Yields of β-hydroxynitrates, dihydroxynitrates, and trihydroxynitrates formed from OH radical-initiated reactions of 2-methyl-1-alkenes

Transcription

1 Yields of β-hydroxynitrates, dihydroxynitrates, and trihydroxynitrates formed from OH radical-initiated reactions of 2-methyl-1-alkenes Aiko Matsunaga a, and Paul J. Ziemann a,b,1 a Air Pollution Research Center, Department of Chemistry, and b Department of Environmental Sciences, University of California, Riverside, CA Edited by Barbara J. Finlayson-Pitts, University of California, Irvine, CA, and approved November 6, 2009 (received for review September 17, 2009) Yields of β-hydroxynitrates, dihydroxynitrates, and trihydroxynitrates, in particles formed from OH radical-initiated reactions of C 9 -C 15 2-methyl-1-alkenes in the presence of NO x were measured by using a thermal desorption particle beam mass spectrometer coupled to a high-performance liquid chromatograph with a UVvisible (UV-vis) detector. Yields of β-hydroxynitrates and dihydroxynitrates increased with carbon number primarily due to enhanced gas-to-particle partitioning before reaching plateaus at C 14 -C 15, where the compounds were essentially entirely in the particle phase. Plateau yields of β-hydroxynitrates, dihydroxynitrates, and trihydroxynitrates were , , and , and, after normalization for OH radical addition to the C ¼ C double bond, were , , and The fractions of 1-hydroxy and 2-hydroxy β-hydroxynitrate isomers were Yields measured here and in our previous study of reactions of linear internal alkenes and linear 1-alkenes indicate that, for these alkene classes, the relative branching ratios for forming tertiary, secondary, and primary β-hydroxyalkyl radicals by OH radical addition to the C ¼ C double bond are , and the branching ratios for forming β-hydroxynitrates from reactions of tertiary, secondary, and primary β-hydroxyperoxy radicals with NO are 0.25, 0.15, and The effects of H 2 O vapor and NH 3 on yields were also explored. aerosol atmospheric chemistry organice nitrate particles Hydrocarbons are emitted to the atmosphere from anthropogenic and biogenic sources. Globally, more than half are biogenic alkenes including isoprene (C 5 ), monoterpenes (C 10 ), and sesquiterpenes (C 15 ) (1). These compounds have a variety of linear, branched, and cyclic structures containing one or more C ¼ C double bonds and in the atmosphere react with OH radicals, O 3, and NO 3 radicals to form oxidized products (2, 3). In urban and polluted rural areas where peroxy radical intermediates react primarily with NO (4), one important class of products is organic nitrates (2, 3). For example, hydroxynitrates formed from OH radical-initiated reactions of isoprene have been identified in urban air (5) and impact O 3 formation and NO x removal (6) as well as secondary organic aerosol (SOA) formation (7). It has been estimated that, globally, about two-thirds of hydroxyperoxy radicals formed from isoprene oxidation react with NO (8), a value that probably applies to other terpenes. In spite of their atmospheric importance, little is known about the chemistry of hydroxynitrates. This is primarily because of a lack of analytical methods and reference compounds (5, 9). In a recent study, Paulot et al. (10) employed chemical ionization mass spectrometry to monitor hydroxynitrates formed during isoprene photooxidation, and then used this information to help constrain a detailed chemical mechanism. Even here, however, significant uncertainties exist because many hydroxynitrate isomers could not be distinguished. Improving models of terpene oxidation will require a greater knowledge of reaction products than is currently available, and represents a considerable challenge because of the large variety of terpene structures and reaction products. The approach we have taken to this problem is to investigate reactions of simpler model alkenes that contain some of the same structural features as terpenes but whose products can be more easily analyzed. The hope is that knowledge gained about the chemistry of these compounds can be used to build a detailed, quantitative understanding of the reactions of more complex molecules. In a previous publication (11), we presented a general methodology for identifying and quantifying particulate organic nitrates and applied this to hydroxynitrates (including isomers) and dihydroxynitrates in aerosol formed from OH radical-initiated reactions of linear internal alkenes and 1-alkenes in the presence of NO x. We investigated similar reactions of 2-methyl-1-alkenes. The reactive site in these molecules (a C ¼ C double bond with an attached methyl group) is present in isoprene and α-pinene, the two most abundant terpenes, as well as in many other monoterpenes and sesquiterpenes (1). Yields of β- hydroxynitrates, dihydroxynitrates, and trihydroxynitrates have not been previously reported for this class of alkenes. Furthermore, by combining these and previous measurements (11), we are now able to generalize our results to a variety of alkene structures to estimate the branching ratios for two key reaction steps: addition of an OH radical to the C ¼ C double and reaction of NO with β-hydroxyperoxy radicals to form either β-hydroxynitrates or β-hydroxyalkoxy radicals. The approach should help to improve models of atmospheric oxidation of organic compounds and SOA formation. Results and Discussion Reaction Mechanism. The mechanism of the OH radical-initiated reaction of 2-methyl-1-alkenes in the presence of NO x is shown in Fig. 1 (2, 3, 11), where R represents an alkyl group. In Fig. 1, α i is the branching ratio for the reaction of a specie by pathway i, defined as α i ¼ r i Σr i, where r i is the reaction rate for pathway i and the sum is over all pathways for the specie and equal to one. The focus here is on quantifying branching ratios α C¼C and α 1 -α 6, so only these are discussed. The reaction is initiated by the addition of an OH radical to the C ¼ C double bond (α C¼C )orby abstraction of an H atom (1 α C¼C ). The reactions that follow H-atom abstraction are the same as those that occur in reactions of alkanes (2). For the OH radical addition pathway, addition occurs at either C atom of the C ¼ C double bond to form a pair of β-hydroxyalkyl radical isomers that then react with O 2 to form β-hydroxyperoxy radicals. The branching ratios for the combined addition of OH radicals and O 2 to form the 1-hydroxy-2-peroxy Author contributions: A.M. performed research and analyzed data; A.M. and P.Z. wrote the paper; and P.Z. designed research. The authors declare no conflict of interest. This article is a PNAS Direct Submission. 1 To whom correspondence should be addressed: paul.ziemann@ucr.edu. This article contains supporting information online at /DCSupplemental PNAS April 13, 2010 vol. 107 no. 15

2 CHEMISTRY ENVIRONMENTAL SCIENCES SPECIAL FEATURE Fig. 1. Mechanism of the OH radical-initiated reaction of 2-methyl-1-alkenes in the presence of NO x. and 2-hydroxy-1-peroxy radicals are α 1 and α 2, respectively, with α 1 þ α 2 ¼ α C¼C. It is assumed here that the yields of the two isomers are determined by the OH radical addition step (12), so that α 1 and α 2 are also the branching ratios for forming 1-hydroxy- 2-alkyl and 2-hydroxy-1-alkyl radicals. The β-hydroxyperoxy radicals react with NO to form β-hydroxynitrates [P1, P2] with branching ratios α 3 and α 4 or β-hydroxyalkoxy radicals with branching ratios α 5 and α 6. Note that α 3 þ α 5 ¼ 1 and α 4 þ α 6 ¼ 1. The term same pathways in Fig. 1 refers to species derived from 2-hydroxy-1-alkoxy radicals by reactions similar to those shown for 1-hydroxy-2-alkoxy isomers. The β-hydroxyalkoxy radicals react with O 2, decompose, or isomerize via a six-member ring transition state. Reaction of the 2-hydroxy-1-alkoxy radical with O 2 forms a α-hydroxycarbonyl [P4], but this pathway is not accessible to the 1-hydroxy-2-alkoxy radical because there is no H atom available on the second C atom for abstraction by O 2. Decomposition followed by reaction with O 2 forms formaldehyde [P5], a ketone [P6], and isomerization followed by a reaction with O 2 forms dihydroxyperoxy radicals that react (similar to β-hydroxyperoxy radicals) with NO to form dihydroxynitrates [P7, P8] or dihydroxyalkoxy radicals. The dihydroxyalkoxy radicals isomerize in two ways. Those derived from reactions of 2-hydroxy-1-alkoxy radicals primarily undergo a rapid reverse isomerization, in which the alkoxy group abstracts an H atom from the terminal C atom, and then reaction with O 2 forms a dihydroxycarbonyl [P13]. A minor isomerization pathway involves abstraction of an H atom from a CH 2 group further along the carbon chain followed by reaction with O 2 to form trihydroxyperoxy radicals that then react by pathways similar to those described above for dihydroxyperoxy radicals to form trihydroxynitrates [P10] and trihydroxycarbonyls [P12]. Reverse isomerization cannot occur for dihydroxyalkoxy radicals derived from reactions of 1-hydroxy-2-alkoxy radicals because there is no H atom to abstract from the second C atom. Isomerization therefore continues along the carbon chain leading to trihydroxynitrates [P9] and trihydroxycarbonyls [P11]. SOA Product Identification. In the remaining discussion, the following notation is used for structures of β-hydroxynitrates (β-hn), dihydroxynitrates (DHN), and trihydroxynitrates (THN) that are all substituted 2-methylalkanes: Hydroxy and nitrooxy groups are designated H and N, the C-atom on which a group is located is designated by a number before the letter, and the number of C atoms in the molecule is indicated by a subscript following the letter C. For example, 1,2-dihydroxy-4-nitrooxy- 2-methyltetradecane is designated 1; 2H4NC 15. The results of HPLC-UV-thermal desorption particle beam mass spectrometer (TDPBMS) analysis of nitrate-containing SOA products formed from the reaction of 2-methyl-1-tetradecene are shown in Fig. 2. The UV chromatogram in Fig. 2A was taken under reverse-phase conditions, so compound polarity decreases with increasing retention time. Peaks 1 3 are therefore assigned to THN, DHN, and β-hn, the three expected nitratecontaining products. The structures were verified by 1 H NMR analysis as shown in Table S1. The small, partially resolved peak at the front of peak 3 indicates the two β-hn isomers have very different abundances. On the basis of 1 H NMR analysis, the major isomer was determined to be 1H2NC 15, and, from the Matsunaga and Ziemann PNAS April 13, 2010 vol. 107 no

3 chromatographic peak areas, the fractions of 1H2NC 15 and 1N2HC 15 isomers were determined to be The mass spectra shown in Fig. 2B and C for peaks 1 and 2 are from mixtures of THN and DHN isomers, respectively. Fig. 2D shows the mass spectrum for peak 3; it is from a mixture of β-hn isomers, although the spectrum should be mostly from 1H2NC 15, the dominant isomer. The primary electron ionization fragmentation channel for compounds having two adjacent functional groups is scission of the C-C bond between the groups (11). The major ions observed for DHN and THN are consistent with this channel. The molecular weight (M) of the DHN is 305, and m z 211 is likely an ½M ð31 þ 63ÞŠ þ ion formed by loss of CH 2 OH and HNO 3. The m z 193 ion is formed by additional loss of H 2 O. For THN, M ¼ 321 and m z 272 is likely an ½M ð31 þ 18ÞŠ þ ion formed by loss of CH 2 OH and H 2 O. For β-hn, M ¼ 289 and fragmentation of 1H2NC 14 gives the m z 213 ion that is likely an ½M ð30 þ 46ÞŠ þ ion formed by loss of CH 2 O and NO 2. Transfer of an H atom from the CH 2 OH scission product to the ion forms the more stable CH 3 ðch 2 Þ 11 CðCH 3 Þ¼OH þ ion and formaldehyde. Fig. 2. Analysis of SOA formed from the OH radical-initiated reaction of 2-methyl-1-tetradecene in dry air in the presence of NO x.(a) HPLC-UV chromatogram. (B D) HPLC-TDPBMS mass spectra of trihydroxynitrates, dihydroxynitrates, and β-hydroxynitrates, respectively. β-hydroxynitrate Yields. The molar yields (moles of product/moles of alkene reacted) of β-hn in the particle phase are presented in Table S2. Losses of β-hn by reactions with OH radicals were neglected because model calculations that included gas particle partitioning indicated losses were <5% for the high-carbon number products that are of primary interest here. The table contains yields with and without normalization for the fraction of OH radical reaction that occurs by addition to the C ¼ C double bond, α C¼C ¼ k add ðk add þ k abs Þ, where k abs and k add are the rate constants for H-atom abstraction and OH radical addition, respectively. Values of k abs and k add were calculated by using equations for 2-methyl-1-alkenes (13, 14): k abs ¼ 2.6 þ 1.4 ðc n 6Þ and k add ¼ 51 þ 16 ½1 expð 0.35 ðc n 3ÞÞŠ, where C n is the alkene carbon number and units of k add and k abs are cm 3 molecule 1 s 1. Values of k add and k abs increase with increasing C n, but because k abs increases more than k add for C 9 -C 15 alkenes, α C¼C decreases from 0.91 at C 9 to 0.81 at C 15 and normalization increases yields by 10 20%. Values of α C¼C are larger for 2-methyl-1-alkenes than for linear 1-alkenes (13), indicating a methyl group increases α C¼C. Molar yields of β-hn isomers in the particle phase, with and without normalization, and their sums are presented in Fig. 3. Values increase with increasing carbon number primarily because vapor pressures decrease so compounds partition more to the particle phase where they are measured (11). The branching ratio for β-hn formation from the reaction of β-hydroxyperoxy radicals with NO also increases with carbon number due to enhanced stabilization of the β-hydroxyperoxy radical-no complex, but the effect on yields is minor for C 9 -C 15 compounds (11). Total β-hn yields (Fig. 3A) appear to reach a plateau at C 14 or C 15. We believe that, at the plateau, these compounds are essentially entirely in the particle phase, so that our measurements of particle-phase concentrations of β-hn can be used to determine yields of β-hn at the plateau. Determining yields before the plateau requires measurements of both gas and particle phases. In our previous study of β-hn formed from reactions of linear 1-alkenes (11), a plateau was also reached at C 14, and it was possible to observe that the plateau continued to C 17 because larger 1-alkenes were available. The C 15 β-hn yields measured here are and with and without normalization for OH radical addition, with uncertainties of one standard deviation. Yields at carbon numbers below the plateau can be estimated by using an approach presented previously to account for the reduced stability of β-hydroxyperoxy radical-no complexes at smaller carbon numbers (11). To our knowledge, the only measured β-hn yield is a value of for the reaction of 2-methyl-1-propene (15). This is about half the plateau value measured here, as expected because of the small Matsunaga and Ziemann

4 Fig. 3. Molar yields of β-hydroxynitrates in SOA formed from OH radical-initiated reactions of 2-methyl-1-alkenes in dry air in the presence of NO x. (A) Total β-hydroxynitrate (1H2NC n and 1N2HC n ) isomers. (B) 1H2NC n isomers. (C) 1N2HC n isomers. Yields are presented with and without normalization for the fraction of the OH radical reaction that occurred by addition to the C ¼ C double bond. The error bars for C 14 ðn ¼ 3Þ and C 15 ðn ¼ 2Þ are SD. Other yields are based on single experiments. Dashed curves were drawn to aid the eye. carbon number (11). The 1H2NC n isomer comprises 90% of the total β-hn, so the plot of yields (Fig. 3B) is similar to that of the total β-hn (Fig. 3A) and reaches a plateau at C 14 or C 15. Yields of the 1N2HC n isomer (Fig. 3C) are more difficult to interpret. They may reach a plateau at C 15 but this could not be verified because larger 2-methyl-1-alkenes were not available. On the basis of vapor pressures, the yield of the 1N2HC 15 isomer should be close to a plateau. Vapor pressures of 1N2HC n isomers formed from reactions of linear 1-alkenes are 3 times larger than those of 1H2NC n isomers (11), equivalent to one less CH 2 unit. Partitioning of the 1N2HC 15 isomer should therefore be similar to that of the 1H2NC 14 isomer that, because it has nearly the same yield as the 1H2NC 15 isomer appears to be almost entirely in the particle phase. Yields with and without normalization for OH radical addition are and for the 1H2NC 15 isomer and and for the 1N2HC 15 isomer, and the ratio is The plateau value of for OH addition-normalized yields of β-hn formed from 2-methyl-1-alkene reactions is higher than the value of measured for reactions of linear 1-alkenes, but lower than the value of 0.3 expected for secondary alkyl nitrates formed from reactions of n alkanes (11, 16). The lower yields of β-hn are due to weakening of the O-O bond in the β-hydroxyperoxy radical-no complex, RO-O-NO, by hydrogen bonding between hydroxy and peroxy groups (9). The methyl group on the same C atom as the peroxy group (the structure for 90% of the β-hydroxyperoxy radicals formed from 2-methyl- 1-alkenes) appears to partially compensate for the effect of the hydroxy group, presumably by strengthening the O-O bond by electron donation. A more stabile complex has more time to rearrange to a β-hn before it can decompose to a β-hydroxyalkoxy radical and NO 2 (17). Branching Ratios. As shown in Fig. 1, the yield (Y) of a compound is equal to the product of the branching ratios along the pathway that it is formed. For the β-hn isomers, 1H2NC n [P1] and 1N2HC n [P2], this means Y P1 ¼ α 1 α 3 and Y P2 ¼ α 2 α 4,soðα 1 α c¼c Þ¼ðY P1 α c¼c Þ α 3 and ðα 2 α c¼c Þ¼ðY P2 α c¼c Þ α 4, where Y P1 α c¼c and Y P2 α c¼c are OH addition-normalized yields as reported above. In our study of reactions of linear 1- alkenes (11), these equations were solved for values of the four branching ratios by assuming α 3 α 4, the ratio of branching ratios for forming β-hn from reactions of NO with secondary and primary β-hydroxyperoxy radicals, is 1 or 1.5 (18). We combined measured product yields for reactions of linear internal alkenes, linear 1-alkenes, and 2-methyl-1-alkenes to estimate branching ratios for forming primary, secondary, and tertiary β-hydroxyalkyl radicals by OH radical addition to a C ¼ C double bond, and branching ratios for forming β-hn from reactions of the corresponding β-hydroxyperoxy radicals with NO. Starting with linear internal alkenes, for which two secondary β-hydroxyperoxy radicals are formed, ðα 1 α c¼c Þ¼ðα 2 α c¼c Þ¼0.5 and α 3 ¼ α 4 ¼ðY P1 α c¼c ÞþðY P2 α c¼c Þ¼0.15. By using α 3 ¼ 0.15 for secondary β-hydroxyperoxy radicals formed from linear 1-alkenes gives ðα 1 α c¼c Þ¼ðY P1 α c¼c Þ α 3 ¼ ¼ 0.65, so for primary β-hydroxyperoxy radicals formed from this reaction ðα 2 α c¼c Þ¼1 ðα 1 α c¼c Þ¼0.35 and α 4 ¼ðY P2 α c¼c Þ ðα 2 α c¼c Þ¼ ¼ Using these values for primary β-hydroxyperoxy radicals formed from 2-methy-1-alkenes gives ðα 2 α c¼c Þ¼ðY P2 α c¼c Þ α 4 ¼ ¼ 0.19, so for tertiary β-hydroxyperoxy radicals formed in this reaction ðα 1 α c¼c Þ¼1 ðα 2 α c¼c Þ¼0.81 and α 3 ¼ðY P1 α c¼c Þ ðα 1 α c¼c Þ¼ ¼ The branching ratios α tert, α sec, and α prim for forming tertiary, secondary, and primary β-hydroxyalkyl radicals by OH radical addition to a C ¼ C double bond are as follows: α sec α sec ¼ for linear internal alkenes, α sec α prim ¼ for linear 1-alkenes, and α tert α prim ¼ for 2-methyl- 1-alkenes. Combining these values, we obtain the relative branching ratios α tert α sec α prim ¼ , where α prim is set equal to 1.0. The results indicate that OH radical addition preferentially forms β-hydroxyalkyl radicals in the order tertiary > secondary > primary. This trend is also observed in site-specific OH radical addition rate constants used in a structure-activity relationship developed from rate constants for reactions of alkenes with OH radicals (12) that predict α tert α sec α prim ¼ The chemical basis for this trend is uncertain (12). Branching ratios have also been measured for reactions of alkenes with OH radicals by using other techniques. Values of α sec α prim ¼ , identical to those measured here, were obtained by gas chromatographic analysis of products of the reaction of 1 propene in the absence of NO x (3), and values of α sec α prim ¼ , α tert α prim ¼ , and α tert α sec ¼ SPECIAL FEATURE CHEMISTRY ENVIRONMENTAL SCIENCES Matsunaga and Ziemann PNAS April 13, 2010 vol. 107 no

5 that agree fairly well with the values , , and measured here, were obtained from mass spectrometric analysis of β-hydroxyalkyl radicals formed from reactions of 1-butene, 2-methylpropene, and 2-methyl- 2-butene in the absence of O 2 (12). The branching ratios of 0.25, 0.15, and 0.12 for β-hn formation from reactions of tertiary, secondary, and primary β- hydroxyperoxy radicals with NO indicate the stability of β- hydroxyperoxy radical-no complexes follow the same order as β-hydroxyalkyl radicals: tertiary > secondary > primary. The values α sec α prim ¼ ¼ 1.3 and α tert α prim ¼ ¼ 2.1 bracket the ratios of 1.5 and 1.5 measured for the corresponding alkylperoxy radicals (18), with our values indicating an enhancement in branching ratios for tertiary compared to secondary β-hydroxyperoxy radicals. Dihydroxynitrate and Trihydroxynitrate Yields. Total (sum of isomers) molar yields of DHN and THN are given in Table S2. Total yields are presented because isomers could not be resolved, and for THN only C 14 and C 15 compounds could be analyzed because peaks of smaller ones overlapped with the solvent peak. As shown in Fig. 4, yields of DHN increase with increasing carbon number to a plateau at C 14 or C 15. Yields of C 15 DHN, with and without normalization for OH radical addition, are and Yields decrease with decreasing carbon number less slowly than those of β-hn because lower vapor pressures enhance partitioning to particles. Average yields of C 14 and C 15 THN with and without normalization for OH radical addition are and These values are expected to be at a plateau like those of the more volatile DHN. Effects of Humidity and Ammonia on Yields. Experiments were also performed to investigate effects of H 2 O vapor and NH 3 on yields. In previous studies of reactions of 1-tetradecene yields of β-hn and DHN were reduced by 75% by NH 3, but were not affected by H 2 O vapor (11). Yields of β-hn, DHN, and THN measured for the reaction of 2-methyl-1-tetradecene are shown in Table 1. Yields of DHN and THN were the same under all conditions, whereas yields of 1N2HC 15 and 1H2NC 15 appear to be reduced by about 15% and 8% by NH 3 and the yield of 1H2NC 15 increased by 20% in the reaction at 50% relative humidity (R.H.) (compared to dry air). The methyl group on the second C atom reduces the effect of NH 3 on both β-hn and DHN yields (compared to the 75% decrease for the 1-tetradecene reaction), most likely by reducing hydrogen bonding between hydroxy and peroxy DHN molar yield without normalization with normalization carbon number Fig. 4. Molar yields of dihydroxynitrates in SOA formed from OH radicalinitiated reactions of 2-methyl-1-alkenes in dry air in the presence of NO x. Yields are presented with and without normalization for the fraction of the OH radical reaction that occurred by addition to the C ¼ C double bond. Error bars for C 14 ðn ¼ 3Þ and C 15 ðn ¼ 2Þ are SD. Other yields are based on single experiments. Dashed curves were drawn to aid the eye. Table 1. Effects of H 2 O vapor and NH 3 on product yields from the reaction of 2-methyl-1-tetradecene Condition* 1H2NC 15 1N2HC 15 DHN THN Dry H 2 O NH * Dry ¼< 1% R.H.; H 2 O ¼ 50% R.H.; NH 3 ¼ 20 ppmv NH 3 DHN ¼ dihydroxynitrate isomers 1; 2H4NC 15 and 1; 2H5NC 15. THN ¼ trihydroxynitrate isomers 1; 2; 4H7NC 15 and 1; 2; 5H8NC 15. groups. It is worth noting that, in reactions of 1-tetradecene and 2-methyl-1-tetradecene, only the yield of 1H2NC 15 isomer from the latter reaction was higher, at 50% R.H., whereas all other products were unaffected because only this isomer has a methyl group on the same C atom as the peroxy group. This indicates that the methyl group impacts the yield through an electronic rather than steric effect; apparently, by electron donation that enhances clustering of H 2 Otoaβ-hydroxyperoxy radical-no complex, RO-O-NO, and strengthens the O-NO bond. Conclusions In this study, β-hn, DHN, and THN formed from OH radicalinitiated reactions of C 9 -C 15 2-methyl-1-alkenes in the presence of NO x were identified and quantified and the results used to calculate yields with and without normalization for OH radical addition to the C ¼ C double bond. Comparison with results from our previous study of reactions of linear 1-alkenes (11) shows that the additional methyl group affected relative yields of β-hn isomers, increased OH addition-normalized yields of β-hn, DHN, and THN, with the latter product not being observed previously, and also altered the effect of H 2 O vapor and NH 3 on yields. Yields not normalized for OH radical addition were further increased because of enhanced addition relative to H-atom abstraction (13). The effects are, apparently, due to electron donation by the methyl group that stabilizes both the tertiary β-hydroxyalkoxy radical formed when an OH radical adds to the terminal C atom, leading to a strong preference for that addition pathway, and the β-hydroxyperoxy radical-no complex, thereby enhancing β-hn formation. Electron donation may also increase yields of DHN by reducing hydrogen bonding between hydroxy and peroxy groups that otherwise tends to reduce yields (9). THN are formed in these reactions because replacement of an H atom by a methyl group forces dihydroxyalkoxy radicals to isomerize by an alternate pathway. Calculations made by using β-hn yields for reactions of 2-methyl-1-alkenes, linear internal alkenes, and linear 1-alkenes (11) indicate that for C n C 14 the relative ratios for forming tertiary, secondary, and primary β-hydroxyalkyl radicals by OH radical addition to the C ¼ C double bond are , and branching ratios for forming β-hn from reactions of tertiary, secondary, and primary β-hydroxyperoxy radicals with NO are 0.25, 0.15, and 0.12, corresponding to relative ratios of These values should be applicable to other systems, although effects of double-bond conjugation and ring strain on OH radical addition should be considered (12). Values can be extrapolated to carbon numbers below C 14 by using the approach presented previously (11). For example, the sum of the yields of secondary and tertiary β-hn predicted for the reaction of α-pinene is 0.22, in reasonable agreement with a measured total organic nitrate yield of (19). We have shown previously for reactions of linear alkenes (20) that yields of β-hn, DHN, and carbonyls can be used with structure-reactivity relationships to develop a complete reaction mechanism that can be used with gas particle partitioning theory to model SOA formation. These results can be used similarly Matsunaga and Ziemann

6 Materials and Methods Environmental Chamber Method. Reactions of 2-methyl-1-alkenes with OH radicals in the presence of NO x were conducted in a 5,900 L polytetrafluoroethylene environmental chamber at room temperature ( 25 C) and atmospheric pressure. The chamber, which has black lights covering two walls, was filled with clean, dry air (<5 ppb hydrocarbons and <1% R.H.) from an Aadco clean air system. In a typical experiment, 1 part per million by volume (ppmv) alkene, 5 ppmv each of methyl nitrite and NO, and 200 μgm 3 of dioctyl sebacate seed particles generated by an evaporation condensation apparatus were added to the chamber. The NO suppressed O 3 and NO 3 radical formation and concentrations were sufficiently high that all peroxy radical intermediates reacted with NO, thereby meeting the major criteria for simulating OH radical-initiated reactions of alkenes in a polluted atmosphere. Reactions with 2-methyl-1-tetradecene were also performed at 50% R.H. and with 20 ppmv NH 3 in dry air. Reactions were initiated by turning on the black lights that generate OH radicals by photolyzing methyl nitrite (21). The chamber was irradiated for 6 min. During that time, 40 50% of the alkene reacted. The average OH radical concentration estimated from the amounts of alkenes reacted and OH radical rate constants (13) was cm 3. NO and NO x were measured by using a Thermo Environmental Instruments, Inc. 42C NO-NO 2 -NO x analyzer. The chemicals used in these experiments are described in SI Text. Aerosol Mass and Alkene Analysis. Alkene concentrations were measured before and after reaction by using gas chromatography with flame ionization detection (FID) to analyze samples collected on Tenax TA solid adsorbent (11). Analyses of replicate samples taken at 30 min intervals agreed within 5%, and FID signals before reaction were within 5% of values expected for a 1 ppmv chamber concentration (based on solution calibration curves), indicating no wall losses. Aerosol volume concentrations were measured every 2 min during reactions and filter sampling with a scanning mobility particle sizer. Aerosol mass concentrations were calculated by using these values and a density of 1.1 gcm 3 measured by using a microliter syringe and balance and dried SOA filter extract from reaction of 2-methyl-1-tridecene without seed aerosol. β-hydoxynitrate, Dihydroxynitrate, and Trihydroxynitrate Analysis. Particulate organic nitrates were collected and analyzed as described previously (11). After each reaction, filter samples were collected for 2 h on Millipore filters (1.0 μm pore size, Fluoropore FALP, 47 mm, without pretreatment) at 15 Lmin 1, with the flow controlled by a calibrated critical orifice located between the filter holder and vacuum pump. The collection efficiency of these filters was the same as for 0.45 μm pore size filters, but had a smaller pressure drop. This minimized corrections for the flow reduction (<5 10%) that occurred as the filter accumulated sample. Filter samples were extracted immediately with ethyl acetate or stored at 20 C, and stored samples were extracted within 1 week because β-hn, DHN, and THN were all stable for at least this period. Samples extracted after 1 mo showed losses of β-hn (but not DHN and THN) due to decomposition. Tests by using different solvents, sonication, multiple extractions, and filter spiking indicated recovery was %. Extracts were analyzed by using an Agilent 1100 Series HPLC coupled to a UV-vis diode array detector at 210 nm, a wavelength that nitrates absorb strongly. The HPLC contained a mm Zorbax 5 μm Eclipse XDB- C18 column operating at room temperature. The HPLC method employed gradient elution by using water and acetonitrile, with the contribution of acetonitrile maintained at 50% for 10 min and then increased linearly to 100% in 50 min. The flow rate was 1 ml min 1, and the sample injection volume was 10 μl. The β-hn, DHN, and THN were quantified by using calibration curves prepared from purified mixtures of isomers of each compound class, formed from a 2-methyl-1-tridecene reaction. Purified mixtures were obtained by collecting aliquots of the appropriate mixture after it had passed through the UV-vis detector. Purities of standards were verified by HPLC analysis. Molar yields in particles were calculated as (moles of compound on filter/volume of air sampled)(aerosol mass concentration after reaction/ average aerosol mass concentration during filter sampling)/(moles of alkene reacted per volume of air). The second term corrects for wall loss of particles during filter sampling. Molar absorptivities of the authentic β-hn, DHN, and THN standards relative to 2-ethylhexyl nitrate were 0.412, 0.964, and 0.608, respectively. Products were identified from particle mass spectra obtained by coupling the HPLC to a TDPBMS via a Collison atomizer. Effluent from the HPLC column was atomized, the solvent was removed as the aerosol passed through activated charcoal and silica gel diffusion dryers, and the residual aerosol particles were analyzed in the TDPBMS. Product identities were verified by 1 H NMR analysis of purified compounds by using a Varian Inova 500 MHz instrument. ACKNOWLEDGMENTS. We thank Roger Atkinson for helpful discussions. This material is based on work supported by the National Science Foundation under Grant ATM SPECIAL FEATURE CHEMISTRY ENVIRONMENTAL SCIENCES 1. Guenther A, et al. (1995) A global model of natural volatile organic compound emissions. J Geophys Res, 100: Atkinson R, Arey J (2003) Atmospheric degradation of volatile organic compounds. Chem Rev, 103: Calvert JG, et al. (2000) The Mechanisms of Atmospheric Oxidation of the Alkenes. (Oxford Univ Press, New York). 4. Sillman S (1999) The relation between ozone, NO x, and hydrocarbons in urban and polluted rural environments. Atmos Environ, 33: Werner G, Kastler J, Looser R, Ballschmiter K (1999) Organic nitrates of isoprene as atmospheric trace compounds. Agnew Chem Int Edit, 38: Chameides WL, Lindsay RW, Richardson J, Kiang CS (1988) The role of biogenic hydrocarbons in urban photochemical smog: Atlanta as a case study. Science, 241: Surratt JD, et al. (2006) Chemical composition of secondary organic aerosol formed from the photooxidation of isoprene. J Phys Chem A, 110: Paulot F, et al. (2009) Unexpected epoxide formation in the gas-phase photooxidation of isoprene. Science, 325: O Brien JM, Czuba E, Hastie DR, Francisco JS, Shepson PB (1998) Determination of the hydroxy nitrate yields from the reaction of C 2 -C 6 alkenes with OH in the presence of NO. J Phys Chem A, 102: Paulot F, et al. (2009) Isoprene oxidation: New insights into the production of acids and organic nitrates. Atmos Chem Phys, 9: Matsunaga A, Ziemann PJ (2009) Yields of β-hydroxynitrates and dihydroxynitrates in aerosol formed from the OH radical-initiated reactions of linear alkenes in the presence of NO x. J Phys Chem A, 113: Peeters J, Boullart W, Pultau V, Vandeberk S, Vereecken L (2007) Structure-activity relationship for the addition of OH to (poly)alkenes: Site-specific and total rate constants. J Phys Chem A, 111: Nishino N, Arey J, Atkinson R (2009) Rate constants for the gas-phase reactions of OH radicals with a series of C 6 -C 14 alkenes at 299 2K. J Phys Chem A, 113: Kwok ESC, Atkinson R (1995) Estimation of hydroxyl radical reaction rate constants for gas-phase organic compounds using a structure-reactivity relationship: An update. Atmos Environ, 29: Tuazon EC, Aschmann SM, Arey J, Atkinson R (1998) Products of the gas-phase reactions of a series of methyl-substituted ethenes with the OH radical. Environ Sci Technol, 32: Arey J, Aschmann SM, Kwok ESC, Atkinson R (2001) Alkyl nitrate, hydroxyalkyl nitrate, and hydroxycarbonyl formation from the NO x -air photooxidations of C 5 -C 8 n-alkanes. J Phys Chem A, 105: Zhang J, Dransfield T, Donahue NM (2004) On the mechanism for nitrate formation via the peroxy radical þno reaction. J Phys Chem A, 108: Cassanelli P, Fox DJ, Cox RA (2007) Temperature dependence of pentyl nitrate formation from the reaction of pentyl peroxy radicals with NO. Phys Chem Chem Phys, 9: Nozière B, Barnes I, Becker K-H (1999) Product study and mechanisms of the reactions of -pinene and of pinonaldehyde with OH radicals. J Geophys Res, 104: Matsunaga A, Docherty KS, Lim YB, Ziemann PJ (2009) Composition and yields of secondary organic aerosol formed from OH radical-initiated reactions of linear alkenes in the presence of NO x : Modeling and measurements. Atmos Environ, 43: Atkinson R, Carter WPL, Winer AM, Pitts JN Jr. (1981) An experimental protocol for the determination of OH radical rate constants with organics using methyl nitrite photolysis as an OH radical source. J Air Pollut Control Assoc, 31: Matsunaga and Ziemann PNAS April 13, 2010 vol. 107 no