Formation and evolution of molecular products inα-pinene secondary organic aerosolXuan Zhanga, Renee C. McVayb, Dan D. Huangc, Nathan F. Dalleskaa, Bernard Aumontd, Richard C. Flagana,b,and John H. Seinfelda,b,1

aDivision of Engineering and Applied Science, California Institute of Technology, Pasadena, CA 91125; bDivision of Chemistry and Chemical Engineering,California Institute of Technology, Pasadena, CA 91125; cDepartment of Chemical and Biomolecular Engineering, Hong Kong University of Science andTechnology, Hong Kong, China; and dLaboratoire Interuniversitaire des Systèmes Atmosphériques, UMR CNRS 7583, Université Paris Est Créteil andUniversité Paris Diderot, 94010 Créteil, France

Contributed by John H. Seinfeld, September 30, 2015 (sent for review September 4, 2015; reviewed by Thorsten Hoffmann and Murray V. Johnston)

Much of our understanding of atmospheric secondary organicaerosol (SOA) formation from volatile organic compounds derivesfrom laboratory chamber measurements, including mass yield andelemental composition. These measurements alone are insuffi-cient to identify the chemical mechanisms of SOA production. Wepresent here a comprehensive dataset on the molecular identity,abundance, and kinetics of α-pinene SOA, a canonical system thathas received much attention owing to its importance as an organicaerosol source in the pristine atmosphere. Identified organic spe-cies account for ∼58–72% of the α-pinene SOA mass, and are char-acterized as semivolatile/low-volatility monomers and extremelylow volatility dimers, which exhibit comparable oxidation statesyet different functionalities. Features of the α-pinene SOA forma-tion process are revealed for the first time, to our knowledge, fromthe dynamics of individual particle-phase components. Althoughmonomeric products dominate the overall aerosol mass, rapid pro-duction of dimers plays a key role in initiating particle growth. Con-tinuous production of monomers is observed after the parentα-pinene is consumed, which cannot be explained solely by gas-phase photochemical production. Additionally, distinct responsesof monomers and dimers to α-pinene oxidation by ozone vs. hy-droxyl radicals, temperature, and relative humidity are observed.Gas-phase radical combination reactions together with condensedphase rearrangement of labile molecules potentially explain thenewly characterized SOA features, thereby opening up further av-enues for understanding formation and evolution mechanisms ofα-pinene SOA.

secondary organic aerosol | particulate matter | air quality | climate

Secondary organic aerosol (SOA), comprising a large numberof structurally different organic oxygenates, is a dominant

constituent of submicrometer atmospheric particulate matter(1). Molecular characterization of SOA has been a major re-search goal in atmospheric chemistry for several decades (2),owing to the importance of organic aerosol in air quality andEarth’s energy budget. Both biogenic (e.g., isoprene, monoter-penes) and anthropogenic (e.g., aromatics, large alkanes)organic compounds are well-established precursors to SOA.Knowledge of the SOA molecular composition is crucial forelucidation of its underlying formation mechanisms.The most abundant monoterpene in the troposphere is α-pinene

(3). The oxidation of α-pinene by ozone has become a canonicalSOA system (4–12). Identification of multifunctional particle-phase products has been reported, including monomers with car-boxylic acid moieties (4, 6) and high-molecular-weight compounds(7, 8, 12), although molecular structures and formation pathwaysof oligomers remain uncertain (5). Recently, a class of extremelylow-volatility gas-phase organic compounds (ELVOCs) has beenidentified as an important component in the α-pinene ozonolysischemistry (13). Identification of the ELVOCs in the particle phaseand elucidation of the mechanism of their formation remain keymissing pieces in closing the α-pinene SOA system (14).

We report here, for the first time, to our knowledge, time-resolved molecular characterization of the abundance, forma-tion, and evolution of organic species in α-pinene-derived SOA.Identified classes of species account for (∼58–72) ± (∼34–39)%of the overall α-pinene SOA mass, with volatilities spanning fromthe semivolatile to extremely low-volatility range and molecularstructures characterized as multifunctionalized monomers anddimers. These organic species exhibit distinct characteristics interms of oxidation states, chemical structures, initial growth rates,evolution patterns, and responses to variations in temperature (T),relative humidity (RH), and oxidant type.

Results and DiscussionDistribution of α-pinene SOA Constituents. The α-pinene-derivedSOA was generated in the Caltech Environmental Chamber; seeMaterials and Methods. The Particle-into-Liquid Sampler (PILS)integrated with Ultra Performance Liquid Chromatography/Electrospray Ionization Quadrupole Time-of-Flight Mass Spec-trometry (UPLC/ESI-Q-ToFMS) is used to characterize tem-poral profiles of particulate molecular constituents (15). Aspectrum of monomers and dimers, with molecular formulasC8–10H12–16O3–6 and C14–19H24–28O5–9, respectively, is observedin α-pinene SOA. Chemical structure elucidation of these or-ganic molecules is based on the interpretation of chromato-graphic and spectrometric behaviors of the corresponding ionsupon electrospray ionization (SI Appendix, Molecular StructureElucidation). Based on the fragmentation pattern of the relevantparent ions upon collision-induced dissociation in MS/MS spectra,monomers are generally (di)-carboxylic acids. Ester-containing

Significance

Secondary organic aerosol (SOA) plays a pivotal role in climateand air quality. Characterizing the molecular makeup of SOAhas been a major research goal for several decades, yet thechemical dynamics of most anthropogenic and biogenic SOAsystems remain poorly resolved. We report here the time-resolved molecular characterization of SOA derived from thecanonical α-pinene system, one of the most abundant bio-genic emissions in the troposphere. We reveal distinct fea-tures of SOA components in terms of molecular structure,abundance, growth rates, evolution patterns, and responsesto variations in temperature, relative humidity, and oxidanttype. Our findings provide a comprehensive analysis of pro-cesses governing α-pinene SOA formation and aging.

Author contributions: X.Z., R.C.F., and J.H.S. designed research; X.Z., D.D.H., and N.F.D.performed research; X.Z., R.C.M., and B.A. analyzed data; and X.Z. and J.H.S. wrote thepaper.

Reviewers: T.H., Johannes Gutenberg University; and M.V.J., University of Delaware.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: seinfeld@caltech.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1517742112/-/DCSupplemental.

14168–14173 | PNAS | November 17, 2015 | vol. 112 | no. 46 www.pnas.org/cgi/doi/10.1073/pnas.1517742112

Dow

nloa

ded

by g

uest

on

Oct

ober

5, 2

020

structures are prevalent in dimers, with C8H11O4 (m/z 171) andC9H13O4 (m/z 185) as primary building blocks.Fig. 1 summarizes identified gas- and particle-phase products

mapped onto the carbon number versus oxidation state (nC −OSC)space at the point when >99% α-pinene is consumed during ozo-nolysis in the absence of an OH scavenger at 298 K and <5%relative humidity. Products with mass saturation concentration(C*) in the range of <3 × 10−4, 3 × 10−4 to 0.3, 0.3–300, and 300 to3× 106 μg m−3 are designated as extremely low-volatility, low-volatility,semivolatile, and intermediate-volatility organic compounds(ELVOCs, LVOCs, SVOCs, and IVOCs), respectively. A numberof monomers in the SVOC category (C8H12O4, C8H14O5, C9H14O3,C9H14O4, C10H16O3, and C10H16O4) are detected both in gas andparticle phases. Two monomers (C10H14O5 and C10H16O6) fallin the LVOC range, whereas all of the dimers are categorized asELVOCs. At the completion of the oxidation of α-pinene, massconcentrations of individual oxidation products, represented bythe size of the circular symbols in Fig. 1, range from ∼0.1 μg m−3

to ∼26 μg m−3 and account for 58 ± 34% of the total organicparticulate mass. The mean carbon oxidation state of the iden-tified SOA molecular constituents (−0.68 ± 0.27) agrees essen-tially identically with the average level derived from the AerosolMass Spectrometer (AMS) measurement (−0.72 ± 0.43). Note thatthe monomer and dimer units have comparable O:C ratios.Fig. 2 shows the temporal profiles of the three volatility cat-

egories, characterized as the mass loading and fraction (Fig. 2 Aand C), as well as the growth dynamics of individual species(Fig. 2 B and D). At first glance, SVOC products dominate theα-pinene SOA, whereas the LVOC and ELVOC products in total

account for less than 45% of the overall organic mass. The growthrates of ELVOCs are comparable to or even exceed those forSVOCs and LVOCs. As α-pinene oxidation proceeds, ELVOCproduction is inhibited, whereas SVOC and LVOC accumulation isfavored. By examining the growth dynamics of individual oxidationproducts, key processes that significantly contribute to α-pineneSOA formation and evolution can be deduced. Prompt formationof ELVOCs at the beginning of the ozonolysis of α-pinene iscrucial to initiate organic particle growth. The observation thatmost ELVOC dimers grow faster than SVOCs and LVOCs indi-cates that there is no intrinsic kinetic barrier to the ELVOC dimerformation (16). The role of aging in the α-pinene SOA evolution isreflected, to a certain degree, by the slow but continuous growth ofLVOCs over the 4-h course of an experiment.

Effect of Temperature, Relative Humidity, and Oxidant. Currentunderstanding regarding the effect of temperature on SOA yieldis that lower temperature favors partitioning into the particlephase and hence enhances organic aerosol production owing todecreasing compound vapor pressure. For instance, mass fractionsof terpenylic acid (C8H12O4), pinalic acid (C9H14O3), pinic acid(C9H14O4), oxopinonic acid (C10H14O4), and hydroxy-pinonic acid(C10H16O4), increase by 33%, 119%, 10%, 59%, and 111%, re-spectively, after 3 h of reaction (>99% of α-pinene is consumed) whenthe chamber temperature is decreased from 25 °C to 12 °C; see Fig. 3and SI Appendix, Fig. S6. The particle-phase mass fractions of mostdimers, on the other hand, decrease at lower temperature, indi-cating that additional pathways that are temperature-sensitive

Fig. 1. Distribution of gas/particle-phase products from the ozonolysis of α-pinene in the absence of an OH scavenger in the oxidation state versus carbonnumber (OSC −nC) space. The gray circles represent all possible combinations of OSC and nC for stable organic molecules. Experimental details are given in SIAppendix, Table S1. Chemical structures of particle-phase species, as denoted by filled circles, are given in SI Appendix, Table S2. Here, structures are shownonly for compounds that have been reported in the literature. Particle-phase mass concentration of each compound (micrograms per cubic meter), as denotedby the marker size, is a 5-min average quantification at the point when >99% of α-pinene is consumed via reaction with O3 and, to a lesser extent, OH radicals.Molecular formulas and potential structures of gas-phase species, as denoted by filled triangles, are given in SI Appendix, Table S3. The average carbonoxidation state (OSC) derived from AMS measurement is represented by the horizontal gray line. The saturation mass concentration of each species (log C*) isestimated via an empirical model developed by Donahue et al. (34).

Zhang et al. PNAS | November 17, 2015 | vol. 112 | no. 46 | 14169

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Oct

ober

5, 2

020

govern dimer production other than purely gas–particle partitioning.We discuss this observation in more detail subsequently.The impact of RH on the growth and evolution of individual

oxidation products is shown in Fig. 3 and SI Appendix, Fig. S7. AsRH increases, SOA yield increases slightly, potentially as a resultof decreasing the particle-phase diffusion timescale and themean molecular weight of the organic particulate matter. Promptand significant increase in the mass concentration of certaincompounds in the aerosol is also observed, potentially a result ofenhanced reactive uptake of hydrophilic compounds and/or re-actions involving water. However, the presence of liquid watercan lead to a decrease in the mass yield of certain products. Asshown in SI Appendix, Fig. S7, the continuous dynamics observedfor species C10H16O6 (m/z 231) and C19H29O9 (m/z 399) over thecourse of SOA formation reflects the competition between

production and removal pathways governing the accumulation ofthese two compounds in the particle phase.Significant differences in the product distribution and abun-

dances are observed in the SOA generated from the O3-initiatedvs. OH-initiated oxidation of α-pinene. The identities of car-boxylic acid monomers in both systems are quite similar; see SIAppendix, Fig. S1. Gas-phase products with peroxide func-tionalities are prevalent in the OH system (SI Appendix, TableS3), which is expected because the RO2 + HO2 pathway isdominant under the current experimental conditions. Impor-tantly, none of the covalent ester dimers is observed in the OHsystem, although the major building blocks, i.e., C8H11O4 (m/z 171)and C9H13O4 (m/z 185), are still present. This provides strongevidence that these dimers are not a product of particle-phaseesterification (17).

Fig. 2. Temporal profiles of identified organic molecules lumped as SVOCs (blue hues), LVOCs (green hues), and ELVOCs (red hues) in α-pinene SOA.(A) Particle-phase mass concentration of three volatility categories as a function of reaction time. (B) A 2D pie chart representing the mass fraction of 23molecules in the particle phase after 10 min and 180 min of α-pinene reaction with O3. Note that molecular formulas are shown for compounds with massfractions exceeding 5%. (C) Mass fraction of three volatility categories in the particle phase as a function of reaction time. (D) Growth rates of individualspecies in the particle phase, defined as the ratio of the mass of each species normalized by its highest mass value to the reacted mass of α-pinene normalizedby its highest mass value after ∼83 min of reaction.

Fig. 3. Effect of temperature and relative humidity on the mass fraction of particle-phase components (defined as the ratio of PILS + UPLC/ESI-Q-ToFMScharacterized mass concentration of individual species to the SMPS measured total particulate organic mass) when >99% α-pinene is reacted (predominantlywith O3). Corresponding temporal profiles of these compounds are given in SI Appendix, Figs. S6 and S7.

14170 | www.pnas.org/cgi/doi/10.1073/pnas.1517742112 Zhang et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

5, 2

020

Mechanism of Extremely Low-Volatility Dimer Formation. We pro-pose that a combination of two well-established mechanisms,that is, combination of acetylperoxy radicals yielding diacylperoxides and their subsequent decomposition in the condensedphase after partitioning, is an important pathway for the pro-duction of ELVOC dimers. This mechanism is illustrated for theproduct, C17H26O6 (m/z 325), in Fig. 4A. Addition of ozone tothe double bond of α-pinene produces two carbonyl-substitutedCriegee biradicals, which either are collisionally stabilized orisomerize via 1,4-H-shift, yielding an alkenoxy radical (C10H15O4·).The C10H15O4· radical reacts further with an RO2 radical andundergoes subsequent 1,7-H shift, due to the presence of a labile Hatom at the aldehydic carbon, producing an acetylperoxy radical(C9H13O5·). The homogeneous recombination of two C9H13O5·radicals via elimination of O2 produces a covalently bound diacylperoxide. Note that this reaction pathway was proposed as a radicalchain termination step from oxidation of aldehydes in both gasand liquid phases (18, 19). The covalent dimer homologs havebeen observed in the gas phase from the ozonolysis of α-pineneranging from ∼500 parts per trillion (ppt) to ∼30 parts perbillion (ppb) (13, 20). Alternatively, the C9H13O5· radical reactswith RO2/HO2, yielding pinic acid.The decomposition of diacyl peroxides in the condensed phase

generally proceeds through the formation of ion/radical pairintermediates, which undergo decarboxylation, with an ester as aprimary product, as well as carboxylic acids and alcohols (21–27).The ester yield could be increased markedly by increasing thesolvent polarity and the addition of common ions (25–27). Thee-folding lifetime of diacyl peroxides with respect to the first-orderdecomposition ranges from essentially instantaneous to severalhours at room temperature (23, 24, 27), and the decompositionrate is accelerated with increasing solvent polarity and temperature(22–24, 27). Based on measurement of decomposition rates ofsecondary/tertiary alkyl and phenyl diacyl peroxides in pure organicsolvents (e.g., acetone), we expect that the decomposition of diacylperoxide homologs produced in the α-pinene+O3 system pro-ceeds rapidly (shorter than the SOA formation timescale) in theorganic, water, and ammonium sulfate aerosol mixture.

Processes Altering the α-pinene SOA Nature. The gas-phase pro-duction of diacyl peroxides via acetylperoxy radical self/cross-combination and its subsequent decomposition upon solvationcan rationalize a series of important experimental observations.First, it has been suggested that the highly functionalized andextremely low-volatility vapors (O:C > 0.7) produced via the gas-phase autooxidation of peroxy radicals contribute, on average,>10% of the total α-pinene SOA and dominate the organic massat the onset of the SOA growth (13). However, the AMS mea-sured average O:C ratio ranges from ∼0.45 to ∼0.55 over theentire course of α-pinene SOA formation; see SI Appendix, Fig.S5. This indicates that processes involving the loss of oxygen sig-nificantly alter the oxygenated nature of compounds contributingto α-pinene SOA. Here the RO2 + RO2 → ROOR reaction anddecomposition of diacyl peroxides result in a net loss of four ox-ygen atoms, substantially decreasing the O:C ratio of the ELVOCdimer produced in the particle phase. Second, as shown in Fig. 3and SI Appendix, Figs. S6 and S7, the yield of ester dimers in-creases as the aerosol water content and temperature increase,providing further evidence that the decomposition of diacyl per-oxide is temperature and solvent polarity sensitive, consistent withprevious observations (24–27). Third, none of the ELVOC dimersis observed from the OH-initiated oxidation of α-pinene, althoughthe monomeric building blocks, such as pinic acid, are present inthe OH system. This demonstrates that traditionally cited particle-phase esterification is not important in dimer production. It is alsoconsistent with the general α-pinene degradation chemistry in thegas phase, where the diacyl peroxide precursors, acetylperoxyradicals, are one of the major products from the ozonolysisvinylhydroperoxide channel, but rather limited from the OH oxi-dation pathway (6). Additionally, the mass fraction of the ELVOCdimers decreases significantly as the ozonolysis reaction proceeds(Fig. 2 B and C), because the continuous production of freeradicals enhances their collision probability, thus decreasing thelifetime of free radicals and inhibiting the isomerization channel.Fourth, gas-phase production coupled with gas–particle parti-tioning cannot completely explain the observed behavior ofpinic acid, which is the dominant monomeric product and

Fig. 4. (A) Proposed gas- and particle-phase mechanism for the formation of the dimer C17H26O6 and pinic acid (C9H14O4). Model simulations (solid line) andexperimental observations (plus sign) of the growth dynamics of (B) C9H14O4 and (C) C17H26O6.

Zhang et al. PNAS | November 17, 2015 | vol. 112 | no. 46 | 14171

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Oct

ober

5, 2

020

accounts for up to ∼23% of the overall α-pinene SOA mass. Asshown in Fig. 4B and SI Appendix, Fig. S8, a continuous growthof pinic acid is observed after >99% of α-pinene is consumed,indicating an additional production pathway, which will bediscussed in SOA Formation Kinetics in the a-pinene System.

SOA Formation Kinetics in the α-pinene System. Successive chemicalreactions yielding one covalent ester dimer (C17H26O6), as listedin SI Appendix, Table S4, are incorporated into the Vapor–ParticleDynamics Model, which accounts for a series of comprehensiveprocesses during SOA formation in chamber experiments (SIAppendix, Vapor–Particle Dynamics Model). By adjusting free pa-rameters in the model framework, simulation of the temporalprofile of the C17H26O6 ester dimer agrees well with the experi-mental observation (see Fig. 4C), confirming that the mechanisminvolving diacyl peroxide as an intermediate is a plausible expla-nation for the formation of the identified ELVOC dimers inα-pinene SOA. Optimal molar yield of the C18H26O8 diacyl per-oxide from C9H13O5· radical combination is 2%. Despite a smallyield, the irreversible partitioning of diacyl peroxides onto aerosolsmakes these species an important contributor to α-pinene SOA.The best-fit first-order decomposition rate of the C18H26O8 diacylperoxide in the condense phase is 10−1 s−1. This timescale needs tobe rather short to match the rapid growth of the C17H26O6 esterdimer in the particle phase.Although our model framework captures the observed rapid

growth of the C17H26O6 ester dimer in the particle phase, thepredicted pinic acid concentration is up to ∼50% lower than thatmeasured; see Fig. 4B. This is expected because, in the currentmechanism, only one diacyl peroxide decomposition reaction isconsidered, whereas 14 structurally different ELVOC dimers areidentified. Note that the model captures the initial growth trendof pinic acid, reflecting the role of gas-phase production coupledwith gas–particle partitioning. The later gap between simulationsand observations potentially represents an upper bound of thecontribution of diacyl peroxide decomposition to pinic acid for-mation. Although other potential reaction channels for thecontinuous production of pinic acid may exist, the fact that thepinic acid concentration continues to increase after α-pinene isconsumed strongly suggests an additional particle-phase pro-duction pathway.Optimal fitting of model simulations to the observed temporal

profiles of individual particulate species yields the effective vapor–particle accommodation coefficient (αp,i). Here “effective” indi-cates that αp,i represents gas–particle interfacial mass transfer aswell as particle bulk phase diffusion. The best-fit value of αp,iranges from 0.01 to 0.1, indicative of rather slow diffusion in theparticle phase, which is consistent with recent observationsreporting a semisolid phase state of α-pinene SOA (28). Thisprovides insights into the limiting timescale that ultimatelycontrols the aerosol growth.To assess the effect of the relatively high experimental α-pi-

nene mixing ratio of 150 ppb, we simulated the C17H26O6 esterdimer formation under more atmospherically relevant mixingratios of 10 ppb α-pinene and 20 ppb O3. The predicted dimeryield increases by ∼6 times at ∼30 μg m−3 total SOA massloadings (SI Appendix, Fig. S9). Although the timescale with re-spect to the decomposition of diacyl peroxide depends on ambi-ent conditions (e.g., RH and T), changes in initial reactantconcentrations impact the lifetime of free radicals. At low α-pinenelevels, the longer lifetime of radicals favors the isomerizationpathway and consequently leads to enhanced ELVOC dimerproduction.

Atmospheric ImplicationsThe identified SVOC/LVOC monomers and ELVOC dimers,with carboxylic acids and esters as the prevalent functionalities,are found to account for a significant fraction of total α-pinene

SOA mass (∼58–72%). The remaining carbon mass likelycomprises highly oxidized multifunctional organic compounds(HOMs) that have been successfully characterized in the gasphase, with an O:C ratio of >0.7 and a molar yield of 7% (13).HOMs production is ultimately controlled by two competingprocesses, i.e., RO2 autooxidation via H shift vs. reaction withRO2/HO2. The initial α-pinene mixing ratio in the presentchamber experiment was ∼150 ppb, which favors RO2 radicalcombination rather than self-isomerization, compared with moreatmospherically relevant conditions (a few parts per billion). Thisincrease in α-pinene level is estimated to lead to a negligible tomoderate decrease in HOMs yield (SI Appendix, Fig. S10). ThePILS+UPLC/ESI-Q-ToFMS technique used here is sensitive torelatively polar compounds, whereas the HOMs are mostlycarbonyl-/peroxide-containing compounds. That these HOMsmight undergo rapid chemical transformation after partitioninginto particles is supported by the AMS measured average O:Cratio of α-pinene SOA, ranging from 0.45 to 0.55 over thecourse of 4-h ozonolysis.Prompt and significant production of the covalent ester dimers

accounts for up to 40% of the α-pinene SOA mass at the earlystage of particle growth. We propose these dimers are producedfrom the particle-phase decomposition of diacyl peroxide, whichhas been suggested as an important nucleating agent of aerosolformed from reactions of cyclic alkenes and ozone (29). Althoughthe role of organic peroxides in SOA production in pristine at-mospheres has been recognized, their actual characterization inaerosols is attended by substantial difficulties due to their ther-mally labile nature. We provide here indirect evidence for theabundance of organic peroxides and their rapid transformation tocovalent esters upon solvation in organic aerosol mixtures. Sig-nificant and continuous production of SVOC monomers under-scores the importance of gas-phase photochemistry coupled withgas–particle partitioning, as well as particle-phase reaction path-ways, e.g., diacyl peroxide decomposition. Although the importanceof accretion reactions in oligomer formation has been widelyacknowledged, the role of rapid rearrangement/decompositionof large labile molecules in SOA growth and evolution isestablished here.

Materials and MethodsExperiments were carried out in the 24-m3 Teflon reactor in the CaltechEnvironmental Chamber. The α-pinene (∼150 ppb) was oxidized by O3

(∼200 ppb) or OH radicals (∼2 × 106 molecules cm−3) in the presence ofammonium sulfate seed particles at concentrations of NOx typical ofpristine conditions. The ozonolysis experiments were conducted in theabsence of an OH scavenger, resulting in an initial OH molar yield of 0.74(30). As a consequence, ∼20% of the α-pinene mass is expected to reactwith OH, and first-generation ozonolysis products are subject to oxida-tion by OH as well; see SI Appendix, Chamber Experiments. The gas-phasecomposition of oxidation products was monitored by a Chemical Ioniza-tion Mass Spectrometer. The particle size distribution and number con-centration were characterized using a custom-built Scanning MobilityParticle Sizer (SMPS). The particle-phase elemental composition wasmeasured by an Aerodyne High Resolution Time-of-Flight AMS; see SIAppendix, Instrument Operation and Data Analysis Protocols.

Temporal profiles of α-pinene SOA components were characterized byusing the PILS+UPLC/ESI-Q-ToFMS technique (15). The overall PILS collectionefficiency toward α-pinene-derived SOA is estimated to be >85%, based onan empirical correlation of water solubility and average O:C ratio of theaerosol ensemble (15). The time resolution of the PILS+UPLC/ESI-Q-ToFMStechnique is 5 min, thus providing information on the particle-phase dy-namics during ∼4 h of SOA formation experiments. Mass concentrations oforganic molecules in the particle phase (micrograms per cubic meter) arecalculated based on mass conservation balance upon phase transfer (SI Ap-pendix, Particle-Phase Components Mass Concentration Retrieval). Thenegative ESI sensitivity of pinonic acid product in the PILS collected liquidsample is calculated based on the counts/concentration calibration curve of acommercially available standard (98% purity; Sigma-Aldrich). Owing to thelack of authentic standards, the relative ionization efficiency of otheridentified products toward that of cis-pinonic acid in the ESI negative mode

14172 | www.pnas.org/cgi/doi/10.1073/pnas.1517742112 Zhang et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

5, 2

020

is calculated using a linear model with the acid dissociation constant atlogarithmic scale (pKa) of individual compounds, pH of the mobile phase,and weighted average positive sigma of different molecules as the input(31). These three parameters are computed using the Conductor-likeScreening Model for Real Solvents (32) implemented into the Amster-dam Density Functional (ADF 2014.07) molecular modeling suite (33).

Computational details are given in SI Appendix, Electrospray IonizationEfficiency Estimation.

ACKNOWLEDGMENTS. We thank John Crounse and Paul Wennberg foruseful discussions. This work was supported by National Science FoundationGrant AGS-1523500.

1. Jimenez JL, et al. (2009) Evolution of organic aerosols in the atmosphere. Science326(5959):1525–1529.

2. Hallquist M, et al. (2009) The formation, properties and impact of secondary organicaerosol: Current and emerging issues. Atmos Chem Phys 9(14):5155–5236.

3. Guenther A, et al. (1995) A global model of natural volatile organic compoundemissions. J Geophys Res 100(D5):8873–8892.

4. Claeys M, et al. (2009) Terpenylic acid and related compounds from the oxidation ofα-pinene: Implications for new particle formation and growth above forests. EnvironSci Technol 43(18):6976–6982.

5. Hall WA, Johnston MV (2011) Oligomer content of α-pinene secondary organicaerosol. Aerosol Sci Technol 45(1):37–45.

6. Jenkin ME, Shallcross DE, Harvey JN (2000) Development and application of a possiblemechanism for the generation of cis-pinic acid from the ozonolysis of α- and β-pinene.Atmos Environ 34(18):2837–2850.

7. Kristensen K, et al. (2014) Dimers in α-pinene secondary organic aerosol: Effect ofhydroxyl radical, ozone, relative humidity and aerosol acidity. Atmos Chem Phys 14(8):4201–4218.

8. Muller L, Reinnig MC, Warnke J, Hoffmann T (2008) Unambiguous identification ofesters as oligomers in secondary organic aerosol formed from cyclohexene andcyclohexene/α-pinene ozonolysis. Atmos Chem Phys 8(5):1423–1433.

9. Pathak RK, et al. (2007) Ozonolysis of α-pinene: Parameterization of secondary or-ganic aerosol mass fraction. Atmos Chem Phys 7(14):3811–3821.

10. Presto AA, Donahue NM (2006) Investigation of α-pinene + ozone secondary organicaerosol formation at low total aerosol mass. Environ Sci Technol 40(11):3536–3543.

11. Shilling JE, et al. (2009) Loading-dependent elemental composition of α-pinene SOAparticles. Atmos Chem Phys 9(3):771–782.

12. Yasmeen F, et al. (2010) Terpenylic acid and related compounds: Precursors for dimersin secondary organic aerosol from the ozonolysis of α- and β-pinene. Atmos ChemPhys 10(19):9383–9392.

13. Ehn M, et al. (2014) A large source of low-volatility secondary organic aerosol. Nature506(7489):476–479.

14. Mutzel A, et al. (2015) Highly oxidized multifunctional organic compounds observedin tropospheric particles: A field and laboratory study. Environ Sci Technol 49(13):7754–7761.

15. Zhang X, et al. (2015) Time-resolved molecular characterization of organic aerosols byPILS + UPLC/ESI-Q-TOFMS. Atmos Environ, 10.1016/j.atmosenv.2015.08.049.

16. Heaton KJ, Dreyfus MA, Wang S, Johnston MV (2007) Oligomers in the early stage ofbiogenic secondary organic aerosol formation and growth. Environ Sci Technol41(17):6129–6136.

17. DePalma JW, Horan AJ, Hall WA, Johnston MV (2013) Thermodynamics of oligomerformation: Implications for secondary organic aerosol formation and reactivity. PhysChem Chem Phys 15(18):6935–6944.

18. Clinton NA, Kenley RA, Traylor TG (1975) Autoxidation of acetaldehyde. 3. Oxygen-labeling studies. J Am Chem Soc 97(13):3757–3762.

19. McDowell CA, Sifniades S (1963) Oxygen-18 tracer evidence for termination mecha-nism in photochemical oxidation of acetaldehyde. Can J Chem 41(2):300–307.

20. Rissanen MP, et al. (2015) Effects of chemical complexity on the autoxidation mech-anisms of endocyclic alkene ozonolysis products: From methylcyclohexenes towardunderstanding α-pinene. J Phys Chem A 119(19):4633–4650.

21. Cooper RA, Ward HR, Lawler RG (1972) Mechanistic implications of chemically in-duced dynamic nuclear polarization spectra derived from radical-pair interactions inaliphatic diacyl peroxide decompositions. J Am Chem Soc 94(2):545–552.

22. Lamb RC, Pacifici JG, Ayers PW (1965) Organic peroxides. 4. Kinetics and products ofdecompositions of cyclohexaneformyl and isobutyryl peroxides. BDPA as a free-radical scavenger. J Am Chem Soc 87(17):3928–3935.

23. Leffler JE, More AA (1972) Decomposition of bicyclo [2.2.2]-1-formyl and pivaloylperoxides. J Am Chem Soc 94(7):2483–2487.

24. Linhardt RJ, Murr BL (1979) Mechanism of decomposition for diacyl peroxides andrelated carboxy inversion compounds in aqueous acetone. Tetrahedron Lett 20(12):1007–1010.

25. Linhardt RJ, Murr BL, Montgomery E, Osby J, Sherbine J (1982) Mechanism for diacylperoxide decomposition. J Org Chem 47(12):2242–2251.

26. Taylor KG, Govindan CK, Kaelin MS (1979) Radical and ionic reaction pathways in thethermal decomposition of diacyl peroxides. J Am Chem Soc 101(8):2091–2099.

27. Walling C, Waits HP, Milovanovic J, Pappiaonnou CG (1970) Polar and radical paths indecomposition of diacyl peroxides. J Am Chem Soc 92(16):4927–4932.

28. Kidd C, Perraud V, Wingen LM, Finlayson-Pitts BJ (2014) Integrating phase andcomposition of secondary organic aerosol from the ozonolysis of α-pinene. Proc NatlAcad Sci USA 111(21):7552–7557.

29. Ziemann PJ (2002) Evidence for low-volatility diacyl peroxides as a nucleating agentand major component of aerosol formed from reactions of ozone with cyclohexeneand homologous compounds. J Phys Chem A 106(17):4390–4402.

30. Aumont B, Szopa S, Madronich S (2005) Modelling the evolution of organic carbonduring its gas-phase tropospheric oxidation: Development of an explicit model basedon a self generating approach. Atmos Chem Phys 5(9):2497–2517.

31. Kruve A, Kaupmees K, Liigand J, Leito I (2014) Negative electrospray ionization viadeprotonation: Predicting the ionization efficiency. Anal Chem 86(10):4822–4830.

32. Klamt A (1995) Conductor-like screening model for real solvents: A new approach tothe quantitative calculation of solvation phenomena. J Phys Chem 99(7):2224–2235.

33. Pye CC, Ziegler T, van Lenthe E, Louwen JN (2009) An implementation of theconductor-like screening model of solvation within the Amsterdam density func-tional package - Part II. COSMO for real solvents. Can J Chem 87(7):790–797.

34. Donahue NM, Epstein SA, Pandis SN, Robinson AL (2011) A two-dimensionalvolatility basis set: 1. Organic-aerosol mixing thermodynamics. Atmos Chem Phys11(7):3303–3318.

Zhang et al. PNAS | November 17, 2015 | vol. 112 | no. 46 | 14173

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Oct

ober

5, 2

020