Unimolecular Decomposition Kinetics of the Stabilised Criegee Intermediates CH 2 OO and CD 2 OO

Decomposition kinetics of stabilised CH 2 OO and CD 2 OO Criegee intermediates have been investigated as a function of temperature (450 – 650 K) and pressure (2 – 350 Torr) using flash photolysis coupled with time-resolved cavity-enhanced broadband UV absorption spectroscopy. Decomposition of CD 2 OO was observed to be faster than CH 2 OO under equivalent conditions. Production of OH radicals following CH 2 OO decomposition was also monitored using flash photolysis with laser-induced fluorescence (LIF), with results indicating direct production of OH in the v=0 and v=1 states in low yields. Master equation calculations performed using the Master Equation Solver for Multi-Energy well Reactions (MESMER) enabled fitting of the barriers for the decomposition of CH 2 OO and CD 2 OO to the experimental data. Parameterisations of the decomposition rate coefficients, calculated by MESMER, are provided for use in atmospheric models and implications of the results are discussed. For CH 2 OO, the MESMER fits require an increase in the calculated barrier height from 78.2 kJ mol -1 to 81.8 kJ mol -1 using a temperature-dependent exponential down model for collisional energy transfer with <(cid:507) E> down = 32.6 (T/298 K) 1.7 cm -1 in He. The low- and high-pressure limit rate coefficients are k 1,0 = 3.2 × 10 -4 (T/298) -5.81 exp(-12770/T) cm 3 s -1 and k 1,∞ = an investigation of the reactions of CH 2 OO with SO 2 and water vapour at A further investigation of the decomposition of stabilised CH 2 OO was performed using a free-jet flow reactor, in which CH 2 OO was produced by ozonolysis of ethene in air and monitored by sampling from the flow reactor and titration to H 2 SO 4 through reaction with SO 2 . The design of the free-jet flow reactor reduces wall losses, and the pressure used in the reactor inhibits losses through diffusion, thus minimising the contributions to CH 2 OO loss from physical processes. The study reported a rate coefficient for stabilised CH 2 OO decomposition of (0.19 ± 0.07) s -1 at 297 K and 1 bar, with quantum chemical calculations at the CCSD(T)/aug-cc-pVTZ level giving a calculated barrier height to decomposition of 78.9 kJ mol -1 . 23 A high pressure limiting rate coefficient of 0.25 s -1 was determined from master equation calculations at 297 K, 23 similar to that obtained in an earlier theoretical study, 24 with fall-off behaviour predicted at the pressure of the experiment and predictions for the rate coefficient at 1 bar between 0.037 s -1 and 0.12 s -1 . 23 Improved agreement with the experimental value was achieved by a reduction in the calculated barrier height to 76.8 kJ mol -1 , resulting in an increase in the predicted high pressure limiting rate coefficient to 0.58 s -1 at 297 K and predictions for the rate coefficient at 1 bar and 297 K between 0.08 s -1 and 0.26 s -1 . 23 of CH and pressure. Full assessment of the impacts of stabilised CH 2 OO decomposition in the atmosphere, where it may be in competition with CH 2 OO + SO 2 at low SO 2 concentrations given the range of CH 2 OO decomposition rate coefficients reported in the literature (0.037 – 11.6 s -1 ), and in chamber studies of ozonolysis reactions, is therefore hindered by a lack of measurements of stabilised CH 2 OO kinetics over a wide range of temperatures and pressures. In this work we report a detailed study of the decomposition kinetics of stabilised CH 2 OO (k 1 ) and CD 2 OO (k 2 ) as a function of temperature and pressure. Flash photolysis of CH 2 I 2 /O 2 /He and CD 2 I 2 /O 2 /He gas mixtures coupled with time-resolved cavity enhanced broadband UV absorption spectroscopy was used to monitor changes in CH 2 I 2 /CD 2 I 2 , CH 2 OO/CD 2 OO and IO to determine the kinetics at pressures between 2 and 350 Torr and temperatures between 450 and 650 K, thereby increasing the decomposition rate and minimising effects of physical losses of SCI. The production of OH radicals from CH 2 OO decomposition was also investigated using flash photolysis of CH 2 I 2 /O 2 /N 2 coupled with laser-induced fluorescence (LIF) spectroscopy at temperatures between 500 and 600 K and pressures in the range 10 to 95 Torr. We discuss the results from the UV experiments in which CH 2 OO and CD 2 OO are directly monitored, and compare the results to probe any potential kinetic isotope effects in the decomposition mechanism. We then discuss the LIF experiments in which the OH products from stabilised CH 2 OO are probed, and compare the results to the UV experiments. Finally, we discuss the results from Master equation calculations, using the Master Equation Solver for Multi-Energy well Reactions (MESMER), 25 which were performed to fit the barrier height for CH 2 OO decomposition to the CH 2 OO decomposition kinetics determined in the UV experiments, thus providing a theoretical framework for the reaction and a full parameterisation as a monitored by K-type thermocouples situated along the length of the reaction cell. Experiments were performed in He (Matheson, 99.9999 %) at pressures between 2 and 350 Torr and at temperatures in the range 450 to 650 K, with CH 2 I 2 (Aldrich, 99 %)/CD 2 I 2 (Aldrich, 98 %) concentrations in the range 8 × 10 12 to 8 × 10 13 cm -3 and O 2 (Matheson, 99.9999 %) concentrations varied between 1 × 10 16 and 7 × 10 18 cm -3 . Gases and chemicals were used as supplied. The decomposition kinetics of CH 2 OO Criegee intermediate have been investigated at temperatures between 450 K and 650 K and pressures in the range 2 – 350 Torr of He using flash photolysis of CH 2 I 2 in O 2 and a combination of time-resolved cavity enhanced broadband UV absorption spectroscopy, for direct monitoring of CH 2 OO, and laser-induced fluorescence, for monitoring of OH decomposition products. Kinetics of CD 2 OO decomposition were also investigated using flash photolysis of CD 2 I 2 with time-resolved cavity enhanced broadband UV absorption spectroscopy. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of the USDOE or the United States Government. LS was supported by the Division of Chemical Sciences, Geosciences and Biosciences, BES/USDOE, through the Argonne-Sandia Consortium on High Pressure Combustion Chemistry.


Introduction
Atmospheric oxidation initiated by ozone (O3) is a key removal mechanism for unsaturated hydrocarbons and volatile organic compounds (VOCs) emitted into the atmosphere, and proceeds via the addition of ozone to a carbon-carbon double bond in an ozonolysis reaction. 1,2 Such reactions lead to the production of Criegee intermediates (R2COO), and are associated with high exothermicities (typically ~250 kJ mol -1 ) 1 . Nascent Criegee intermediates produced in ozonolysis reactions thus contain an excess of internal energy, which can promote unimolecular decomposition, leading to production of key atmospheric species including OH, HO2 and CO, or can be quenched through collisional energy transfer to surrounding gas molecules, leading to the production of stabilised Criegee intermediates (SCIs) which can undergo further chemistry in the atmosphere.
Recent developments have identified photolytic sources of stabilised Criegee intermediates, facilitating laboratory studies of SCI reaction kinetics. 3,4 Subsequent experimental studies have largely focused on the bimolecular reactions of SCIs, often finding higher reactivity than previously expected, with relatively few studies placing an emphasis on SCI unimolecular decomposition reactions. [5][6][7][8] Decomposition reactions of SCIs are potentially important in their own right; furthermore, analysis of SCI decomposition may provide insight to the decomposition of nascent excited CIs produced in atmospheric ozonolysis reactions. In addition, production of the CH2OO Criegee intermediate in the combustion of the biofuel dimethyl ether (DME, CH3OCH3) has been proposed, with unimolecular decomposition of CH2OO likely to be important under combustion conditions. 9,10 Because CH2OO decomposition involves possible radical and closed-shell product pathways, the knowledge of its rate coefficient and product branching may explain why no evidence of Criegee intermediates has yet appeared in the combustion studies of DME or larger ethers.
Investigation of the potential energy surface (PES) for CH2OO decomposition (R1) using quantum chemical calculations at the CCSD(T)/aug-cc-pVTZ//B3LYP/aug-cc-pVTZ level indicated the presence of two main reaction channels that produce formic acid and dioxirane, with barriers of 67. 4 and 78.2 kJ mol -1 , respectively, and with the formic acid channel proceeding via a roaming-like transition state. 11 Both channels potentially lead to the final products H2 + CO2, H2O + CO, or HCO + OH.

CH2OO
H2 + CO2 (R1a) H2O + CO (R1b) However, these calculations apparently underestimated the barrier for the formic acid channel owing to spin-contamination. 12 Higher level multi-reference calculations 12,13 predict a similar barrier of 79.5 kJ mol -1 for the dioxirane channel, but a much higher one for the formic acid channel (~205 kJ mol -1 ), rendering the formic acid pathway uncompetitive with the dioxirane channel. 12 If active, the formic acid channel would result in a strong kinetic isotope effect between CH2OO and CD2OO.
Theoretical studies of the decomposition of vibrationally excited nascent CH2OO produced in ethene ozonolysis, using high-accuracy calculations performed with HEAT-345(Q), have also revealed potential decomposition pathways involving dioxirane (with a barrier of 79.9 kJ mol -1 and ultimately leading to H2 and CO2) and formic acid (with final products H2 + CO2 or H2O + CO). An alternative pathway to that involving dioxirane was also reported, leading to production of OH and HCO, but with a calculated barrier of 133.1 kJ mol -1 and thus unlikely to compete with the dioxirane pathway. 14,15 Experimental evidence for OH production from CH2OO also comes from studies in which the behaviour of OH radicals has been probed following the photolysis of CH2I2 in O2/Ar gas mixtures. 16,17 These studies have successfully used OH measurements as proxies to determine the kinetics of CH2OO reactions, including with SO2. This finding implicates the decomposition of stabilised CH2OO as the OH radical source, although decomposition kinetics were not directly probed. 16,17 Similarly, a study of SCIs generated by alkene ozonolysis has also identified the decomposition of stabilised CH2OO as a potential source of OH, but kinetic information was limited and a low OH yield was reported. 18 Ozonolysis experiments performed in atmospheric simulation chambers have been used to infer the kinetics of stabilised CH2OO decomposition, but with large uncertainties, since first-order losses of CH2OO by physical processes such as wall reactions are difficult to distinguish from loss through unimolecular decomposition. A recent series of chamber experiments at atmospheric temperature and pressure reported an upper limit of 4.2 s -1 for the first-order loss of stabilised CH2OO, which includes a contribution from the decomposition reaction. 19,20 This value was determined by measuring the ratio of the rate coefficient for first-order losses to that for the reaction of CH2OO with SO2, and the decomposition rate was indistinguishable from zero within the measurement uncertainties. 19,20 Measurements of photolytically generated CH2OO using cavity ringdown spectroscopy (CRDS) have also been used to infer the unimolecular decomposition kinetics in N2 at 293 K in the pressure range 7 to 30 Torr, giving an upper limit to the decomposition rate coefficient of (11.6 ± 8.0) s -1 . Similarly to the results obtained in the chamber experiments, the first-order removal of CH2OO in the CRDS measurements contains not only the contribution from the unimolecular decomposition but also contributions from physical processes such as diffusion and wall loss. 21 A similar upper limit to the rate coefficient for decomposition of (9.4 ± 1.7) s -1 was reported from an investigation of the reactions of CH2OO with SO2 and water vapour at 293 K and atmospheric pressure. 22 A further investigation of the decomposition of stabilised CH2OO was performed using a free-jet flow reactor, in which CH2OO was produced by ozonolysis of ethene in air and monitored by sampling from the flow reactor and titration to H2SO4 through reaction with SO2. 23 The design of the free-jet flow reactor reduces wall losses, and the pressure used in the reactor inhibits losses through diffusion, thus minimising the contributions to CH2OO loss from physical processes. The study reported a rate coefficient for stabilised CH2OO decomposition of (0.19 ± 0.07) s -1 at 297 K and 1 bar, with quantum chemical calculations at the CCSD(T)/aug-cc-pVTZ level giving a calculated barrier height to decomposition of 78.9 kJ mol -1 . 23 A high pressure limiting rate coefficient of 0.25 s -1 was determined from master equation calculations at 297 K, 23 similar to that obtained in an earlier theoretical study, 24 with fall-off behaviour predicted at the pressure of the experiment and predictions for the rate coefficient at 1 bar between 0.037 s -1 and 0.12 s -1 . 23 Improved agreement with the experimental value was achieved by a reduction in the calculated barrier height to 76.8 kJ mol -1 , resulting in an increase in the predicted high pressure limiting rate coefficient to 0.58 s -1 at 297 K and predictions for the rate coefficient at 1 bar and 297 K between 0.08 s -1 and 0.26 s -1 . 23 Thus, while there have been several attempts to determine the decomposition kinetics of stabilised CH2OO Criegee intermediates, the results have large uncertainties, and the most reliable measurements obtained using the free-jet flow reactor have only been reported at a single temperature and pressure. Full assessment of the impacts of stabilised CH2OO decomposition in the atmosphere, where it may be in competition with CH2OO + SO2 at low SO2 concentrations given the range of CH2OO decomposition rate coefficients reported in the literature (0.037 -11.6 s -1 ), and in chamber studies of ozonolysis reactions, is therefore hindered by a lack of measurements of stabilised CH2OO kinetics over a wide range of temperatures and pressures. In this work we report a detailed study of the decomposition kinetics of stabilised CH2OO (k1) and CD2OO (k2) as a function of temperature and pressure.
Flash photolysis of CH2I2/O2/He and CD2I2/O2/He gas mixtures coupled with time-resolved cavity enhanced broadband UV absorption spectroscopy was used to monitor changes in CH2I2/CD2I2, CH2OO/CD2OO and IO to determine the kinetics at pressures between 2 and 350 Torr and temperatures between 450 and 650 K, thereby increasing the decomposition rate and minimising effects of physical losses of SCI. The production of OH radicals from CH2OO decomposition was also investigated using flash photolysis of CH2I2/O2/N2 coupled with laser-induced fluorescence (LIF) spectroscopy at temperatures between 500 and 600 K and pressures in the range 10 to 95 Torr. We discuss the results from the UV experiments in which CH2OO and CD2OO are directly monitored, and compare the results to probe any potential kinetic isotope effects in the decomposition mechanism. We then discuss the LIF experiments in which the OH products from stabilised CH2OO are probed, and compare the results to the UV experiments. Finally, we discuss the results from Master equation calculations, using the Master Equation Solver for Multi-Energy well Reactions (MESMER), 25 which were performed to fit the barrier height for CH2OO decomposition to the CH2OO decomposition kinetics determined in the UV experiments, thus providing a theoretical framework for the reaction and a full parameterisation as a function of temperature and pressure.

Experimental UV absorption
Time-resolved cavity-enhanced broadband UV absorption spectroscopy experiments were performed at Sandia National Laboratory, Livermore, USA, using experimental apparatus described in detail in previous publications. 26-28 Precursor gas mixtures containing CH2I2 (or CD2I2), entrained in He, O2, and He buffer gas were admitted into a quartz flow cell using calibrated mass flow controllers (MKS Instruments). Chemistry in the reaction cell was initiated by the photolysis of the di-iodo precursor at a wavelength of 266 nm, which was generated by the 4 th harmonic of an Nd:YAG laser (Continuum Surelite III) with a typical fluence of ~19 mJ cm -2 , with O2 concentrations such that the production of CH2OO (or CD2OO) was rapid compared to the subsequent decay. Transient absorptions in the reaction cell were monitored using a Xe arc lamp (Newport Corp.), which was reflected between two concave high reflectivity mirrors (JDSU Inc.) forming an optical resonator cavity 1.6 m in length, operating over probe wavelengths between 300 and 450 nm simultaneously with total effective absorption path length of 40-56 m. Light exiting the optical cavity was directed into a time-resolved spectrometer, consisting of a rotating mirror, synchronized with the photolysis laser, that rapidly sweeps the probe beam vertically, followed by a ruled grating, which provides spectral dispersion in the horizontal plane. Spectral and temporal information contained in the probe beam are thus spatially mapped onto the horizontal and vertical position, respectively, at the focal plane of the spectrometer. The time evolution of the entire broadband cavity output is recorded by a TEC-cooled 1024×1024element CCD camera for every laser shot, and averaged on-chip (300 -600 shots in the present work), prior to transfer to a computer for data analysis.
Transient absorption spectra were computed by Beer's Law from the difference between probe light intensities with (ION) and without (IOFF) the photolysis laser: ln . The mirror rotation was adjusted between 1 and 10 Hz (corresponding to total observation times between 13.5 and 1.35 ms, respectively) as needed to capture the kinetics under investigation. The experimental resolution of this spectrometer is ultimately determined by spatial focusing of the probe beam on the CCD sensor: ~7 pixels (FWHM), corresponding to spectral resolution of ~1.5 nm and temporal resolution of ~9 -90 s, depending on the mirror rotation frequency. The total flow rate through the reaction cell was adjusted with changes in pressure and laser repetition rate to ensure a fresh sample of gas in the cell for each photolysis shot.
The pressure in the reaction cell was maintained by a roots pump and actively controlled by a butterfly valve throttling the exit of the cell. Temperatures in the reaction cell were controlled by a series of ceramic heaters (Watlow) surrounding the cell and monitored by K-type thermocouples situated along the length of the reaction cell. Experiments were performed in He (Matheson, 99.9999 %) at pressures between 2 and 350 Torr and at temperatures in the range 450 to 650 K, with CH2I2 (Aldrich, 99 %)/CD2I2 (Aldrich, 98 %) concentrations in the range 8 × 10 12 to 8 × 10 13 cm -3 and O2 (Matheson, 99.9999 %) concentrations varied between 1 × 10 16 and 7 × 10 18 cm -3 . Gases and chemicals were used as supplied.
Concentrations of CH2OO were determined by fitting reference spectra for the CH2I2 precursor, CH2OO and IO (generated by secondary chemistry within the system) to the observed total absorbance between 300 and 440 nm for each time point throughout the reaction. Typical absorbance signals of 10 -3 -10 -4 were measured in this work, which correspond to changes in concentration of 0.001-0.0001 % (assuming 100 % photodissociation on absorption of a photon), which is insignificant compared to the changes in concentration owing to reaction. Figure 1 shows a typical concentrationtime profile for CH2OO. Details regarding the fitting procedure are given in the Supplementary Information.

Laser-induced fluorescence
Laser-induced fluorescence (LIF) experiments were performed at the University of Leeds, UK, in a slow flow reactor which has been described in detail in previous work. 29-31 Precursor gas mixtures (CH2I2/O2/N2) were prepared in a glass gas manifold and passed into a stainless steel six-way cross at known flow rates determined by calibrated mass flow controllers (MKS Instruments). Photolysis of CH2I2, leading to rapid production of CH2OO, was achieved at a wavelength of 355 nm using the 3rd harmonic of an Nd:YAG laser (Continuum Powerlite 8010). Experiments were typically performed at a repetition rate of 10 Hz, although the lasers were also operated at lower repetition rates to ensure that there were no interferences from photolysis products. The laser fluence was typically ~20 mJ cm -2 .
The pressure in the reaction cell was monitored by a capacitance manometer, and was maintained by a rotary pump throttled by a needle valve on the exhaust line. Heating of the reaction cell was achieved by a series of cartridge heaters surrounding the cell, with temperatures monitored by K-type thermocouples situated close to the reaction zone.
OH radicals produced in the system were monitored by off-resonance laser-induced fluorescence following either A 2 (v'=1) X 2 (v''=0) excitation at a wavelength of 282 nm for detection of OH in the ground vibrational state, OH(v''=0), or A 2 (v'=1) X 2 (v''=1)) excitation at 288 nm for detection of OH in its first vibrationally excited state, OH(v''=1). The 532 nm output of a Nd:YAG laser (Continuum Powerlite 8010) was used to pump a dye laser (Spectra Physics PDL-3) operating on either Rhodamine-6-G or pyromethene 597 dye, with the dye output frequency-doubled to generate light at 282 or 288 nm, respectively. For both excitation wavelengths, the off-resonant OH fluorescence at ~308 nm was passed through an interference filter (Barr Associates, (308 ± 5) nm) and monitored by a channel photomultiplier (CPM, Perkin-Elmer C1943P) mounted perpendicular to the plane of photolysis and probe laser beams. The CPM signal was digitised and integrated on an oscilloscope (LeCroy LT262) prior to being passed to a computer for data analysis. The time delay between the photolysis and probe laser pulses was controlled by a digital delay generator (SRS DG535) and varied to enable monitoring of the OH profiles as a function of time following photolysis of the gas mixture. Kinetic traces typically consisted of 200 time points, with each time point averaged 5-10 times.

Results
UV absorption CH2OO Figure 1 shows a typical concentration-time profile for CH2OO determined from the observed absorbance between 300 and 440 nm. The CH2OO profiles were fitted to a first-order loss process, convolved with a Gaussian instrument response function, which was determined by the spatial profile of the incident probe light on the CCD detector. 26 Fits to a mixed first-and second-order loss process were also examined, but were insensitive to any second-order loss and gave first-order losses within 5 % of those obtained from the fits considering first-order loss only. The results obtained from the first-order fits are shown in Figure 2 as a function of temperature and pressure, and are given in Table  1. Experiments in which the pulse repetition rate of the photolysis laser was varied did not yield any significant differences in the fitted rate coefficients describing the CH2OO decays.
At temperatures below 500 K, there is little variation in the observed rate coefficients as a function of pressure, although there is an increase from 450 K to 475 K. At temperatures of 500 K and above, the rate coefficients increase with increasing temperature and pressure. The loss of CH2OO thus appears to contain contributions from two processes, a pressure-and temperature-dependent term, k(p,T), and a pressure-independent temperature-dependent term, k(T). Given the PES for CH2OO decomposition, 11,12,14,23 we attribute the pressure-dependent term to CH2OO decomposition, and the pressure-independent term to other background losses of CH2OO, such that the observed rate coefficient, k1,obs, is given by the sum of k1(p,T) and kbg(T). A global fit using data at all temperatures and pressures was performed to determine k1(p,T) and kbg(T), with k1(p,T) described by the basic Troe equation 32 as shown in Equation 1: where k1,0(T) is the low-pressure limiting rate coefficient for CH2OO decomposition, k1,∞(T) is the high-pressure limiting rate coefficient for CH2OO decomposition, M is the total number density, Fc is the broadening factor, and k1,bg(T) is the pressure-independent rate coefficient for secondary background removal processes for CH2OO.  Table 1. While this parameterisation can be used to provide a value for k1 at 298 K and 760 Torr, the extrapolation is subject to significant uncertainties since the experiments do not cover a sufficiently broad range of pressures in the fall-off regime. Instead, the parameterisation is performed primarily to determine the contributions to the total loss from decomposition and background losses, with the pressure and temperature dependence of the decomposition best described by the Master Equation treatment discussed below.
The pressure-independent background losses of CH2OO, kbg(T), demonstrate the presence of removal processes other than CH2OO decomposition, including wall losses and secondary chemical loss of CH2OO. Given the magnitude and temperature dependence of the pressure-independent contribution to the loss, chemical reactions of CH2OO are likely to be the dominant factor. Results from mixedorder fits to the observed decays indicated little sensitivity to second-order processes, and thus a negligible contribution from CH2OO self-reaction. At 450 K, the data suggest a contribution from reaction between CH2OO and CH2I2 (see Supplementary Information), with a bimolecular rate coefficient of (8.2 ± 1.7) × 10 -12 cm 3 s -1 . The reaction has also recently been observed by Liu et al., 17 with a rate coefficient of (5.2 ± 2.6) × 10 -14 cm 3 s -1 at 298 K. At temperatures above 450 K, concentrations of CH2I2 were not varied over a sufficient range to fully assess the role of CH2OO + CH2I2; however, the results overall are consistent with the pressure-independent loss term at all temperatures coming largely from the pseudo-first-order loss of CH2OO through reaction with CH2I2.

CD2OO
Experiments were also performed in which CD2I2 was photolysed in the presence of excess O2 to generate CD2OO at temperatures between 450 and 650 K. Figure 3 shows the normalised absorption spectra for CH2OO and CD2OO determined from experiments reported in this work at 295 K and 10 Torr (see Supplementary Information for details regarding characterisation of the spectra and fitting to determine normalised concentrations). The spectra are broadly similar in both shape and the position of the peak absorption cross-section, in agreement with a recent report of the CD2OO spectrum. 35 While previous studies of the CH2OO 26,36,37 and CD2OO 35 spectra have shown the presence of vibronic structure at wavelengths above 340 nm, the resolution of the experiments reported here was insufficient to resolve the vibronic structure for either CH2OO or CD2OO.
Decays for CD2OO were fit to first-order loss kinetics, convolved with a Gaussian instrument function, and the observed rate coefficients are shown in Figure 4 and Table 2. Similarly to CH2OO, the rate coefficients describing the decays of CD2OO exhibit pressure dependence at temperatures of 500 K and above, but not at 450 K, indicating contributions from both the pressure-dependent CD2OO decomposition (k2) and pressure-independent secondary background losses (k2,bg). The observed rate coefficients for the CD2OO decays were thus fit to an analogous expression to Equation 1 given for CH2OO. We propose that the background loss of CD2OO is, at least in part, a result of reaction with CD2I2. However, constraining k2,bg(T) to Arrhenius behaviour gave poor fits to data at 450 K, where the background loss dominates the observed decay, potentially owing to fewer data points compared to CH2OO. The fits with k2,bg unconstrained to Arrhenius behaviour give k2,0 = (1.5±4.0) × 10 -11 exp(-(4640±1800)/T) cm 3 s -1 and k2,∞ = (6.4±86.7) × 10 15 exp(-(14750±7800)/T) s -1 with Fc fixed at a value of 0.6. Values for k2,bg are summarised in Table 2 and can be approximated by the expression k2,bg = (2.4±6.9) × 10 4 exp(-(2080±1570)/T) s -1 . Similarly to the parameterisation for CH2OO decomposition, the uncertainties in the individual fit parameters for k2,obs are deceptively large, owing to correlations between the fit parameters. Again, the fits are performed largely to determine k2,bg, rather than to extrapolate k2 to 298 K and 760 Torr. Consideration of these correlations between the fit parameters in the uncertainty analysis, as described in the Supplementary Information for CH2OO, indicates a median total uncertainty of 21 % in the fits to k2,obs. The total uncertainties in the fits to CD2OO decays are larger than for CH2OO, and the fits display greater variability, since there are fewer data points for CD2OO compared to CH2OO, particularly at the lower pressures. Fit results and uncertainties for CD2OO are given in Table 2.
Laser-induced fluorescence OH(v=0) Figure 5 shows a typical OH(v=0) time profile following the photolysis of CH2I2/O2/N2 mixtures. The OH(v=0) signal exhibited a near-instant (photolytic) production of OH(v=0), with a further rapid growth followed by decay. The amplitudes of the photolytic signal and of the subsequent rapid growth were both observed to display a linear dependence on the initial CH2I2 concentration, while the rate of the rapid growth was also observed to depend on the total pressure and on the concentration of O2. These observations are consistent with production of OH radicals in the ground vibrational state (v=0) and excited vibrational states (v>0), either directly through photolysis or through rapid reactions of species generated photolytically, followed by relaxation of the OH(v>0) to the ground vibrational state. The near-instant production of OH in v=0 and v>0 states potentially occurs directly from the reaction of excited CH2I * with O2, or from the rapid decomposition of excited CH2OO * . 38,39 The rate of the OH(v=0) decay was dependent on the concentration of CH2I2, indicating removal of OH(v=0) through the expected reaction of OH with CH2I2. However, the observed loss of OH was not well described by a single exponential decay. Instead, it was better described by a biexponential function, indicating a slower growth of OH(v=0) in the system on a timescale similar to the loss through reaction with CH2I2. The slow growth of OH(v=0) in the system was attributed to production through the decomposition of CH2OO. The production and loss of OH in the system was thus assigned to the mechanism in reactions R1 and R3-R7: In the reaction scheme above we assume that the relaxation of higher-lying OH vibrational states is much faster than that of OH(v=1) to OH(v=0). Reactions R1 and R3-R7 are all either first-order or occur under pseudo-first-order conditions, and thus an analytical solution can be obtained to describe the temporal behaviour of OH(v=0) in the system following rapid production of CH2OO and any OH radicals resulting from photolysis (Equation 2).
where SOH,t is the OH(v=0) fluorescence signal at time t, SCH2OO is the amplitude of the OH(v=0) signal deriving from CH2OO decomposition, SOH(v>0),t=0 is the amplitude of the OH(v=0) signal deriving from vibrational relaxation of all photolytically generated OH(v>0) states, k1 is the rate coefficient describing the decomposition of CH2OO, k'6 is the pseudo-first-order rate coefficient for vibrational relaxation of OH(v=1) (i.e. k'6 = k6[M]), and k'7 is the pseudo-first-order rate coefficient for loss of OH(v=0), primarily through reaction with CH2I2 (i.e. k'7 = k7[CH2I2]).
While good fits to the observed OH(v=0) signals could be achieved, as shown in Figure 5, the fits displayed poor sensitivity to individual rate coefficients. The complexity of the mechanism controlling OH(v=0) in the system thus led to difficulties in obtaining reliable CH2OO decomposition kinetics, although the temperatures and pressures over which the slow growth of OH(v=0) in the system was apparent are consistent with those where UV absorption experiments observed CH2OO decomposition.
The OH(v=0) signal attributed to production from CH2OO indicates a low yield of OH from stabilised CH2OO decomposition. Previous experiments indicate that photolysis of CH2I2 at a wavelength of 248 nm leads to near-instant production of HCHO, via generation of excited CH2I* or CH2OO*, followed by subsequent growth of HCHO produced via chemistry of CH2OO. 40,41 These experiments indicated eventual 100 % yield of HCHO from CH2OO (through reactions including CH2OO + CH2OO, CH2OO + I and CH2OO + SO2), with the near-instant yield of HCHO representing approximately 5-10 % of the total HCHO, or CH2OO, yield. Assuming that the near-instant OH signal observed in this work is produced via a similar mechanism to the near-instant HCHO signal observed previously (i.e. via generation of excited CH2I* or CH2OO*), and with similar yields to the nearinstant HCHO signal, we can estimate that the near-instant yield of OH also represents only 5-10 % of the total CH2OO in the system. The OH signals observed in this work were typically dominated by the near-instant signal, comprising both the instant OH(v=0) signal and the relaxation of OH(v>0), with the OH(v=0) produced via CH2OO decomposition being only a fraction of the total OH signal. Thus, the yields of OH(v=0) from CH2OO decomposition are low. For example, for the data shown in Figure 5, the fits to Equation 2 indicate that SCH2OO is ~(46±5) % of the total OH(v=0) signal (i.e. SCH2OO + SOH(v=1),t=0 + SOH,t=0). If we estimate that the near-instant OH signal (SOH,t=0 and SOH(v=1),t=0 combined) represents only 5-10 % of the total CH2OO produced in the system, the yield of OH(v=0) from CH2OO decomposition is ~4-8 %. However, similarly to the kinetic analysis, a fully quantitative analysis of the OH yields in the system is not possible owing to the complexity of the mechanism and poor sensitivity of the fits to individual processes, and the data allow only a qualitative analysis of the OH yields from stabilised CH2OO decomposition.

OH(v=1)
Measurements of OH(v=1) data were initially performed in order to confirm assignment of the OH(v=0) data. A typical time profile for OH(v=1) following the photolysis of CH2I2/O2/N2 is shown in Figure 6. The OH(v=1) profile displays a near-instant growth, owing to rapid production from CH2OO* or CH2I* + O2, as discussed for OH(v=0), followed by growth owing to relaxation of OH(v≥2) states, which are co-produced by the same process involving CH2OO* or CH2I* + O2. The collisional relaxation of OH(v=1) to OH(v=0) was expected to lead to a single exponential decay for OH(v=1). However, fits to the data indicated that the OH(v=1) decays were more suitably described by a biexponential function which accounts for a slow growth of OH(v=1) in the system. Figure 6 shows the fits to the data using Equation 3. These fits were started at sufficiently long delay times (typically 100 µs), to ensure complete collisional relaxation of OH(v≥2) states to OH(v=1), since the low OH(v≥2) yields made fitting the initial rise in OH(v=1) signal difficult 19,20 .
Here SOH(v=1),t is the OH(v=1) fluorescence signal at time t, k'6 is the pseudo-first-order rate coefficient for vibrational relaxation of OH(v=1) (i.e. k'6 = k6[M]), and Sg is the amplitude of the signal arising from slow growth of OH(v=1) which occurs with rate coefficient kg.
The kinetics of the fast component of the OH(v=1) decay were consistent with collisional relaxation to OH(v=0), principally by O2, 42 and are provided in the Supplementary Information. The kinetics of the slow component to the decay displayed a dependence on temperature and total pressure similar to that observed for CH2OO decomposition in the UV experiments, as shown in Figure 7. Thus, we propose that the apparent biexponential decay of OH(v=1) results from a combination of OH(v=1) relaxation to OH(v=0) and direct production of OH(v=1) from decomposition of CH2OO. Thus, Sg in Equation 3 represents the amplitude of the OH(v=1) signal arising from decomposition of CH2OO and kg is equivalent to k1, the rate coefficient for CH2OO decomposition. The results for k1 determined from the OH(v=1) experiments are summarised in Table 3 and compare well to those obtained in the UV experiments in which CH2OO was monitored directly.
The yields of OH(v=1) from decomposition of CH2OO are thus low, since there is little perturbation to the OH(v=0) signal that we attribute to OH(v=1) relaxation. For the data shown in Figure 6, the fitted yield of OH(v=1) from CH2OO decomposition is approximately 30 % of the total OH(v=1) signal. Examination of all fits for v=0 and v=1 OH signals leads us to conclude that if the the v=0 signal is ~4-8 % of the total CH2OO then v≥1 is on the order of 1 %.

Master Equation Analysis
Master equation calculations for the decomposition of CH2OO were performed using the Master Equation Solver for Multi-Energy well Reactions (MESMER), which has been described in detail in previous work. 25,43,44 The energies of each species, including reactants, transition states and products, are divided into a number of levels, known as grains, which contain a defined number of states. These grains are assigned populations, average energies, and, where appropriate, average values of microcanonical rate coefficients, forming the basis for the master equation analysis. Changes in the population distribution among the grains occur through collisional energy transfer via interactions with a thermal bath gas or via transfer from one species to another via reactions governed by the microcanonical rate coefficients in the system.
The equation of motion of the grain population probabilities is represented by: dp/dt = Mp (Equation 4) where p is a vector containing the populations of the energy grains and M is a matrix that contains transition rates between grains and determines the evolution of the grain population distribution owing to collisional activation/deactivation or reaction. The reactive processes are described by Rice, Ramsperger, Kassel and Marcus (RRKM) theory, with a temperature-dependent exponential down model (Equation 5) used to describe collisional transfer energy: where < E>down,T represents the average energy transferred in a downward direction on collision with the bath gas at temperature T, and n is the exponent used to parameterise the temperature dependence.
For the master equation calculations presented in this work, geometries, frequencies and rotational constants for CH2OO, transition states to decomposition and the decomposition products were provided by the calculations of Nguyen et al. 11 at the CCSD(T)/aug-cc-pVTZ//B3LYP/aug-cc-pVTZ level of theory, with the roaming channel leading to formic acid excluded, as suggested by recent improved calculations. 12 If the roaming channel were active, a strong kinetic isotope effect might be expected between CH2OO and CD2OO, which is not supported by the experimental data or the calculations reported in this work. Geometries, frequencies and rotational constants for CD2OO, the transition state to decomposition and the initial intermediate leading to decomposition products were calculated using the Gaussian 09 suite of programs 45 at the M06-2X/aug-cc-pVTZ 46-51 level of theory. The barrier to decomposition was improved via single point energy computations (SPE) of the stationary structures using coupled cluster calculations with single, double and pertubative triple excitations (CCSD(T)). 52 The SPEs were extrapolated to the complete basis set limit (CBS) with the use of correlation-consistent basis sets (aug-cc-pVXT, X=D,T,Q) [47][48][49][50][51] and the extrapolation scheme presented by Peterson et al. 48 Barrier heights and stationary point energies were corrected for zero point energies (ZPEs), and although the deuterated reactant does have a lower ZPE compared to the non-deuterated reactant, the same effect is observed in the respective transition states, such that the barrier height is similar between the deuterated and non-deuterated systems. The barrier calculated at the CCSD(T)/CBS//M06-2//aug-cc-pVTZ level of theory (81.04 kJ mol -1 ) is in agreement with the barrier of 78.24 kJ mol -1 obtained by Nguyen et al. 11 for CH2OO.
Pressure dependent rate coefficients for CH2OO and CD2OO were calculated by MESMER using an inverse Laplace transformation to determine microcanonical rate coefficients (k(E)), with molecular densities of states calculated by a rigid rotor-harmonic oscillator approximation. 43 A grain size of 100 cm -1 was used in the calculations described here. The molecular constants and further details regarding the calculations are given in the Supplementary Information.
The master equation calculations were optimised by varying the parameters < E>down,298K and n in Equation 5, as well as the barrier height to decomposition. A fit to the rate coefficients for CH2OO and CD2OO determined from the UV experiments, was performed using a Levenburg-Marquardt algorithm to minimise the merit function 2 , as defined by Equation 6: (Equation 6) where kexp(Ti,pi) and kmodel(Ti,pi) are the experimental and modelled rate coefficients at temperature Ti and pressure pi, respectively, i 2 is the experimental uncertainty at temperature Ti and pressure pi, N is the total number of experimental measurements, and kexp(Ti,pi) is the experimentally determined value of k1 (or k2) after subtraction of k1,bg (or k2,bg). The temperatures and pressures used in the fits to Equation 5 for CH2OO and CD2OO are highlighted in Tables 1 and 2, respectively. In both cases, only data at temperatures which show a clear pressure dependence and which have a positive value for k1 or k2 after the subtraction of k1,bg or k2,bg from k1,obs or k2,obs, respectively, are included in the fits. Thus, data at temperatures below 500 K are excluded from the fits, and some of the data at low pressures are excluded owing to the uncertainties in separating the decomposition from the background loss when the decomposition is slow compared to the background losses resulting from reaction with the precursor and diffusion.
Since the decomposition of CH2OO is thought to proceed via a single barrier, 11,12,14,23 optimisation of < E>down,298K, n and the barrier to decomposition can be achieved through consideration of the simplified potential energy surface shown in blue in Figure 8, consisting of only CH2OO, the first transition state (TS2), and the cyclic intermediate (dioxirane). An analogous PES was used for CD2OO, in which a further simplification was made such that it considers only the energies of CD2OO and the initial transition state which ultimately leads to product formation. Figure 8 shows the results of the optimisation of the barrier to decomposition of CH2OO, with the comparison between the experimentally observed rate coefficients for CH2OO decomposition and the output from the MESMER optimisation given in Figure 2 and Table 1. The MESMER fits to the data yield < E>down = (32.6 ± 13.7) (T/298 K) (1.7 ± 0.4) cm -1 , and require an increase of 3.6 kJ mol -1 in the calculated barrier height from 78.2 kJ mol -1 to 81.8 kJ mol -1 , giving k1 = × 10 -3 s -1 in He at T = 298 K and p = 760 Torr. For the experimental conditions surveyed in this work, the value for < E>down ranges from 65 cm -1 at T = 450 K to 121 cm -1 at T = 650 K. The optimised parameters are given in Table 4. Although the increase in the barrier height is greater than the estimated uncertainty of ~2 kJ mol -1 in calculations of this nature, 7 the optimised barrier in MESMER is also subject to uncertainties of several kJ mol -1 , and the calculations may be influenced by multireference effects which could result in additional uncertainty. Optimisation of < E>down in N2 was also performed using the data obtained from measurements of OH (shown in Table 3) with the barrier to decomposition constrained to the value of 81.8 kJ mol -1 as indicated by the UV experiments. Figure  7 shows the results of the optimisation, which gave < E>down = (125.4 ± 32.2) (T/298 K) (0.5 ± 0.4) cm -1 and k1 = 0.01 s -1 at 298 K and 760 Torr in N2. Uncertainties in the value of k1 in N2 at 298 K and 760 Torr determined from the OH experiments, determined by propagation of errors in the MESMER fits, are on the order of ~200 %. However, as shown in Figure 7 the optimisation tends to overpredict the observed rate coefficients, and the results may be subject to larger uncertainties than indicated by the statistical error propagation owing to the complexity of the mechanism controlling the production and loss of OH in the system.
The results of Berndt et al. 23 required a decrease in the calculated barrier height from 78.9 kJ mol -1 to 76.8 kJ mol -1 to improve the agreement between the master equation calculation and the measured rate coefficient for decomposition of (0.19 ± 0.07) s -1 at 298 K and 760 Torr in N2 using the free-jet flow reactor. The higher barrier height determined in this work from the UV observations of CH2OO in He lead to a lower value of k1 = 1.1 × 10 -3 s -1 at 298 K and 760 Torr compared to the work of Berndt et al. 23 The difference in the barrier heights is significant, but it is worth noting that the barrier height determined by Berndt et al. was fitted to a single measurement of k1, while that determined in this work fitted over a range of temperatures and pressures, providing greater constraint in the fit to Equation 6. The experiments reported in this work also use direct detection of CH2OO, while the experiments of Berndt et al. rely on titration of CH2OO to H2SO4, with subsequent ionisation and detection of H2SO4.
Simulations in MESMER using the full PES by Nguyen et al., 11 shown in Figure 8, with the optimised values for < E>down and TS2 energy, determined from the UV experiments, were performed at p = 1 -3040 Torr and T = 400 -1200 K to investigate the product distribution. There was little variation in the product distribution over the pressure and temperature ranges investigated, with yields of 63.7 % for H2 + CO2 and 36.0 % for H2O + CO, on average. The yields of OH + HCO is predicted to be 0.3 %, on average, and is lower than the estimates based on the OH measurements reported in this work and those indicated by the use of OH as a proxy to CH2OO in experiments by Liu et al. 16 and Li et al. 17 The optimised TS2 energy and < E>down were also used in MESMER simulations to calculate k1 at temperatures between 200 K and 850 K and pressures up to 10 atm. The calculated rate coefficients were subsequently parameterised using the Troe expression for broad falloff curves 53 (Equations 7-9) for use in kinetic models: Analogous results for CD2OO give < E>down = (39.6 ± 7.8) (T/298 K) (1.3 ± 0.2) cm -1 and a barrier to decomposition of 80.1 kJ mol -1 , a decrease of 0.9 kJ mol -1 from the calculated barrier of 81.0 kJ mol -1 , with < E>down thus ranging from 67 cm -1 at 450 K to 109 cm -1 at 650 K. The fits give a value of k2 = × 10 -3 s -1 in He at T = 298 K and p = 760 Torr. The comparison between the experimentally observed rate coefficients and the MESMER output is given in Figure 4 and Table 2, with the optimised parameters summarised in Table 4. Fits to Equations 7-9, using the optimised parameters for CD2OO in MESMER to calculate k2 at temperatures between 200 K and 850 K and pressures up to 10 atm, give k2,0 = 5.2 × 10 -5 (T/298) -5.28 exp(-11610/T) cm 3 s -1 , k2,∞ = 1.2 × 10 13 (T/298) 0.06 exp(-9800/T) s -1 and Fc = 0.427.
The optimised PES thus indicates that there is no significant change in the barrier height to decomposition upon deuteration of CH2OO. However, comparison of Figures 2 and 4 shows that CD2OO decomposes faster than CH2OO under equivalent conditions, which is also confirmed by the MESMER calculations. Given the similar electronic barriers to decomposition for CH2OO and CD2OO, such differences likely result from an increased density of states in CD2OO near the transition state, which promotes the high pressure limit at lower pressures. A similar effect has been observed in a recent study of deuterated Criegee intermediate kinetics, in the reactions of (CH3)3COO and (CD3)3COO with SO2, 54 and was attributed to the potential impact of increased collisional stabilisation of the deuterated association complex between (CD3)3COO and SO2 compared to (CH3)3COO and SO2 owing to the increased density of vibrational states in the deuterated system.

Conclusions
The decomposition kinetics of CH2OO Criegee intermediate have been investigated at temperatures between 450 K and 650 K and pressures in the range 2 -350 Torr of He using flash photolysis of CH2I2 in O2 and a combination of time-resolved cavity enhanced broadband UV absorption spectroscopy, for direct monitoring of CH2OO, and laser-induced fluorescence, for monitoring of OH decomposition products. Kinetics of CD2OO decomposition were also investigated using flash photolysis of CD2I2 with time-resolved cavity enhanced broadband UV absorption spectroscopy.
The decomposition of CH2OO is expected to be slow under ambient conditions, and thus is not a significant sink for CH2OO in the atmosphere or in chamber experiments of ozonolysis reactions, despite reports in previous work. Master equation fits using MESMER give k1 = (1.1 ± 1.5) × 10 -3 s -1 at 298 K and 760 Torr in He, using an exponential down model to describe the collisional energy transfer, where < E>down = (32.6 ± 13.7) (T/298 K) (1.7 ± 0.4) cm -1 , and requiring an increase in the calculated barrier height to decomposition from 78.2 kJ mol -1 to 81.8 kJ mol -1 . Product yields, determined from MESMER simulations using the increased barrier height to decomposition, are predicted to be 63.7 % for H2 + CO2, 36.0 % for H2O + CO and 0.3 % for OH + HCO. For CD2OO, the master equation fits give k2 = (5.5 ± 9.2) × 10 -3 s -1 at 298 K and 760 Torr in He, and give values of < E>down = (39.6 ± 7.8) (T/298 K) (1.3 ± 0.2) cm -1 and a barrier height of 80.1 kJ mol -1 compared to the calculated value of 81.0 kJ mol -1 . We observed no kinetic isotope effect between the decomposition kinetics of CH2OO and CD2OO.
Results from this work provide a detailed description of CH2OO decomposition kinetics that can be applied to the analysis of the decomposition and stabilisation of nascent CH2OO Criegee intermediates produced in ozonolysis reactions, and to assess the contributions of wall losses to larger SCI species produced by ozonolysis in chamber experiments, which decompose more rapidly under ambient conditions owing to the existence of alternative decomposition pathways. Measurements reporting combined kinetics of CH2OO decomposition and wall loss are likely dominated by wall losses, [19][20][21][22] and can therefore provide an estimate for SCI wall loss rates that could be applied to other SCI species, enabling separation of the wall loss rate and decomposition rate in chamber experiments.
The low yield of OH radicals observed indicates that decomposition of CH2OO cannot be responsible for any potential OH interferences in field instruments measuring ambient OH concentrations using the LIF-FAGE (laser-induced fluorescence with fluorescence assay by gas expansion) technique, as has been postulated in the literature. 18 Under combustion conditions, decomposition of CH2OO will be rapid, with a fraction of decomposition leading to production of OH and HCO radicals, and thus contributing to chain-branching processes. The role of CH2OO in combustion, however, has yet to be fully established.