CSIRO PUBLISHINGResearch Paper

L. Chiappini et al., Environ. Chem. 2006, 3, 286–296. doi:10.1071/EN06037 www.publish.csiro.au/journals/env

Gaseous and Particulate Products from the Atmospheric Ozonolysisof a Biogenic Hydrocarbon, Sabinene

Laura Chiappini,A,E Nathalie Carrasco,B Brice Temine,C Benedicte Picquet-Varrault,A

Régine Durand-Jolibois,A John C. Wenger,D and Jean-François DoussinA

A Laboratoire Interuniversitaire des Systèmes Atmosphériques, Universités de Paris 12 et 7,Créteil Cedex 94010, France.

B Laboratoire de Chimie Physique, Université Paris Sud, Orsay Cedex 91405, France.C Laboratoire de Chimie et Environnement, Université de Provence, Marseille Cedex 13331, France.D Department of Chemistry and Environmental Research Institute, University College, Cork, Ireland.E Corresponding author. Email: chiappini@lisa.univ-paris12.fr

Environmental Context. Volatile organic compounds (VOCs) are a source of ozone and secondary organicaerosols, which have significant effects in the lower troposphere and on human health. The emission rateof VOCs from plants exceeds anthropogenic emissions by a factor of ten. In order to understand how theseplant-derived compounds influence global ozone budgets, studies into the atmospheric reactions of thesecompounds are needed. This study investigates the ozonolysis of sabinene, a VOC abundantly emitted bytrees in Europe.

Abstract. This work investigates both the gaseous and particulate phase products from the ozonolysis of sabinenein smog chamber experiments. The gaseous phase was analyzed in situ by FTIR. The particulate phase was analyzedafter sampling with a supercritical fluid extraction technique directly coupled to gas chromatography and massspectrometry (SFE-GC-MS) and to an in situ derivatization method. Sabinaketone, formaldehyde, and formicacid have been detected in the gaseous phase. More than 30 products have been observed in the secondary organicaerosol formed from sabinene oxidation and among them 10 have been identified as compounds containing carbonyl,hydroxyl and carboxyl groups. Hypotheses concerning reaction formation pathways have been proposed for eachidentified product in gaseous and particulate phases.

Keywords. atmospheric chemistry — ozonolysis — plant emissions — secondary organic aerosols — volatileorganic compounds

Manuscript received: 23 June 2006.Final version: 10 August 2006.

Introduction

Thousands of trace level organic compounds are emitted intothe atmosphere where they have strong effects such as influ-encing ozone budgets or forming particles.[1] Thus, theirreactivity and their ability to be converted into particles haveto be studied in order to include them in models and bet-ter assess their atmospheric chemistry, climate and healthimplications.[2] Among this variety of volatile organic com-pounds (VOCs), some are emitted in large quantities into theatmosphere from vegetation: biogenic volatile organic com-pounds (BVOCs). On a global scale, emissions of BVOCsexceed those of anthropogenic non-methane organic com-pounds by a factor of ten.[3] The world-wide emission rateof non-methane BVOCs is estimated to 1150 Tg of Cyear−1.[4] Of the estimated 84 Tg of C year−1 of BVOCsemitted in the USA, 20% is due to monoterpene com-pounds (C10H16).[5] As atmospheric oxidation of terpenes,

and more particularly their ozonolysis,[2,6,7] significantlycontributes to the formation of secondary organic aerosols(SOAs) and to the production of tropospheric ozone, theirreactivity has been extensively studied.[1,8–10] Among thelarge number of monoterpenes observed from plant emis-sions, sabinene was reported to be abundantly emitted byvery common tree species in Europe such as European beech(Fagus sylvatica),[11] evergreen holm oak (Quercus ilex),[12]and silver birch (Betula pendula).[13] The concentration ofsabinene in the atmosphere ranges from a few ppt to severalppb.[14] However, very few studies have been performed onits reactivity compared with those of α- or β-pinene, for exam-ple.As a consequence, the oxidation mechanisms of sabineneand the resulting products are poorly known. Indeed, only onekinetic investigation is available in the literature on its reac-tion with OH radicals and ozone.[15] Its lifetimes towardsOH and ozone were estimated to be 1–2 h for OH and 4–8 h

© CSIRO 2006 286 1448-2517/06/040286

Gaseous and Particulate Products from Ozonolysis of Sabinene

for ozone byAtkinson et al.[15] The high reactivity of sabinenedue to its double bond and the large extent of its emissionsby some common trees suggest that sabinene could play animportant role in the lower troposphere chemistry on a localscale. Information on both gaseous and particulate productsof sabinene ozonolysis is important to thoroughly under-stand sabinene oxidation mechanisms and aerosol formationprocesses. Sabinene’s gas phase chemistry has been studiedby Reissell et al.[16] and Hakola et al.[9] Reissell et al.[16]reported the production of acetone, whereas Hakola et al.[9]measured the production of sabinaketone.A great difficulty inthe identification and the quantification of sabinaketone is thelack of commercially available standards.A major outcome ofthis work lays in the synthesis and the FTIR calibration of themajor semi-volatile product from sabinene ozonolysis, sabi-naketone. Griesbaum and Miclaus[17] also studied sabineneozonolysis. However, their experiments are not comparable tothe experiment presented in this article as their concentrationswere much higher than those used in this study.

As for the particulate phase, sabinene has been shown tobe a good SOA precursor.[10,18,19] Some studies have beenperformed to elucidate the composition of the particulatematter formed from sabinene ozonolysis.[10,18,20–22] Kochet al.[18] identified sabinic acid by gas chromatography–massspectrometry (GC-MS) and BF3 derivatization, and Glasiuset al.[21] observed sabinic acid and norsabinic acid. However,the most exhaustive work was made by Yu et al.,[10] whichis, to the best of our knowledge, the only study reportingresults on products formed in both gas and particulate phases.This investigation, using GC-MS and a double derivatiza-tion technique, has resulted in the detection in the particulatephase of compounds containing carbonyl (e.g. sabinaketone),hydroxyl (e.g. hydroxysabinaketone), and dicarboxylic (e.g.sabinic acid) functional groups. However, the characteriza-tion of SOA molecular composition remains a real analyticalchallenge because of its complexity. SOAs, which consist oftrace level compounds displaying a wide range of functionali-ties, solubilities, and polarities, are still poorly characterized.Moreover, the formation of SOA is still not completely under-stood. In the present work, the SOAs formed from sabineneozonolysis in smog chamber experiments are analyzed by anon-line supercritical fluid extraction–gas chromatography–mass spectrometry (SFE-GC-MS) technique coupled to aN,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) derivati-sation method.This direct technique allows detection of someof the functionalized compounds detected byYu et al.,[10] theassumption of the structure of some other products on thebasis of their chemical ionization (CI) and electron impact(EI) mass spectra, and the description of some proposedformation pathways for each product.

Experimental

Smog Chamber Experiments — Gas Phase Analysis

Experiments were carried out using two different atmospheric simu-lation chambers, at LISA (Créteil, France) and at University CollegeCork (UCC, Ireland). As both chambers have been described in detailby Doussin et al. (1997)[23] and Thuener et al.[24] only relevant infor-mation will be provided here. The LISA chamber is a 977-L cylindrical

Pyrex evacuable reactor, equipped with a multiple-reflection opticalsystem coupled to a Bomem DA8-ME FTIR spectrometer.[23] The pathlength of the chamber was adjusted to 156 m. The chamber was filledwith synthetic air (80% N2, 20% O2) at atmospheric pressure and thetemperature was kept at 295 ± 2 K throughout the experiment. Duringthe reaction, the gaseous reaction mixture was continuously monitoredby FTIR spectroscopy. Infrared spectra were obtained at a resolution of0.5 cm−1 using a mercury cadmium telluride (MCT) detector and werederived from the collation of 200 scans collected over 5 min.

The chamber at UCC is a cylinder made of FEP (fluorine-ethene-propene) foil (4.1 m long, 1.1 m diameter and 0.127 mm thickness) witha volume of 3.91 m3. It was operated at 295 ± 2 K using purified dry air at0.1–1.0 mbar above atmospheric pressure.The chamber is equipped witha multiple reflection optical arrangement coupled to a FTIR spectrome-ter (BioRad Excalibur) to monitor gas phase chemical composition. Theoptical path length was (229.6 ± 0.6) m. Infrared spectra were obtainedat a resolution of 1 cm−1 using a narrow-band MCT detector and werederived from the collation of 200 scans collected over 4 min. Around1 ppm (see Table 2 for initial concentrations) of sabinene was first intro-duced in the simulation chamber through a very gently heated (∼50◦C)inlet system by a flow of purified air. Ozone was generated by flowinghigh purity O2 through a silent discharge ozonizer (Kaufmann Umwelt-technik, GmBH, in LISA; Ozone Services GE60/M5000 in UCC) andits concentrations were in the same range as sabinene. As it has beenshown that the reaction of sabinene with ozone leads to the formationof OH radicals,[25] an OH radical scavenger was used to isolate onlythe ozonolysis products of sabinene. Carbon monoxide, the capacityof which to act as an OH radical scavenger has been demonstrated byGutbrod et al.,[26] was chosen to perform this function. Indeed, despitesome safety problems, not only is it easy to handle, but its IR spectra alsoexhibits little interference with other absorbing compounds. Moreover,its oxidation product is CO2, which is an advantage when studying SOAformation compared with the use of other scavengers such as cyclohex-ane, which is likely to produce species that can be found in the particulatephase. Using k(OH+ CO) = 2.4 × 10−13 cm3 molecule−1 s−1, deter-mined by Gutbrod et al.,[26] and k(OH+ sabinene) = 1.2 × 10−10 cm3

molecule−1 s−1, determined by Atkinson et al.,[15] it can be calculatedthat between 85% and 95% of the OH radicals formed from sabineneozonolysis are scavenged with CO concentrations ranging from 5000 to13 000 ppm.

For the ozonolysis experiments, as the water content of the gas phaseis an important parameter,[27–29] careful attention was paid to the drynessof the air introduced in the chamber. However, as the UCC reactor ismade of a thin, 0.127-mm Teflon film, water permeation could not beexcluded and water concentrations in the chamber, as measured by arelative humidity recorder, ranged from ∼100 to 300 ppm. These valuesare given in Table 2. At LISA, because the Pyrex reactor is evacuableand can be filled with synthetic air, the water concentrations were lowerand were estimated to be less than 100 ppm. This water can be due tooffgas from the walls.

Quantitative analyses were performed by subtraction of calibratedreference IR spectra of known compounds and subsequent integrationof selected absorption bands for sabinene and the reaction products.Sabinaketone was previously synthesized by sabinene ozonolysis in theliquid phase as described in Carrasco et al.[30] and calibrated by FTIRspectroscopy. Its IR spectrum is presented (Fig. 1). Integrated bandintensities (IBI) of products used during the subtraction procedures aregiven inTable 1. Formation yields of oxidation products were determinedby calculating the initial slopes of the plots displaying the concentrationof products as a function of reacted sabinene (see Fig. 2). Uncertain-ties take into account twice the standard deviation and uncertainties oncalibrations.

Particle Formation Sampling and Analysis

In the UCC chamber, particles were formed from the reaction ofsabinene with ozone (seeTable 2 for experimental conditions).The num-ber, concentration, and size distribution of aerosol particles producedduring the experiments were determined by a scanning mobility particlesizer (SMPS) consisting of a condensation particle counter (CPC, TSI

287

L. Chiappini et al.

1250 1200 1150 1100 1050 1000 950 900 850 800

Wavenumber [cm�1]

(b)

(a)

(c)

1136 cm�1 1016 cm�1 937 cm�1

Fig. 1. FTIR spectra. (a) Gaseous mixture spectrum 50 min afterozone injection. (b) Residual spectrum after subtraction of prod-ucts from spectrum A (subtracted compounds: sabinene, sabinaketone,ozone, HCHO, HCOOH). (c) Sabinaketone reference spectrum.

Table 1. IBI of the main IR absorption bandsValues are given in loge

Compound Main absorption IBI [cm molecule−1] Referenceband [cm−1]

Sabinene 2800–3110 (5.440 ± 1.5) × 10−17 This workAcetone 1260–1150 (1.02 ± 0.02) × 10−17 [47]Formaldehyde 3000–2630 (1.33 ± 0.06) × 10−17 [47]Sabinaketone 1700–1800 (3.7 ± 0.4) × 10−17 This work

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0 0.2 0.4 0.6 0.8 1

[Sabinene]0 � [Sabinene] [ppm]

[Pro

duct

] [pp

m]

Sabinaketone

Formaldehyde

Formic acid

Fig. 2. Formation of sabinaketone, formaldehyde, and formic acidduring a reaction between sabinene and ozone (LISA, 27 August 2004).

3022A) and a differential mobility analyzer (DMA, TSI 3081). Particlesize distributions were collected with a 3 min time resolution. During theexperiment, once the particles were fairly stable in size and number asmonitored by the SMPS, they were collected on a glass microfibre filter

(47 mm, Whatman).The sampling lasted 20 min at 20 L min−1 through astainless steel filter holder. Prior to use, the filters were soxhlet extractedfor 24 h with distilled dichloromethane (DCM). After sampling, thefilters were immediately stored in glass jars with Teflon lids at −4◦C.The jars and the filter sampler were prepared before use by rinsing withdistillated water and distilled DCM. Several blank filters were anal-ysed for quality control. The analytical method used to analyze the SOAchemical composition is described in detail by Chiappini et al.[31]

The extraction unit was a Carlo Erba SFC 3000 (developed forSupercritical Fluid Chromatography), connected to a Varian 1200quadrupole GG-MS equipped with an analytical column (CP-Sil 8 CBlow bleed/MS, 30 m × 0.25 mm i.d. from Varian). The extraction cellwas filled with the sampling filter impregnated with 15 µL of a 1%BSTFA solution in DCM. The filter was then allowed to equilibrate andthe derivatization could occur in supercritical CO2 at 300 bar and 60◦Cfor 30 min. The analytes were then extracted for 15 min with supercrit-ical CO2 flowing throughout the extraction cell and trapped at the headof the cooled analytical column. Meanwhile, the CO2 was vented to theatmosphere. Once the extraction step was complete, the GC oven tem-perature was raised from −20◦C to 250◦C at 15◦C min−1 and heldat 250◦C for 15 min. Detection was achieved with a quadrupole massspectrometer (Varian 1200) either by EI (70 eV electron energy) or CI,using methane in positive ionisation mode (methane pressure 4 Torr,150 eV electron energy). The switching between EI and CI is per-formed by changing the ion volume in the ionization source on themass spectrometer. The scanned mass ranged from 50 to 800 amu.

Chemicals

Chemicals were obtained from the following sources and usedwithout further purification: sabinene (Interchim, >99%), CO (BOCgases, >99%). No identified products from sabinene ozonolysis werecommercially available.As a consequence, we chose as standards for theparticulate phase analysis compounds with similar structure and com-prising the most characteristic functionalities: pinic acid and norsabinicacid as surrogates to C9 and C8 compounds bearing two carboxylicfunctionalities respectively, cis-pinonic acid for C9 and C8 compoundsbearing one oxocarboxylic function, and 2-hydroxycyclohexanone as asubstitute of hydroxysabinaketone. All standards were purchased fromSigma-Aldrich. Sabinaketone was synthesized in our group.[32] CO2(SFE-grade) was provided by Air liquide. BSTFA (which contains 1%trimethylchlorosilane as a catalyst) was supplied by Sigma-Aldrich.Dichloromethane (HPLC grade) was provided by Prolabo.

Results and Discussion

Gas Phase Chemical Composition

Product yields from the reaction of sabinene with ozone weremeasured in the UCC chamber and in the LISA chamber byFTIR (see Table 2 for experimental details). Formaldehyde,sabinaketone, and formic acid were observed and quantified(see FTIR spectra on Fig. 1 for sabinaketone).A typical plot ofproduction yields observed during the different experimentsis shown in Fig. 2. These plots are straight lines, whichdemonstrates that sabinaketone, formaldehyde, and formicacid are primary products. Their molar yields are given inTable 2. Table 2 shows that formaldehyde and sabinaketoneare the main products formed and that most of the reac-tive carbon is sabinaketone. It must also be noted that somesabinaketone has been formed in the particulate phase (seenext section) and that consequently, the overall yield for thiscompound is probably slightly higher. The production yieldsare averaged and compared to previous literature data inTable 3. The sabinaketone yields observed by Hakola et al.[9]and Yu et al.[10] were slightly higher than the yield found

288

Gaseous and Particulate Products from Ozonolysis of Sabinene

Table 2. Initial conditions of the experiments of sabinene ozonolysis, amount of reacted sabinene, aerosol formed and overallaerosol yield (Y)

Experiments UCC 17.xi.03 UCC 20.xi.03 LISA 27.viii.04 UCC 12.x.04 UCC 13.x.04

Initial concentrations [ppm] [Sabinene]0 0.81 1.67 1.96 1.17 1.29[O3]0 1.54 1.43 0.82 1.32 0.76[CO]0 5000 5000 10 000 5000 13 000

HO2 [ppm] n.a. n.a. n.a. 100–300 100–300

Products in the gas phase Formaldehyde 0.55 ± 0.09 0.50 ± 0.06 0.49 ± 0.08 0.49 ± 0.08 0.57 ± 0.10(molar yield) Sabinaketone 0.37 ± 0.12 0.23 ± 0.07 0.31 ± 0.10 0.43 ± 0.14 0.40 ± 0.13

Formic acid 0.21 ± 0.07 0.23 ± 0.08 0.12 ± 0.04 n.d. 0.12 ± 0.04�HC (µg m−3) n.a. n.a. n.a. 492 481

Aerosol M0 (µg m−3) n.a. n.a. n.a. 3000 3500Y n.a. n.a. n.a. 0.16 0.13

n.d., not detected; n.a., not available.

Table 3. Comparison of gas phase product yields from sabinene ozonolysis with values from the literature

This work Yu et al.[10] Hakola et al.[9] Reissell et al.[16]

Sabinaketone 0.35 ± 0.14 0.47 ± 0.24 0.50 ± 0.09 –Acetone Detected – – 0.03 ± 0.02Formaldehyde 0.52 ± 0.09 – – –Formic acid 0.17 ± 0.07 – – –Hydroxy-sabinaketone – 0.07 ± 0.04 – –Pinic acid – 0.11 ± 0.06 – –Sabinic acid – 0.08 ± 0.04 – –Norasabinic acids + isomers – 0.04 ± 0.02 – –Carbon balance 0.38 ± 0.15 0.69 ± 0.35 0.45 ± 0.08 0.009 ± 0.006Aerosol molar yield 0.13 0.003 – –

in the present work. The yields obtained for sabinaketonewere variable. This can be attributed on the one hand tothe IR spectra complexity (sabinaketone measurment is per-formed in a region where many other compounds absorb, seeFig. 1) and on the other hand to sabinaketone partitioningbetween the gas and the particulate phase (which can bevariable and depend on experimental parameters such astemperature homogeneity and water concentration). Never-theless, all three studies are in agreement considering theoverall combined uncertainties. It must be pointed out thatthis study is the first time sabinaketone has been mea-sured using a synthesized standard. Our FTIR response wascarefully calibrated using the synthesized standard, whereasYu et al.[10] and Hakola et al.[9] detected and quantifiedthis compound by a derivatization technique involving O-(2,3,4,5,6-pentafluorobenzyl)hydroxylamine hydrochloride(PFBHA)-GC-MS and by GC-FID, respectively, assumingthat the sabinaketone response factor was equal to the one ofa similar compound, nopinone.

Acetone yield (R) has been previously studied by Reissellet al.[16] leading to the value Racetone = 0.03 ± 0.02. How-ever, in the present study IR spectra are too complex toallow acetone detection by FTIR analysis. Indeed, its IRspectrum signature can hardly be detected because of itsweak absorption among other products absorbing in the samezone (1260–1150 cm−1). Yu et al.[10] observed the forma-tion of sabinic acid, 2-(2-isopropyl)-2-formyl-cyclopropyl-methanoic acid, hydroxysabinaketones, and norsabinic acidand isomers in the gaseous phase as a result of PFBHA and

BSTFA derivatizations after sampling and extraction througha denuder. We did not detect such compounds. This study iscomplementary to the previous ones, using another analy-sis method, FTIR, to detect sabinaketone. It allowed us tomeasure other products in the gas phase, such as formalde-hyde and formic acid. Nevertheless, the residual spectrum onFig. 1b shows that some absorbing bands are still uncharac-terized. Many compounds potentially formed from sabineneozonolysis would display C-O and O-O groups and absorbin the same region (1200–900 cm−1).These bands could cor-respond to the products observed by Yu et al.,[10] however,because of the lack of commercially available standards, thesespecies could not be identified. Moreover, the carbon balancetaking into account the three products is only 0.38 ± 0.15 andis still highly incomplete, which is a good reason to investigatethe particulate phase.

Particle Formation

Particle formation from the ozonolysis of sabinene was quasi-immediate, a few seconds after the injection of O3 into thechamber. The particle size diameter was initially ∼100 nm.As the SMPS scanning time resolution (120 s) was lowerthan the rate of particle formation (a few seconds), particlessmaller than 100 nm diameter could not be observed. Withinthe experiment, particles grew in size from 100 to 280 nm bycondensation of the semivolatile molecules on their surfaceand coagulation while they were decreasing in number, notonly because of coagulation but also because of loss on thechamber walls.

289

L. Chiappini et al.

Table 4. Characteristic m/z ions from the BSTFA derivatives

CI ions Pseudo-molecular ions (m/z) EI ions Fragments (m/z)

M + 29 [M + C2H5]+ 73 [Si(CH3)3]+M + 1 [M + 1]+ 75 [HO=Si(CH3)2]+M − 15 [M − CH3]+ 117 [COOSi(CH3)2]+M − 89 [M − OSi(CH3)3]+ 147 [(CH3)2Si=OSi(CH3)3]+M − 117 [M − COOSi(CH3)3]+M − 147 [M − (CH3)2Si=OSi(CH3)3]M − 133 [M − (CH3 + COOSi(CH3)3 + H)]+M − 207 [M − (OSi(CH3)3 + COOSi(CH3)3 + H)]

Table 2 presents the overall aerosol yield (Y) for twoexperiments of sabinene ozonolysis (12.10.04 and 13.10.04).Odum et al.[33] defined Y as the fraction of M0, the organicmass aerosol concentration (µg m−3), on �HC, the amountof VOC that has reacted (µg m−3). The aerosol yields mea-sured for both experiments were very similar and their value(∼15%) was higher than the value found by Griffin et al.[19](∼3%). The precursor concentration these authors used wasmuch lower than the concentration used in the present study,which can explain the lower aerosol yields they measured.Theyield values obtained in the present paper provide informationon the ability of sabinene to form SOAs and can be used forfurther modelling studies of chemical reactivity and aerosolproduction. Caution must be taken before direct applicationto the real atmosphere.

Aerosol Composition

The SFE-GC-MS method associated with a BSTFA deriva-tization technique allowed the detection of some highlypolar SOA compounds formed by sabinene oxidation. TheBSTFA derivatization technique is necessary to detect com-pounds bearing -OH and -COOH groups as they are poorlyextracted by supercritical CO2, which is a good solvent forlow polar compounds, and as they are poorly detected andidentified by GC-MS because of the strong interaction withthe GC column. The trimethylsililation of -OH and -COOHgroups lowers the polarity of the analytes and, therefore,enhances their solubility in the supercritical CO2 and allowstheir elution onto the GC column. The resulting derivativescould then be resolved by the GC column and identifiedby their EI and CI mass spectra; these spectra give com-plementary information for characterizing the compounds.The derivatization method gives rise to specific ion frag-ments at m/z 73 [(CH3)3Si]+ and 75 [HO=Si(CH3)2]+ inEI mass spectra. EI mass spectra also show a fragment atm/z 117 [COOSi(CH3)3]+ for monocarboxylic acids. Asfor compounds bearing two active H atoms, they exhibit am/z 147 corresponding to [(CH3)2Si=OSi(CH3)3]+. We thusmanaged to differentiate compounds with -OH and -COOHfrom other organics.[34] CI mass spectra exhibit a unique setof pseudo-molecular ions, corresponding to the derivatizedmolecular ions, allowing for the molecular weight determi-nation of the studied compounds.[34,35] Table 4 summarizesthe characteristic fragments of both the CI and EI spectra.

Figure 3 shows two chromatograms resulting from thesame analysis of sabinene SOA. The bottom chromatogram

0.00E�00

2.00E�09

4.00E�09

6.00E�09

8.00E�09

1.00E�10

1.20E�10

17 18 19 20 21 22

Time [min]

Ion

curr

ent

S1

S2

S3S4

S5

S7S3

S5

S6

S4

S8

S6

S3

Fig. 3. Chromatogram of products from the sabinene + O3 reac-tion in particulate phase. Top chromatogram: total ion current. Bot-tom chromatogram: product containing OH/COOH groups (m/z =73 + 75 + 117 + 147). The structures of the detected compounds, S1to S8 are reported in Table 3 and their mass spectra in Fig. 4.

is a reconstruction of the m/z 73 + 75 + 117 + 147 ions, dis-playing all products with a hydroxyl group, whereas thetop chromatogram shows the total ion current. The deriva-tized compounds, that is to say the compounds bearing -OHand -COOH groups, could thus be differentiated from otherorganics.[34] Table 5 shows the identified products with theirmolecular weight (MW), derivatized molecular weight (derMW), the pseudo-molecular ions in CI with BSTFA, theirstructure, and whether they have already been observed inprevious works by other groups. The CI or EI mass spectraof the identified compounds are shown on Fig. 4.

Compound S1 was easily identified as sabinaketonebecause of its authentic standard. The two peaks assignedto S1 cannot be explained but have also been observed byYu et al.[10] The structure of S2 was assumed from the CImass spectrum, which shows ions at m/z 243, 215, 199, 125,and 97 corresponding to M + 29, M + 1, M − 15, M − 89and M − 117. M + 1 and M + 29 are pseudo-molecular ionscorresponding to [M + H]+ and [M + C2H5]+. The ionsat M − 15, M − 89 and M − 117 are fragments resultingfrom the loss of CH3, OSi(CH3)3 and C(O)OSi(CH3)3.This latest fragment loss is presumed to indicate the pres-ence of a monocarboxylic group. Moreover, the M − 43ion on EI spectra could stand for an isopropyl group.A postulated structure for S2 is given in Table 3. Threepeaks are assigned to S3 which is tentatively identifiedas hydroxysabinaketone, as its CI mass spectrum showsa molecular weight of 226 for its BSTFA derivative.Moreover, the general frame of its mass spectra is very

290

Gaseous and Particulate Products from Ozonolysis of Sabinene

Table 5. Identified products of sabinene ozonolysis

Formula Pseudo-molecular ions (CI mass spectra) Structure

S1: C9H14O Not derivatizedSabinaketone Observed previously[10]

MW: 138

O

S2: C8H14O2 M + 29 = 243 M − 89 = 1251-Methyl-3-cyclopropyl-pentanoic acid M + 1 = 214 M − 117 = 97MW: 142 M − 15 = 199der MW: 214 Not observed previously

O

OH

S3: C9H14O2 M − 15 = 211Hydroxysabinaketone M − 89 = 137MW: 154 Observed previously[10]der MW: 226

O

OH

O

OH

S4: C8H12O3 M + 29 = 257 M − 89 = 1393-Formyl-1-isopropyl-cyclopropan- M + 1 = 229 M − 117 = 111

1-formic acid M − 15 = 213MW: 156 Observed previously[10]der MW: 228

O

O

OHH

H

O

OOH

S5: C9H14O2 Not derivatized6-Isopropyl-2-oxa-3 and 6-isopropyl- Observed previously in gas phase[17]3-oxa-2-oxobicyclo[4.1.0]heptane

MW: 154

OO

O

O

S6: C9H14O3 M + 1 = 243 M − 117 = 125Sabinalic acid M − 15 = 227 M − 133 = 109MW: 170 M − 89 = 153der MW: 242 Observed previously[10]

O

OOH

H

S7: C8H12O4 M + 1 = 317 M − 207 = 109Norsabinic acid M − 15 = 301MW: 172 M − 89 = 227der MW: 316 Observed previously[10,21]

O

O

OHOH

S8: C9H14O4 M − 15 = 315 M − 147 = 183Sabinic acid M − 117 = 213MW: 186 M − 133 = 197der MW: 330 Observed previously[10,18,21]

O

OOH

OH

similar to its surrogate standard, 2-hydroxycyclohexanone.The two possible isomers of hydroxysabinaketone corre-sponding to the two possible OH group positions on the cycleexplain the presence of two peaks for S3. An explanationfor the third peak could be the possibility of an OH groupattached to the secondary carbon, to be above or under thecycle. A structure for S4 is proposed based on the set of CIfragments M + 29, M + 1, M − 15, M − 89, and M − 117,which suggests a molecular weight of 228 for the derivatizedcompound and indicates the presence of a monocarboxylicgroup.The assumptions concerning the S5 structure are based

on its CI mass spectra, which indicates a molecular weightof 154. S6, which has already been observed before by Yuet al.,[10] has been tentatively identified based on the set ofCI fragments M − 89, M − 117, and M − 133, suggesting amolecular weight of 242 for the derivatized compound andindicating the presence of two carboxylic groups. As for S7

and S8, they are tentatively identified by comparing the setof ions in their CI mass spectrum to those of the surrogatestandards (trans-norpinic acid and pinic acid respectively).Their previous detection by other groups was an additionalclue in confirming their identification.[10,21]

291

L. Chiappini et al.

100

75

50

50 75 100 125

25

0

Rel

ativ

e in

tens

ity [%

]

5567

8191

109123

138

O

m/z

100

75

50

25

0

Rel

ativ

e in

tens

ity [%

]

m/z

50 75 100 125 150

O O

95

109

137

155

100

75

50

25

0

Rel

ativ

e in

tens

ity [%

]

m/z

50 100 150 200 250

M�117

M�117

M�89 M�89

M�133

M�15

M�1

O

O

O HSi

109

73

125

153

183 227

243

100

75

50

25

0

Rel

ativ

e in

tens

ity [%

]

m/z

50 100 150 200 250 300

M�117

M�

117

M�

89 M�

15

M�1

227 301

317

M�207M�147109

169

73 137

199

345

M�89

M�29

M�15OO

O

OSi

Si

100

75

50

25

0

Rel

ativ

e in

tens

ity [%

]

m/z

50 100 150 200 250 300 350

M�117

M�117M�29

M�89M�89M�15

M�

1

OO

OO

Si

Si

213151

241

M�15315

73

331359

100

75

50

25

0

Rel

ativ

e in

tens

ity [%

]

m/z

50 100 150 200

M�89 M�15

M�89

M�1

137153

227

109

73

M�15211O

O Si 100

50

0

Rel

ativ

e in

tens

ity [%

]

m/z

50 100 150 200 250

M�117M�89

M�15

229

257

O

OSi

H

111

139

213

S1

S3

S5

S6

S7 S8

100

75

50

50 100 150 200

25

0

Rel

ativ

e in

tens

ity [%

]

m/z

M�117 M�89

M�15

SiO

O

199

M�

15M

�1

M�

117

M�

89

M�29

215

243

S2

57

73 97

125

S4

Fig. 4. EI (for S1) and CI (for the other compounds) mass spectra of identified compounds.

Yu et al.[10] and Glasius et al.[20] also identified pinic acidin the particulate phase formed from sabinene ozonolysis.In both cases, the mechanism leading to pinic acid from thereaction of sabinene with ozone remains unclear and an ana-lytical artefact (e.g. isomerization) cannot be excluded. In our

case, we found no evidence of the presence of pinic acid inSOAs from sabinene ozonolysis.

To conclude, sabinaketone (S1) has been identified withcomplete certainty because of the synthesis of an authen-tic standard. Five compounds (S3, S4, S6, S7, S8) have been

292

Gaseous and Particulate Products from Ozonolysis of Sabinene

tentatively identified thanks to their mass spectra and bycomparison with surrogate standards bearing very similarstructure and functionalities. For the other compounds (S2

and S5), we assumed a possible structure on the basis of theirEI and CI mass spectra. Despite the fact that many SOAsfrom sabinene constitutive compounds are still not charac-terized, mostly because of the lack of available standards,the SFE-GC-MS method coupled to BSTFA derivatizationproved to be a valuable technique; it enabled the investigationof SOAs from sabinene ozonolysis, eliminating the manual,time-consuming, and error-prone sample pretreatment steps.As an illustration of the importance of this investigation, anevaluation of the aerosol molar yield is given inTable 3.Thesevalues, deduced from the aerosol mass yields (15%), are cal-culated on the basis of the molecular weights of compoundsidentified in SOAs from sabinene ozonolysis, balanced withtheir peak area in gas chromatograms. The particulate phasemolar weight was then estimated to be 150 g mol−1. Consid-ering the value of 0.13 for the aerosol molar yield, it appearsthat the particulate phase can explain a part of the missingcarbon in sabinene ozonolysis reaction.

Mechanism for the Formation of Detected Compounds

In this experiment it was found that sabinaketone, formalde-hyde and formic acid were formed in the gas phase and severalpolyfunctional compounds were formed in the particulatephase. No data are available from the literature concern-ing formaldehyde, and formic acid production yields fromsabinene ozonolysis. Therefore, based on the results of thecurrent investigation, we tried to elucidate the mechanismof sabinene ozonolysis and identify the product formationpathways. Figure 5 depicts the possible routes to the for-mation of the identified products. Sabinaketone, which isthe main product detected, is formed from the decomposi-tion of the primary ozonide produced after the reaction ofO3 on the sabinene exocyclic double bound (reaction 1).It then partitions between the gaseous and the particulatephase.

Concurrently with sabinaketone formation, a small vibra-tionally excited Criegee biradical is formed (reaction 2a).Formaldehyde is formed following a similar process wherea large Criegee biradical is also formed (reaction 2b). As forformic acid, it is formed from the chemical evolution of thesmall excited Criegee radical. It must be indicated that somesecondary formic acid (not observed in the present work) canbe formed from the reaction of formaldehyde with HO2. Itcan also be produced indirectly from the reaction between thesmall Criegee and H2O. The role of water in the production oforganic acids from the reaction of Criegee radical have beenintensively discussed.[27,36–40] Some authors propose a reac-tion between the Criegee radical and water which decomposesinto organic acid. As an example, Neeb et al.[37] propose thefollowing reaction from one carbon Criegee radical:

CH2OO + H2O −→ HOCH2OOH −→ HCOOH + H2O

In this case, hydroxymethyl hydroperoxide (HMHP), decom-poses into HCOOH with a life time of several minutes leading

to a secondary formation of formic acid. Other authors suchas Orzechowska et al.,[38] describe a direct formation ofHCOOH, independently of the relative humidity.

In our case, formic acid was positively detected as pri-mary product (see Fig. 2). In spite of systematic searching, noHMHP was found in experimental FTIR spectra.These obser-vation seem to support the existence of the direct HCOOHformation pathway, which can be explained by the verylow relative humidity in the chamber (less than 300 ppm).The large Criegee biradical can either become collisionallystabilized or decompose following different pathways: thehydroperoxide channel (reaction 3) or the ester channel (reac-tion 4).[27] This latest, as described by Winterhalter et al.,[41]begins with a ring closure to form the dioxirane intermedi-ate whose O–O bond dissociation leads to the formation ofa biradical which evolves to give the two identified lactones,and which correspond to S5. These molecules have alreadybeen observed by Griesbaum and Miclaus,[17] when study-ing the gas phase of sabinene ozonolysis. The hydroperoxidechannel consists of the isomerization of the excited Criegeevia hydrogen shift to the unsaturated hydroperoxide whichis to decompose to an alkenoxy radical and an OH radical(reaction 5a). The alkenoxy radical reacts with O2 to give theperoxy radical (reaction 6) which can react with HO2 to giverise to one of the two hydroxysabinaketones denoted S3. Theformation of the double bond via the shift of the hydrogenattached to the tertiary carbon does not seem to be possiblebecause of the ring strain caused by the two cycles. A reac-tion with a hydrogen abstraction by an OH radical (reaction5b) to explain the formation of S3, was proposed. This per-oxy radical can also react with any peroxy radical denoted asRO2 and thus lead to the alkoxy radical (reaction 8). Thislatest can fragment either via a ring opening (reaction 9)or the loss of CO2 (reaction 10). The ring opening pathwaygives rise to two acylperoxy radicals, described by Jenkin[42]as key intermediates in the formation of dicarboxylic acids(S8). These dicarboxylic acids have been assigned as themost likely product from terpene ozonolysis leading to SOAformation.[18,41,43] Jenkin et al.[42] explained the acylperoxyradical evolution either by an isomerisation after reaction witha HO2 radical (reaction 11b), first proposed by Winterhalteret al.,[41] or reaction with RO2 (reaction 11a). Dicarboxylicacids could also be formed from the oxidation of the oxoacidsS6. However, Jenkin et al.,[43] when studying pinene ozonol-ysis, described pinic acid as a first generation product on thebasis of the aerosol formation timescale determined by Kochet al.[18] The radical formed from reaction 9 can undergoa radical delocalization, react with O2 (reaction 12) andevolve to form a dialdehyde, the oxidation of which canexplain the formation of S4 and S7. This reaction (13) hasalready been described by Yu et al.[10] Reaction 10 couldbe a possible route to obtain S2 whose structure remainsuncertain.

Thus, even if some products remain unknown, the analysisof the particulate phase of sabinene ozonolysis enabled us topropose some reaction mechanisms for biogenic hydrocarbonoxidation, and provided us with some clues concerning itsatmospheric behaviour in response to oxidation.

293

L. Chiappini et al.

O

OOO

C

OO

OOH

OOH

O O

OO

OHO

O

OH

O

OC

O

OC

O

O

C

OO

O

O

H

OOO

OOO

OCO H

CO

OO

O

OHO

O

O

O

O

H

O

OO

OH O

O

OHH

O

O

HOH

O

O

O

OH

H

O

OOH

OH

O

O

OH

O

OO

OHH

O

O

OH

O

O

H H

O

O

H OH

O

O

OHOH

O

O

OHH

O

O

OHOO

O

O

OHOO

OOH

H

O

OH

O

CO

O O

*

�

HCHO

�

OH �

�C•H2OO•HCOOH

�OH

(1)

(2a)

(2b)

(4)

(3)

(5a) (6)

(8)

(10)

�CO2

(9)(12)O2

�CO2

(13)HO2or

ox*

HO2HO2

or ox*or ox*

HO2or

RO2

HO2or

RO2

HO2HO2RO2 RO2 (11b)

(11a)

(11b)

(11b)(11b)

Isom

IsomIsom

Isom

HO2or

RO2

Radicaltransfer

HO2

orRO2

(5b)

O2

HO2or

RO2

S5

S3

S3

S1

S4

S4S7

S8 S2

S6 S6

Fig. 5. Reaction mechanism of sabinene ozonolysis (ox*, oxidation in particulate phase; in circles, products observed in gaseousphase; in squares, products observed in particulate phase).

Conclusions and Future Experiments

The analysis of the gas phase of sabinene ozonolysis led to theidentification of some compounds such as formic acid, whichhad not been observed before. However, some functionalizedproducts (sabinic acid, sabinalic acid etc.) detected by Yuet al.[10] were not observed in the present work by FTIR.

Further experiments involving gas phase analysis with anadditional gas phase analytical technique like the doublederivatization method developed by Yu et al.[34] should becarried out.

Our investigation of the particulate phase led to the iden-tication of highly functionalized molecules such as sabinic

294

Gaseous and Particulate Products from Ozonolysis of Sabinene

acid, norsabinic acid, sabinalic acid, and sabinaketone, andincluded two compounds that have not been previouslyreported in the particulate phase (the monoacid S2, and thelactone S5). Quantification of these compounds in the partic-ulate phase, associated with a simultaneous quantification inthe gaseous phase, should also be carried out to (i) determinethe main routes leading to SOA formation, and (ii) calcu-late partitioning coefficients between the gas and particulatephase of compounds formed from volatile organic compoundoxidation. Nevertheless, identification of compounds in bothgas and particulate phases enabled the elucidation of thesabinene ozonolysis mechanism. In particular, we proposeformation pathways for sabinonic and sabinic acid, explain-ing how these low volatile products could influence SOAformation and growth in the atmosphere.

The particulate phase has been shown to account foraround 13% of the amount of reacted sabinene and is not ableto explain the all of the missing carbon.This may be attributedto the fact that at this stage, the technique is not adapted to theanalysis of high molecular weight compounds (oligomers)which have recentlty been observed in SOA.[44–46] In thefuture, this method may be modified (e.g. through the useof supercritical fluid chromatography) to handle those kindsof compounds.

Acknowledgements

The authors would like to thank INERIS (Institut Nationalde l’Environnement et des RISques), EGIDE, a French lead-ing agency for international mobility and EUROCHAMP,an Integrated Infrastructure Initiative of the 6th frameworkEuropean program, for financial support.

References

[1] J. Williams, Environ. Chem. 2004, 1, 125. doi:10.1071/EN04057[2] M. Kanakidou, J. H. Seinfeld, S. N. Pandis, I. Barnes,

F. J. Dentener, M. C. Facchini, R. Van Dingenen, B. Ervens,Atmos. Chem. Phys. 2004, 5, 1053.

[3] B. Lamb, A. Guenther, D. Gay, H. Westberg, Atmos. Environ.1987, 21, 1695. doi:10.1016/0004-6981(87)90108-9

[4] A. Guenther, N. C. Hewitt, D. Erickson, R. Fall, C. Geron,T. Graedel, P. Harley, L. Klinger, J. Geophys. Res. 1995, 100,8873. doi:10.1029/94JD02950

[5] A. Guenther, C. Geron, T. Pierce, B. Lamb, P. Harley,R. Fall, Atmos. Environ. 2000, 34, 2205. doi:10.1016/S1352-2310(99)00465-3

[6] M. Claeys, B. Graham, G. Vas, W. Wang, R. Vermeylen,V. Pashynska, J. Cafmeyer, P. Guyon, Science 2004, 303, 1173.doi:10.1126/SCIENCE.1092805

[7] J. Joutsensaari, M. Loivamäki, T. Vuorinen, P. Miettinen,A.-M. Nerg, J. K. Holopainen, A. Laaksonen, Atmos. Chem.Phys. 2005, 5, 1489.

[8] R. Atkinson, J. Arey, Acc. Chem. Res. 1998, 31, 574. doi:10.1021/AR970143Z

[9] H. Hakola, J. Arey, S. M. Aschmann, R. Atkinson, J. Atmos.Chem. 1994, 18, 75. doi:10.1007/BF00694375

[10] J. Yu, D. R. Cocker III, R. J. Griffin, R. C. Flagan, J. H. Seinfeld,J. Atmos. Chem. 1999, 34, 207. doi:10.1023/A:1006254930583

[11] L. Tollsten, M. Müller, Phytochemistry 1996, 43, 759.doi:10.1016/0031-9422(96)00272-5

[12] H. Hakola, J. Rinne, T. Laurila, Atmos. Environ. 1998, 32, 1825.doi:10.1016/S1352-2310(97)00482-2

[13] S. Owen, C. Boissard, R. A. Street, S. C. Duckham, O. Csiky,C. N. Hewitt, Atmos. Environ. 1997, 31, 101. doi:10.1016/S1352-2310(97)00078-2

[14] J. Kesselmeier, M. Staudt, J. Atmos. Chem. 1999, 33, 23.doi:10.1023/A:1006127516791

[15] R. Atkinson, S. M. Aschmann, J. Arey, Atmos. Environ. 1990,24A, 2647.

[16] A. Reissell, C. Harry, S. M. Aschmann, R. Atkinson, J. Arey,J. Geophys. Res. 1999, 104, 13869. doi:10.1029/1999JD900198

[17] K. Griesbaum, V. Miclaus, Environ. Sci. Technol. 1998, 32, 647.doi:10.1021/ES970602R

[18] S. Koch, R. Winterhalter, E. Uherek, A. Kolloff, P. Neeb, G. K.Moortgat, Atmos. Environ. 2000, 34, 4031. doi:10.1016/S1352-2310(00)00133-3

[19] R. J. Griffin, I. D. R. Cocker, R. C. Flagan, J. H. Seinfeld,J. Geophys. Res. 1999, 104, 3555. doi:10.1029/1998JD100049

[20] M. Glasius, M. Lahaniati, A. Calogirou, D. Di Bella,N. R. Jensen, J. Hjorth, M. Duane, D. Kotzias, et al., inProc. Fifth EUROTRAC Symp. (Eds P. M. Borell, P. Borell)1998, p. 618 (WIT Press: Southampton).

[21] M. Glasius, M. Lahaniati, A. Calogirou, D. Di Bella,N. R. Jensen, J. Hjorth, D. Kotzias, B. R. Larsen, Environ.Sci. Technol. 2000, 34, 1001. doi:10.1021/ES990445R

[22] B. Warscheid, T. Hoffmann, Rapid Commun. Mass Spectrom.2001, 15, 2259. doi:10.1002/RCM.504

[23] J. F. Doussin, D. Ritz, R. Durand-Jolibois, A. Monod, P. Carlier,Analusis 1997, 25, 236.

[24] L. P. Thuener, P. Bardini, G. J. Rea, J. C. Wenger, J. Phys.Chem. 2004, 108, 11019.

[25] A. A. Chew, R. Atkinson, J. Geophys. Res. 1996, 101, 28649.doi:10.1029/96JD02722

[26] R. Gutbrod, S. Meyer, M. Muhibu Rahman, R. N. Schindler,Int. J. Chem. Kinet. 1997, 29, 717. doi:10.1002/(SICI)1097-4601(1997)29:9<717::AID-KIN9>3.0.CO;2-X

[27] B. Bonn, G. Schuster, G. Moortgat, J. Phys. Chem. 2002, 106,2869.

[28] D. R. Cocker III, S. L. Clegg, R. C. Flagan, J. H. Seinfeld,Atmos. Environ. 2001, 35, 6049. doi:10.1016/S1352-2310(01)00404-6

[29] A. S. Hasson, M. Y. Chung, K. T. Kuwata, A. D. Converse,D. Krohn, S. E. Paulson, J. Phys. Chem. 2003, 107, 6176.

[30] N. Carrasco, J. F. Doussin, P. Carlier, Atmos. Environ. 2006, 40,2011. doi:10.1016/J.ATMOSENV.2005.11.042

[31] L. Chiappini, E. Perraudin, R. Durand-Jolibois, J. F. Doussin,Anal. Bioanal. Chem. 2006 (In press).

[32] N. Carrasco, M. T. Rayez, J. C. Rayez, J. F. Doussin, Phys.Chem. Chem. Phys. 2006 (In press).

[33] J. Odum, T. Hoffman, F. Bowman, D. Collins, R. Flagan,J. H. Seinfeld, Environ. Sci.Technol. 1996, 30, 2580. doi:10.1021/ES950943+

[34] J. Yu, R. C. Flagan, J. H. Seinfeld, Environ. Sci. Technol. 1998,32, 2357. doi:10.1021/ES980129X

[35] E. O. Edney, T. E. Kleindienst, T. S. Conver, C. D. McIver,E. W. Corse, W. S. Weathers, Atmos. Environ. 2003, 37, 3947.doi:10.1016/S1352-2310(03)00461-8

[36] P. Aplincourt, M. F. Ruiz-Lopez, J. Am. Chem. Soc. 2000, 122,8990. doi:10.1021/JA000731Z

[37] P. Neeb, F. Sauer, O. Horie, G. K. Moortgat, Atmos. Environ.1997, 31, 1417. doi:10.1016/S1352-2310(96)00322-6

[38] G. E. Orzechowska, H. T. Nguyen, S. E. Paulson, J. Phys.Chem. 2005, 109, 5366.

[39] A. M. Jonsson, M. Hallquist, E. Ljungström, Environ. Sci.Technol. 2006, 40, 188.

[40] H. J. Tobias, K. S. Docherty, D. E. Beving, P. J. Ziemann,Environ. Sci. Technol. 2000, 34, 2116. doi:10.1021/ES991057S

[41] R. Winterhalter, P. Neeb, D. Grossmann, A. Kolloff, O. Horie,G. K. Moortgat, J. Atmos. Chem. 2000, 35, 165. doi:10.1023/A:1006257800929

[42] M. E. Jenkin, Atmos. Chem. Phys. 2004, 4, 1741.

295

L. Chiappini et al.

[43] M. E. Jenkin, D. E. Shallcross, J. N. Harvey, Atmos. Environ.2000, 34, 2837. doi:10.1016/S1352-2310(00)00087-X

[44] S. Gao, M. Keywood, N. L. Ng, J. Surratt, V. Varutbangkul,R. Bahreni, R. C. Flagan, J. H. Seinfeld, J. Phys. Chem. A 2004,108, 10147. doi:10.1021/JP047466E

[45] M. Kalberer, D. Paulsen, M. Sax, M. Steinbacher, J. Dommen,A. S. H. Prevot, R. Fisseha, E. Weigartner, Science 2004, 303,1659. doi:10.1126/SCIENCE.1092185

[46] U. Baltensperger, M. Kelberer, J. Dommen, D. Paulsen,M. R. Alfarra, H. Coe, R. Fisseha, A. Gascho, Faraday Discuss.2005, 130, 265. doi:10.1039/B417367H

[47] B. Picquet-Varrault, J. F. Doussin, R. Durand-Jolibois, O. Pirali,P. Carlier, C. Fittschen, Environ. Sci. Technol. 2002, 36, 4081.doi:10.1021/ES0200138

296