Hydroxyl Radical Regeneration in Isoprene Oxidation: the Upgraded Mechanism LIM1 J. Peeters, S. V. Nguyen, T. L. Nguyen University of Leuven T. Stavrakou and J.-F. Müller Belgian Institute for Space Aeronomy ACCENT Meeting, Urbino, September 2013

 LEUVEN ISOPRENE MECHANISM (LIM) INTRODUCTION (1) In recent campaigns in isoprene-rich areas: OH concentration often 5 (up to 10) times higher than model predictions HO2 also frequently underestimated Peeters et al. (PCCP, 2009, 2010)1-3 proposed and theoretically quantified a new isoprene oxidation mechanism for low- and moderate NO levels that results in HOx radical regeneration  LEUVEN ISOPRENE MECHANISM (LIM) 1. J. Peeters, T.L. Nguyen, L. Vereecken, PCCP 11, 5935-5939, 2009 2. J. Peeters and J.-F. Müller, PCCP 12, 14227-14235, 2010. 3. T. Stavrakou, J. Peeters, J.-F. Müller, ACP 10, 9863-9878, 2010

INITIAL STEPS chemistry of OH-isoprene adducts I INTRODUCTION (2) INITIAL STEPS chemistry of OH-isoprene adducts I 1,5-Hs OH + MVK + HCHO Here is chemistry of major OH-isoprene adduct I (similar for OH-isoprene adduct II) The new features are the redissociation of peroxy radicals from E-delta to Beta and to Z-delta And the 1,6-H shift of Z-delta to form an alpha-hydroxy radical, which rapidly react with O2 to eliminate HO2 radical and forming HPALD. E-delta-OH-peroxys: low stability, and present in only low fractions, moreover: low reactivities similar to betas

INTRODUCTION (3) SECONDARY CHEMISTRY : HPALD photolysis recycling OH (Peeters and Müller, PCCP 2010; EGU meeting 2011) - fast photolysis : sigmas as MACR (enone), but much higher Quantum Yield due to avoided crossing Potential Energy surfaces of excited enone and excited (repulsive) hydroperoxide functions - internal energy nascent oxy photoproduct: estimated 20 kcal/mol as measured for similar cases - barrier 1,5-H shift only ca 3 kcal/mol barriers 1,6-H shift in acylperoxy and 1,5-H shift in 3-oxo-vinylperoxy both only 14 kcal/mol; internal energy of this peroxy: ca 45 kcal/mol product of the last 1,5-H shift HOOC(CH3)=CH-C°=O has resonant mesomer HOOC°(CH3)-CH=C=O that is unstable, as all alfa-hydroperoxy-alkyl-type radicals: fast loss of OH and carbonyl formation

NEW ELEMENTS : UPGRADED MECHANISM LIM1 Crounse et al. 2011: confirm HPALD formation, but report lower yield4 than LIM0 predictions Wolfe et al. 2012: confirm fast HPALD photolysis with unit quantum yield5 Hofzumahaus et. al. 2009,6 Whalley et. al. 2011:7 invoke fast conversion HO2 + X  OH + … in order to reproduce field observations, with X unknown However, no (O)VOC known to react fast with HO2 ! 4. Crounse, J. D. et al., Phys. Chem. Chem. Phys. 13, 13607, 2011. 5. Wolfe, G. M. et al., Phys. Chem. Chem. Phys. 14, 7276, 2012. 6. Hofzumahaus, A. et al., Science 324, 1702, 2009. 7. Whalley, L. K. et al., Atmos. Chem. Phys. 11, 7223, 2011.

UPGRADED MECHANISM LIM1 (re-)Computation of isoprene peroxy isomerisation rate at much higher levels of theory Branching fraction to initial OH-Adduct I or II overall isom-rate for pool of peroxys (or “bulk” peroxy rate) steady-state fraction of Z-δ peroxy isomers isomer-specific rate coefficient 1,6-H shift of Z-δ peroxy isomers We have to take the weighted sum for the two initial OH-adducts, taking into account the branching fractions.

Multi-conformer Partition functions ΣQTS, ΣQZδ, ΣQβ: Isomerisation rate of Z-δ peroxys Fraction of Z-δ peroxys at low NO : Multi-conformer Partition functions ΣQTS, ΣQZδ, ΣQβ: from M06-2x/6-311++G(3df,2p) level of QC theory accounts properly for London-dispersion! (while B3LYP method used for LIM0 does not) Energies of lowest conformers ETS, EZδ, Eβ: from very high-level CCSD(T)/aug-cc-pVTZ based on QCISD/6-311G(d,p) geometries (the best level available: “gold standard”) expected error only ~ 0.3 kcal/mol κ(T) tunneling factor: ZCT approach on asymmetric Eckart barrier (for similar H-shifts in peroxy radicals: results close to high-level SCT approach, Zhang & Dibble, 2011) NOTE: LONDON DISPERSION especially important for HYDROGEN BONDED SYSTEM

RESULTS (a) 21.81 2.23 0.00 -2.86 Lowest TS conformer: H-bonded (London-dispersion!) RESULTS (a) OH-adduct I TS E 21.81 Eb = ETS - EZδ = 19.58 kcal/mol 2.23 Z-δ 0.00 β -2.86 Z,Z’-product radical H-bonded Reaction Coordinate

RESULTS: overall k(isom) by Z-δ 1,6 H shift at 295 K OH-adduct I OH-adduct II f(Zδ) 1.3 x10−2 0.44 s−1 2.6 x10−3 4.9 s−1 Overall = 7.5×10−3 s−1 at 295K Compare to: k(peroxys→HPALDs) = (2±1)×10−3 s−1 of Crounse; but: isomerisation yields other products besides HPALDs Note: k(isom) : strongly T-dependent: 4.5×109 e−8000/T s−1 You see here the rate coefficients for the 1,6-H shifts of the Z-delta peroxys, for the two kinds of OH-adducts, at 295 K, and also the fractions of the Z-delta peroxys at low NO (Crounse). The f(Zdelta) are small, in particular for the Case II OH-adducts. The predicted overall isomerisation rate at low NO, in the conditions of the Crounse experiments, at 295 K, is higher than the experimental overall rate coefficient for HPALD production from the peroxys of Crounse, but as we we will see, we find that the isomerisation leads to another product channel besides that to HPALDs, that is at least equally important . Note: isomersation rate is strongly T-dependent. At a temperature of 303 K, the overall k(isom) 1.5 E-2 per second, which is of the same magnitude as the combined rate of the traditional peroxy reactions at low NO at 303 K: k(isom) =1.5×10−2 s−1, i.e. about half the combined rate of traditional peroxy reactions at low NO

RESULTS: 1,5 H shift rate of β-hydroxyperoxys (  OH+CH2O+MVK or MACR ) OH-adduct I OH-adduct II

Results : Product yields Peroxy radical sink due to traditional (i.e. bimolecular) reactions, e.g. RO2 + NO

PRODUCT CHANNELS Following Z-δ 1,6-H shift Stabilized allylic product radical : adds O2 Two main pathways: † O H OO † O OOH Eb ~ 11 kcal/mol HO2 + fast O H 2 ~ 0.5 HPALDs 1 1st peroxy + O2 (~ 50%) O H † ~ 0.5 O H † These stabilized product radicals, shown here again, will then add O2. But since these are allylic radicals, we have two addition sites, and hence two O2-addition pathways. We estimate their fractions at respectively 0.4 and 0.6, mainly on the basis of the spin densities at the two sites. The first peroxy will then very rapidly eliminate the HO2 radical to form HPALD. The critical barrier is computed to be only 10 kcal/mol, such that there can be no redissociation of this peroxy, which would face a barrier of more than 20 kcal/mol. The overall branching fraction to HPALDs would then be around 35%. The second peroxy, or its more stable conformer here, was computed to undergo an even faster reaction: an enolic 1,6-H shift, that faces a barrier of only some 4 kcal/mol. So, here too, no possibility for redissociation back to the initial radical (to the left here) and re-addition of O2 to the other site. Also, we could not find any other routes that might interconvert these two peroxys. Now, we have to look at the fate of this “B2” radical , which could account for an overall branching fraction of some 50%. enolic 1,6-H Z,Z’-product major conformer! Eb ~ 9 kcal/mol very fast most stable conformer “B2” 2nd peroxy (~ 50%) refs. : GK21 Boulder June 2012, and IGAC Beijing, Sept 2012 ACS meeting Philadelphia, Aug. 2012

PRODUCT CHANNELS (3) Fate of “B2†” radical: Possibly chemically activated reactions (likely minor, still being explored) b) Collisionally stabilized B2: adds O2 1,4-H shift O H † (1) O H OO + O2 (2) + NO (3) + HO2 “B2” “B2-O2”

PRODUCT CHANNELS (4) “B2-O2” (1): 1,4-H shift - CO, - OH HOO O H (1): 1,4-H shift - CO, - OH Eb~ 23 kcal/mol fast di-hydroperoxycarbonyl “B2-O2”

PRODUCT CHANNELS (4) O O (2): + NO O diss. + OH + OCH-CH2OOH O H O NO2 O O H MGLY Hydroperoxy Acetaldehyde 3" OH (3): + HO2 O H HOO O2 3’ Hydroperoxy-acetaldehyde and counterpart from OH-adduct II: observed by Crounse et al. (Note in PCCP Sept. 2011) in yield of ~ 25% of HPALDs. Since routes (2) and (3”) from the 2nd peroxy should be minor in their conditions: channels through 2nd peroxy and B2• / B2OO• at least as important as HPALD formation Secondary chemistry of products from B2OO° reactions: complex sequences of competing photolysis and OH-reactions

Main fate of oxoketenes ? OH REGENERATION BY SUBSEQUENT CHEMISTRY This slide was already shown, We must point out that this secondary chemistry of HPALD regenerate OH but also leads to OXOKETENES formation. And these OXOKETENES can also on their turn regenreate OH by Oxoketen with HO2 reactions. Which is briefly shown in the next slide. Main fate of oxoketenes ? OXOKETENE + HO2

Very fast α-oxoketenes+HO2 reactions Recycling OH Through pre-reactive complex and submerged transition states HO2 + OKET I: k(300K, 1 atm) ≈ 1.2×10-11 cm3s-1 HO2 + OKET II: k(300K, 1 atm) ≈ 0.8×10-12 cm3s-1 i.e. 104 to 105 x faster than e.g. CHOCHO + HO2 Including secondary chemistry: overall OKET I/II + HO2  2 OH + CH3CO + CO + CO2 ideal candidates for HO2 + X  OH + … of Hofzumahaus and Whalley

Comparison with Crounse et al. 2011 LIM1 reproduces reasonably well the product yields observed by Crounse et al. 2011 at three different temperatures (factor of ~1.8 for HPALD). Note that The reported HPALD measurement error is 50% (Crounse et al.) The HPALD channel ratio is also uncertain   Exp. # 1 T = 295 K Exp. # 2 T = 310 K Exp. # 3 T = 318 K Obs. LIM1 H2O2 2.33 1.49 3.61 2.20 5.21 2.99 ISOPOOH 4.27 3.93 3.78 3.74 3.10 3.44 ISONO2 0.53 0.54 0.36 0.46 0.16 0.37 MVK+MACR 7.53 6.30 5.31 6.49 4.76 6.83 HPALD 1.02 1.79 2.78 4.65 4.06 7.26 NO = 19 ppb  peroxy lifetime only ~0.2  the beta <=> Z-delta equilibrium cannot be attained Product growth rates in pptv min-1

Crounse et al. 2011 provide strong evidence for RO2 interconversion Relative yields :   Without interconversion β-OH-RO2 Z-δ-RO2, the total product yield from 1,6-H shift would be limited to the ~25% Z-δ-RO2 initially formed

Impact of 1,6-H shift on HOx in the IMAGESv2 CTM Impact on PBL OH Impact on PBL HO2 Globally averaged isomerisation yield = 28% in LIM1 (LIM0: 60%; Crounse et al. 2011: 10% HPALD yield)

Evaluation against field campaigns S1: no isomerisation; S2: LIM1 GABRIEL, Oct. 2005 (Lelieveld et al., 2008) PBL-averages, 9-17 LT OP3-1, Borneo, Apr. 2008 (Whalley et al., 2011) Averages, 9-17 LT INTEX-A, Jul.-Aug. 2004 (Ren et al., 2008) PBL-averages, 9-17 LT, [ISOP]obs > 300 pptv   Obs. S1 S2 NO (pptv) 20 22 OH (106 cm-3) 5.6 1.24 2.82 HO2 (108 cm-3) - 3.1 7.1 HO2* (108 cm-3) 10.5 5.84 10.6 ISOP (ppbv) 2.0 2.1 MVK+MACR (ppb) ~1 1.0 1.1 O3 (ppbv) 18 .5 17.1 19.4   Obs. S1 S2 NO (pptv) ~40 59 40 NO2 (pptv) 152 136 OH (106 cm-3) 1.5 0.54 1.37 HO2 (108 cm-3) - 1.96 3.8 HO2* (108 cm-3) 1.7 3.7 6.6 ISOP (ppbv) 2.3 2.4 1.9   Obs. S1 S2 NO (pptv) 114 93 79 NO2 (pptv) 477 532 466 OH (106 cm-3) 6.5 3.8 5.0 HO2 (108 cm-3) - 4.8 6.0 HO2* (108 cm-3) 9.9 7.4 8.6 ISOP (ppbv) 0.94 1.27 0.93 (MACR+MVK) /ISOP (*) 2.2 2.1 2.0 O3(ppbv) 53 63 62 (*) NEAQS, New England (Fuchs et al., 2012) Fuchs, H. et al. Atmos. Meas. Tech. 5, 1611, 2012. Lelieveld, J. et al. Nature Ren, X. et al. J. Geophys. Res. 113, doi:10.1029/2007JD009166, 2008. Whalley, L.K. et al., Atmos. Chem. Phys. 11, 7223, 2011.

CONCLUSIONS Main features of LIM mechanism upgrade : - equilibrium ratio Z-δ-OH/β-OH peroxys reduced by factor ~5; - isomerisation rate of Z-δ-OH-peroxy reduced by factor ~1.5 - new routes besides HPALD formation - overall k(isom) compatible with observations Crounse et al. 2011 - still, isomerisation yield = ca. 28% globally - secondary chemistry generates more OH, while keeping [HO2] down α-Oxoketenes react very fast with HO2 and convert it efficiently into OH  prime candidates for reactions X + HO2 → OH invoked in recent studies Several secondary mechanisms and rates still remain uncertain, requiring refined theoretical quantification as already done for the primary steps