Gas-Phase Reactions of Isoprene and Its Major Oxidation Products

Mar 9, 2018 - ... at the Joint Center for Earth Systems Technology of University of Maryland Baltimore County and NASA Goddard Space Flight Center...
3 downloads 0 Views 9MB Size
Review Cite This: Chem. Rev. 2018, 118, 3337−3390

pubs.acs.org/CR

Gas-Phase Reactions of Isoprene and Its Major Oxidation Products Paul O. Wennberg,*,†,‡ Kelvin H. Bates,*,§,∥ John D. Crounse,† Leah G. Dodson,§,⊥ Renee C. McVay,§,# Laura A. Mertens,§,∇ Tran B. Nguyen,†,○ Eric Praske,§ Rebecca H. Schwantes,†,◆ Matthew D. Smarte,§ Jason M. St Clair,†,¶ Alexander P. Teng,†,▲ Xuan Zhang,‡,◆ and John H. Seinfeld‡,§ †

Division of Geological and Planetary Sciences, ‡Division of Engineering and Applied Science, and §Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125, United States

Chem. Rev. 2018.118:3337-3390. Downloaded from pubs.acs.org by IOWA STATE UNIV on 01/05/19. For personal use only.

S Supporting Information *

ABSTRACT: Isoprene carries approximately half of the flux of non-methane volatile organic carbon emitted to the atmosphere by the biosphere. Accurate representation of its oxidation rate and products is essential for quantifying its influence on the abundance of the hydroxyl radical (OH), nitrogen oxide free radicals (NOx), ozone (O3), and, via the formation of highly oxygenated compounds, aerosol. We present a review of recent laboratory and theoretical studies of the oxidation pathways of isoprene initiated by addition of OH, O3, the nitrate radical (NO3), and the chlorine atom. From this review, a recommendation for a nearly complete gas-phase oxidation mechanism of isoprene and its major products is developed. The mechanism is compiled with the aims of providing an accurate representation of the flow of carbon while allowing quantification of the impact of isoprene emissions on HOx and NOx free radical concentrations and of the yields of products known to be involved in condensed-phase processes. Finally, a simplified (reduced) mechanism is developed for use in chemical transport models that retains the essential chemistry required to accurately simulate isoprene oxidation under conditions where it occurs in the atmosphereabove forested regions remote from large NOx emissions.

CONTENTS 1. Introduction 2. Mechanism Development 2.1. Alkene + OH Reactions 2.1.1. Kinetics 2.1.2. Mechanism 2.2. RO2 + NO Reactions 2.2.1. Kinetics 2.2.2. Mechanism 2.3. RO2 + HO2 Reactions 2.3.1. Kinetics 2.3.2. Mechanism 2.4. Hydrogen Shifts to Peroxy Radicals 2.5. Additional Rate Parameterizations 2.6. Implementing the Full and Condensed Mechanisms 3. Reaction of OH with Isoprene 3.1. Kinetics 3.2. Location of OH Addition 3.3. Addition of O2 to Hydroxy Allylic Radicals 3.3.1. cis/trans Allylic Radical Distribution 3.3.2. Initial (Kinetic) Distribution of Isoprene Hydroxy Peroxy Radicals (ISOPOO) 3.3.3. Reversibility of O2 Addition 3.3.4. Constraint on Absolute Rates of O2 Loss from β-ISOPOO

© 2018 American Chemical Society

3.3.5. Representing the ISOPOO Isomer Distribution 3.4. Fate of ISOPOO 3.4.1. Reaction with NO 3.4.2. Reaction with HO2 3.4.3. Reaction with RO2 3.4.4. H-Shift Isomerization 4. Reaction of O3 with Isoprene 4.1. Kinetics 4.2. Mechanism 4.3. Reactions of the Stabilized Criegee Intermediate 4.3.1. Kinetics 4.3.2. Mechanism 5. Reaction of NO3 with Isoprene 5.1. Kinetics 5.1.1. Mechanism: Formation of INO2 5.2. INO2 Reaction with HO2 5.2.1. Kinetics 5.2.2. Mechanism 5.3. INO2 Reaction with NO and NO3 5.3.1. Kinetics 5.3.2. Mechanism 5.4. INO2 Reaction with RO2

3338 3339 3339 3339 3339 3340 3340 3340 3341 3341 3341 3342 3342 3344 3344 3344 3344 3344 3344 3345 3345 3346

3346 3347 3347 3349 3351 3351 3353 3353 3354 3355 3355 3356 3356 3357 3357 3357 3357 3357 3358 3358 3358 3358

Received: July 21, 2017 Published: March 9, 2018 3337

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews 5.4.1. Kinetics 5.4.2. Mechanism 5.5. INO2 H-shift Isomerization 5.6. Fate of INO 6. Isoprene + Cl 6.1. Kinetics 6.2. Mechanism 6.3. Reactions of IClOO 6.3.1. Kinetics 6.3.2. Mechanism 7. Photochemistry of Major Oxidation Products of Isoprene 7.1. Methyl Vinyl Ketone (MVK) 7.1.1. Kinetics: Reaction with OH 7.1.2. Kinetics: Other Oxidants 7.1.3. Mechanism: Reaction with OH 7.2. Methacrolein (MACR) 7.2.1. Kinetics: Reaction with OH 7.2.2. Kinetics: Other Oxidants 7.2.3. Mechanism: OH addition 7.2.4. Mechanism: H Abstraction 7.2.5. Peroxymethacryl Radical 7.3. Isoprene Hydroxy Hydroperoxide (ISOPOOH) 7.3.1. Kinetics: Reaction with OH 7.3.2. Mechanism: Reaction with OH 7.4. Isoprene Expoxydiols (IEPOX) 7.4.1. Kinetics: Reaction with OH 7.4.2. Mechanism: Reaction with OH 7.5. Isoprene Nitrates: IHN, ICN, and IPN 7.5.1. Reaction with OH 7.5.2. Reaction with O3 and NO3 7.5.3. Photolysis of Organic Nitrates 7.6. Methacryloyl Peroxynitrate (MPAN) 7.6.1. Kinetics: Reaction with OH 7.6.2. Mechanism: Reaction with OH 7.7. δ-Hydroperoxy Aldehydes (HPALD) 7.7.1. Photolysis Rate 7.7.2. Mechanism: Photolysis 7.7.3. Kinetics: Reaction with OH 7.7.4. Mechanism: Reaction with OH 7.8. Hydroxymethyl Hydroperoxide (HMHP) 8. The Challenges That Remain 8.1. OH Addition at C2 and C3 of Isoprene 8.2. Formation and Fate of Organonitrates 8.3. RO2 + HO2 Reaction Rates and Products 8.4. RO2 + RO2 Reaction Rates and Products 8.5. Intramolecular RO2 Chemistry (Autoxidation): H-Shift Rates and Products 8.6. Epoxide Formation 8.7. Photolysis of Multifunctional Oxygenated Compounds 8.8. Unknown Unknowns of Heterogeneous Chemistry 9. Conclusion Associated Content Supporting Information Author Information Corresponding Authors ORCID Present Addresses Notes Biographies Acknowledgments

Review

References

3358 3359 3359 3359 3359 3359 3359 3360 3360 3361

3381

1. INTRODUCTION Volatile organic compounds (VOCs) emitted from vegetation significantly impact atmospheric photochemistry. This biogenic carbon flux is dominated by a single compound, isoprene (C5H8, 2-methyl-1,3-butadiene). The global budget of isoprene has been estimated by several approaches,1,2 typically constrained using surface flux measurements3,4 and/or satellite observations of one of its major oxidation products, formaldehyde.5 The most up-todate modeling framework suggests that, at roughly 500 Tg year−1, the emissions of isoprene alone are comparable to those of methane and comprise about half the total biogenic emissions of non-methane VOCs worldwide.6 This emission originates from a broad taxonomic distribution of plants (e.g., mosses, ferns, and trees). The majority of the massive flux of isoprene originates from a light-dependent de novo synthesis in plants using carbon from the Calvin cycle; dark production from microorganisms, plants, and animals is only a minor contribution.7 The question of why plants emit isoprene is more complex. A recent unified hypothesis suggests that isoprene is synthesized in plant tissue to mitigate the compounded effects from several environmental stress factors that produce reactive oxygen species in vivo,8 e.g., extreme temperatures, light, water deficiency, soil salinity, air pollution, and mechanical damage. Following biosynthesis, isoprene is lost to the atmosphere through stomata and does not significantly accumulate in the leaves.9 As a result, nearly the entire flux of isoprene occurs during daytime.10,11 Further information on the biogenesis of isoprene and historical context of studies is provided by a number of reviews.12−16 Here, we focus on isoprene’s atmospheric reactions and those of its immediate oxidation products. Unlike methane, which has an atmospheric lifetime of nearly a decade and so is relatively well mixed through the atmosphere, isoprene does not travel far from its source before being oxidized. As a five carbon conjugated diene, isoprene reacts rapidly with atmospheric oxidants, especially the hydroxyl radical (OH). This high reactivity limits its accumulation in the atmosphere such that, despite its large flux, only moderate mixing ratios are observed (0−10 ppbv)17 and even then only near its source. Under typical atmospheric daytime conditions, the reactive conversion of isoprene into more oxidized VOCs (OVOCs) occurs with a time constant of approximately 1−2 h.18 At 300 K and 1000 hPa, reaction with OH is the dominant sink (τOH = 1.4 h at [OH] = 2 × 106 molecules cm−3, using the IUPAC recommended rate coefficient19), followed by ozone (O3) (τO3 = 16.4 h, [O3] = 50 ppbv, using the IUPAC recommended rate coefficient19), the nitrate radical (NO3) (τNO3 = 17.5 h, [NO3] = 1 pptv, using the IUPAC recommended rate coefficient19), and the chlorine radical (Cl) (τCl = 29 days, [Cl] = 1 × 103 molecules cm−3, using the JPL recommended rate coefficient20). Because isoprene emissions peak in the daytime, and NO3 radical concentrations peak in the nighttime, oxidation by ozone is generally a larger sink for isoprene than that by the nitrate radical. Nevertheless, nitrate radical chemistry is an important source of organic nitrates, and formation of these nitrates represents a major pathway for loss of NOx from the atmosphere21,22 with significant impacts on [OH] and O3 formation.23−25 Due to the cascade of chemistry following its oxidation, these effects can be felt far from the source regions of isoprene.22

3362 3362 3362 3362 3362 3363 3363 3363 3364 3365 3365 3366 3366 3367 3368 3368 3368 3368 3369 3374 3374 3374 3374 3375 3375 3376 3376 3376 3377 3377 3377 3377 3378 3378 3378 3378 3379 3379 3379 3379 3379 3379 3380 3380 3380 3380 3380 3380 3381 3338

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

studies, we provide the greatest detail on the first (section 3) and subsequent (section 7) generations of this chemistry. In our mechanism, we attempt to provide reaction recommendations for all products formed in ≥1% overall yield from isoprene in typical atmospheric conditions, with the exception of ubiquitous small fragmentation products (e.g., hydroxyacetone, glyoxal, formaldehyde (CH2O), and PeroxyAcetyl Nitrate (PAN)), for which we defer to previous recommendations. We also provide a detailed description and mechanism for the chemistry following the oxidation of isoprene by NO3 (sections 5 and 7.5), which has recently been elucidated in multiple studies. While we review the mechanisms of oxidation of isoprene by ozone (section 4) and chlorine (section 6), in our full mechanism we only provide a cursory treatment of the ozone chemistry, and we omit chlorine chemistry altogether. To assist in the visualization of the complex mechanisms described in this review, our full mechanism recommendations are illustrated in each section’s corresponding figures, with product yields reported at T = 298 K and P = 1013 hPa., and recommendations for the condensed mechanism are described in the figure captions. Branching ratios in the figures are shown in bold font when they are constrained by experimental observations, in italics when they are derived from the generic parameterizations described in sections 2.1−2.4, and in plain font for all other cases (e.g., inferred but not directly measured in experiments, or extrapolated from similar compounds). Finally, to avoid cluttering the schemes further, where molecular oxygen is produced, it is not shown. More detailed treatments of the reactions can be found in a separate model repository (http://dx. doi.org/10.22002/D1.247), where the full and reduced mechanisms are made available as computer-readable codes for communal use and development. Finally, when describing the myriad isomers of products formed in the oxidation of isoprene, we frequently employ an abbreviated numbering scheme for the naming of compounds with multiple substituents. To denote the positions of isoprene’s reactions and substituents in the figures and text, we assign numbers to the carbons of isoprene as follows: carbons 1−4 comprise the conjugated butadiene backbone, with the methyl substituent (carbon 5) connected to carbon 2 of the backbone. Throughout the text, we refer to these carbons as “C#” without subscripts (e.g., “C2”); subscripted numbers (e.g., “C2”) are used instead to refer to the total number of carbon atoms in a molecule. In the names of functionalized isoprene oxidation products, we drop the “C” when describing substituent positions; for example, (1-OH,2-ONO2)-isoprene hydroxy nitrate (IHN) has a hydroxy group at C1 and a nitrooxy group at C2.

The large flux and rapid oxidation of isoprene conspire to make its chemistry, and the chemistry of the associated reaction products, of high global importance. The reactions have broad implications for regional air quality24,26,27 and global climate.28 For example, isoprene chemistry significantly affects tropospheric ozone production by perturbing the nitrogen oxide free radical cycle (NOx = NO + NO2 + NO3),29 the rate of oxidation of many compounds by impacting HOx (OH + HO2),30,31 and the Earth’s radiative balance and human health through the production of secondary organic aerosols (SOA).32,33 In addition, the precise impact of isoprene chemistry depends on the environment in which it is oxidized. In particular, the levels of anthropogenically emitted pollutants such as NOx and SOx will affect the oxidation pathways of isoprene and its oxidation products. Finally, the mechanism for degradation of isoprene has been subject to significant scrutiny due to efforts to square photochemical model estimates and measurements of the abundance of OH in regions with large isoprene fluxes.31,34 Quantification of the global impacts of isoprene chemistry requires, at minimum, an accurate description of oxidation mechanisms and VOC emission fields.22,35−38 Over the past 10 years, renewed experimental and theoretical efforts in elucidating the detailed photochemical and dark oxidation mechanisms of isoprene have advanced our understanding of the chemistry substantially (references herein). However, there are still large gaps in our knowledge that preclude the inclusion of an accurate description of isoprene chemistry into atmospheric chemical transport models (CTMs). These lingering kinetic and mechanistic uncertainties translate to uncertainties in the simulations of global air quality and climate feedback. In the spirit of past reviews on the isoprene oxidation mechanism,39−43 this work44 seeks to reduce such uncertainties by compiling a state-of-the-science mechanism with a particular focus on incorporation of recent laboratory and computational studies.

2. MECHANISM DEVELOPMENT This work has three primary goals: (1) to review the results of recent laboratory and computational studies and to place them within the context of previous knowledge on the dark and photochemical reactions of isoprene; (2) to provide a recommendation for a comprehensive mechanism describing the atmospheric reactions of isoprene for use in detailed chemical models; (3) to synthesize a condensed mechanism that adequately captures the primary features of the isoprene chemistry but with a footprint small enough to be implemented in global CTMs. In order to make our mechanisms relevant to CTMs and other broad applications within atmospheric chemistry, we prioritize retaining carbon balance and providing accurate representations of the impact of isoprene emissions on HOx and NOx free radical concentrations, as well as the formation of products that may contribute to organic aerosol formation. The following sections of this review catalog these mechanisms in detail, describing recent advances in the measurements of reaction rates and product yields. The remainder of section 2 describes generic parameterizations of rate constants and branching ratios used for broad classes of reactions throughout the review, including reactions of alkenes + OH, peroxy radicals (RO2) + NO/HO2, H-shift isomerizations, and pressure-dependent rate formulas. Sections 3−7 then proceed systematically through the steps of isoprene oxidation. As oxidation by OH accounts for ∼90% of the atmospheric fate of isoprene and has been the focus of most recent experimental

2.1. Alkene + OH Reactions

2.1.1. Kinetics. The rates of OH addition to alkenes have been measured for a wide range of functionalized compounds; we therefore do not parametrize these rate coefficients in our mechanism. Rather, for alkenes without measured OH addition rates, or without measured temperature dependences, we use the measured rate coefficients and/or temperature dependences of structurally similar compounds. 2.1.2. Mechanism. For linear and branched alkenes, OH preferentially adds to the less substituted carbon, forming the more substituted (and therefore more stable) alkyl or allylic radicals. The ratios of primary:secondary, secondary:tertiary, and primary:tertiary OH addition have been measured for a number of unsaturated hydrocarbons,45−52 typically by detection of stable products. Peeters et al.53 incorporated measurements into 3339

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Figure 1. Nitrate branching ratio (α, on color scales) as a function of temperature and pressure for a peroxy radical with n (=number of non-hydrogen atoms, not including the peroxy moiety) = 6 (left) and n = 11 (right), using the parameterization in eqs 2−4. The thick black lines show typical altitude profiles of temperature and pressure over a tropical continent.

approach to use the sum of all heavy atoms rather than just carbon, consistent with the recommendation of Teng et al.,52 who measured the hydroxy nitrate branching ratios (α) from a variety of alkenes at 293 K and 993 hPa to derive the formula

a structure−activity relationship (SAR), which has agreed well with subsequent measurements. Most recently, Teng et al.52 constrained the branching ratios for locations of OH addition by measuring the stable hydroperoxides formed from the subsequent reaction of HO2 with alkene-derived hydroxy peroxy radicals. Here, we use experimentally constrained branching ratios for locations of OH addition wherever possible. For isoprene, as discussed below (see section 3.2), we assume that OH addition only forms the allylic radicals. For other alkenes where the site of OH addition has not been studied experimentally, we use a parameterization that draws upon the values reported by Peeters et al.53 and Teng et al.:52 0.75:0.25 for primary:secondary OH addition, 0.70:0.30 for secondary:tertiary OH addition, and 0.90:0.10 for primary:tertiary OH addition. These values typically fall within the error bounds of measurements made for relevant alkenes; where measurements deviate substantially from the parameterization for isoprene-derived compounds, these departures are noted in the text.

α0(n) =

(2)

for n ≥ 3, where n represents the number of heavy atoms excluding the peroxy moiety. Taking this formula as a baseline branching ratio (α0(n)), we add the temperature and pressure dependence from Arey et al.56 to derive the following parameterization: αRONO2(T , M , n) =

A(T , M , n) A(T , M , n) + A 0(n)

1 − α0(n) α0(n)

(350 K/ T )

)e

3

cm molecule

(3) −3

where A0(n) = A(T = 293 K, M = 2.45 × 10 molecules cm , n), and A(T, M, n) is given by the Arey et al.56 formula: 19

2.2.1. Kinetics. The rate coefficients for reactions between NO and RO2 (excluding acyl peroxy radicals) do not vary significantly with the structure of the peroxy radical;54,55 we therefore use a single temperature-dependent rate coefficient formula for all non-acyl peroxy radical + NO reactions: k RO2 + NO = (2.7 × 10

k RO2 + NO,total

= (0.045 ± 0.016)n − (0.11 ± 0.05)

2.2. RO2 + NO Reactions

−12

k RO2 + NO → RONO2

γ en[M ]

A(T , M , n) =

γ en[M ] 0.43(T (K) / 298)−8

1+

n

−8

× 0.41(1 + [log(γ e [M ]/0.43(T(K)/298)

−1 −1

−22

)]2 )−1

(4)

−1

where γ = 2 × 10 cm molecule , [M] is the number density of air (molecules cm−3), and n is again the number of heavy atoms excluding the peroxy moiety. To incorporate these complex parameterizations into the mechanism, we denote the rate coefficients of nitrate-forming reactions with “k_nitrate[X,Y,Z,n]” or “NIT(X,Y,Z,n)”, where X and Y are the preexponential and exponential factors in the Arrhenius equation for the overall reaction rate, Z is the normalization term for the nitrate branching ratio (equal to A0(n)(1 − α0)/α0 in eq 3), and n is the number of heavy atoms excluding the peroxy moiety. Thus, as in eq 3, the nitrate formation rate coefficient is given by the equation

s

(1)

2.2.2. Mechanism. The reaction of NO and RO2 yields both alkoxy radicals plus NO2 and organic nitrates. Accurate simulation of organic nitrate formation in these reactions is essential for properly describing the influence of isoprene chemistry on NOx, and therefore on HOx and ozone formation as well.22,24,25 Because different nitrate isomers can have significantly different subsequent chemistries, it is generally necessary to capture the isomer distribution. The branching ratios to form organic nitrates from reactions of NO with peroxy radicals are complicated functions of pressure, temperature, molecular size, and structure. While it is not possible to encompass this diversity in a simple set of rules, we take as a starting point the parameterization recommended by Arey et al.,56 which is an update of earlier formulations by Carter and Atkinson.57,58 Arey et al.56 constrained their parameterization with data only for simple alkanes. We modify their

3

⎛ A(T , [M ], n) ⎞ k nitrate = (X e−Y / T )⎜ ⎟ ⎝ A(T , [M ], n) + Z ⎠

(5)

Similarly, we denote non-nitrate-forming RO2 + NO reactions with “k_alkoxy[X,Y,Z,n]” or “ALK(X,Y,Z,n)”, where the same parameters are used with only a minor adjustment: 3340

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews ⎛ ⎞ Z kalkoxy = (X e−Y / T )⎜ ⎟ ⎝ Z + A(T , [M ], n) ⎠

Review

k RO2 + HO2 = (2.82 × 10−13)e(1300 K/ T )[1 − e(−0.231n)] (6)

cm 3 molecule−1 s−1

Using n = C + O + N − 2, this fit (shown in Figure 2) has an adjusted correlation coefficient (R2) of 0.81 and root-mean-

While these formulas are significantly more complex than other rate coefficients in the model, the dependence of the branching on temperature and pressure is necessary to accurately describe radical cycling through organic nitrates in the atmosphere. However, for some applications of the model, the full parameterization may not be necessary; in these cases, we recommend calculating branching ratios at the desired temperature and pressure before running the simulation, and scaling the Arrhenius parameterization for the rate coefficients by the relative yields of each pathway. The pressure and temperature dependences of the nitrate branching ratio resulting from the parameterization described here are shown in Figure 1 for peroxy radicals with n = 6 (e.g., isoprene + OH + O2) and n = 11 (e.g., monoterpenes + OH + O2). For isoprene, and indeed for all molecules with n = 4−7, the pressure and temperature dependence of the yields largely compensate so that the branching calculated using the modified Arey approach at conditions throughout the lower atmosphere is not significantly different from that at the surface (Figure 1). Laboratory measurements of the nitrate branching ratios for multifunctional compounds often fall below this recommendation, possibly reflecting a shorter lifetime for the hot OONO molecule formed in the initial addition of NO to the peroxy radical. In our mechanism, we use insight gleaned from these studies to modify eq 3 to account for the dependence of the branching ratio on molecular structure. For example, we reduce nitrate yields from primary peroxy radicals by 25%, following substantial experimental evidence,52,56−58 and increase those from tertiary peroxy radicals by 25%, following studies by Teng et al.52 Additionally, we reduce by 90% the nitrate yields from reactions of α-carbonyl peroxy radicals + NO, following insights from observed nitrate branching ratios of methyl vinyl ketone (MVK) and methacrolein (MACR);59 we similarly reduce nitrate yields from β-hydroxy, β-hydroperoxy, and β-nitrooxy peroxy radicals by 10, 60, and 90%, and apply reductions of half these magnitudes for the same functional groups in γ positions (β for carbonyls). The effects of multiple functional groups are combined multiplicatively. These adjustments are poorly constrained, however, and so where laboratory measurements of nitrate branching ratios have been made, these are explicitly noted, and are generally used in place of the formula described above (and shown in bold font in figures).

Figure 2. Individual measurements and parameterizations of kRO2+HO2 as a function of n, where n represents the number of carbon atoms (black) or the number of heavy atoms excluding the peroxy moiety (red).

square error (RMSE) of 0.23, compared to an R2 of 0.71 and RMSE of 0.29 for the fit with n = C. We use eq 8 for all alkyl peroxy radical + HO2 reactions in our mechanism. For reactions of acyl peroxy radicals with HO2, we generally use the IUPAC recommendation for CH3C(O)O2 + HO2. While many experimental studies have investigated the reaction between CH3C(O)O2 and HO2, older studies62−64 may be subject to considerable bias, as they did not account for the OHgenerating product channel. More recent studies65−67 and reevaluations of the historical data68 suggest a value of kCH3C(O)O2+HO2 = (2.0−2.4) × 10−11 cm3 molecule−1 s−1 at 298 K, and a reduced temperature dependence. IUPAC recommends a rate coefficient of kCH3C(O)O2+HO2 = (3.14 × 10−12)e(580 K/T) cm3 molecule−1 s−1. While larger acyl peroxy radicals may react more quickly if acyl peroxy + HO2 reactions follow the same size dependence as alkyl RO2 + HO2 reactions, the rate of CH3C(O)O2 + HO2 is near the asymptotic limit of the sizedependent alkyl peroxy + HO2 rate (2.2 × 10−11 cm3 molecule−1 s−1 at 298 K). 2.3.2. Mechanism. Reactions of organic peroxy radicals with HO2 are generally expected to form hydroperoxides, terminating the radical chain. They can, however, also result in radical propagation via a channel that forms an alkoxy radical and OH or carbonyl plus OH and HO2 (see Orlando and Tyndall55). These channels were initially identified for acyl peroxy radicals67,69−72 and alkyl peroxy radicals containing halogens. They have recently, however, also been shown to be important in other peroxy radical species, with higher yields for α-carbonyl and more highly substituted species.59,72−74 Insight from theoretical studies suggests that the radical recycling pathways may proceed through a singlet tetroxide intermediate (ROO−OOH) that, when stabilized by hydrogen bonding between the − OOH hydrogen and other functional groups on the RO2 radical, will decompose to either RO, O2, and OH75−78 or HO2, OH, and a carbonyl compound.55,59 For acyl peroxy radicals, the reaction of CH3C(O)O2 with HO2 has been studied in detail, and provides a model for analogous acyl peroxy radicals throughout our mechanism. The

2.3. RO2 + HO2 Reactions

2.3.1. Kinetics. The reaction rate coefficients of nonacyl peroxy radicals with HO2 have been shown to follow a simple pattern that depends both on temperature and on the number of carbon atoms in the peroxy radical:41,60 k RO2 + HO2 = (2.91 × 10−13)e(1300 K/ T )[1 − e(−0.245n)] cm 3 molecule−1 s−1

(8)

(7)

where n represents the number of carbon atoms in the peroxy radical molecule. We suggest, similar to the organic nitrate yield parameterization, that the rate scales more closely with the number of heavy (non-hydrogen) atoms. Refitting the HO2 + RO2 data, using IUPAC recommended rates where available and rates from Boyd et al.61 for additional species, with n = C + O + N − 2 (subtracting off the peroxy moiety) gives an improved fit to the experimental data, with a slightly different formula: 3341

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

of the major methacrolein hydroxy peroxy radical (section 7.2 and Crounse et al.88), and the 1,5 and 1,6 α-hydroxy H-shifts in hexanol.87 Throughout our mechanism, we recommend intramolecular H-shift rate coefficients (including temperature dependences) analogous to the measured examples described above whenever such isomerizations are likely to occur on time scales similar to bimolecular reactions. These include 1,4, 1,5, and 1,6 H-shifts from aldehydic hydrogens, to which we assign rates at 298 K of 0.5, 20, and 5 s−1 respectively; 1,5 and 1,6 α-hydroxy H-shifts, to which we assign rates at 298 K of 5, 0.1, and 0.05 s−1 for allylic, secondary, and primary hydrogens respectively (the 1,5 and 1,6 shifts do not differ substantially87); and 1,5 and 1,6 αhydroperoxy H-shifts, to which we assign rates equal to half those of the α-hydroxy H-shifts based on unpublished data from chamber studies at Caltech. The α-carbonyl H-shifts are assigned temperature dependences of −(5000 K)/T, by analogy with methacrolein, while the α-hydroxy and α-hydroperoxy H-shifts are assigned temperature dependences of −(10000 K)/T. αNitrooxy hydrogens are not considered labile to hydrogen shifts. In our full and condensed mechanisms, the 1,6 α-hydroxy Hshifts described in section 3.4.4 are modeled with a strongly negative T-dependent tunneling factor, best expressed as an Arrhenius rate parameterization with an additional exponential term that includes T−3 in the argument. We denote these rates with “TUN(A,B,C)”, where A, B, and C can then be applied in the following formula:

reaction can proceed via three channels, yielding peroxyacetic acid via channel 9, acetic acid and ozone via channel 10, and propagating OH radical via channel 11: CH3C(O)O2 + HO2 → CH3C(O)OOH + O2

(9)

CH3C(O)O2 + HO2 → CH3C(O)OH + O3

(10)

CH3C(O)O2 + HO2 → CH3C(O)O + OH + O2

(11)

Numerous studies have measured the products of channels 9 and 10; together, they suggest a ratio of 2.8:1 for the relative contributions of the two pathways. Indirect evidence for channel 11 was first presented by Hasson et al.67 and Jenkin et al.;70 more recent studies directly detected OH production confirming a large contribution from this pathway (0.5).65,66,71 While the contributions of each individual pathway may be temperaturedependent,79 and data from recent studies could be consistent with a small pressure dependence65,71 (though the studies do not report one, as the variation with pressure is within estimated error bounds), we recommend a constant branching ratio of 0.37:0.13:0.50 for channels 9:10:11 for all acyl peroxy radicals. For alkyl peroxy radicals, we use measured hydroperoxide yields wherever possible to constrain the branching between radical terminating and recycling pathways; where such data do not exist, we extrapolate from related compounds using the observations compiled in Orlando and Tyndall55 and in more recent studies showing evidence of radical propagation from this channel,59,80−83 which tend to show that increased functionalization of the peroxy radical (particularly α-carbonyl and β-nitrate substituents, and other functional groups enabling intramolecular hydrogen bonding of the tetroxide intermediate) increases the radical yields. We use these observations to create a rule similar to that described above for nitrate formation from RO2 + NO reactions. From a base hydroperoxide branching ratio of 1, we subtract 0.1 for every β-hydroxy group and 0.5 for every α-carbonyl, β-nitrate, or β-hydroperoxide group; additionally, we subtract half these amounts for the same functional groups in γ positions (β for carbonyls). The effects of multiple functional groups are additive up to a maximum of 20% radical propagation for primary peroxy radicals, 75% for secondary peroxy radicals, and 100% for tertiary peroxy radicals. These substituent effects are, however, highly uncertain and are likely to vary substantially from molecule to molecule, depending on the availability of hydrogen bonding and the conformations toward which molecules tend. With the exception of MVK oxidation (section 7.1), we assume that radical recycling (1-branching fraction to ROOH) occurs via formation of alkoxy radicals. The alternative channel via formation of a carbonyl + OH + HO2 is likely active in many of the schemes detailed here, but is poorly constrained at this time.

k tunneling = Ae−B / T eC / T

3

(12)

We also include intramolecular H-shifts between hydroperoxy hydrogens and peroxy radicals, which have not been explicitly measured but have been calculated to occur rapidly.89,90 The rates of these shifts are thought to depend predominantly on availability of intramolecular hydrogen bonding, which makes their estimation from structure−activity relationships difficult. Given the lack of experimental or theoretical constraints on such shifts in most isoprene-derived hydroperoxy peroxy radicals, we assign high, temperature-independent rate coefficients, weighted by the substitution at the carbon attached to the peroxy radical formed in the H-shift (tertiary > secondary > primary). The recommendations described above should be viewed as simple placeholders until more comprehensive investigations are available that better constrain the rates of these reactions. In particular, it is known that the structures of the molecules undergoing unimolecular chemistry are critical in determining the rates. Hydrogen bonding and steric effects, for example, can alter the rates by orders of magnitude. 2.5. Additional Rate Parameterizations

Most rate coefficients in our mechanism are formulated with the traditional Arrhenius exponential parameterization for temperature dependence (k = Ae−r/T). Those that are not directly addressed in the following sections are either taken from IUPAC recommendations or MCM v3.3.143 or, when no experimental data have previously been reported in the literature, extrapolated from analogous reactions. In addition to the exceptions described above for RO2 + NO reactions, RO2 + HO2 reactions, and H-shift isomerizations, our full mechanism employs two rate constant formulas with more complex parameterizations: the pressureand temperature-dependent Troe modification of a Lindemann−Hinshelwood rate expression,91 and the pressuredependent expression for epoxide formation rates formulated by Jacobs et al.,83 both described below. The reduced mechanism

2.4. Hydrogen Shifts to Peroxy Radicals

We have incorporated into our mechanism intramolecular hydrogen shifts of peroxy radicals wherever we expect their rates to be competitive with bimolecular chemistry in the environment. The rate coefficients of these H-shifts, which play an important role in the rapid formation of highly oxidized VOCs via autoxidation, have seldom been measured. They are known to be strongly temperature dependent and influenced by quantum tunneling and intramolecular H-bonding.31,84−87 We must, therefore, rely on theoretical calculations and on extrapolation from the few existing measurements, which include the 1,6 αhydroxy H-shifts of the Z-δ isoprene hydroxy peroxy radicals (section 3.4.4 and references therein), the 1,4 α-carbonyl H-shift 3342

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Figure 3. Dynamics of the isoprene + OH + O2 system. In our recommendations, 1-OH addition (top) accounts for 63% of isoprene + OH reactivity, while the 4-OH addition system accounts for the remaining 37%. For the full mechanism, we explicitly treat all forward and backward reaction rates. For the reduced mechanism, we treat the peroxy radical (ISOPOO) pool as a steady-state distribution, using bimolecular chemistry only for the β- and E-δISOPOO isomers, and parametrizing the isomerization loss of the Z-δ-ISOPOO isomers, as described in section 3.3. These reductions are appropriate for atmospheric conditions where the ISOPOO bimolecular reactivity, kbimolecular, is less than 1 s−1 and at temperatures greater than 280 K and for pressure between 500 and 1000 hPa.

additionally includes a number of unique rate formulas as part of the simplification of the initial isoprene hydroxy peroxy radical dynamics (section 3.3.5), which are not discussed in detail here. The modified Lindemann−Hinshelwood rate coefficient91 is applied only in two reactions within our mechanism: the formation and decomposition of methacryloyl peroxynitrate (MPAN). These rates are denoted with “TROE(A,B,C)”, where A, B, and C, respectively, signify k0, k∞, and Fc in the equation k Troe =

k 0[M ] 1+

k 0[M ] k∞

concentration in the rate coefficient to skip a step and remove an intermediate compound from the mechanism. In the full mechanism, OH addition to a double bond forms an alkyl radical, which can then either add O2 or form an epoxide from a β-hydroperoxy or -nitrooxy group. Thus, the branching ratios of O2 addition and epoxide formation are dependent upon the O2 concentration. Alternatively, the reaction can be written in a single step (OH addition to directly form an epoxide) if the O2 concentration is included in the rate coefficient; these rates are denoted “EPO(A,B,C)”, where A, B, and C can then be applied in the following formula:

2 −1

Fc(1 + [log10(k 0[M ]/ k∞)] )

(13)

kepoxide = A e−B / T [1/(C × [O2 ] + 1)]

Our representation of the branching ratio to epoxide + OH (NO2) from hydroperoxide (-nitrooxy) groups found β to alkyl radicals as a simple pressure-dependent yield (see, for example, section 6.3) can be understood as a means of including the O2

(14)

This parameterization allows for the removal of a species from the mechanism (the intermediate alkyl radical), which may benefit models limited by computational capacity, but at the cost 3343

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

of making the rate formulations more complicated. We provide both parameterizations, such that either can be used depending on the priorities and limitations of the modeling platform. The full mechanism includes the intermediate alkyl radical species and their relevant reactions for every epoxide-forming step in the mechanism (including those of isoprene hydroxy hydroperoxide (ISOPOOH), isoprene hydroxy nitrates, and other functionalized nitrates formed from NO3 chemistry). A separate module is also included in the model with the reactions required for the [O2]-dependent rate coefficient parameterization, as well as a list of reactions to remove from the full mechanism.

Our recommendation for the complex peroxy radical chemistry that follows addition of OH to isoprene follows the structure described by Peeters et al.31,85 The dynamic nature of the isomer-specific chemistry results in variable product yields that depend on atmospheric conditions. In particular, the reaction pathways change depending on the bimolecular lifetime of the hydroxy peroxy radicals (ISOPOO) and temperature. The recommended reaction rate coefficients, thermochemistry, and product yields needed to describe this chemistry are derived from Teng et al.95 One essential aspect of our recommendation is that the chemistry following addition of OH to either C1 or C4 (see numbering in Figure 3) is sufficiently distinct that these systems must be treated separately. Finally, in addition to the explicit rates and yields as detailed in Teng et al.,95 we provide a recommendation for simplified parameterizations that capture this highly complex chemistry for most atmospheric conditions.

2.6. Implementing the Full and Condensed Mechanisms

To assist in implementing the schemes illustrated here into photochemical models, full and condensed versions of the mechanism are provided in a separate repository (http://dx.doi. org/10.22002/D1.247). Included are the reactions described in this paper along with rate constants and product distributions. Two different versions of the condensed mechanism are delineated: one that only includes the reactions described in this review, and another that includes additional reactions of terminal species (largely adapted from MCM v3.3.143) such that all species can react to eventually form CO2. The mechanisms are in a format suitable for use with the Kinetic Preprocessor tool (KPP),92−94 which includes a “.def” file (the model parameters and initial conditions), a “.spc” file (a list of species, which can be defined as fixed or variable), and a “.eqn” file (the chemical equations for the mechanism). We also include the pictorial representation of the schemes shown in both the main manuscript and in the SI in their native format. A “README” file included with the mechanisms provides a longer description of their contents. While the full mechanism treats all isomers of the oxidation products of isoprene separately if they are known to have different subsequent reaction rates and products, the condensed mechanism generally makes use of isomer grouping (e.g., “δhydroxy nitrate” to represent E-δ-(1-OH,4-ONO2)-, Z-δ-(1OH,4-ONO2)-, E-δ-(1-ONO2,4-OH)-, and Z-δ-(1-ONO2,4OH)-isoprene hydroxy nitrates) to minimize the number of species and reactions included in the mechanism while retaining an accurate description of the oxidative fate of those grouped species. The condensed mechanism also uses a steady-state approximation for the initial isoprene hydroxy peroxy radical distribution (see section 3.3.5), enabling the removal of many intermediate species and reactions. This approximation is quite accurate for atmospheric simulations, but it may be inappropriate for many laboratory experiments where radical abundances are elevated. The mechanism is further reduced by lumping particularly minor pathways (20 hPa) has been measured numerous times, and the JPL20 and IUPAC19 reviews provide consistent recommendations: k OH + isoprene = (3.0 × 10−11)e(360 K/ T ) cm 3 molecule−1 s−1 [JPL] k OH + isoprene = (2.7 × 10−11)e(390 K/ T ) cm 3 molecule−1 s−1 [IUPAC]

As the bulk of isoprene is oxidized in the atmosphere over a rather small temperature range (e.g., 280−315 K), either expression can be used (these parameterizations are within 1%). The estimated uncertainty in this rate coefficient (T = 300 K) is less than 10%.19,20 Here, we use the IUPAC expression. 3.2. Location of OH Addition

The reaction of OH with isoprene at temperatures relevant for atmospheric chemistry proceeds mainly via addition to the unsaturated backbone. Despite the allylic resonance, abstraction of the methyl hydrogen is likely less than 1% of the total reaction under atmospheric conditions (cf. recent calculations on propene96 and isoprene97). OH can add at any of four positions on the conjugated carbon chain (Figure 3). Consistent with both theory and experiments, the majority of the addition occurs at the terminal (primary) carbons. On the basis of theoretical calculations by Greenwald et al.,98 IUPAC recommends a ratio of 0.67:0.02:0.02:0.29 for isomers 1, 2, 3, and 4. Work in Simon North’s laboratory has suggested that, following internal addition, isomerization via a cyclopropane-like intermediate can lead to the formation of an unsaturated ketone (via C2 addition) or aldehyde (via C3 addition), as shown in Figure 3.99,100 Recent laboratory studies at Caltech suggest that formation of β-ISOPOO radicals via OH addition at C2 and C3 is less than 1% of the total.95 Given the lack of experimental evidence, we suggest that the internal addition channels be ignored in atmospheric modeling for now, though their importance should be revisited with an additional focus on potential for aerosol formation via H-shift autoxidation.101 We recommend yields of 0.63:0.00:0.00:0.37 for OH addition to C1:C2:C3:C4.95

3. REACTION OF OH WITH ISOPRENE Reaction with OH represents the largest loss pathway for isoprene in the atmosphere, owing both to their synchronous diurnal cycles of production and to their fast reaction rate coefficient. Globally, oxidation by OH is estimated to account for ∼85% of the reactive fate of isoprene.22

3.3. Addition of O2 to Hydroxy Allylic Radicals

3.3.1. cis/trans Allylic Radical Distribution. Following addition of OH to the unsaturated backbone, two separate pools 3344

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

of allylic radicals are established: the 1-OH and 4-OH adducts (Figure 3). In each pool, the rotation of the internal carbon bond between cis and trans OH−isoprene adducts is rapid upon OH addition but slows upon thermalization.31 The initial distribution of cis to trans adducts is set, therefore, by the steady-state distribution upon collisional thermalization. Peeters et al.31 calculated the distribution of cis:trans allylic radicals of 0.46:54 for the 1-OH adducts and 0.69:0.31 for the 4-OH adducts; there are no experimental constraints on these ratios. In our mechanism, we assign these as 0.50:0.50 and 0.70:0.30, respectively. In the atmosphere, where the peroxy radical pools generally live for longer than 1 s, this initial branching ratio does not substantially impact the final isomer distribution of isoprene hydroxy peroxy radicals (ISOPOO). 3.3.2. Initial (Kinetic) Distribution of Isoprene Hydroxy Peroxy Radicals (ISOPOO). Following collisional thermalization, each OH-isoprene allylic radical will add oxygen at either the β or δ position to form three distinct ISOPOO radicals (Figure 3). The overall distribution of ISOPOO is initially set, therefore, by both the trans/cis allylic distribution and the oxygen forward addition rates for the β and δ positions. Isomer-specific O2 addition rates are not experimentally constrained, and therefore provide little insight into the initial ISOPOO distribution. Previous estimates for the initial distribution of peroxy radicals have been based on computational estimates102 or bulk product analysis.103 Using a speciated distribution of

Table 3. Kinetic ISOPOO Isomer Distributions at 297 K distribution

k1 k2 k3 k4

relative rate

1-OH-trans + O2 → E-δ-(1-OH,4-OO)-ISOPOO 1-OH-trans + O2 → β-(1-OH,2-OO)-ISOPOO 1-OH-cis + O2 → β-(1-OH,2-OO)-ISOPOO 1-OH-cis + O2 → Z-δ-(1-OH,4-OO)-ISOPOO

0.48 1 1 0.18

Table 2. O2 Addition Rates for 4-OH Isoprene Adducts, Relative to the Average of 4-OH-cis and 4-OH-trans + O2 → β(3-OO,4-OH)-ISOPOO Assuming a cis/trans ratio of 0.7/ 0.3a relative rate

4-OH-trans + O2 → E-δ-(4-OH,1-OO)-ISOPOO 4-OH-trans + O2 → β-(4-OH,3-OO)-ISOPOO 4-OH-cis + O2 → β-(4-OH,3-OO)-ISOPOO 4-OH-cis + O2 → Z-δ-(4-OH,1-OO)-ISOPOO

0.75 1 1 0.32

(1-OH,2-OO)E-δ-(1-OH,4-OO)Z-δ-(1-OH,4-OO)(4-OH,3-OO) E-δ-(4-OH,1-OO)Z-δ-(4-OH,1-OO)-

0.479 0.102 0.049 0.259 0.048 0.063

0.415 0.044 0.169 0.263 0.014 0.095

dation is slightly modified from that presented in Teng et al.95 as we adopt isomer-dependent branching ratios for the formation of hydroxy nitrates from ISOPOO + NO reactions, as discussed below. Assuming the ratio of cis:trans allylic radicals calculated by Peeters et al.31 is accurate, the isomer distribution of the organic nitrates produced from ISOPOO implies a similar ratio of the cis:trans oxygen addition rate at the δ position in both 1-OH and 4-OH isoprene allylic radicals.95 There exists, however, significant disagreement between the experimentally derived O2 addition rate coefficient recommendations here and recent theoretical calculations. Peeters et al.,85 for example, suggested that the O2 addition rate to the cis δ position is significantly faster than to the trans δ position in both 1-OH and 4-OH systems. For both the theory and experiments, the absolute recommendations for individual isomer O2 addition rates are constrained only by the assumed bulk O2 addition rates. Here, we recommend an overall O2 addition rate constant of 2.0 × 10−12 cm3 molecule−1 at 298 K,104 though within the framework for our recommendation, this choice has little impact on the product yields under atmospherically relevant conditions, provided that self-consistent kinetics are used for the loss of O2 (section 3.3.3). 3.3.3. Reversibility of O2 Addition. Peroxy radicals formed β to olefinic carbon centers, such that electronic resonance is reestablished upon loss of molecular O2, have relatively small R− OO bond dissociation energy ( 1 × 1013 molecules cm−3), the mechanism we present herein gives MVK and MACR branching fractions of 41.2 and 22.5% respectively at 298 K. bCorrected for O3P reaction as communicated to Paulson et al.39 c This is a lower limit because for the conditions of this study a significant fraction of the peroxy radicals isomerize via 1,6 H-shift following OH addition at C4. dCorrected for the 7% of ISOPOO that do not react with NO.80

Peeters et al.125 predicted that the −OH group lowers the barrier to decomposition by a full 8 kcal mol−1, making decomposition much more favorable relative to the competing reaction with O 2 . 125−129 As detailed in Table 7, yields determined experimentally under NO dominated conditions range from 30 to 45% for MVK and 20−30% for MACR. In the presence of oxygen, formaldehyde is produced in a yield equal to the combined yield of MVK and MACR.118,119,130 Because the fraction of ISOPOO that are β isomers varies with peroxy radical lifetime, the yield of MVK and MACR from isoprene oxidation is not fixed. As shown in Table 7, the NO concentrations in many of the chamber experiments (except for those of Karl et al.132 and Liu et al.80) are much higher than those typically found in the troposphere. As discussed in section 3.3, in the atmosphere the ISOPOO isomer distribution favors the thermodynamically more stable β-substituted isomers. Since these isomers (those leading to MVK and MACR) are typically present at a lower fraction in chamber experiments than in the atmosphere, such experiments systematically underpredict MVK yields in the atmosphere, and may also incorrectly represent MACR yields, which are further complicated by the rapid isomerization of Z-δ-(4-OH,1-OO)-ISOPOO. Karl et al.132 and Liu et al.,80 who used NO concentrations comparable to the urban troposphere (2 × 1010 and 5 × 109 molecules cm−3, respectively), found MVK and MACR yields higher than many other studies that employed higher NO concentrations, with the notable exception of Sprengnether et al.119 (Table 7). The estimated MACR yield by Karl et al.132 is likely a lower limit as they were unaware that a significant fraction of the hydroxy peroxy radicals from the 4-OH system were undergoing 1,6 Hshift isomerization or reaction with HO2. In addition, subsequent studies have illustrated that MVK and MACR measurements can often be impacted by analytical challenges when ISOPOOH is present.108 Here, for the β-ISOPOO + NO reactions, we take the branching ratio of MACR and MVK to be (1 − α), where α is the branching ratio to form IHN. 1,4- and 4,1-δ-ISOPOO react with NO to form δ-ISOPO radicals (Figure 6). The Z-δ-ISOPO radicals undergo a 1,5 3349

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Figure 7. Reactions of isoprene−CO−OH−OOH−OO isomers produced by δ-ISOPOO + NO, as shown in Figure 6. Branching ratios are for 298 K and 1 atm; the relative contributions of the nitrate and alkoxy pathways vary with both temperature and pressure. For the reduced mechanism, we assume that the 1,5 and 1,6 H-shifts from the aldehyde outcompete all bimolecular chemistry.

Figure 8. Reactions of ISOPOO isomers with HO2. Branching ratios for the products of β-ISOPOO + HO2 are derived from Liu et al.80 For the reduced mechanism, we ignore the Z-δ isomers, as they contribute less than 1% of the bimolecular chemistry for atmospheric conditions.

Table 8. Products of ISOPOO + HO2 branching ratio (%) our recommendations products

Liu et al., 2013

ISOPOOH MVK + OH + HO2 + CH2O MACR + OH + HO2 + CH2O

80,a

ISOPOO

93.7 ± 2.1 3.8 ± 1.3 2.5 ± 0.9

1,2-ISOPOO

4,3-ISOPOO

δ-ISOPOO

93.7 6.3 0

93.7 0 6.3

100 0 0

a At RO2 lifetime of ∼4 s (HO2 = 1000 ppt). MVK and MACR branching fractions may be lower to the extent that the laboratory results may have been impacted by ISOPOO + ISOPOO chemistry.

ISOPOOH, with branching ratios estimated to be 88 ± 12%146 or 93.7 ± 2.1%.80 Six possible isomers of ISOPOOH can be formed from this series of reactions (Figure 8), with relative abundances determined by the distribution of the precursor RO2. Under most atmospherically relevant conditions, the two β isomers(1,2)- and (4,3)-ISOPOOHcomprise the majority (∼95%) of the total, with only ∼5% in the E-δ isomers (4,1) and (1,4) for the mechanism shown here. MVK and MACR have been suggested as minor products from ISOPOO + HO2 for the

coefficient of Boyd et al. at 298 K,61 we scale our standard recommendation: kISOPOO + HO2 = (2.12 × 10−13)e(1300 K/ T ) cm 3 molecule−1 s−1

with an uncertainty of ±20% from the experimental rate. 3.4.2.2. Mechanism. The main product of the reaction of ISOPOO with HO2 is the unsaturated hydroxy hydroperoxide, 3350

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Table 9. Self-Reaction Rate Constants (×10−14 cm3 molecule−1 s−1) for the β-ISOPOO Radicalsa

(1,2)-ISOPOO and (4,3)-ISOPOO, respectively.71,80,146 Liu et al.80 determined a branching ratio to ISOPOOH of 93.7% from ISOPOO + HO2, which was calculated from MVK, MACR, and isoprene observations (1−MVK−MACR) while accounting for non-HO2 chemistry (Table 8). Liu et al.80 measured branching ratios of MVK (3.8 ± 1.3%) and MACR (2.5 ± 0.9%) separately at an RO2 lifetime of ∼4 s (e.g., [HO2] = 1000 ppt). This result implies that the MACR yield from (4,3)-ISOPOO is slightly higher than the MVK yield from (1,2)-ISOPOO. This seems unlikely and suggests, perhaps, that some fraction of the MVK and MACR observed in the study may have been produced via ISOPOO + RO2. Here, we recommend (1,2)-ISOPOO + HO2 → MVK + OH + HO2 + CH2O with a branching ratio of 6.3% and, equivalently, (4,3)-ISOPOO + HO2 → MACR + OH + HO2 + CH2O with the same branching ratio, 6.3%. The balance yields ISOPOOH. 3.4.3. Reaction with RO2. Peroxy radical self- and crossreactions are important in low-NOx environments. These reactions have been the subject of several review papers.55,147−150 3.4.3.1. Kinetics. Very few direct experimental measurements have been completed on the self- and cross-reactions for ISOPOO. Overall rate constants, as well as the accessible product channels, vary widely with the type of peroxy radical involved. The cocktail of ISOPOO formed during OH-initiated oxidation of isoprene makes deconvolution of the individual isomerspecific rates and branching ratios complex. Jenkin et al.151 used a SAR to infer the self- and cross-reaction rates and branching fractions for ISOPOO radicals. The SAR results were improved later with constraints from OH-initiated oxidation experiments on 1,3-butadiene and 2,3-dimethyl-1,3-butadiene, which differ from isoprene only by the presence or absence of a methyl group on the carbon backbone. These experiments were conducted at high radical concentrations ((2.0−6.8) × 1013 molecules cm−3 of OH) in a small (∼175 cm3) quartz reaction cell.139 The resulting experimental rate constants are much faster than those proposed with SAR arguments using smaller peroxy radicals, but appear to adequately describe the kinetics observed in isoprene oxidation experiments. MCM combines all peroxy radical rate constants into one value for each isomer by using a rate constant which is twice the geometric mean of the self-reaction rate constant for that isomer and the methyl peroxy self-reaction. The overall reaction rate constants for all isomers are based on their structure (primary > secondary > tertiary peroxy radicals) and functionality (β-hydroxy peroxy radicals are significantly activated; allylic peroxy radicals are slightly activated). Here, we consider only the reactions of the two β-ISOPOO radicals, as they comprise the large majority of the RO2 radical pool from isoprene. Table 9 lists the self-reaction rate constants inferred from SAR arguments for smaller peroxy radicals151 with those derived from 1,3-butadiene and 2,3-dimethyl-1,3-butadiene.139 Here we recommend the rates determined by Jenkin et al., 1998.139 Table 10 contains similar data for the cross-reaction between the two β-ISOPOO isomers, and between ISOPOO and CH3O2. Again, we recommend the rates determined by Jenkin et al., 1998.139 3.4.3.2. Mechanism. Peroxy radical (RO2) self-reactions occur via three channels: RO2 + RO2 → RO + RO + O2

(15)

RO2 + RO2 → ROH + R′HO + O2

(16)

RO2 + RO2 → ROOR + O2

(17)

ISOPOO isomer

Jenkin et al., 1995151

Jenkin et al., 1998139

β-(1-OH,2-OO)β-(4-OH,3-OO)-

1.65 139

6.92 ± 1.38 574 ± 57

a Values from Jenkin et al., 1995,151 were estimated by SAR, while those from Jenkin et al., 1998,139 were measured experimentally for 1,3-butadiene or 2,3-dimethyl-1,3,-butadiene, which are structurally similar to isoprene. We recommend the rates determined by Jenkin et al., 1998.139

The branching between these channels controls the contributions of each reaction to the formation of additional radical species, or in some cases low-volatility products. The alkoxy radicals (RO) formed in reaction 15 undergo further reactions described in section 3.4.1.3, propagating the radical chain. Reaction 16 is chain-terminating, forming alcohol and aldehyde products through a hydrogen-bonded transition state involving a hydrogen atom α to the peroxy radical. This product channel does not occur for tertiary peroxy radicals. Reaction 17 has never been directly observed for ISOPOO, but the ROOR product would likely have a low vapor pressure and would be important to the particle phase. In Table 11, we compare the findings of Jenkin et al.139,151 for the branching fraction to form alkoxy radicals by reaction 15. The yield of channel 17 (formation of ROOR) is assumed to be zero, and the branching fraction for channel 16 is the remainder of the rate constant. The tertiary peroxy radical, β-(1-OH-2-OO)ISOPOO, cannot undergo reaction 16, so the branching fraction is 1, by definition. As discussed previously, the alkoxy radicals will likely rapidly decompose to either MVK or MACR with unity yield. Similarly, the cross-reactions have been shown to undergo three channels (ignoring the analogous ROOR channel) where R1O2 and R2O2 are different peroxy radicals: R1O2 + R2O2 → R1O + R2O + O2

(18)

R1O2 + R2O2 → R1OH + R2′HO + O2

(19)

R1O2 + R2O2 → R1′HO + R2OH + O2

(20)

Like the self-reaction, if either R1O2 or R2O2 is a tertiary peroxy radical, the corresponding aldehyde product channel is not accessible. Given the potential influence of reactions 15 and 18 on radical propagation in low-NO environments, the chain branching of the self- and cross-reactions must be considered when developing an overall mechanism for ISOPOO. The treatment of cross-reactions is complex in the case of ISOPOO. Individual isomers are expected to react with any available peroxy radical, including the methyl peroxy radical. In Table 10, we list the rate constants and branching fractions for only the β isomers; a more detailed treatment of the crossreactions can be found in the model itself. In the case of primary or secondary peroxy radicals, channels 19 and 20 are assumed to make up equal portions of the non-alkoxy-forming branching pathway (like channel 16 for the self-reaction system). We also list the rate constant for ISOPOO reactions with CH3O2, which is assumed to be the same for all isomers. 3.4.4. H-Shift Isomerization. Based on theoretical calculations, two HOx recycling pathways involving H-shift isomerizations of the ISOPOO radicals have been proposed.31,85,152 3.4.4.1. 1,6 H-Shift. The Z-δ-ISOPOO radicals of isoprene undergo a H-shift reaction activated by the presence of the α OH 3351

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Table 10. ISOPOO + RO2 Cross-Reaction Rate Constants and Branching Ratiosa reaction

parameter

β-(1-OH,2-OO) + β-(4-OH,3-OO)

all ISOPOO isomers + CH3O2 β-(1-OH,2-OO) + CH3O2 β-(4-OH,3-OO) + CH3O2

−14

3

−1 −1

rate (×10 cm molecule s ) yield: MVK + MACR + 2HO2 + 2CH2O yield: ISOP3CO4OH + ISOP1OH2OH rate (×10−14 cm3 molecule−1 s−1) yield: MVK + CH3O + HO2 + CH2O yield: ISOP1OH2OH + CH2O yield: MACR + CH3O + HO2 + CH2O yield: ISOP3OH4OH + CH2O yield: ISOP3CO4OH + CH3OH

Jenkin et al., 1995151

Jenkin et al., 1998139

15.1 0.875 0.125

308 0.9 0.1 200 0.5 0.5 0.5 0.25 0.25

Individual rate constants (×10−14 cm3 molecule−1 s−1) and expected product branching ratios are indicated for each radical pair. Values from Jenkin et al., 1995,151 were estimated by SAR, while those from Jenkin et al., 1998,139 were measured experimentally for 1,3-butadiene or 2,3-dimethyl-1,3,butadiene, which are structurally similar to isoprene. We recommend the rates and branching ratios from Jenkin et al., 1998.139 a

Table 11. Branching Ratios from the Self-Reactions of the β ISOPOO Radicalsa ISOPOO isomer β-(1-OH,2-OO) β-(4-OH,3-OO) a

Jenkin et al., 1995151

Jenkin et al., 1998139

products

1 0.75 0.25

1 0.8 0.2

2MVK + 2HO2 + 2CH2O 2MACR + 2HO2 + 2CH2O ISOP3CO4OH + ISOP3OH4OH

k(1,6‐H), Z ‐ δ ‐ (1‐OH,4‐OO) 8 3

3

= (5.05 × 1015)e(−12200 K/ T )e(1 × 10 K / T ) s−1

(21)

k(1,6‐H), Z ‐ δ ‐ (4‐OH,1‐OO) 8 3

3

= (2.22 × 109)e(−7160 K/ T )e(1 × 10 K / T ) s−1

(22)

95

Teng et al. estimate that these rates are only certain within a factor of 3.5, which puts their values at ambient temperatures (0.36 and 3.7 s−1, respectively, at 297 K) well within the range of those calculated by Peeters et al.85 (0.49 ± 0.32 and 5.4 ± 3.3 s−1, respectively). The odd Arrhenius parameterization for reaction 21 (i.e., the A-factor makes little sense for a unimolecular process) may reflect the very limited temperature range of the experimental data95 along with the substantial uncertainty in the isomer specific rates of the 1,6 and 1,5 H-shifts. Teng et al.95 suggest alternatively that it may result from the transition state for loss of O2 from this isomer being energetically submerged substantially below the energy of the 1,6 H-shift transition state. If so, however, the apparent agreement between the Peeters et al. calculation of the room temperature rate coefficient22 and the experimental data95 would be fortuitous. In any case, although the experimentally determined absolute rate for the 1,6 H-shift to the Z-δ hydroxy peroxy radicals is highly uncertain due to large uncertainty in the fraction of ISOPOO present in this isomer, uncertainty in the bulk rate coefficient within each system (OH addition at C1 or C4) is much smaller provided self-consistent application of the thermochemistry is employed (see Teng et al.95). 3.4.4.3. Mechanism (Figures 9 and 10). Teng et al.95 determined the yields of δ-HPALDs and other products from the 1,6 H-shifts of the Z-δ-ISOPOO radicals, and proposed a tentative mechanism for the formation of the observed products. We modify them slightly in this work. There remains considerable uncertainty in the yields of HPALDs and other products from the 1,6 H-shifts. Here, we recommend the δHPALD yield measured by Teng et al.95 of 25% from each 1,6 Hshift channel. Teng et al.95 also measured a 15% yield of a species isobaric with δ-HPALD, which remains unidentified. We tentatively assign this mass to two C5 β-HPALDs, the subsequent chemistry of which is proposed in Figure S3. While the balance of the isomerization products remains uncertain, we route the remainder through O2 addition at the β-hydroperoxy position. Teng et al.95 and Crounse et al.86 observed approximately half of the non-HPALD products produced via the 1,6 H-shift chemistry at masses consistent with hydroperoxy acetone and hydroperoxy ethanal. Here, we recommend that these products are formed via

We recommend the branching ratios from Jenkin et al., 1998.139

group and the formation of an allylic radical (Figures 9 and 10). Following the 1,6 H-shift to the peroxy radical, reaction with O2 and elimination of HO2 leads to the formation of unsaturated hydroperoxy aldehydes (HPALDs). Peeters et al.31 suggested that these HPALD compounds photolyze rapidly to generate OH, sustaining HOx, even in the face of high levels of isoprene and low levels of NOx. The significance of H-shift isomerization chemistry is very sensitive to environmental conditions and to the isomer system. Using the kinetics described here, for example, we calculate that for the conditions in the Ozarks (a high isoprene emission region of Missouri) sampled during the summer of 2013 (surface temperature = 25 °C; [NO] ≈ 50 ppt; [HO2] ≈ 20 ppt) 41% of the peroxy radicals formed following OH addition at C4 exit via 1,6 H-shift isomerization, but only 7% do so following OH addition at C1. This fraction falls rapidly with altitude due to the decrease in temperature. At the top of the atmospheric boundary layer (1 km, 17 °C), these fractions decrease to 28 and 3%, respectively. 3.4.4.2. Kinetics. Laboratory studies by Crounse et al.86 determined that the bulk HPALD formation rate is substantially slower than originally calculated by Peeters et al.,31 but still fast enough to have importance for atmospheric isoprene chemistry. Improved calculations of the isoprene RO2 equilibrium constants (section 3.3) and H-shift rates,85 along with the recognition that the HPALD yield following the 1,6 H-shift is not unity,85,86,95 bring the laboratory and theoretical determinations for the individual peroxy radical isomerization rates well within their combined uncertainties. Here, we recommend the 1,6 H-shift rates of the Z-δ-ISOPOO radicals determined by Teng et al.,95 adjusted for the revised isomer-specific IHN branching ratios described above (section 3.4.1). These parameterizations assume the 1,5 H-shift rates of βISOPOO radicals (see below) and tunneling temperature 8 3 3 dependences (e1×10 K /T ) reported in Peeters et al.85 3352

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Figure 9. Reactions and products following the 1,6 H-shift of the Z-(1-OH,4-OO)-ISOPOO radical. Yields are for 298 K and 1 atm; in reactions of peroxy radicals with NO, the relative contributions of the nitrate and alkoxy pathways vary with both temperature and pressure. Although we show possible bimolecular reactions of the E-enol peroxy radicals (lower left) for completeness, we assume the initial addition and removal of oxygen occurs quickly enough to route all the non-HPALD products via the Z-enol peroxy radical (center). For the reduced mechanism, we ignore the epoxy hydroperoxy aldehyde formed in 3% overall yield, and we assume the H-shifts from the dihydroperoxy aldehyde peroxy radicals outcompete all bimolecular chemistry, immediately forming DHP−MVK.

bimolecular chemistry and accompanied by HOx recycling. We assign the remainder of the carbon to the C4 dihydroperoxy (DHP) carbonyls (DHP−MVK and DHP−MACR) proposed by Peeters et al.85 These products were not observed by Teng et al.;95 if they are formed, they may be rapidly lost to walls of chambers or tubing, and/or too fragile to detect at specific product masses with the CF3O− ion chemistry used in the Teng et al.95 study. In the atmosphere, they may contribute to SOA and/or undergo rapid photolysis,59,85 leading to further HOx regeneration. We propose photolysis mechanisms for DHP− MVK and DHP−MACR in Figure S4. Other experimental evidence for HOx recycling153 and HPALD formation86,154 is consistent with the determinations of Teng et al.,95 though they do not reduce the uncertainty in the isomerization rates or product yields. 3.4.4.4. 1,5 H-Shift. The rate of the 1,5 H-shift of the O−H hydrogen in the major β-ISOPOO radicals (Figure 11) has been calculated by several groups;31,85,152 there are no experimental constraints. Near 300 K, the calculated rates from Peeters et al. are similar to those determined by da Silva for the 1-OH system; nearly a factor of 2 faster for the 4-OH system. The products will include MVK, MACR, CH2O, and OH. Here, we use the rate coefficients reported in Peeters et al.,85 consistent with the experimental constraints of Crounse et al.;86 in very remote conditions, the fraction of the ISOPOO in the 1-OH system undergoing the 1,5 H-shift is estimated to be nearly as large as the fraction undergoing a 1,6 H-shift (Figure 4).

k(1,5‐H) ,( 1‐OH,2‐OO) = (1.04 × 1011)e[(−9746 K)/ T ] s−1 k(1,5‐H) ,( 4‐OH,3‐OO) = (1.88 × 1011)e[(−9752 K)/ T ] s−1

4. REACTION OF O3 WITH ISOPRENE Although reaction with OH constitutes the dominant isoprene loss process, reaction with ozone is estimated to account for approximately 10% of isoprene removal in the atmosphere.19,155 Through a complex series of reactions and intermediates, ozonolysis of isoprene provides a source of OH radicals, organic peroxy radicals, Criegee intermediates, stable gaseous products, and secondary organic aerosol (SOA) to the atmosphere (see, e.g., Kamens et al.,156 Biesenthal et al.,157 and Nguyen et al.158). In isoprene-rich regions, ozonolysis of isoprene can also represent an important ozone loss mechanism.159 4.1. Kinetics

A number of studies have reported the rate coefficients for the reaction of ozone with isoprene (Table 12). Most utilized environmental chambers and included addition of compounds such as cyclohexane, n-octane, or carbon monoxide to remove the OH generated from ozonolysis. The measured reaction rate constants are generally in good agreement. Adeniji et al.160 reported an abnormally high value, likely because they did not use any OH scavenger.161 The causes of other smaller discrepancies are unclear. The IUPAC preferred value of kisop+O3 = (1.03 × 10−14)e(−1995 K/T) cm3 molecule−1 s−1 is adopted here. 3353

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Figure 10. Reactions and products following the 1,6 H-shift of the Z-(4-OH,1-OO)-ISOPOO radical. Branching ratios are for 298 K and 1 atm; in reactions of peroxy radicals with NO, the relative contributions of the nitrate and alkoxy pathways vary with both temperature and pressure. Although we show possible bimolecular reactions of the E-enol peroxy radicals (lower left) for completeness, we assume the initial addition and removal of oxygen occurs quickly enough to route all the non-HPALD products via the Z-enol peroxy radical (center). For the reduced mechanism, we ignore the epoxy hydroperoxy aldehyde formed with a 3% overall branching ratio, and we assume the H-shifts from the dihydroperoxy aldehyde peroxy radicals outcompete all bimolecular chemistry, immediately forming DHP−MACR.

4.2. Mechanism

The mechanism and products of isoprene ozonolysis have also been the subject of numerous investigations, with more widely varying results. Here, we outline a reaction scheme developed from that recommended by Nguyen et al.172 (Figure 12). The first step of ozonolysis involves the cycloaddition of ozone at either double bond in isoprene to form one of two possible primary ozonides (POZ).169 In our mechanism, we use the POZ

Figure 11. Reactions and products following 1,5 H-shift of β-ISOPOO.

Table 12. Chronological Estimates of Isoprene + O3 Reaction Rate Constants methodology

T (K)

rate constant (×10−17 cm3 molecule−1 s−1)

citation

chamber, excess reactant, no scavenger chamber, excess reactant, no scavenger chamber, excess reactant, no scavenger outdoor chamber, no scavenger relative rate technique, n-octane scavenger chamber, excess reactant, no scavenger chamber, cyclohexane scavenger chamber, cyclohexane scavenger relative rate technique, He diluent gas theoretical study relative rate technique, trimethylbenzene scavenger chamber, CO scavenger chamber, no scavenger IUPAC/our recommendation IUPAC/our recommendation

295 ± 1 294 ± 2 278−323 291 296 ± 2 242−323 293 ± 2 291 ± 2 242−363 300 278−353 293 ± 2 286 K 240−360 298

1.27 1.65 1540e[(−2153±430 K)/T] 1.1 1.16 ± 0.02 780e[(−1913±139 K)/T] 0.895 ± 0.025 1.13 ± 0.32 1090e[(−1998±63 K)/T] 1.58 1360e[(−2015±149 K/T] 1.19 ± 0.09 0.96 ± 0.07 1030e(−1995 K/T) 1.27

Arnts et al., 1979162 Adeniji et al., 1981160 Atkinson et al., 1982163 Kamens et al., 1982156 Greene et al., 1992164 Treacy et al., 1992165 Grosjean et al., 1993166 Grosjean et al., 1996167 Khamaganov et al., 2001168 Zhang et al., 2002169 Avzianova et al., 2002170 Klawatsch-Carrasco et al., 2004161 Karl et al., 2004171 Atkinson et al., 200619 Atkinson et al., 200619

3354

DOI: 10.1021/acs.chemrev.7b00439 Chem. Rev. 2018, 118, 3337−3390

Chemical Reviews

Review

Figure 12. Reactions and products following ozonolysis of isoprene. Many of the Criegee decomposition products remain unidentified; we therefore generally assume that they decompose to small, stable products. For a more detailed treatment of the complete ozonolysis mechanism, see Nguyen et al.172

oxide, also known as the activated Criegee intermediate (CI*, where the asterisk denotes the unstable, activated state).174,175 A fraction of these CI* decompose while the remainder are stabilized (SCI) and can subsequently undergo bimolecular or further unimolecular chemistry. 4.3. Reactions of the Stabilized Criegee Intermediate

The C1 SCI (CH2OO) is formed from the ozonolysis of all terminal alkenes, and has therefore been studied in much greater detail than the C4 SCIs. Because CH2OO lacks alkyl substituents and cannot have a syn conformation, it has low unimolecular reactivity.176 Instead, CH2OO is expected to react with a number of gaseous molecules, the most relevant of which are shown in Figure 13. 4.3.1. Kinetics. The principal reaction partners of CH2OO in the atmosphere are water and its dimer, (H2O)2. The rate coefficients of CH2OO with water and water dimer have been experimentally determined by relative rate kinetics, model simulations of environmental chamber data, and measurements of the decay of laser-generated Criegee intermediates at reduced pressure. Rate coefficients at 298 K for the reaction of CH2OO with the water monomer are in the range