Surprisingly long lifetime of methacrolein oxide, an isoprene derived Criegee intermediate, under humid conditions

Ozonolysis of isoprene, the most abundant alkene, produces three distinct Criegee intermediates (CIs): CH2OO, methyl vinyl ketone oxide (MVKO) and methacrolein oxide (MACRO). The oxidation of SO2 by CIs is a potential source of H2SO4, an important precursor of aerosols. Here we investigated the UV-visible spectroscopy and reaction kinetics of thermalized MACRO. An extremely fast reaction of anti-MACRO with SO2 has been found, kSO2 = (1.5 ± 0.4) × 10−10 cm3 s−1 (±1σ, σ is the standard deviation of the data) at 298 K (150 − 500 Torr), which is ca. 4 times the value for syn-MVKO. However, the reaction of anti-MACRO with water vapor has been observed to be quite slow with an effective rate coefficient of (9 ± 5) × 10−17 cm3 s−1 (±1σ) at 298 K (300 to 500 Torr), which is smaller than current literature values by 1 or 2 orders of magnitude. Our results indicate that anti-MACRO has an atmospheric lifetime (best estimate ca. 18 ms at 298 K and RH = 70%) much longer than previously thought (ca. 0.3 or 3 ms), resulting in a much higher steady-state concentration. Owing to larger reaction rate coefficient, the impact of anti-MACRO on the oxidation of atmospheric SO2 would be substantial, even more than that of syn-MVKO.


S10
We used the standardized spectra of MACRO, IO, and I 2 in our analysis to extract their individual contributions. The standard MACRO spectrum is from the Gauss fit of our experimental results (see Figure 3). Supplementary Figure 4 shows the standard spectra of IO and I 2 , which are derived from our experimental spectra to include the instrumental functions.
This is important for IO, of which the shape of the sharp peaks depends on the resolution of the used spectrometer. We obtained the spectra of IO and I 2 at long photolysis-probe delay times (where the MACRO contribution is gone). We estimated the effect of the CI reactions with water vapor in the isoprene ozonolysis system in a manner similar to Newland et al. 9 We separate the stabilized CIs produced in isoprene ozonolysis into two groups, CH 2 OO and C4-SCI (MVKO and MACRO). We found that we can fit the results of their Figure Table 3. Kinetic parameters used to fit the data of Figure 4 of Newland et al. 9 Where k H2O , k (H2O)2 , and k SO2 are the rate coefficients of SCI reactions with water monomer, water dimer, and SO 2 , respectively; k other includes the unimolecular decay and other loss processes;  is the relative yield of the SCI in the isoprene ozonolysis. e This work.

Supplementary
f This value is too small to be sensitive to the fit.
g Assumed the same as that for CH 2 OO.
h Copied from Table 2  This practice indicates that the complicated situation of isoprene ozonolysis can be interpreted with more than one set of kinetic parameters. Note that we do not have full knowledge about the experiments of Newland et al. 9 , thus, we should not claim that our fitting can interpret their results. We just like to demonstrate that their situation is complicated and can have more than one interpretation.
We further plot the f value (the fraction of the SCI produced that reacts with SO 2 , as  13 1.6x10 13 Model 1

Supplementary Note 3: Theoretical Calculations
To obtain the rate coefficients for the unimolecular decomposition of MACRO and the reactions of MACRO with 1 and 2 water molecules, we optimized the reactant and transition state geometries on the singlet ground electronic state using Becke's three parameter hybrid functional (B3LYP) 13,14 with 6-311+G(2d,2p) basis set. 15 Using these geometries, we corrected the electronic energies by quadratic configuration interaction singles, and doubles with perturbative triples (QCISD(T)) method 16 with Dunning's Basis sets 17 extrapolated to complete basis set (CBS) limit. 18 We used the aug-cc-pVXZ (X = D, T, Q) basis sets, and the Hartree−Fock energy was extrapolated using , and the correlation energy calculated by QCISD(T) was extrapolated using , where X is the cardinal number of the basis set and , , and are optimization parameters. All density functional theory calculations were done with the Gaussian09 program 19 while all QCISD(T) calculations were performed using the MOLPRO program. 20 The rate coefficients were calculated using the conventional transition state theory method using rigid rotor harmonic oscillator approximations using the THERMO program in the Multiwell suite. [21][22][23] We used the tunneling correction based on the 2 nd order vibrational perturbation theory method implemented in the Multiwell suite. 24 This method was used previously for C1 to C3 CIs and has shown good predictability. 25,26 Supplementary Figure

S22
We also comment on using the E b values of the water monomer reactions to estimate the E b values of the water dimer reactions, which was used in previous studies. 27 As given in Supplementary Figure 9, we can see a correlation between them. However, one important aspect is that the trend for the simplest CI CH 2 OO (black dashed line) is quite different from the correlation involving unsaturated CIs (red dashed line). In addition, the deviations from the linear fit (Supplementary Figure 9) are much larger than those in the case when we use Figure 7). Therefore, we think it is more efficient and accurate to use the

Justification for the theoretical methods
Concerning the accuracy of the present QCISD(T)//B3LYP method, we have confirmed that the QCISD(T) gives energies that are consistent with CCSD(T) and multireference methods for the CH 2 OO+H 2 O reaction (difference ~ 1 kcalmol -1 , see Table S3 of Lin et al. 28  cm 1 at most for the low frequency modes (<1000 cm -1 ). This difference in harmonic frequencies can give a variation of 0.08 kcal mol -1 in the free energy at 298 K, which is smaller than the variation in the quantum chemistry energies.
For MACRO, the harmonic approximation may not be valid for the CH 3 internalrotation mode. Therefore, we have calculated the potential energy along the CH 3 torsional angle for MACRO and the TS of MACRO+H 2 O 1b reaction. We find that the torsional potential energy curves at both geometries are nearly identical, and thereby the free energy correction for this mode will also cancel out when we calculate the free energy difference.

Error of the theoretical k uni
Electronic energy. Concerning the error of the rate coefficient of the unimolecular reaction, similar to the reaction with water, we can assign the main error source to the uncertainties in the electronic energies obtained by QCISD(T). Here we refer to our previous paper by Yin and Takahashi 25 concerning the theoretical calculation of unimolecular decomposition rates for different Criegee intermediates. For anti-MACRO, the unimolecular reaction proceeds through the OO bending channel forming dioxirane 3,27,29 , similar to that of CH 2 OO. 25,30 In Hindered rotor partition function (free energy). We have calculated the free energy by directly counting the eigenstates for the methyl-rotor Schrodinger equation obtained from the potential energy curve calculated using B3LYP/6-311+(2d,2p), given in Figure S10.

Supplementary
Compared to the harmonic approximation, the one obtained from the direct counting is smaller by 0.04 kcal mol -1 , which is negligible compared to the error in the electronic energy.

Tunneling correction.
In the present calculation of the unimolecular rate coefficient for MACRO, we have included the tunneling correction, but as shown below, the effect is minor. At room temperature, it will only lead to a correction of 1.2 times for the most stable anti-trans conformer.