ARTICLES PUBLISHED ONLINE: 23 MARCH 2014 | DOI: 10.1038/NCHEM.1890

Extremely rapid self-reaction of the simplest Criegee intermediate CH2OO and its implications in atmospheric chemistry Yu-Te Su1, Hui-Yu Lin1, Raghunath Putikam1, Hiroyuki Matsui1, M. C. Lin1 * and Yuan-Pern Lee1,2 * Criegee intermediates, which are carbonyl oxides produced when ozone reacts with unsaturated hydrocarbons, play an important role in the formation of OH and organic acids in the atmosphere, but they have eluded direct detection until recently. Reactions that involve Criegee intermediates are not understood fully because data based on their direct observation are limited. We used transient infrared absorption spectroscopy to probe directly the decay kinetics of formaldehyde oxide (CH2OO) and found that it reacts with itself extremely rapidly. This fast self-reaction is a result of its zwitterionic character. According to our quantum-chemical calculations, a cyclic dimeric intermediate that has the terminal O atom of one CH2OO bonded to the C atom of the other CH2OO is formed with large exothermicity before further decomposition to 2H2CO 1 O2(1Dg). We suggest that the inclusion of this previously overlooked rapid reaction in models may affect the interpretation of previous laboratory experiments that involve Criegee intermediates.

C

riegee intermediates are extremely important in the reactions of ozone (O3) with unsaturated hydrocarbons in the atmosphere. These reactions have been investigated extensively because they are responsible for the removal of both O3 and unsaturated hydrocarbons, and for the production of OH and organic acid that might lead to the formation of particulate material in the troposphere1–3. The reactions of O3 with unsaturated hydrocarbons are initiated with the cycloaddition of O3 to the C¼C double bond to form a primary cyclic ozonide with a C–C single bond. As a result of the large exothermicity of this reaction, rapid cleavage of this C–C bond and an O–O bond of the ozonide occurs to form a carbonyl molecule and a carbonyl oxide, commonly referred to as the Criegee intermediate4–6. Internally excited Criegee intermediates might isomerize or decompose to produce H, OH, CH3 , CO, CO2 and other products. An extensive search for Criegee intermediates was carried out after their proposal in 19497, but direct detection of the gaseous Criegee intermediates was reported only recently8–12. Theoretical investigations of the structure and reactivity of the Criegee intermediates have been extensive13–17. The simplest Criegee intermediate, CH2OO (formaldehyde oxide), is produced in the atmosphere from the reaction of O3 with ethene (C2H4) and a number of 1-alkenes via formation and fragmentation of alkene ozonide. The high reactivity of CH2OO meant that it eluded direct detection until Taatjes and co-workers produced it from reactions of CH3SOCH2 þ O2 (ref. 9) and CH2I þ O2 (ref. 10) in a flow reactor and detected its cation with vacuum ultraviolet photoionization and a mass spectrometer. They confirmed that the Criegee intermediate, rather than other isomers such as dioxirane or formic acid (HCOOH), was found based on the observed photoionization threshold 10 eV that conforms to theoretical predictions of 9.98 eV (ref. 13), which is smaller than the values of 10.8 and 11.3 eV for dioxirane13 and HCOOH18, respectively. Beames et al. employed the CH2I þ O2 reaction to prepare CH2OO in a supersonic jet and reported a broad ultraviolet spectrum of CH2OO with an absorption maximum near 335 nm;

1

the spectrum was obtained from ultraviolet-induced depletion of the signal of CH2OOþ cations produced on photoionization10. The observation that the reaction of CH2I with O2 produces CH2OO also has important atmospheric implications. CH2I2 is one of the main organic iodine compounds emitted by various types of phytoplankton into the ocean surface. As sunlight can efficiently induce the photolytic production of CH2I þ I from CH2I2 because of its intense absorption band in the near-ultraviolet region19,20, photochemical transformation of I atoms into aerosol and ozone destruction caused by CH2I2 have been shown to be significant in the marine boundary layer21,22. Thus, the examination of the role of CH2OO that is produced directly from CH2I þ O2 in the atmospheric becomes an important subject. In our laboratory, Su et al. employed a step-scan Fourier-transform infrared (FTIR) spectrometer to record the infrared spectrum of CH2OO on irradiation at 248 nm of a gaseous mixture of CH2I2 and O2 (ref. 11); the five features observed at 1,435, 1,286, 1,241, 908 and 848 cm21 provided definitive identification of this intermediate when compared with the spectra of possible products in the reaction CH2I þ O2 simulated according to high-level quantum-chemical calculations. The vibrational wavenumbers observed indicate a significant zwitterionic contribution to this singlet biradical compound because of a strengthened C–O bond and a weakened O–O bond. The infrared detection of CH2OO also proves useful as a direct probe in kinetic and mechanistic investigations. As shown previously, these new infrared absorption bands of CH2OO appeared within the first 12.5 ms on irradiation at 248 nm of a flowing mixture of CH2I2/N2/O2 (1/20/760, 94 Torr) at 343 K (ref. 11). These bands decayed rapidly and diminished after 100 ms. The lifetime of 50 ms of CH2OO observed in these experiments is much less than that, 2 ms, reported by Welz et al.9. The main difference in experimental conditions is the concentration of CH2I. Welz et al. employed [CH2I]0  9 × 1011molecule cm23 in their photoionization experiments9, but Su et al.11 employed a much greater concentration,

Department of Applied Chemistry and Institute of Molecular Science, National Chiao Tung University, 1001 Ta-Hsueh Road, Hsinchu 30010, Taiwan, Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei 10617, Taiwan. * e-mail: [email protected] (M.C.L.); [email protected] (Y.P.L.)

2

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

1

ARTICLES

NATURE CHEMISTRY

[CH2I]0  4 × 1013 molecule cm23, because the sensitivity of infrared absorption is much less than that of mass detection. To understand the significant difference in the lifetimes and its atmospheric implications, we performed both laboratory experiments and quantum-chemical computations to investigate the decay kinetics of CH2OO under varied partial pressures of O2 and N2. We found that the extremely rapid self-reaction of CH2OO is the major channel for its decay.

0.0

1.2 TS2

–13.2

Decay kinetics of CH2OO. Figure 1 shows two representative plots of [CH2OO]21 versus reaction period t. Trace 1 was recorded with [N2] ¼ [O2] ¼ 2.8 × 1017 molecule cm23 (10 Torr at 343 K), whereas trace 2 was recorded with [N2] ¼ 2.5 × 1018 molecule cm23 (90 Torr at 343 K) and [O2] ¼ 2.8 × 1017 molecule cm23; CH2OO was produced on photolysis of a mixture of CH2I2/O2/N2 at 355 nm. Both plots exhibit a linear relationship between [CH2OO]21 and t, which indicates secondorder behaviour. The effective second-order rate coefficients are keff ¼ (3.1+0.1) and (4.5+0.2) × 10210 cm3 molecule21 s21, respectively, approaching the gas kinetic collision value; the listed uncertainties reflect only the standard deviation in fitting. The experimental conditions and fitted keff are summarized in Supplementary Table 1. Reaction mechanism and rate coefficients. Several reactions might contribute to the decay of [CH2OO]: CH2 OO + CH2 OO  products

(1)

CH2 OO + I  products

(2)

CH2 OO + CH2 I  products

(3)

CH2 OO + O2  products

(4)

We performed high-level theoretical calculations to investigate the energies of possible reaction channels for each reaction. Simplified potential-energy diagrams that show the most important reaction paths are presented in Figs 2 and 3 for reactions (1) and (2),

–27.3

[CH2OO]–1 (10–14 cm3 molecule–1)

1 5

150

Time (µs)

Figure 1 | Plots of [CH2OO]21 versus reaction time. Trace 1, [CH2I2]0 ¼ 0.34 Torr, [O2] ¼ 10 Torr, [N2] ¼ 10 Torr. The data are shown as circles and the fitted effective second-order rate coefficient, keff ¼ (3.1+0.1) × 10210 cm3 molecule21 s21, is shown as a solid line. The error limits in keff represent one standard deviation in fitting. Trace 2, [CH2I2]0 ¼ 0.21 Torr, [O2] ¼ 10 Torr, [N2] ¼ 90 Torr. The data are shown as triangles and the fitted keff ¼ (4.5+0.2) × 10210 cm3 molecule21 s21 is shown as a solid line. Error bars shown for [CH2OO]21 only reflect integration errors, not systematic errors, as discussed in the text. 2

TS1

LM 2

–75.9 –92.4

2H2CO + 1O2

(CH2OO)2

Figure 2 | Schematic energy diagram for the CH2OO 1 CH2OO reaction paths. The paths were computed at the CCSD(T)//B3LYP/aug-cc-pVTZ-pp level with corrections of vibrational ZPE. Relative energies at 0 K are given in kilocalories per mole. Transition states are labelled as TS and reaction intermediates as LM. The minimal energy path, reaction (5a), proceeds via (CH2OO)2 and TS1 to form 2H2CO þ 1O2 in which the latter represents O2 in its electronically excited 1Dg state. The path via TS2 is unimportant because of the higher barrier involved.

respectively. Key structural parameters for the reactant, intermediates, transition states and products in these two figures are shown in Supplementary Fig. 1. The T1 diagnostic values, available in Supplementary Table 2, were obtained for all species involved in the potential-energy surfaces (PESs) of the CH2OO self-reaction and the CH2I þ O2 reaction at the CCSD(T)/aug-ccpVTZ-pp level of theory. Most T1 values of the reactants and products are less than 0.02, whereas those of the transition states are in the range 0.03–0.04, which indicates some multiconfigurational characters, as reported in the literature23. As shown in Fig. 2, the self-reaction of CH2OO proceeds with three major channels that lead to products in two sets:   (5a) CH2 OO + CH2 OO  2H2 CO + O2 1 Dg  M  CH2 OO 2

10

100

–34.3 –41.7

2

50

TS3

LM 1

15

0

TS4

–25.2

CH2OO + CH2OO

Results and discussion

DOI: 10.1038/NCHEM.1890

(5b)

in which M indicates the third body and M ¼ O2 was used in the calculations. The formation of dimeric CH2OO has no barrier and shows an exceptionally large exothermicity, 292.4 kcal mol21, at the CCSD(T)//B3LYP/aug-cc-pVTZ-pp level of theory. For comparison, the enthalpy (DH 0298) of recombination of methyl peroxyl radicals CH3OO þ CH3OO to form (CH3OO)2 is only 10–15 kcal mol21. Exothermic and the dimeric bonding is mainly between two terminal O atoms24. The large exothermicity of the bimolecular association reaction reflects the nature of the head-to-tail zwitterionic recombination. The structure of (CH2OO)2 is cyclic with each of the terminal oxygen atoms (with partial negative charge) of CH2OO binding to the carbon atom (with partial positive charge) of the other CH2OO molecule, as shown in Fig. 4. Stabilization of (CH2OO)2 , reaction (5b), is unimportant because the internally excited (CH2OO)2 readily decomposes into 2H2CO þ O2(1Dg) via a transition state TS1 with energy 34 kcal mol21 less than that of the reactants. A second channel for the formation of 2H2CO þ O2(1Dg) from CH2OO þ CH2OO proceeds via a transition state, TS2, having a C...C interaction between two CH2OO molecules with a barrier of 1.2 kcal mol21 and is considered to be less important than the first channel. We investigated the effects of spin-symmetry breaking on the singlet reaction CH2OO þ CH2OO at the NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY

ARTICLES

DOI: 10.1038/NCHEM.1890

5.9

H

TS7 0.0

2.4

1.9

TS5

–0.6

TS6

LM 3 CH2OO +

H

0.6 2CH I 2

1.081

+ 3O2

1.254 C

1.083 H

2I (2P ) 3/2

115.4

O

O

11 9. 2

O

.089

1

H

1.41

O

59

1.4

ϕ COOC = 62.6°

(CH2OO)2

–42.2 H2CO + 2IO

Figure 3 | Schematic energy diagram for the CH2OO 1 I(2P3/2) reaction path. The paths were computed at the CCSD(T)//B3LYP/aug-cc-pVTZ-pp level with corrections of vibrational ZPE. Relative energies at 0 K are given in kilocalories per mole. The major path proceeds via ICH2OO, which is either stabilized (reaction (6a)) or decomposes further to form CH2I þ 3O2 (reaction (6b)) in which the latter represents the ground 3S2 g electronic state of O2. The path for the formation of H2CO þ IO (reaction (6c)) is unimportant because of the higher barrier involved.

CCSD(T)/cc-pVTZ level. The results shown in Supplementary Table 3 indicate that the energies obtained agree within +1 kcal mol21 for the favoured low-energy path via TS1, and within 8 kcal mol21 for TS2. A similar effect was observed in the reaction CH2OO þ SO2 , as reported by Vereecken et al.25. As shown in Fig. 3, reactions of CH2OO with I generate products in three sets: M

O

C

H

–26.2 ICH2OO

105.5

110.9

1.41

1.351

CH2OO

H

C

O

CH2 OO + I  ICH2 OO    CH2 I + O2 3 S− g

(6b)

 H2 CO + IO

(6c)

(6a)

ICH2OO is formed either by direct attack at the C atom of CH2OO by the I atom with no barrier or via an intermediate LM3 (I atom attached to the terminal O atom) that lies 20.6 kcal mol21 relative to CH2OO þ I, followed by a rearrangement to form stable ICH2OO via TS5 with a barrier height of 3.0 kcal mol21. Internally excited ICH2OO might further decompose to form CH2I þ O2(3S2 g ), which has an energy 0.6 kcal mol21 above CH2OO þ I. Abstraction of the terminal O atom of CH2OO by the I atom to form H2CO þ IO proceeds from decomposition of LM3 via

Figure 4 | Geometries of CH2OO and (CH2OO)2 optimized at the B3LYP/aug-cc-pVTZ-pp level. Lengths are in a˚ngstro¨ms, angles are in degrees and f is the dihedral angle. The monomer structure indicates a significant zwitterionic character with a strengthened (shorter) C2O bond and a weakened (longer) O2O bond. The structure of (CH2OO)2 is cyclic with each of the terminal oxygen atoms (with partial negative charge) of CH2OO binding to the carbon atom (with partial positive charge) of the other CH2OO molecule.

TS6 with a barrier height of 2.5 kcal mol21; the other direct abstraction channel via TS7 with a barrier of height 5.9 kcal mol21 is unimportant. Reaction (3) has three possible product channels:   − (7a) CH2 OO + CH2 I  C2 H4 I + O2 3 Sg  2H2 CO + I M

 ICH2 CH2 OO

(7b) (7c)

but only channel (7a) is important; this reaction proceeds via decomposition of the adduct ICH2CH2OO in which the carbon atoms of CH2OO and CH2I become bonded. The reaction of CH2OO with O2 , reaction (4), has several channels to produce O2CH2O2 , CO, CO2 OH and HO2 , but all these channels have barriers greater than 10 kcal mol21. The potential-energy diagrams and the structures of intermediates and transition states for reactions (3) and (4) are presented in Supplementary Figs 2–5. Rate coefficients predicted for the major channels at 298 K and 343 K are summarized in Table 1; predicted rate coefficients of all channels discussed in the preceding sections are listed in Supplementary Table 4. The rate coefficient for reaction (5a), k5a ¼ 2.2 × 10210 cm3 molecule21 s21 at 343 K, is independent of pressure for P ¼ 20–100 Torr because the energy of the transition state TS1 is 34 kcal mol21 less than the energy of the reactants. It is more than 4,000 times that of reaction (5b) for P ¼ 20–100 Torr.

Table 1 | Rate coefficients (in cm3 molecule21 s21) predicted for important reactions of CH2OO. Reaction

Pressure (Torr)

Rate coefficient T 5 298 K

T 5 343 K

(5a) CH2OO þ CH2OO  2H2CO þ 1O2 (6a) CH2OO þ I  ICH2OO

20–100 20 40 60 80 100 20 40 60 80 100 20–100 20–100

2.44 × 10210 1.02 × 10211 1.32 × 10211 1.53 × 10211 1.69 × 10211 1.82 × 10211 8.11 × 10211 7.92 × 10211 7.78 × 10211 7.66 × 10211 7.55 × 10211 5.54 × 10214 6.28 × 10211

2.17 × 10210 6.57 × 10212 8.75 × 10212 1.03 × 10211 1.14 × 10211 1.24 × 10211 9.25 × 10211 9.12 × 10211 9.02 × 10211 8.93 × 10211 8.85 × 10211 1.02 × 10213 4.78 × 10211

(6b) CH2OO þ I  CH2I þ 3O2

(6c) CH2OO þ I  H2CO þ IO (7a) CH2OO þ CH2I  C2H4I þ 3O2

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

3

ARTICLES

NATURE CHEMISTRY b [CH2OO] (1012 molecule cm−3)

[CH2OO] (1013 molecule cm−3)

a 3

2

1

DOI: 10.1038/NCHEM.1890

12 10

0

8 6 4 2 0

0

20

40

60

80

0

100

20

40

c

7

d

[CH2OO] (1013 molecule cm−3)

6

[CH2OO] (1013 molecule cm−3)

5 4 3 2 1 0 20

40

80

100

80

100

3

2

1

0 0

60

Time (µs)

Time (µs)

60

80

100

0

20

40

60

Time (µs)

Time (µs)

Figure 5 | Comparison of experimental decay profiles of CH2OO with those simulated using a kinetic model. a, [CH2I]0 ¼ 2.6 × 1013 molecule cm23, [N2] ¼ 3.5 Torr, [O2] ¼ 16.5 Torr. b, [CH2I]0 ¼ 1.1 × 1013 molecule cm23, [N2] ¼ 5.6 Torr, [O2] ¼ 89.5 Torr. c, [CH2I]0 ¼ 7.1 × 1013 molecule cm23, [O2] ¼ [N2] ¼ 10 Torr. d, [CH2I]0 ¼ 2.6 × 1013 molecule cm23, [N2] ¼ 90 Torr, [O2] ¼ 10 Torr. In all simulations, k5a ¼ (4.1+0.8) × 10210 cm3 molecule21 s21, k6b ¼ (4.1+0.7) × 10211 cm3 molecule21 s21, k8 ¼ 1.5 × 10212 cm3 molecule21 and the other rate coefficients are listed in Table 1. The shaded area covers uncertainties of k6a with one standard deviation in fitting. In b the rise of CH2OO is the most rapid and the decay is slowest among all four figures because [O2] is the largest and [CH2OO]0 is the smallest. In c, the decay of CH2OO is the fastest because [CH2OO]0 is the largest.

Hence, most reactants proceed via TS1 to form 2H2CO þ O2(1Dg) instead of being stabilized to (CH2OO)2. For reactions (6), the dominant channel is the formation of CH2I þ O2(3S2 g ), reaction (6b), with k6b ¼ (9.3–8.9) × 10211 cm3 molecule21 s21 for 20–100 Torr O2 at 343 K; is nearly independent of pressure. The formation of ICH2OO, according to reaction (6a), with k6a ¼ (6.6–12.4) × 10212 cm3 molecule21 s21 for 20–100 Torr O2 at 343 K, is less important. Reaction (7a), with k7a ¼ 4.8 × 10211 cm3 molecule21 s21, is unimportant under our experimental conditions because most CH2I reacts readily with O2 to form CH2OO þ I so that its concentration is small. Rate coefficients of other channels in reactions (1)–(4) are smaller than 1 × 10213 cm3 molecule21 s21. Simulation of decay profiles. Hence, the model for the formation and decay mechanism of CH2OO includes reactions (5a), (6a), (6b), (6c), (7a) and (8): CH2 I + O2  CH2 OO + I

(8)

but only reactions (5a) and (6b) are important for the decay of CH2OO. We simulated the decay profiles of CH2OO according to this model with the CHEMKIN-II program26 to compare with our experimental data because no algebraic solution is feasible from this mechanism. From the sensitivity analysis (available in Supplementary Fig. 6), reaction (5a) is the most important for the decay of CH2OO. We varied values of k5a and k6b , but kept the experimental value of k8  1.5 × 10212 cm3 molecule21 (refs 27,28) and the other rate coefficients, as predicted by theory, 4

unaltered to perform the least-squares fitting. The least-squares fits for ten decay profiles that covered O2 pressures in the range 10–90 Torr and N2 pressures in the range 3.5–90 Torr (available in Supplementary Table 5) yield k5a ¼ (4.1+0.8) × 10210 and k6b  (4.1+0.7) × 10211 cm3 molecule21 s21 at 343 K; the error limits represent one standard deviation for the average of the data listed in Supplementary Table 5. Some representative plots are shown in Fig. 5, in which the grey areas represent the fitting that covers one standard deviation of the average value of k5a. As discussed in Supplementary Section A, the uncertainties in concentration measurements of CH2OO (and consequently of I atoms) might be as large as 60%, which translates directly to the error of the second-order rate coefficients k5a and k6b. Combining possible errors in reaction modelling and kinetic fitting (36% for two standard deviations), the derived rate coefficients for k5a and k6b might have uncertainties as large as factors of two; hence the values k5a ¼ (4+2) × 10210 and k6b  (4+2) × 10211 cm3 molecule21 s21 agree satisfactorily with those predicted theoretically, k5a ¼ 2.2 × 10210 and k6b ¼ (8.9–9.3) × 10211 cm3 molecule21 s21 for P ¼ 20–100 Torr at 343 K. Although the rate coefficient might not be as accurate as desired because of the uncertainties in concentration, the key result is that CH2OO reacts with itself extremely rapidly, with a rate coefficient approaching the gaseous collision rate coefficient. The extremely rapid self-reaction of CH2OO is a result of a barrierless entrance channel, the high exothermicity (292.4 kcal mol21) for the formation of the dimeric intermediate because of the strong head-to-tail zwitterionic interaction, and the small barrier for decomposition of the dimer. The rate coefficient of

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY

ARTICLES

DOI: 10.1038/NCHEM.1890

k5a  (4+2) × 10210 cm3 molecule21 s21 at 343 K is about 1,100 times the rate coefficient of 3.4 × 10213 cm3 molecule21 s21 for CH3OO þ CH3OO (ref. 29) at 298 K. This rate coefficient is also much greater than those of the reactions of CH2OO with SO2 , NO2 , acetone and acetaldehyde, which are reported as (3.9+0.7) × 10211, (7+3) × 10212, (2.3+0.3) × 10213 and (9.5+0.7) × 10213 cm3 molecule21 s21, respectively9,30. Impacts on atmospheric chemistry. Such a rapid self-reaction of CH2OO was overlooked in previous experimental and theoretical investigations, and thus it was not included in the modelling of systems that involved Criegee intermediates. Although abundant studies have been performed to speculate the role of the Criegee intermediates in atmospheric chemistry, the discussion on the reaction mechanism is based mainly on indirect experimental results under conditions with the concentrations of the precursors of CH2OO in the range 1012–1014 molecule cm23. The selfreaction of CH2OO is expected to have significant effects under such experimental concentrations. For example, in the laboratory investigations of the ozonolysis of C2H4 , the current model indicates that reactions of CH2OO with H2CO and HCOOH are the two major reactions of stabilized CH2OO (ref. 31). Although the rate coefficients for the reactions CH2OO þ H2CO and CH2OO þ HCOOH are unreported, they were estimated to be 5 × 10212 cm3 molecule21 s21 (ref. 25), about 80 times smaller than that of the self-reaction of CH2OO. As H2CO, HCOOH and CH2OO might be produced with concentrations on the same order of magnitude from laboratory reactions of O3 þ C2H4 , the self-reaction of CH2OO should play a significant role. The current model estimates yields (with respect to the loss of C2H4) of 0.23, 0.23 and 0.04 for the formation of CO2 , CO þ HCO and HCOOH, respectively, from direct decomposition of internally excited CH2OO, and yields of 0.22 and 0.07 for conversion of CH2OO into HCOOH and OH þ HCO, respectively, from the reaction of stabilized CH2OO with H2CO, and a yield of 0.21 for the formation of hydroperoxymethyl formate (HPMF, CH2(OOH)–O–CHO) and formic acid anhydride (FA, (HCO)2O) from the reactions of CH2OO with HCOOH. We employed a similar mechanism (Supplementary Section C) to simulate their experimental conditions and found that when the selfreaction of CH2OO is included in the model the simulated yield of HPMF þ FA decreased by 12–16% and that of HCOOH increased by 6%. The additional H2CO produced because of the self-reaction of CH2OO accounts for a yield of 0.06–0.07. When we increased the rate coefficient for O3 þ alkene from 1 × 10218 cm3 molecule21 s21 for C2H4 to 1 × 10216 cm3 molecule21 s21 for larger alkenes29, the simulated yield of compounds caused by the reaction of the Criegee intermediate þ HCOOH, which corresponds to HPMF þ FA in O3 þ C2H4 , decreased by 40–50%, whereas that of HCOOH increased by 16–18%. Furthermore, the additional carbonyl compounds produced from the self-reaction of the Criegee intermediates account for a yield of 0.19–0.23. Hence, this simulation explains the observed larger-than-unity stoichiometry ratio for the formation of carbonyl compounds in reactions of O3 with larger alkenes, such as in the case of ozonolysis of 2-butene3. The enhanced effects caused by an increased instantaneous concentration of the larger Criegee intermediates result from a greater rate coefficient for O3 þ larger alkenes. Another unexplored, yet potentially important, issue is the predicted production of singlet O2 from the self-reaction of CH2OO, which might participate actively in the secondary reactions. The rate coefficient of O2(1Dg) þ CH2OO, calculated to be 2 × 10218 cm3 molecule21 s21 in this work, is small, but reactions of O2(1Dg) with other radicals might be important. More detailed investigations are needed to assess this effect.

The rapid self-reaction of CH2OO may also have an impact on the CH2I þ O2 system, which is important in the atmospheric chemistry of the marine boundary layer because of the release of CH2I2 from exposed macroalgae followed by solar photolysis on a timescale of minutes21,22. In the atmosphere, I and IO can participate in the gasphase reaction cycles, which might affect O3 , HOx and NOx levels, and also the formation of particles32. The average concentration of CH2I2 was reported to be ,1 parts per trillion by volume22, which is too small for the self-reaction of CH2OO to be important. For laboratory investigations, the current model uses CH2 I + O2 + M  ICH2 OO + M

(9)

ICH2 OO + ICH2 OO  2ICH2 O + O2   ICH2 O  H2 CO + I rapid decomposition

(10)

ICH2 OO + I  IO + H2 CO + I

(12)

(11)

with k9 ¼ 2.6 × 10214 cm3 molecule21 s21, k10 ¼ 9 × 10211 cm3 molecule21 s21 and k12 ¼ 3.5 × 10211 cm3 molecule21 s21, to explain the laboratory formation of H2CO, IO and additional I atoms, apart from photolysis of CH2I2 (refs 33–35). The rate coefficient of reaction (10), reported by Sehested et al.36 and employed in this model, is much greater than that typically expected for the self-reaction of peroxyl radicals. For example, k ¼ 3.4 × 10213 cm3 molecule21 s21 for CH3OO þ CH3OO (ref. 29) and k10 ¼ (2–61) × 10214 cm3 molecule21 s21 are predicted for ICH2OO þ ICH2OO in the pressure range 10–760 Torr; in the latter case, the major product is the formation of a dimer instead of 2ICH2O þ O2 (Supplementary Table 6 and Supplementary Figs 7 and 8). The ultraviolet absorption spectra of ICH2OO reported by Sehested et al.36 and Gravestock et al.33 probably have a significant contribution from CH2OO, as reported by Beames et al.10. Hence, the reported k10 value is expected to have some contribution from the self-reaction of CH2OO. If the self-reaction of CH2OO is included in the model, the yields of CH2OO, ICH2OO, IO and I from the reaction CH2I þ O2 in the laboratory have to be reanalysed.

Conclusions We monitored the concentration of CH2OO with its infrared absorption in the reaction of CH2I þ O2 at 343 K and observed extremely rapid second-order decay of CH2OO. Using a reaction mechanism assisted by quantum-chemical predictions to simulate the decay profiles, we estimated a rate coefficient of k ¼ (4+2) × 10210 cm3 molecule21 s21 for the self-reaction of CH2OO. This extremely rapid self-reaction of CH2OO, which results from the strong head-to-tail zwitterionic interaction, was overlooked in previous experimental and theoretical investigations. Previous laboratory data might have to be re-evaluated on inclusion of this rapid self-reaction of CH2OO into the chemical reaction model that involves Criegee intermediates.

Methods Experimental. A step-scan FTIR (Bruker, Vertex 80v) spectrometer coupled with a multipass White cell was employed to record the infrared spectra of transient species. The laser beam passed through the White cell and was reflected 6–12 times with two external mirrors to photodissociate a flowing mixture of CH2I2 in O2 to produce CH2I that subsequently reacted with O2 to form CH2OO. The photolysis light was generated from a frequency-tripled neodymium-doped yttrium aluminium garnet (Nd:YAG) laser (Lotis, LS-2137, 11 Hz, 60 mJ pulse21, beam size 0.5 cm2) at 355 nm. The infrared probing light was detected with a HgCdTe detector with which d.c.- and a.c.-coupled signals were recorded. Typically, these signals were averaged over 15 laser shots at each scan step before being sent to the internal 24-bit digitizer of the spectrometer. One hundred data points at 12.5 ms intervals were acquired to cover a period of 1,250 ms after photolysis. We performed undersampling to decrease the size of the interferogram, and hence the duration of

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

5

ARTICLES

NATURE CHEMISTRY

data acquisition. For spectra in the range 021,579 cm21 at a resolution of 4 cm21, 798 scan steps were completed within 30 minutes. Conventional time-resolved difference absorption spectra were derived from Fourier transform of the interferograms that corresponded to varied intervals on photolysis11. To obtain a desirable partial pressure of CH2I2 , the samples and the flow reactor were heated to 343+3 K by heated water circulated from a thermostatted bath through the jacket of the reactor. To investigate the associated chemical kinetics, we must determine the concentration of CH2I and CH2OO in the reactor. As the infrared probe beam does not follow the ultraviolet photolysis beam in our apparatus, the infrared-probed volume differs from the photolysis volume. The concentration of CH2I in the photolysis volume thus differs from the average concentration of the infraredprobed volume. As diffusion is slow under our experimental conditions at T ¼ 343 K, P ≥ 20 Torr and t ≤ 100 ms, for kinetic analysis we used the concentration in the photolysis volume rather than the averaged concentration in the infrared-probed region. We estimated the initial CH2I concentration in the photolysis volume, [CH2I]hv 0 , according to the laser fluence and the absorption cross-section of CH2I2 at the photolysis wavelength and assuming that [CH2OO]0 ¼ [CH2I]hv 0 . We also compared [CH2OO] estimated from the integrated absorbance of observed bands by assuming that calculated infrared intensities of these bands of CH2OO were correct and found that these values are approximately 60% of those derived from photolysis yield. Detailed description of the procedures of experiments and concentration estimation is presented in Supplementary Section A. Computational. The geometries of reactants, products, intermediates and transition states on the PESs of the reactions were optimized with the B3LYP method37,38 using the aug-cc-pVTZ Dunning’s correlation-consistent basis set39. For the iodine atom, we used the aug-cc-pVTZ-pp basis set of Peterson et al.40, which incorporates a relativistic pseudopotential that largely accounts for scalar relativistic effects in iodine. The vibrational wavenumbers were calculated at this level to characterize local minima and the transition state, and to correct for the vibrational zero-point energy (ZPE). To obtain reliable energies and predictions of rate coefficients, we calculated single-point energies at the CCSD(T)/aug-cc-pVTZ41,42 level based on the structures predicted with B3LYP/aug-cc-pVTZ, expressed as CCSD(T)//B3LYP/augcc-pVTZ. Analysis of the intrinsic reaction coordinate was performed to confirm the connection between transition states and designated reactants and products43. All calculations of electronic structure were performed with the Gaussian 0944 and Molpro45 programs. For the association process of CH2OO þ CH2OO with no well-defined transition state, its potential function was computed variationally to cover the range of C...O separations from 1.41 to 5.5 Å at 0.1 Å intervals in the reverse decomposition process with the second-order multireference perturbation theory using a CASPT2//CASSCF(10,10)/6-31 þ G(d,p) basis set and the MOLPRO code45; other geometric parameters were optimized fully. The calculated potential energy curve was fitted to a Morse function with a parameter b ¼ 3.32 Å21. The rate coefficients were predicted according to the following method. We employed the variational transition-state theory for direct abstraction and the variational Rice–Ramsperger–Kassel–Marcus theory46–49 with Eckart tunnelling corrections for association/decompositions processes using the Variflex code50 on solving the master equation for the formation and removal of the excited intermediates. The removal processes include collisional deactivation and various channels of decomposition. The effects of pressure, temperature and kinetic energy carried by CH2I on the rate coefficients were reliably examined. More detailed information is available in the Supplementary Section B.

Received 5 September 2013; accepted 11 February 2014; published online 23 March 2014

References 1. Johnson, D. & Marston, G. The gas-phase ozonolysis of unsaturated volatile organic compounds in the troposphere. Chem. Soc. Rev. 37, 699–716 (2008). 2. Calvert, J. G. et al. The Mechanisms of Atmospheric Oxidation of the Alkenes 172–335 (Oxford Univ. Press, 2000). 3. Horie, O. & Moortgat, G. K. Gas-phase ozonolysis of alkenes. Recent advances in mechanistic investigations. Acc. Chem. Res. 31, 387–396 (1998). 4. Sander, W. Carbonyl oxides: zwitterions or diradicals? Angew. Chem. Int. Ed. Engl. 29, 344–354 (1990). 5. Bunnelle, W. H. Preparation, properties, and reactions of carbonyl oxides. Chem. Rev. 91, 335–362 (1991). 6. Hatakeyama, S. & Akimoto, H. Reactions of Criegee intermediates in the gas phase. Res. Chem. Intermed. 20, 503–524 (1994). 7. Criegee, R. & Wenner, G. Die ozonisierung des 9,10-oktalins. Liebigs Ann. Chem. 564, 9–15 (1949). 8. Taatjes, C. A. et al. Direct observation of the gas-phase Criegee intermediate (CH2OO). J. Am. Chem. Soc. 130, 11883–11885 (2008). 9. Welz, O. et al. Direct kinetic measurements of Criegee intermediate (CH2OO) formed by reaction of CH2I with O2. Science 335, 204–207 (2012). 6

DOI: 10.1038/NCHEM.1890

10. Beames, J. M., Liu, F., Lu, L. & Lester, M. I. Ultraviolet spectrum and photochemistry of the simplest Criegee intermediate CH2OO. J. Am. Chem. Soc. 134, 20045–20048 (2012). 11. Su, Y-T., Huang, Y-H., Witek, H. A. & Lee, Y-P. Infrared absorption spectrum of the simplest Criegee intermediate CH2OO. Science 340, 174–176 (2013). 12. Nakajima, M. & Endo, Y. Determination of the molecular structure of the simplest Criegee intermediate CH2OO. J. Chem. Phys. 139, 101103 (2013). 13. Nguyen, M. T., Ngyuen, T. L., Ngan, V. T. & Ngyuen, H. M. T. Heats of formation of the Criegee formaldehyde oxide and dioxirane. Chem. Phys. Lett. 448, 183–188 (2007). 14. Anglada, J. M., Gonzalez, J. & Torrent-Sucarrat, M. Effects of the substituents on the reactivity of carbonyl oxides. A theoretical study on the reaction of substituted carbonyl oxides with water. Phys. Chem. Chem. Phys. 13, 13034–13045 (2011). 15. Vereecken, L. & Francisco, J. S. Theoretical studies of atmospheric reaction mechanisms in the troposphere. Chem. Soc. Rev. 41, 6259–6293 (2012). 16. Cremer, D., Gauss, J., Kraka, E., Stanton, J. F. & Bartlett, R. J. A CCSD (T) investigation of carbonyl oxide and dioxirane. Equilibrium geometries, dipole moments, infrared spectra, heats of formation and isomerization energies. Chem. Phys. Lett. 209, 547–556 (1993). 17. Fang, D-C. & Fu, X-Y. CASSCF and CASþ1þ2 studies on the potential energy surface and the rate constants for the reactions between CH2 and O2. J. Phys. Chem. A 106, 2988–2993 (2002). 18. Cool, T. A., Wang, J., Nakajima, K., Taatjes, C. A. & McIlroy, A. Photoionization cross sections for reaction intermediates in hydrocarbon combustion. Int. J. Mass Spectrom. 247, 18–27 (2005). 19. Sander, S. P. et al. Chemical Kinetics and Photochemical Data for Use in Atmospheric Studies (Evaluation Number 14, JPL Publication 02-25, 2003). 20. Mo¨ssinger, J. C., Shallcross, D. E. & Cox, R. A. UV-vis absorption cross-sections and atmospheric lifetimes of CH2Br2 , CH2I2 and CH2BrI. J. Chem. Soc. Faraday Trans. 94, 1391–1396 (1998). 21. Vogt, R., Sander, R., Glasow, R. V. & Crutzen, P. J. Iodine chemistry and its role in halogen activation and ozone loss in the marine boundary layer: a model study. J. Atmos. Chem. 32, 375–395 (1999). 22. Saiz-Lopez, A. et al. Atmospheric chemistry of iodine. Chem. Rev. 112, 1773–1804 (2012). 23. Rienstra-Kiracofe, J. C., Allen, W. D. & Schaefer, H. F. III. The C2H5 þ O2 reaction mechanism: high-level ab initio characterizations. J. Phys. Chem. A 104, 9823–9840 (2000). 24. Liang, Y-N., Li, J., Wang, Q-D., Wang, F. & Li, X-Y. Computational study of the reaction mechanism of the methylperoxy self-reaction. J. Phys. Chem. A 115, 13534–13541 (2011). 25. Vereecken, L., Harder, H. & Novelli, A. The reaction of Criegee intermediates with NO, RO2 , and SO2 , and their fate in the atmosphere. Phys. Chem. Chem. Phys. 14, 14682–14695 (2012). 26. Kee, R. J., Rupley, F. M. & Miller, J. A. Chemkin-II: A Fortran Chemical Kinetics Package for the Analysis of Gas-Phase Chemical Kinetics Sandia Report SAND89-8009B (Sandia National Laboratories, 1995). 27. Masaki, A., Tsunashima, S. & Washida, N. Rate constants for reactions of substituted methyl radicals (CH2OCH3 , CH2NH2 , CH2I, and CH2CN) with O2. J. Phys. Chem. 99, 13126–13131 (1995). 28. Eskola, A. J., Wojcik-Pastuszka, D., Ratajczak E. & Timonen, R. S. Kinetics of the reactions of CH2Br and CH2I radicals with molecular oxygen at atmospheric temperatures. Phys. Chem. Chem. Phys. 8, 1416–1424 (2006). 29. Atkinson, R. et al. Summary of Evaluated Kinetic and Photochemical Data for Atmospheric Chemistry, IUPAC Subcommittee on Gas Kinetic Data Evaluation for Atmospheric Chemistry (2006). http://rpw.chem.ox.ac.uk/IUPACsumm_web_latest.pdf. 30. Taatjes, C. A. et al. Direct measurement of Criegee intermediate (CH2OO) reactions with acetone, acetaldehyde, and hexafluoroacetone. Phys. Chem. Chem. Phys. 14, 10391–10400 (2012). 31. Neeb, P., Horie, O. & Moortgat, G. K. The ethane-ozone reaction in the gas phase. J. Phys. Chem. A 102, 6778–6785 (1998). 32. McFiggans, G. et al. Direct evidence for coastal iodine particles from Laminaria marcroalgae – linkage to emissions of molecular iodine. Atmos. Chem. Phys. 4, 701–713 (2004). 33. Gravestock, T. J., Blitz, M. A., Bloss, W. J. & Heard D. E. A multidimensional study of the reaction CH2I þ O2: products and atmospheric implications. ChemPhysChem 11, 3928–3941 (2010). 34. Stone, D., Blitz, M., Daubney, L., Ingham, T. & Seakins, P. CH2OO Criegee biradical yields following photolysis of CH2I2 in O2. Phys. Chem. Chem. Phys. 15, 19119–19124 (2013). 35. Cotter, E. S. N., Booth, N. J., Canosa-Mas, C. E. & Wayne, R. P. Release of iodine in the atmospheric oxidation of alkyl iodides and the fates of iodinated alkoxy radicals. Atmos. Environ. 35, 2169–2178 (2001). 36. Sehested, J., Ellermann, T. & Nielsen, O. J. A spectrokinetic study of CH2I and CH2IO2 radicals. Int. J. Chem. Kinet. 26, 259–272 (1994). 37. Becke, A. D. Density-functional thermochemistry. III. The role of exact exchange. J. Chem. Phys. 98, 5648–5652 (1993).

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY

ARTICLES

DOI: 10.1038/NCHEM.1890

38. Lee, C., Yang, W. & Parr, R. G. Development of the Colle–Salvetti correlationenergy formula into a functional of the electron density. Phys. Rev. B 37, 785–789 (1988). 39. Dunning, T. H. Gaussian basis sets for use in correlated molecular calculations. I. The atoms boron through neon and hydrogen. J. Chem. Phys. 90, 1007–1023 (1989). 40. Peterson, K. A., Shepler, B. C., Figgen, D. & Stoll, H. On the spectroscopic and thermochemical properties of ClO, BrO, IO, and their anions. J. Phys. Chem. A 110, 13877–13883 (2006). 41. Pople, J. A., Head-Gordon, M. & Raghavachari, K. Quadratic configuration interaction. A general technique for determining electron correlation energies. J. Chem. Phys. 87, 5968–5975 (1987). 42. Scuseria, G. E. & Schaefer III, H. F. Is coupled cluster singles and doubles (CCSD) more computationally intensive than quadratic configuration interaction (QCISD)? J. Chem. Phys. 90, 3700–3703 (1989). 43. Gonzalez, C. & Schlegel, H. B. An improved algorithm for reaction path following. J. Chem. Phys. 90, 2154–2161 (1989). 44. Frisch, M. J. et al. GAUSSIAN 09, Revision A02 (Gaussian, Inc., Wallingford Connecticut, 2009). 45. Werner, H-J. et al. MOLPRO, version 2009.1. A package of ab initio programs, http://www.molpro.net (University College Cardiff Consultants, Cardiff, UK). 46. Wardlaw, D. M. & Marcus, R. A. RRKM reaction rate theory for transition states of any looseness. Chem. Phys. Lett. 110, 230–234 (1984). 47. Wardlaw, D. M. & Marcus, R. A. Unimolecular reaction rate theory for transition states of partial looseness. II. Implementation and analysis with applications to NO2 and C2H6 dissociations. J. Chem. Phys. 83, 3462–3480 (1985). 48. Klippenstein, S. J. Variational optimizations in the Rice–Ramsperger–Kassel– Marcus theory calculations for unimolecular dissociations with no reverse barrier. J. Chem. Phys. 96, 367–371 (1992).

49. Klippenstein, S. J. & Marcus, R. A. High pressure rate constants for unimolecular dissociation/free radical recombination: determination of the quantum correction via quantum Monte Carlo path integration. J. Chem. Phys. 87, 3410–3417 (1987). 50. Klippenstein, S. J., Wagner, A. F., Dunbar, R. C., Wardlaw, D. M. & Robertson, S. H. VARIFLEX Version 1.00 (Argonne National Laboratory, Argonne, Illinois, 1999).

Acknowledgements The National Science Council of Taiwan (grants NSC102-2745-M-009-001-ASP and NSC101-2113-M-009-002) and the Ministry of Education, Taiwan (‘ATU Plan’ of the National Chiao Tung University) supported this work. The National Center for HighPerformance Computing provided computer time.

Author contributions Y-T.S. performed the experiments and analysed the data, H-Y.L. performed the kinetic simulations and R.P. performed the calculations. H.M. conceived and designed the kinetic analysis. M.C.L. conceived and designed the calculations. Y-P.L. conceived and designed the experiments and wrote a major part of the paper. H.M. and M.C.L. contributed to writing sections of the paper.

Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to M.C.L. and Y.P.L.

Competing financial interests The authors declare no competing financial interests.

NATURE CHEMISTRY | ADVANCE ONLINE PUBLICATION | www.nature.com/naturechemistry

© 2014 Macmillan Publishers Limited. All rights reserved.

7