of the O:C elemental ratio, as measured by the AMS, on M org,corr , and this dependency may also be explained by the volatility distribution of the products.The O:C ratios ranges from 0.5 to 0.3, corresponding to an average addition of 5 to 8 oxygen atoms to the C 15 structure of β-caryophyllene.This average is consistent with a dominant presence of second-generation products in the particle phase because the first-generation products typically have 1 to 4 oxygen atoms (Winterhalter et al., 2009;Li et al., 2011).The experimental results for M org,corr < 10 µg m −3 show that neither an increase in VOC (Fig. 6a) nor an increase in the steady-state O 3 concentration (Fig. 6b) changes the O:C ratio, at least within measurement uncertainty (O:C ≈0.5 and OS c ≈ 0.5; Table 2).This behavior is consistent with the dominant contribution of low-volatility, fastforming second-generation products to the particle phase.By comparison, for M org,corr > 10 µg m −3 the O:C ratio steadily decreases for increasing VOC and ozone concentrations.This correspondence between decreased O:C ratio and increased contribution of high-volatility products (i.e., C * i = 100 µg m −3 ) conforms to the expected structurefunction relationship between a molecule's oxygen content and its vapor pressure, meaning greater oxygen content of a molecule and an associated decreased vapor pressure (Pankow and Asher, 2008).
The yield data of this study are summarized as a function of M org,corr in Fig. 7, including comparisons to data sets of previous studies.Panel a shows the yield data for M org,corr < 10 µg m −3 , corresponding to atmospheric concentrations.Panel b shows the yield data for a greater range of M org,corr .For M org,corr <10 µg m −3 , prior to the present study no data are known to us for the particle mass yield of β-caryophyllene ozonolysis.For M org,corr > 10 µg m −3 , two studies have previously been carried out for conditions of excess ozone (Jaoui et al., 2003;Lee et al., 2006a) (Table 1).Jaoui et al. (2003) reported a yield of 62 % but did not report M org , preventing a specific comparison to our data set.Lee et al. (2006a) reported a yield of 45 % for M org = 336 µg m −3 .This yield is lower than the yield of 55-100 % reported in the present study for similar M org .The differences between the two studies might be related to the employed relative humidity, which was 40 % in our study and 6 % in Lee et al. (2006a).Winterhalter et al. (2009) showed that increased particle mass yields can be expected for higher relative humidity.Figure 7b also shows yields reported for βcaryophyllene photooxidation.Differences in the chemical mechanisms of ozonolysis and photooxidation notwithstanding, Lee et al. (2006b) reported yields approximately similar to those of the present study.The photooxidation yields reported by Griffin et al. (1999), however, are greater than those of the present study.In the case of early stage photooxidation compared to ozonolysis, a possible shift to lower volatility might occur in the product population because of additional oxidation by OH, though the extent of such shift can be expected to depend strongly on reaction conditions.Fig. 6.Oxygen-to-carbon elemental ratios of particle-phase SOM produced by β-caryophyllene ozonolysis for (a) increasing particlephase organic mass concentration and (b) increasing ozone concentration.The elemental ratios were determined from the particlephase high-resolution mass spectra (Aiken et al., 2007(Aiken et al., , 2008;;Chen et al., 2011).Error bars represent the one-sigma measurement precision (Chen et al., 2011).Data are colored by the reacted concentration of β-caryophyllene.Hoffmann et al. (1997) because the M org exceeds the abscissa scale of the figure.For atmospheric conditions, ozonolysis rather than photooxidation is the dominant degradation pathway of β-caryophyllene (Atkinson and Arey, 2003).Panel (b) shows the data over an extended range of organic particle mass concentration (10-400 µg m −3 ).A material density ρ org of 1300 kg m −3 was assumed to convert the volume-based data sets of Lee et al. (2006a) and Griffin et al. (1999) to mass-based data sets.This density corresponded to M org,corr >10 µg m −3 of the present study (Fig. S3).A vaporization enthalpy of 40 kJ mol −1 was used for temperature compensation from 308 K to 298 K for the data reported by Griffin et al. (1999).The dashed lines show four different parameterizations that have been used in air quality and climate models.

Figure 7b omits a data set of
This study investigated the role of second-generation products in the particle mass yield of β-caryophyllene ozonolysis.
For concentrations that overlapped with those of the atmosphere (i.e., 0.1 to 10 µg m −3 , Chen et al., 2009;Slowik et al., 2010), the particle mass yield increases from 2 to 30 %.The yield is not sensitive to the ozone concentration.The explanation is that for this range of mass concentration there is a dominant contribution of low-volatility, fast-forming secondgeneration products to the particle phase.The O:C elemental ratio of 0.5 indicates that these low-volatility products have 7 to 8 oxygen atoms with estimated vapor pressures of 10 −12 to 10 −14 Pa (Li et al., 2011).The data further show that organic particle mass concentration inversely correlates with the oxidation state and the material density of the particlephase organic material.Across the studied range of organic particle mass concentration (0.5-230 µg m −3 ), the O:C ratio drops from 0.5 to 0.3, corresponding to a decrease in OS c from −0.5 to −0.8.Material density likewise decreases from 1600 to 1300 kg m −3 , corresponding to the decreases in relative oxygen content.
Prior to the present study, no data points for the particle mass yield of β-caryophyllene ozonolysis were available in the atmospheric limit of M org < 10 µg m −3 .In the absence of data, regional and global chemical transport models have instead widely employed extrapolations of the data reported by Griffin et al. (1999) to estimate the contribution of sesquiterpene β-caryophyllene oxidation to atmospheric particle mass concentrations (Chung and Seinfeld, 2002;Sakulyanontvittaya et al., 2008b;Zhang and Ying, 2011) (details are provided in the Sect.C and Table S2 of the Supplement).Because the data reported by Griffin et al. (1999) were collected at 308 K and were volume-based, corrections were needed for model applications typically developed for 298 K.The corrections required estimates of the enthalpy of vaporization H vap and of material density.Different air-quality applications have used different estimates of these quantities, and as a result the parameterizations of mass yield have been different in various models, as represented by the dashed lines in Fig. 7. Using an updated H vap of 40 kJ mol −1 as well as a material density ρ org of 1300 kg m −3 (Bahreini et al., 2005;Offenberg et al., 2006), Carlton et al. (2010) present the state-of-the-art parameterization.The comparison between our data set and the parameterization of Carlton et al. (2010) suggests a possible underestimate by that parameterization of 100 % to 300 % for organic particle mass concentrations less than 3 µg m −3 given that ozonolysis rather than photooxidation is the dominant degradation pathway of β-caryophyllene.The difference could be in the underlying chemistry, keeping in mind that the parameterization is based on the photooxidation data set of Griffin et al. (1999) whereas the data of this study correspond to ozonolysis experiments.
The ultimate particle mass yield of β-caryophyllene ozonolysis was parameterized for the present study by mass-based stoichiometric yields α 0 = 0.17 ± 0.05, α 1 = 0.11 ± 0.17, and α 2 = 1.03 ± 0.30 for corresponding saturation concentrations of 1, 10, and 100 µg m −3 .Terms α 0 and α 1 had low sensitivity to ozone exposure for the investigated range of conditions whereas term α 2 increased from 0.32 ± 0.13 to 1.03 ± 0.30 as ozone exposure was increased.These findings potentially allow for simplified yet accurate parameterizations in air quality and climate models that seek to represent the ozonolysis particle mass yield of certain classes of biogenic compounds.The influence of additional important reaction conditions, such as NO x concentrations, photolysis pathways, and particle aging, needs to be investigated in the future.

Fig. 3 .
Figure3 Figure2shows a typical time series for the transient period prior to steady-state in the CMFR.The time traces show the loss of β-caryophyllene after the introduction of ozone and the appearance of particle-phase secondary organic material.Seed particles and β-caryophyllene were introduced into the CMFR and reached steady-state concentrations prior to the injection of ozone.Once ozone was introduced, the concentration of β-caryophyllene decreased rapidly.After a transient period, feedback control between the measured ozone concentration and the quantity of ozone injected into the bag maintained the ozone concentration at 50 ppbv.The steadystate concentrations of particle-phase SOM were reached after 8 h and were maintained until completion of the experiment after 56 h.During the transient period of Fig.2, the initial consumption of β-caryophyllene was primarily by the reaction of ozone with the endo-cyclic double bond.The ozonolysis rate constant of the endo-cyclic double bond of β-caryophyllene (k endo = 1.16 × 10 −14 molecule −1 cm 3 s −1 ) is 100× greater than that of the exo-cyclic double bond of the first-generation products (k exo = 1.1 × 10 −16 molecule −1 cm 3 s −1 )(Shu and Atkinson, 1995;Winterhalter et al., 2009).The time decay of the β-caryophyllene concentration agreed well with that predicted for the CMFR (i.e., the modeled line in Fig.2) using the rate constant k endo and the measured ozone concentrations.Subsequent to the first transient period, Fig.2shows that a further increase in the ozone concentration to 200 ppbv caused an increase in the mass concentration of particlephase SOM.The remaining exo-cyclic double bond of the first-generation products reacted at the higher ozone concentration, leading to additional low-volatility products that partitioned to the particle phase(Kanawati et al., 2008;Li et al., 2011).The data of Fig.3demonstrate the generalization of this result for a single injected β-caryophyllene concentration to all experiments, showing an increase in yield for 200 compared to 50 ppbv ozone for all injected β-caryophyllene concentrations.The representation in Fig.2can be put into a quantitative context by comparing the ozone-dependent e-folding lifetimes τ endo (O 3 ) and τ exo (O 3 ) of the two double bonds at 50 and 200 ppbv (i.e., τ (O 3 ) = 1/k[O 3 ]) to the mean CMFR residence time τ CMFR .The values are as follows: τ endo (50) = 70 s, τ endo (200) = 20 s, τ exo (50) = 7500 s, τ exo (200) = 2000 s, and τ CMFR = 13 100 s.The relationships τ CMFR /τ endo (50) 10 and τ CMFR /τ endo (200) 10 imply that β-caryophyllene is consumed to nearly 100 % for both ozone concentrations.This result is supported by the PTR-MS measurements that show a negligible residual βcaryophyllene concentration in the CMFR outflow for 50 and 200 ppbv ozone (Fig.2).The population of first-generation products i has a distribution of reactivity τ exo,i with respect to ozonolysis Figure 5
a The reacted VOC concentration depends on ozone or OH concentration.bCriegeeintermediate.cTheSOM densities are measured values in this study and assumed values in other studies.d Minimum detection limit of the NO x analyzer (Thermo 8440E).e Calculated from the reported carbon yield of 39 % by applying an OM : OC ratio of 1.6.f Dry ammonium sulfate particles.g

Table 2 .
Experimental conditions and results for the dark ozonolysis of β-caryophyllene.The terms VOC, (M org