Atmospheric Chemistry and Physics Discussions Interactive comment on “ Oligomer formation during gas-phase ozonolysis of small alkenes and enol ethers : new evidence for the central role of the Criegee Intermediate as oligomer chain unit

The paper describes the identification of oligomers in secondary organic aerosols derived from the ozonolysis of different small alkenes and one enol-ether. Detailed mass spectrometric analysis was used to identify structures or at least the chain carriers of the oligomers. QTOF-MS/MS analyses deliver fragmentation pathways and FT-ICRMS and MS/MS were used to measure the elemental composition of oligomers and fragments. Results of the measurements were discussed very detailed and often compared to a previous paper of the authors. The authors conclude that stabilized Crigee

stituents of SOA from gas-phase ozonolysis of small enol ethers. These oligomers contained repeated chain units of the same chemical composition as the main Criegee Intermediates (CI) formed during the ozonolysis reaction, which were CH 2 O 2 (mass 46) for alkyl vinyl ethers (AVE) and C 2 H 4 O 2 (mass 60) for ethyl propenyl ether (EPE). In the present work, we extend our previous study (Sadezky et al., 2006) to another 10 enol ether (ethyl butenyl ether EBE) and a variety of structurally related small alkenes (trans-3-hexene, trans-4-octene and 2,3-dimethyl-2-butene).
Experiments have been carried out in a 570 l spherical glass reactor at atmospheric conditions in the absence of seed aerosol. SOA formation was measured by a scanning mobility particle sizer (SMPS). SOA filter samples were collected and chemically 15 characterized off-line by ESI(+)/MS-TOF and ESI(+)/MS/MS-TOF, and elemental compositions were confirmed by ESI(+)/MS/MS-FTICR. The results for all investigated unsaturated compounds are in excellent agreement with the observations of our previous study (Sadezky et al., 2006). Analysis of the collected SOA filter samples reveal the presence of oligomeric compounds in the mass range 200 to 800 u as major con-

Introduction
Organic material accounts for a substantial fraction of atmospheric fine particular matter that affects the global climate by direct and indirect effects as well as human health (Pöschl, 2005, and references therein). Secondary organic aerosol (SOA) is formed by gas-to-particle conversion of products of the tropospheric photooxidation of volatile organic compounds (VOC), and its global formation is estimated to range from 12 to 70 Tg y −1 (Kanakidou et al., 2005, and references therein). Understanding the chemical composition and formation processes of SOA is required for a quantitative assessment of its production, properties and environmental effects (Fuzzi et al., 2006).
Other suggested pathways leading to oligomer formation in organic atmospheric aerosol involve aqueous-phase reactions of pyruvic acid, a product of the atmospheric oxidation of isoprene, initiated by OH radicals (Altieri et al., 2006) or photolysis (Guz-5 man et al., 2006) within cloud droplets. Recently, formation of higher-molecular weight species was also observed for photooxidation and ozonolysis of tertiary alkylamines (Murphy et al., 2007).
In our recent study (Sadezky et al., 2006) we reported the discovery of oligomeric compounds by chemical analysis of secondary organic aerosol formed during ozonolysis of enol ethers using the off-line ESI/MS-TOF technique. These oligomers were found to consist of repetitive chain units, which have the same elementary compositions as the main Criegee Intermediates (CI) formed from these ozonolysis reactions: CH 2 O 2 (= CH 2 OO for C 1 -CI) of mass 46 for the alkyl vinyl ethers (AVE) and C 2 H 4 O 2 (= CH 3 CHOO for C 2 -CI) of mass 60 for ethyl propenyl ether (EPE). It is proposed that 15 these oligomers have the following basic structure of an oligoperoxide, -[CH(R)-O-O] n -, where R = H for the AVE and R=CH 3 for the EPE. We suggested a new pathway for secondary organic aerosol and oligomer formation involving gas-phase reactions involving stabilized CI, which lead to formation of oligoperoxidic chains carrying mostly three to four CI-like chain units. 20 Our present work is aimed at investigating the possibility that the correlation between the structure of the main Criegee Intermediate formed during alkene ozonolysis and the composition of formed SOA and oligomers described in our previous work (Sadezky et al., 2006) might apply to a wider range of unsaturated compounds. We therefore extend our previous study (Sadezky et al., 2006) to a wider variety of 25 small unsaturated compounds, among which are another enol ether, ethyl butenyl ether (EBE, C 2 H 5 OCH=CHC 2 H 5 ), and three symmetric hydrocarbon alkenes, trans-3-hexene (C 2 H 5 CH=CHC 2 H 5 ), 2,3-dimethyl-2-butene ((CH 3 ) 2 C=C(CH 3 ) 2 ), and trans-4-octene (C 3 H 7 CH=CHC 3 H 7 ). A gas-phase ozonolysis experiment of a mixture of Introduction EGU ethyl vinyl ether (EVE, C 2 H 5 O-CH=CH 2 ) and trans-3-hexene (C 2 H 5 CH=CHC 2 H 5 ) was also performed in order to investigate a possible formation of mixed oligomers that might contain combinations of chain units corresponding to the main CI formed during both reactions. Gas-phase ozonolysis experiments have been performed in a laboratory-reaction chamber under experimental conditions similar to those of Sadezky 5 et al. (2006). SOA formed during the reactions have been observed by a SMPS system and chemically characterized by a hybrid ESI(+)/Q-TOF and chemical composition was confirmed by accurate mass measurements with an ESI Fourier Transform Ion cyclotron resonance (FTICR) mass spectrometer. FTICR MS offers ultrahigh resolution and high sensitivity for the characterization of complex samples (e.g. Römpp et al.,10 2005).

Experimental
Experiments in the laboratory were performed in a 570-liter spherical glass reactor at room temperature in synthetic air at a total pressure of 730 Torr. A detailed description of this setup has been described in earlier publications (Neeb et al., 1998;Winterhal-15 ter et al., 2000). Ozone was produced by a mercury pen-ray lamp inside the reactor, prior to the addition of the mixture of the unsaturated compound and synthetic air (reaction start). The concentrations of reactants and reaction products were followed by Fourier Transform infrared spectroscopy (FTIR). The aerosol concentration and size distribution was monitored with a scanning mobility particle sizer (SMPS, TSI 3936) Introduction EGU to be generated during the ozonolysis of alkenes (Finlayson et al., 1972), cyclohexane (excess, 300 ppm) was added in some experiments. All chemicals were commercially available (purity >95%) and used without further purification. The aerosol formed in the laboratory experiments was collected during 20-25 min on Teflon (PTFE) filters (45 mm diameter, 0.45 µm pore size), using a flow rate of 5 14 L min −1 . After collection the filters were extracted in a 7 cm 3 glass flask with 3 ml pure methanol (HPLC grade), and stored at -20 • C until analysis. Chemical constituents were detected by a hybrid mass spectrometer (quadrupole and time-of-flight) QSTAR (Applied Biosystems MDS SCIEX) with an electrospray ion source. The extraction solution was directly injected (30 µl/min). The electrospray ion 10 source (TurboIonSpray) was operated in the positive mode at 400 • C and an ionization voltage of +3.4 kV. The declustering potential was 0 to + 30 V, and the focussing potential (focus ring) was + 100 V. For tandem Q-TOF experiments, the collision energy was between 10 and 30 eV with CAD (collision gas) set to 2. Instrument control, spectra treatment and calculations of elemental compositions were done with the software Introduction EGU forms the analogous CIs of the type C 3 H 7 CHOO (C 4 -CI, R'=C 3 H 7 ). The primary carbonyl compounds simultaneously formed are the corresponding aldehydes propanal (trans-3-hexene, R'=C 2 H 5 ) and butanal (trans-4-octene, R'=C 3 H 7 ).
2,3-Dimethyl-2-butene produces CIs of the type (CH 3 ) 2 COO (iso-C 3 -CI), which is 5 an isomer of the C 3 -CI formed from the ozonolyses of trans-3-hexene and EBE. The corresponding primary carbonyl compound is acetone. The CI formed from the decomposition of the primary ozonide are formed in excited states, which then either decompose into various products or become collisionally stabilized.

10
About 50-60% of the excited C 1 -CI are stabilized, while the yields of stabilized C 2 -CI and C 3 -CI are estimated to be between 20 and 40% per reacted alkene or enol ether (Sadezky, 2005;Kroll et al., 2002). The stabilization rate of the excited iso-C 3 -CI, however, is very low, as this type of CI decomposes by nearly 100% via the hydroperoxide channel (e.g. Rickard et al., 1999). Total SOA masses M 0 (µg/m 3 ) measured by SMPS after completion of the reaction, before the beginning of the filter sampling, are given in Table 1. Initial mixing ratios of reactants and of cyclohexane (C 6 H 12 ) added as OH scavenger, and the main types of CI formed from these ozonolysis reactions, as discussed in the previous section, are 20 also given. Measured total SOA masses M 0 formed during gas phase ozonolysis amount to several hundreds of µg/m 3 for most unsaturated compounds, except the 2,3-dimethyl-2-butene, which forms much lower amounts of SOA. The reason for this difference might be the very low stabilization rate (less than 1% of stabilized CI) of the disubsti- 25 tuted iso-C 3 -CI in comparison with the monosubstituted C 3 -CI and C 4 -CI (20% to 40% of stabilized CI) (Kroll et al., 2002;Rickard et al., 1999). A correlation between the amount of SOA formed and the stabilization rate of the CI independently of their num-14049 Introduction EGU ber of carbon atoms -as the disubstituted C 3 -CI and the monosubstituted iso-C 3 -CI are isomers -might indicate a key role of stabilized CI in SOA formation. A significant increase of the total SOA mass M 0 upon addition of an excess of cyclohexane (C 6 H 12 ) as an OH radical scavenger is observed for the two alkenes trans-3-hexene and 2,3-dimethyl-2-butene. A similar influence of C 6 H 12 on SOA yields has 5 been observed by Docherty et Ziemann (2003) for the ozonolysis of β-pinene, while the reverse effect was found for the ozonolysis of alkyl vinyl ethers (Sadezky et al., 2006). In all cases, however, the qualitative results obtained from chemical analysis of the SOA by mass spectrometry do not change in the presence of an OH scavenger. 800 for the enol ether EBE and between m/z 200 and 600 for the three alkenes with the typical regular structures of oligomers. The ion peaks could be grouped in series whose ions display regular differences of ∆ m/z=74 for ethyl butenyl ether (EBE), trans-3-hexene and 2,3-dimethyl-2-butene , and of ∆ m/z=88 for trans-4-octene (Fig. 2d). The results are consistent with those from analogous studies (Sadezky et 20 al., 2006;Sadezky, 2005) (Sadezky et al., 2006). The most intense series observed for each ether is coloured in red, and is assigned as oligomer of type (a). The other observed oligomer series usually appear with much 5 lower intensities and differ from the major oligomer series of type (a) by multiples and sums of ∆ m/z=16 and ∆ m/z=14. Oligomer series with similar differences of ∆ m/z towards the main oligomer series of type (a) are labelled with similar letters and colours in Table 2 and in Figs. 2a-d for the different alkenes and enol ethers. For example, for most enol ethers and alkenes, oligomer ions of type (b) are observed, coloured in green 10 in Table 2 and Figs. 2a-d. They differ from the series of type (a) ions by an additional ∆ m/z=16, possibly corresponding to an oxygen atom. While for trans-3-hexene and ethyl vinyl ether (EVE), only oligomers of type (a) and (b) are observed, a variety of other ions series appear for ethyl butenyl ether (EBE), trans-4-octene, 2,3-dimethyl-2-butene and ethyl propenyl ether (EPE). Mass differences of these oligomer series 15 towards the oligomers of type (a) are, for example, ∆ m/z=14 for oligomers (C) and (D) (with ∆ m/z 14 possibly corresponding to a CH 2 group), ∆ m/z=42 for oligomer (B) (∆ m/z 42=3 ×∆ m/z 14, possibly corresponding to three CH 2 groups), and ∆ m/z=30 for oligomer (E) (∆ m/z 30=∆ m/z 14 + ∆ m/z 16, eventually accounting in total for a formaldehyde-like unit CH 2 O).

20
As described in a later section of this work and in our previous study (Sadezky et al., 2006) Oligomer ions observed for EBE in the present work corroborate the dependence of their chain unit on the vinylic side =CHR' of the double bond of the initial enol ether derived from the analogous study of AVE and EPE (Sadezky et al., 2006). The schematic structure of an enol ether shown in Fig. 3 displays the two alkyl substituents, R' on the 5 vinylic side =CHR' of the double bond and R on its enolic side =CHOR.
R'=C 2 H 5 for EBE (C 2 H 5 OCH=CHC 2 H 5 ), while R'=H for AVE (ROCH=CH 2 ) and R'= CH 3 for EPE (C 2 H 5 OCH=CHCH 3 ). R' differs thus by mass 14 (CH 2 ) between the AVE and EPE as well as between EPE and EBE, a difference that is reflected in the masses of the observed oligomer chain units, which are of ∆m/z=46 for the AVE, ∆ m/z=60=46 10 + 14 for EPE and ∆ m/z=74=60 + 14 for EBE.
As discussed in the Sect. 3.1.1, the major Criegee Intermediate (CI) formed with yields around 80% during enol ether ozonolysis originates from the vinylic side of the enol ethers, thus carrying R' as substituent. The major CI produced from EBE is therefore the C 3 -CI with R'=C 2 H 5 , which is identical to the C 3 -CI formed during ozonolysis  Table 2 show that the oligomers obtained from ozonolysis of EBE and trans-3-hexene consist of chain units with identical mass 74. Moreover, chain units with similar mass 74 are also found for oligomers formed from 2,3-dimethyl-2-butene (Fig. 2c and Table 2), which produce the iso-C 3 -CI, an isomer of the C 3 -CI. Finally, trans-4-octene forms the C 4 -CI, 20 which differs by a CH 2 group of mass 14 from the C 3 -CI produced by EBE and trans-3-hexene. The chain units of the oligomers obtained from ozonolysis of trans-4-octene ( Fig. 2d and Table 2) reflect this difference by their ∆ m/z=88=74+14. These results point towards a decisive role of the CI in the formation of the oligomers observed as chemical constituents of the SOA from the different unsaturated compounds. 25 In our previous study, Sadezky et al. (2006) suggested furthermore that the alkoxy group OR on the enolic side of the initial ether is contained once in each oligomer ion. A comparison of the initial structures of EBE and trans-3-hexene shows that each symmetric side of the double bond of trans-3-hexene is identical to the vinylic side of Introduction  (Sadezky et al., 2006), while n=2 with the chain unit of mass 74 (EBE and trans-3hexene, Figs. 2a and b), and n=1 with the largest chain unit of mass 88 (trans-4-octene, Fig. 2d).
Although the chain units are of similar mass 74, lower degrees of oligomerization 15 are observed for oligomers formed from 2,3-dimethyl-2-butene, with n=1 for the most abundant oligomer molecule (Fig. 2c), than for those produced from EBE and trans-3hexene. This observation is consistent with the low total SOA masses formed during ozonolysis of this compound, and might be correlated with the low stabilization rate of the iso-C 3 -CI in comparison with the C 3 -CI, eventually leading to a lower fraction of CI 20 undergoing oligomerization reactions.

Fragmentation pathways of oligomer ions
The main fragment ions observed in the MS/MS spectra can be classified as being part of fragmentation pathways, formed by successive losses of the chain unit as neutral 5 mass 74 or 88, respectively. Two fragmentation pathways, marked in violet and red colour in Figs. 4a-d, can be identified for the type (a) oligomers from all unsaturated compounds. Two more pathways, marked in dark and light green colour, are observed for 2,3-dimethyl-2-butene ( Fig. 4c) only.
The results are consistent with MS/MS spectra of oligomer ions produced from gas 10 phase ozonolysis of AVE and of EPE, which are presented in our previous study (Sadezky et al., 2006). They show similar fragmentation pathways consisting of successive losses of the chain units 46 for AVE and 60 for EPE, corresponding to the respective oligomer chain unit. However, in comparison with oligomers formed from the unsaturated compounds investigated in the present work, fragmentation spectra of 15 oligomers formed from AVE or EPE are much more rich and complex, with up to nine fragmentation pathways being simultaneously observed for a parent oligomer ion of type (a) (Sadezky et al., 2006). Figure 5a-c schematically displays the main fragmentation pathways identified for type (a) oligomer ions from the three enol ethers EBE (C 2 H 5 OCH=CHC 2 H 5 ) (Fig. 4b), 20 EVE (C 2 H 5 OCH=CH 2 ) and EPE (C 2 H 5 OCH=CHCH 3 ) (Sadezky et al., 2006). Fragmentation pathways of oligomers from different enol ethers, which show apparent analogies among each other, are marked by similar colours, that is, violet, red or dark green. Thus, two fragmentation pathways, marked in violet and red, are identified for type (a) oligomers from all three enol ethers, while a third pathway, marked in dark 25 green, is observed for EVE and EPE only (Sadezky et al., 2006).
Each fragmentation pathway is initiated by loss of a specific neutral from the pseudomolecular ion, followed by successive fragmentations of neutrals corresponding to 14054 Introduction the oligomer (a) parent ion (Fig. 4b), but is deduced by subtraction of mass 74 from the next higher observed fragment ion 201 of the violet-marked pathway. A third pathway, marked in dark green colour and observed for EVE and EPE only, terminates on a fragment ion of m/z 99. For all fragmentation pathways, the mass of the neutral species initially fragmented from the pseudomolecular ion increases by ∆m/z=28 (possibly corresponding to twice a CH 2 group) between EVE and EPE, as well as between EPE and EBE. Simultaneously, the masses of the respective oligomer chain units increase by ∆m/z=14 (CH 2 ) with the substituents R' of the initial enol ethers and major CI (Figs. 1 and 3) comprising an additional CH 2 group.  Each fragmentation pathway differs between trans-3-hexene and trans-4-octene in a similar way as it does between EVE and EPE as well as between EPE and EBE, with the masses of the respective CI and oligomer chain units differing at the same ACPD 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis EGU time by ∆ m/z=14 (CH 2 group). Thus, for the violet-coloured fragmentation pathway, the mass of the initially fragmented neutral species increases by ∆ m/z=28 (possibly corresponding to twice a CH 2 group) from mass 120 (trans-3-hexene) to mass 148 (trans-4-octene). The neutral species whose fragmentation from the pseudomolecular ions initiates 5 the red-marked fragmentation pathways is of mass 162 for EBE and of mass 146 for trans-3-hexene. It carries thus an additional ∆ m/z 16 for EBE, which might represent the enolic oxygen atom that causes a similar difference between the pseudomolecular ions from EBE and trans-3-hexene themselves (Sect. 3.3.1). Fragment ions of the violet-marked fragmentation pathways differ by m/z=16 between trans-3-hexene and 10 EBE, indicating that the enolic oxygen, which is additionnally contained in the pseudomolecular ions formed from EBE, remains in the fragment ions. The neutral species initially fragmented from the pseudomolecular ions for the violet-marked fragmentation pathways are for both EBE and trans-3-hexene of similar mass 120. MS/MS spectra of type (a) oligomers formed from 2,3-dimethyl-2-butene (Figs. 4c 15 and 6b) show two fragmentation pathways, marked in red and violet colours, which are identical to those described for trans-3-hexene. However, in comparison with type (a) oligomers formed from trans-3-hexene, those from 2,3-dimethyl-2-butene show additional fragmentation pathways (Figs. 6a and b): one of these pathways, marked in dark green colour, shows analogies with a third fragmentation pathway identified for type 20 (a) oligomers from enol ethers (Fig. 5), which is observed for EVE and EPE, but not for EBE. This fragmentation pathway terminates on fragment ions of identical m/z 99 for EVE, EPE and 2,3-dimethyl-2-butene, and its fragment ions generally differ by ∆ m/z 14 from those of the red-marked fragmentation pathway of the same parent ion. A fourth pathway, marked in light green colour, is observed for the parent oligomer ions 25 from 2,3-dimethyl-2-butene, with its fragment ions differing by ∆ m/z 14 from those of the violet-marked fragmentation pathway. 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis   Tables 3ac show a strong and systematic preference of the elemental composition C 3 H 6 O 2 (calculated exact mass 74.0367 u) for the chain unit of mass 74. The other elemental 20 compositions for mass 74 which might make chemical sense as neutral molecules in order to represent the chain units, C 4 H 10 O and C 2 H 2 O 3 , usually show much larger mass errors. Similarly, accurate mass measurements yield the elemental composition C 4 H 8 O 2 (exact mass 88.0524 u) for the chain unit of mass 88 of oligomers formed from trans-4-octene (Table 3d). Analogous elemental compositions of CH 2 O 2 (exact mass 25 46.0055 u) for the chain unit 46 and C 2 H 4 O 2 (exact mass 60.0211 u) for the chain unit 60 for oligomers formed from AVE and EPE were found in our previous study (Sadezky et al., 2006). 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis

EGU
Further work is needed to elucidate the detailed structures of parent and fragment ions. For their chemical compositions, the elements C, H and O and a single positive charge are taken into account. The fact that good fragmentation spectra are obtained from pseudomolecular ions during MS/MS experiments disagrees with Na + adducts.
Principal structure of type (a) oligomers 5 Figures 4, 5 and 6 show that the fragmentation pathways characterized by the regular loss of the chain unit typically range from m/z =300 to 400, the masses of the pseudomolecular ions, down to m/z about 100 for the smallest fragment ions. The regularity of the fragmentation pathways and their analogy for oligomers observed from different enol ethers and alkenes indicate that all type (a) oligomers follow the same structural 10 principle. These observations confirm the general linear oligomeric structure that we suggested in our previous work (Sadezky et al., 2006) for type (a) oligomers from the ozonolysis of AVE and EPE. This structure can schematically be represented by a chain having a starting group "X" and an end group "Y" linked by several chain units of mass 46 (AVE), 60 (EPE), 74 (EBE, trans-3-hexene and 2,3-dimethyl-2-butene), or 88 15 (trans-4-octene).
An oligomer that contains n chain units of mass 74 is thus represented as X-(74) n -Y. Ionization by ESI(+) leads to the formation of the oligomer pseudomolecular ion [X-(74) n -Y+H] + . Thus, with n=2, this schematic structure might for example represent the pseudomolecular ions 395 (EBE) or 379 (trans-3-hexene).

20
The fragmentation initiates with a loss of a neutral species X or Y from one side of the pseudomolecular ion [X- (74)  EGU mentation pathway of the pseudo-molecular ions 395 (EBE) and 379 (trans-3-hexene), XH + or YH + is represented by the terminating fragment ions of m/z 85. Regarding the violet-coloured fragmentation pathway, the lowest fragment ions observed in the MS/MS spectra are of m/z 201 (EBE) and of m/z 185 (trans-3-hexene) and are not attributed to XH + or YH + , but to the next higher fragment ion [X-(74)+H] + or [Y-(74)+H] + .

5
The neutral species whose fragmentation from the pseudomolecular ions initiates the red-marked fragmentation pathways is of mass 162 for EBE and of mass 146 for trans-3-hexene. It carries thus an additional ∆ m/z 16 for EBE, which might represent the enolic oxygen atom that causes a similar difference between the pseudomolecular ions from EBE and trans-3-hexene themselves (Sect. 3.3.1). Fragment ions of the 10 violet-marked fragmentation pathways differ by m/z=16 between trans-3-hexene and EBE, indicating that the enolic oxygen, which is additionnally contained in the pseudomolecular ions formed from EBE, remains in the fragment ions. The neutral species initially fragmented from the pseudomolecular ions for the violet-marked fragmentation pathways are for both EBE and trans-3-hexene of similar mass 120.

15
The number of successive losses of the chain unit as a neutral molecule, which are directly observed in the MS/MS spectra, corresponds to the minimum number of chain units n comprised in the oligomer molecule. Values of n are given in Table 2 and in Figs. 2a-d.
In the given schematic representations, the proton ionizing the oligomer molecule is 20 suggested to remain attached to the fragment ions. However, this is an arbitrary decision for illustration purposes, as the determination of the exact fragmentation mechanism needs further investigation.

Isomers
The MS/MS spectra of the type (a) oligomer ions formed from the unsaturated com- 25 pounds studied in the present work are much less complex than those described for the enol ethers AVE and EPE in our previous study (Sadezky et al., 2006). While between two and four fragmentation pathways are observed for each type ( EGU ion in the present work, up to nine fragmentation pathways were simultaneously identified for each parent oligomer ion formed from AVE and EPE. In our previous study by Sadezky et al. (2006), we attributed the richness and complexity of the fragmentation spectra for AVE and EPE to the existence of several isomers of the linear type (a) oligomer parent ion. Therefore, in the MS/MS spectrum of a pseudomolecular ion 5 of a certain m/z, the fragmentation pathways of all its isomers appear simultaneously, leading to the observed large number of different fragmentation pathways. The much smaller number of fragmentation pathways observed in the present work therefore indicates that a much smaller variety of isomers exists for these oligomer parent ions. The reason of this might be the larger size of the oligomer chain units, 10 which are of mass 74 (C 3 H 6 O 2 for EBE, trans-3-hexene and 2,3-dimethyl-2-butene) and mass 88 (C 4 H 8 O 2 for trans-4-octene) in the present work compared to mass 46 (CH 2 O 2 for AVE) and mass 60 (C 2 H 4 O 2 for EPE) in our previous study (Sadezky et al., 2006). With increasing size, the chain units eventually become more sterically hindered, an effect that might exclude the formation of some isomers and moreover 15 reduce the total degree of oligomerization (Sect. 3.3.1).
Interestingly, MS/MS spectra of type (a) oligomers formed from 2,3,-dimethyl-2butene show at least four different fragmentation pathways (Figs. 4c and 6b), while those of type (a) oligomers from both EBE and trans-3-hexene (Figs. 4a and b, 5c and 6a) contain only two fragmentation pathways, marked in red and violet colours. 20 However, the chain units of oligomers from all three compounds are of similar mass 74 and chemical composition C 3 H 6 O 2 . A main difference between the compounds is in the structures of their CI: 2,3-dimethyl-2-butene leads to formation of the iso-C 3 -CI ((CH 3 ) 2 COO), while the major CI produced by both EBE and trans-3-hexene is the C 3 -CI (C 2 H 5 CHOO). When we assume that the oligomer chain units 74 (C 3 H 6 O 2 ) 25 from EBE and trans-3-hexene and those from 2,3-dimethyl-2-butene are isomers in the same way as the respective CI formed during the ozonolysis reaction, then the chain unit that is substituted, like the iso-C 3 -CI, by two CH 3 groups is expected to be more compact and therefore less sterically hindered than the chain unit that is substituted, Introduction EGU like the C 3 -CI, by a linear alkylic chain R'= C 2 H 5 . Therefore, oligomers with chain units of mass 74 substituted corresponding to the iso-C 3 -CI might be able to form a wider variety of isomers than those with chain units of similar mass 74 that are substituted corresponding to the C 3 -CI. For example, the fragmentation pathway marked in dark green colour (Figs. 4, 5, and 6), which gener-5 ally terminates on fragment ions of m/z 99, is observed for the type (a) oligomer ions formed from 2,3-dimethyl-2-butene (iso-C 3 -CI, (CH 3 ) 2 COO) (Figs. 4c and 6b), but not for those from EBE and trans-3-hexene (C 3 -CI, C 2 H 5 CHOO). Furthermore, an analogous fragmentation pathway is observed for type (a) oligomers formed from AVE (C 1 -CI, CH 2 OO) (Fig. 5a) and EPE (C 2 -CI, CH 3 CHOO) (Fig. 5b), but not for those from 10 trans-4-octene (C 4 -CI, C 3 H 7 CHOO) (Figs. 4d and 6c). Assuming that the oligomer chain units are generally substituted in a similar way as the respective major CI, chain units from EBE, trans-3-hexene and trans-4-octene supposedly comprise a linear alkyl substituent R'= C 2 H 5 (EBE and trans-3-hexene) or R'= C 3 H 7 (trans-4-octene). Chain units from 2,3-dimethyl-2-butene, AVE and EPE, however, carry only CH 3 substituents 15 (2,3-dimethyl-2-butene and EPE) or are not at all substituted (AVE). The dark greencoloured fragmentation pathway might thus originate from an isomer of the pseudomolecular ion, which is formed only for small chain units similar to the iso-C 3 -CI or less bulky (C 1 -CI, C 2 -CI), but might not be stable for sterical reasons for chain units comprising a linear alkyl substituent with two or more CH 2 groups (C 3 -CI, C 4 -CI).

Oligoperoxidic structure and role of CI for oligomer formation
These results corroborate the decisive role of the Criegee Intermediates in the formation of oligomers, which was already suggested in our previous study (Sadezky et al., 2006

EGU
(=C 3 H 7 CHOO for C 4 -CI) of mass 88 (trans-4-octene). In our previous study (Sadezky et al., 2006), we suggested the basic structure of an oligoperoxide for the oligomers identified from ozonolysis of the enol ethers AVE and EPE. We extend this suggestion now to the oligomers formed from the diverse unsaturated compounds studied in the present work, as shown in Fig. 7.

5
Oligomers with suggested similar structures are known so far only from liquid-phase ozonolysis reactions (Barton et al., 2005(Barton et al., , 2004Lockley et al., 2001; and references therein). Barton et al. (2004) identified by ESI/MS oligoperoxides with m/z up to 900 u as reaction products from liquid-phase ozonolysis of 2,3-dimethyl-2-butene ("tetramethyl ethylene" TME) in pentane as a nonparticipating solvent at -60 • C. The peroxidic 10 chain link of the mass peaks was of mass 74 and thus corresponded to the CI formed during ozonolysis of 2,3-dimethyl-2-butene, referred to as acetone carbonyl oxide by the authors and as iso-C 3 -CI (CH 3 ) 2 COO in this work. The authors observed up to 12 chain units for different oligoperoxides. These observations are in excellent agreement with our results for ozonolysis reactions of enol ethers and small alkenes in the 15 gas phase described in the present work and in our previous study (Sadezky et al., 2006). By similar experimental methods, Barton et al. (2005) characterized oligoperoxides formed from the liquid-phase ozonolysis of 3-methyl-2-pentene, which consisted of units of the major CI, referred to as butanone carbonyl oxide (C 4 -CI, C 3 H 7 CHOO, mass 88), randomly copolymerized with units of the corresponding primary carbonyl 20 compound, acetaldehyde (mass 44). The observations of oligomeric peroxides composed of CI-like chain units are again in very good agreement with our studies and with Barton et al. (2004). However, a random incorporation of the corresponding primary aldehydes into those oligomer chains, as was observed by Barton et al. (2005), could not be confirmed in our experiments. 25 For gas-phase ozonolysis, various mechanisms involving secondary reactions of Criegee Intermediates have been discussed in the literature as formation pathways of major SOA constituents. Unimolecular decomposition of Criegee Intermediates via the "hydroperoxide" and the "hot acid" channel has been suggested as an important EGU pathway to form low-volatile carboxylic acids, diacids, and peroxidic compounds, such as diacylperoxides, peracids and other multifunctional hydroperoxides, during reactions with ozone of cyclic monoterpenes and cycloalkenes consisting of at least six carbon atoms (Tolocka et al., 2006;Docherty et al., 2005;Keywood et al., 2004;Ziemann, 2002;Koch et al., 2000). Moreover, reactions of stabilized Criegee Intermediates lead 5 to a wide variety of products containing peroxide moieties, several of which were identified in the past as SOA constituents from alkene ozonolysis. The most important reactions among them are those of the stabilized CI with carbonyl compounds to form secondary ozonides and with hydroxylic functionalities of carboxylic acids, alcohols and water to form α-acyloxyalkyl hydroperoxides, α-alkyloxyalkyl hydroperoxides and 10 α-hydroxyalkyl hydroperoxides, respectively. Such products were particularly identified in SOA from ozonolysis reactions of large linear alkenes, such as 1-tetradecene, oleic acid and methyl oleate (Zahardis and Petrucci, 2007, and references therein; Tobias et al., 2000;Tobias and Ziemann, 2000). The presence of secondary ozonides and α-alkyloxyalkyl hydroperoxides has also been reported in SOA formed from cyclic 15 monoterpenes and cycloalkenes containing six to ten carbon atoms (Tolocka et al., 2006;Ziemann, 2003). Some of the peroxidic products have been observed to undergo further association reactions with other molecules in order to form high-molecular weight compounds either in the liquid phase within the SOA particles or heterogeneously on the particle surface. EGU SOA constituents in our experiments. Key species of the mechanisms suggested to lead to SOA formation by unimolecular decomposition of CI are acyl radicals, whose structures are based on the original CI formed from cyclic alkenes and monoterpenes and therefore consist of at least five to six C atoms with additional carbonyl or hydroxy functionalities. These acyl radicals are first transformed into acylperoxy radicals by addition of O 2 , which then either react with HO 2 and RO 2 to form low-volatile acids, diacids, peracids and other peroxides (Docherty et al., 2005;Keywood et al., 2004) or cross-react to form diacylperoxides (Ziemann, 2002). However, for the small linear alkenes and enol ethers investigated in this work and in our previous study (Sadezky et al., 2006), the same mechanisms of unimolecular CI decomposition only lead to acyl 10 radicals of the type R'CO, with R'=H, CH 3 , C 2 H 5 , or C 3 H 7 . By further reaction with O 2 , HO 2 and RO 2 , the acyl radicals R'CO form mainly volatile products consisting of four or less C atoms or immediately decompose, while their recombination by cross-reaction (Ziemann, 2002) leads to dimers only and cannot explain the origin of the observed oligomeric chains.

15
Similarly, peroxidic SOA constituents formed by reactions of the stabilized CI, such as secondary ozonides, α-acyloxyalkyl hydroperoxides and α-alkyloxyalkyl hydroperoxides, involve CI that consist of nine, ten or 13 carbon atoms for the ozonolysis of linear alkenes such as oleic acid (Zahardis and Petrucci, 2007), methyl oleate (Mochida et al., 2006), and 1-tetradecene (Tobias et al., 2000;Tobias and Ziemann, 2000), and 20 of six to ten carbon atoms for cyclic alkenes (Ziemann, 2003) and the monoterpene α-pinene (Tolocka et al., 2006). By their secondary reactions, these stabilized CI often become associated to aldehydes or carboxylic acids of similar sizes, such as nonanoic acid (oleic acid/ozone system, Zahardis and Petrucci, 2007;Zahardis et al., 2006), tridecanal (1-tetradecene ozonolysis, Tobias et al., 2000), or pinonaldehyde (α-pinene 25 ozonolysis, Tolocka et al., 2006), leading to large, low-volatile peroxidic compounds consisting of 18 and more carbon atoms. However, the CI formed from the small unsaturated compounds investigated in our present work and in our previous study (Sadezky et al., 2006) consist of one to four carbon atoms only. Known gas-phase reaction path- EGU ways following ozonolysis do not lead to larger carbonyl, carboxylic or hydroxylic products either. Therefore, analogous peroxidic products of reactions of the stabilized CI with aldehydes and carboxylic acids formed in the alkene/ozone reaction system are mainly expected to be volatile. Peroxyhemiacetals formed by subsequent reaction of these compounds with aldehydes within the same reaction system, as described by Zahardis and Petrucci (2007), and references therein, and by Docherty et al. (2005), and references therein, do not account for the observed oligomers with regular CI-like peroxidic chain units either. Moreover, oligomerization of these reaction products by mechanisms similar to those described for the oleic acid/ozone reaction by Zahardis and Petrucci (2007) and Zahardis et al. (2006) should be very limited for the ozonolysis 10 of the compounds studied in our work. These mechanisms require the presence of free carbonyl or hydroxylic functionalities on the peroxidic products, which might allow them to further react with other stabilized CI. However, the CI incorporated into the oligomeric molecules carry hydrogen atoms and alkyl groups, with R'=H, CH 3 , C 2 H 5 , or C 3 H 7 . Given the linearity, small sizes and simple structures of the original unsatu-15 rated compounds, the main carbonyl, carboxylic and hydroxylic products formed in the reaction system also comprise a single functionality and a non-reactive group derived from R' (or from OR for the enol ethers) rather than being multifunctional. Therefore, oligomerization mechanisms based on the participation of additional reactive functional groups on R' (or OR) cannot explain the observed formation of oligomeric chains con-20 sisting of CI-like units in our present work. Still there is mechanistic evidence that the chemistry of stabilized CI plays a decisive role in oligomer and SOA formation in our reaction system. CI formed from the ozonolysis of small simple alkenes in the gas phase and stabilized by collision with inert gas molecules have been shown to undergo a variety of reactions similar to those 25 known since a long time from liquid-phase chemistry (Calvert et al., 2000, and references therein;Bunnelle, 1991). Reactions of the stabilized C 1 -CI and C 2 -CI have been thoroughly investigated for gas-phase ozonolysis of alkyl vinyl ethers (AVE) and ethyl propenyl ether (EPE) (Sadezky, 2005), ethene (Neeb et al., 1998(Neeb et al., , 1997, 2-butene ACPD 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis  (Horie et al., 1997) and propene (Neeb et al., 1996). A variety of peroxidic products analogous to those described above for larger linear alkenes were identified in these studies: thus, the stabilized C 1 -CI and C 2 -CI react with formaldehyde and acetaldehyde to form secondary ozonides and with the hydroxylic functionalities of water and formic acid to produce the hydroperoxides hydroxymethyl and hydroxyethyl hydroper-5 oxide, as well as hydroperoxymethyl and hydroperoxyethyl formate, respectively. All these products are however highly volatile due to their low molecular weights and were identified only in the gas phase of the alkene/ozone reaction systems. We reported in our previous studies (Sadezky et al., 2006;Sadezky, 2005) that SOA yields from ozonolysis of ethyl vinyl ether (EVE) and ethyl propenyl ether (EPE) were drastically reduced, eventually suppressed, in the presence of an excess of gaseous HCOOH. Simultaneously, oligomers were not detected onto the filter samples taken from these experiments (Sadezky et al., 2006;Sadezky, 2005). SOA formation was also shown to be strongly reduced in the presence of water vapour (Sadezky, 2005). HCOOH, and to a lesser extent H 2 O, are known to act as efficient scavengers for stabilized CI through 15 the above-mentioned formation of hydroperoxy formates and hydroxy hydroperoxides. Consequently, the key reaction mechanisms leading to SOA and oligomer formation are in direct concurrence with the known reactions of the stabilized CI with HCOOH or H 2 O to form low-molecular, volatile hydroperoxides. The observations point towards a direct implication of stabilized CI in the SOA and oligomer formation.

20
Therefore, we suggested in our previous study (Sadezky et al., 2006) that oligoperoxide formation is another, so far unknown gas-phase reaction of stabilized CI, which contributes to SOA formation. In our present work, we show that the observed formation of oligomers consisting of repeated CI-like chain units as main SOA constituents is not limited to enol ethers, but occurs in an analogous way for short-chain hydrocarbon 25 alkenes.
To our knowledge, besides our previous study by Sadezky et al. (2006), formation of stable reaction products containing several linked CI-like units has not been reported previously for gas-phase ozonolysis. So far, such reactions have been observed for ACPD 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis EGU alkene-ozone reactions in the liquid phase only. Thus, recombination of two CI leading to the formation of cyclic geminal diperoxides (1,2,4,5-tetroxanes) was described as a minor reaction of stabilized CI during the ozonolysis of oleic acid (Zahardis and Petrucci, 2007;Reynolds et al., 2006;Zahardis et al., 2005). Oleic acid-ozone reactions in these studies take place in the condensed phase either in bulk solution with high 5 reactant concentrations (Reynolds et al., 2006)  Linear oligomerization of CI leading to stable long-chain oligoperoxides was so far only reported by Barton et al. (2005) and Barton et al. (2004) for liquid-phase ozonolysis of the hydrocarbon alkenes 2,3-dimethyl-2-butene ((CH 3 ) 2 C=C(CH 3 ) 2 ), 3-methyl-2-20 pentene, and trans-2-hexene. As mentioned before, oligoperoxides identified as reaction products of the ozonolysis of 2,3-dimethyl-2-butene in the liquid phase by Barton et al. (2004) and in the gas phase in our present work show the same repetitive chain link of mass 74, which corresponds to the only Criegee Intermediate (CH 3 ) 2 COO (iso-C 3 -CI) formed by this alkene. Oligoperoxides identified by Barton et al. (2005) and 25 Barton et al. (2004) are mainly cyclic hexamers or pentamers, with minor products consisting of linear chains endcapped by -OOH or -OH groups. Moreover, these oligomeric products were found by Barton et al. (2004) to be thermally stable when extracted and refluxed in methanol for a few hours, a behaviour that agrees with their Introduction EGU observed stability in methanol used as solvent for the filter extraction in our work. Furthermore, Barton et al. (2004) observed a strong competition among the described formation of oligoperoxides consisting of CI-like units and well known liquid-phase reactions of the CI with compounds carrying hydroxylic functionalities to form small hydroperoxides (see above; Bunnelle, 1991). The authors reported oligoperoxide forma-5 tion during liquid-phase ozonolysis of 2,3-dimethyl-2-butene to take place only in nonprononated solvents, such as pentane, while in methanol, the corresponding hydroperoxide, 2-methoxy-2-propyl hydroperoxide, was formed instead through reaction of the CI, (CH 3 ) 2 COO (iso-C 3 -CI), with MeOH. These observations of liquid-phase chemistry are in analogy to the suppression of SOA and oligomer formation during gas-phase 10 enol ether ozonolysis in the presence of excess HCOOH or H 2 O, as described in our previous work (Sadezky et al., 2006). The mechanistic analogies indicate that oligoperoxide formation takes place in similar ways in the liquid phase and in the gas phase, thus also confirming our suggestion by Sadezky et al. (2006) that the stabilized form of CI might be involved in the gas-phase reaction.
Addition of a CI to the double bond of the initial alkene, leading to formation of an 1,2dioxolane, represents another reaction pathway of CI known from liquid-phase alkene ozonolysis (Bunnelle, 1991), in particular for vinyl ethers (Keul et al., 1985). A similar mechanism as a minor reaction pathway of stabilized CI has been attributed to products identified in the oleic acid/ozone and methyl oleate/ozone reaction systems (Mochida 20 et al., Zahardis et al., 2006). Decomposition of the 1,2-dioxolane then leads to an aldehyde derived from the CI and a ketone representing the oxygenated initial alkene. For the small unsaturated ethers and alkenes studied in our work, these products are expected to be volatile, and consequently, this reaction pathway is not expected to contribute to SOA formation. 25 As mentioned in Sects. 3.2 and 3.3.1 of the present work, a comparison of the SOA mass and oligomer chain lengths obtained from ozonolysis of trans-3-hexene and EBE on the one hand, and 2,3-dimethyl-2-butene on the other hand, points towards a key role of stabilized CI in the SOA formation as well. The very low stabilization rate of the ACPD 7, 2007 Oligomer and SOA formation from gas-phase alkene ozonolysis EGU iso-C 3 -CI ((CH 3 ) 2 COO), which is less than 1% in the gas phase, correlates with much lower SOA mass and lower degrees of oligomerization from 2,3-dimethyl-2-butene in comparison with those from EBE and trans-3-hexene, which involve the isomer C 3 -CI (C 2 H 5 CHOO) with stabilization rates between 20% and 40% (Kroll et al., 2002;Rickard et al., 1999).
However, the influence of excess cyclohexane added as OH radical scavenger on SOA mass described in Sect. 3.2 indicates also a contribution of radical chemistry to SOA formation, in addition to the chemistry of stabilized CI. The increase of SOA mass for the symmetric hydrocarbon alkenes trans-3-hexene and 2,3-dimethyl-2-butene in the presence of excess cyclohexane observed in our present work agrees with similar tendencies of formed SOA mass from β-pinene reported by Docherty and Ziemann (2003). Small vinyl ethers investigated in our previous work (Sadezky et al., 2006) and cyclohexene studied by Keywood et al. (2004), however, follow the opposite trend. Keywood et al. (2004) and Docherty and Ziemann (2003) explain the influence of cyclohexane as an OH radical scavenger by a decrease of the ratio [hydroperoxy radi-15 cals]/[organic peroxy radicals] in the reaction system, which, according to Keywood et al. (2004), especially affects the production pathways and chemistry of acyl radicals considered as key species in the formation of low-volatile compounds during cyclohexene and β-pinene ozonolysis. However, for the alkenes and enol ethers studied in our present and previous work (Sadezky et al., 2006), acyl radicals are not expected to 20 contribute to the formation of high-molecular oligomers by any of their reaction pathways known so far, as discussed before in this section.
The detailed reaction mechanism of the observed oligomer formation in the gas phase, as well as the structures and precise identities of the starting and end groups "X" and "Y", remain so far an enigma and need further investigation. The observations 25 of our present work allow us to identify the oligomer chain units and thus to confirm the systematic correlation between their chemical composition and that of the major CI formed during ozonolysis of the respective alkene. The entire oligomer structures, however, appear to be more complicated than those identified from liquid-phase ozonolysis ACPD 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis  Barton et al. (2004) and Barton et al. (2005), which are either linear with -OOH or -OH as starting and end groups "X" and "Y", or cyclic without any fragments accounting for other than the repeated chain units. The numerous common features among the structures and chemical compositions of the oligomers originating from the different alkenes and enol ethers concerning not only the CI-like repeated chain units, but 5 also the fragment ions and neutrals representing the starting and end groups X and Y (see Sect. 3.3.2), indicate a common mechanism of oligomer formation in the gas phase. The observations in the present work and in our previous study (Sadezky et al., 2006) cannot be explained by any classical reaction mechanisms established so far and must involve some new chemistry. The obvious key role of stabilized CI in SOA 10 and oligomer formation, in combination with the observed influence of cyclohexane as OH radical scavenger on SOA yields, might, for example, point towards a mechanism based on so far unknown CI-radical reactions.

Copolymerization of chain units from different ozonolysis reactions
In order to investigate the role of the CI in relation to the role of the initial unsaturated 15 compounds in the formation of oligomers, simultaneous gas-phase ozonolysis of trans-3-hexene and EVE was performed in the atmospheric simulation chamber. The MS spectrum of the SOA collected during this experiment is shown in Fig. 8. Several of the most intense ion peaks are identified as oligomer ions consisting of CI-like chain units. These ion peaks are marked by red and green arrows in Fig. 8.

20
Oligomer ions marked by red arrows are of m/z similar to those of oligomer (a) ions formed during ozonolysis of trans-3-hexene solely ( Fig. 2b and Table 2). According to our results presented in the Sect. 3.3, these ions might therefore comprise only chain units of mass 74 (C 3 H 6 O 2 ), with n=1 and n=2 for the two most intense ions 305 and 379 in Fig. 8. Oligomer ions marked by green arrows are deduced from ions marked EGU 305 + 2×∆ m/z 46), and m/z 425 (m/z 425=m/z 379 + ∆ m/z 46) (Fig. 8). Some ions between m/z 150 and 300 are identified from MS/MS experiments as fragment ions of higher parent ions and are listed separately in Fig. 8.
MS/MS spectra of the pseudomolecular ions 305 (marked by a red arrow in Fig. 8) and 351 (marked by a green arrow in Fig. 8) are exemplarily given in Figs. 9a and b, and 5 the observed main fragmentation pathways of these two parent ions are schematically displayed in Figs. 10a and b.
The MS/MS spectrum of the parent ion 305 (Figs. 9a and 10a) confirms that its structure is indeed similar to those of the type (a) oligomer ions formed during gasphase ozonolysis of trans-3-hexene solely, which are described in Sect. 3.3 of this work. Its fragmentation pathways are identical to those of its higher homologue 379 observed from ozonolysis of trans-3-hexene solely (Figs. 4a and 6a), and show the presence of one chain unit 74 (C 3 H 6 O 2 ), as expected ( Fig. 2b and Table 2).
The MS/MS spectrum of the parent ion 351 (Figs. 9b and 10b) reproduces the fragmentation pathways observed for the parent ion 305 (Figs. 9a and 10a), but includes 15 an additional loss of a neutral of mass 46. This result shows that the parent ion 351, and more generally all pseudomolecular ions marked by green arrows in Fig. 8, correspond to ions initially formed from ozonolysis of trans-3-hexene (marked by red arrows in Fig. 8) that contain additional chain units of mass 46.
The formation of such ions signifies that copolymerization of the two CI-like chain 20 units 74 (C 3 H 6 O 2 , C 3 -CI from trans-3-hexene) and 46 (CH 2 O 2 , C 1 -CI from EVE) takes place. The two chain units arise from CI that are simultaneously present in the reaction system, but are formed independently during gas-phase ozonolysis of trans-3-hexene and EVE. The observed copolymerization of the two chain units under these experimental conditions indicates that oligomer formation might be based on reactions of the 25 free CI itself after complete dissociation of the primary ozonide. Copolymerization of CI-like chain units formed from different alkenes was also observed by Barton et al. (2005) for the simultaneous ozonolysis of 2,3-dimethyl-2butene and 3-methyl-2-pentene in the liquid phase. The authors identified a variety EGU of oligomeric molecules, which follow similar structural principles as those obtained by separate ozonolysis of each alkene (see Sect. 3.3.3), but contain a mixture of the chain units formed by both unsaturated compounds, that is, acetone carbonyl oxide (iso-C 3 -CI, (CH 3 ) 2 COO) from 2,3-dimethyl-2-butene, and butanone carbonyl oxide (C 4 -CI, C 3 H 7 CHOO) and acetaldehyde from 3-methyl-2-pentene.

4 Conclusions
In the present work, we investigate secondary organic aerosol ( (Sadezky et al., 2006). The results of our present work corroborate our findings of Sadezky et al. (2006), in particular the essential role of Criegee Intermediates (CI), initially formed in the gas-phase, in oligomer and SOA formation: oligomer chain units are shown to systematically have the same chemical composition as the respective main Criegee Intermediate (CI) formed 20 by the gas-phase ozonolysis reaction. Fragmentation spectra (MS/MS) of oligomer pseudomolecular ions furthermore reveal strong structural analogies among oligomers formed from ozonolysis of different alkenes and enol ethers, thus indicating a common structural principle and formation mechanism.
We suggested in our previous study (Sadezky et al., 2006) that CI in their stabilized 25 form undergo oligomerization in the gas-phase according to a new, so far unknown reaction mechanism that leads to the formation of oligoperoxidic chains. In our present EGU work, we show that our observations by Sadezky et al. (2006) are not limited to enol ethers, but analogously apply to a variety of short-chain hydrocarbon alkenes. Therefore, we propose oligoperoxide formation as a new, general reaction pathway of stabilized CI in the gas phase, eventually with participation of CI-radical reactions. The detailed reaction mechanism, however, is so far unknown and needs further investiga-5 tion. Moreover, we report an example for copolymerization of different CI independently produced during simultaneous gas-phase ozonolysis of two different unsaturated compounds, leading to the formation of oligomers with mixed CI-like chain units. These observations further point towards a direct participation of free CI in the oligomerization 10 mechanism, which is independent of their formation process from the original unsaturated compound.
The smallest unsaturated organic compound shown to produce SOA during its gasphase ozonolysis under atmospheric conditions is methyl vinyl ether (MVE) with three carbon atoms (Klotz et al., 2004). Before, it was widely accepted in the literature, that 15 seven or more C atoms are required from non-cyclic unsaturated hydrocarbons, six for cyclic unsaturated species to initiate SOA formation from ozonolysis (Seinfeld and Pankow, 2003). The classical reaction pathways of gas-phase ozonolysis established for such large compounds, which lead to low-volatile products identified as SOA constituents, however, cannot explain SOA formation from small and structurally simple 20 unsaturated reactants, such as MVE.
While Klotz et al. (2004) could not give any clues about the chemical composition and formation mechanism of the SOA formed from ozonolysis of methyl vinyl ether (MVE), the oligomeric compounds newly discovered and characterized in our previous work (Sadezky et al., 2006)  EGU reaction pathway that mainly depends on the nature of the Criegee Intermediate (CI). The high-molecular oligoperoxides formed in the gas phase by the new mechanism, which we suggested to involve the stabilized form of CI (Sadezky et al., 2006), might then initiate nucleation, and thus lead to the observed formation of new SOA particles. It thus represents a new way to SOA formation especially from gas-phase ozonolysis 5 of unsaturated compounds with few carbon atoms and simple structures.
According to Klotz et al. (2004), new SOA particle formation from methyl vinyl ether (MVE) occurs only during its gas-phase ozonolysis, but not by its reactions with OH or NO 3 radicals, the two other major pathways of atmospheric degradation. This observation further supports the key role of the CI in SOA formation as an intermediate 10 species specific to ozonolysis. In the troposphere, stabilized CI are expected to mainly react with water vapour, a reaction that has been shown to be in direct concurrence with oligomer and new SOA particle formation during the ozonolysis of enol ethers (Sadezky et al., 2006;Sadezky, 2005). The relevance of the described oligoperoxide formation for SOA formation in the atmosphere has therefore to be further investigated.  Table 1. Total SOA masses M 0 (µg/m 3 ) formed, initial mixing ratios of reactants and cyclohexane (C 6 H 12 ), and types of major CI formed in the gas-phase ozonolysis reactions of the unsaturated compounds studied in this work.           ACPD 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis  Fig. 4c. SOA formed during the gas phase ozonolysis of 2,3-dimethyl-2-butene (initial mixing ratios: 8 ppm ozone, 15 ppm 2,3-dimethyl-2-butene, 300 ppm cyclohexane): ESI(+)/MS/MS-TOF spectrum of the parent ion 379 of oligomer (a). 7,2007 Oligomer and SOA formation from gas-phase alkene ozonolysis