Kinetics of the simplest Criegee intermediate reaction with ozone studied using a mid-infrared quantum cascade laser spectrometer

Abstract

The kinetics of the reaction of CH₂OO with ozone has been studied by monitoring CH₂OO using time-resolved infrared (IR) absorption spectroscopy, which utilized the fast chirped IR pulse train from a quantum cascade laser [J. Chem. Phys., 2017, 146, 244302]. CH₂OO was prepared by photolyzing a gas mixture of CH₂I₂/O₂/O₃ at 352 nm; the photolysis wavelength was chosen to minimize the photodissociation of O₃. The measured rate coefficient at 298 K and 30 Torr is (6.7 ± 0.5) × 10⁻¹⁴ cm³ s⁻¹, independent of pressure from 30 to 100 Torr. The result indicates that previous ab initio calculations either underestimated or overestimated this reaction rate by one order of magnitude or more. The result also implies that, the reaction of the Criegee intermediate with ozone may play a role in laboratory studies of ozonolysis of alkenes. However, this reaction would not compete with other CH₂OO sinks in the atmosphere.

---

Introduction

In the atmosphere, ozonolysis of alkenes produces highly reactive Criegee intermediates (CIs).^1–4 The produced CIs may have excess internal energy and undergo unimolecular processes, such as isomerization or decomposition to form OH radicals; some of the CIs may be collisionally stabilized and then may react with other atmospheric species. In laboratory studies of ozonolysis, direct detection of CIs is difficult due to low steady-state concentrations. Direct spectroscopic and kinetic studies of CIs only became feasible after the work by Welz et al.,⁵ which demonstrated the efficient preparation of the simplest CI, CH₂OO, via the reaction CH₂I + O₂ → CH₂OO + I. Many recent studies on prototypical CIs have found that CIs play an important role in atmospheric chemistry, including the formation of OH radicals^6–9 and oxidation of atmospheric species, like SO₂, NO₂, organic and inorganic acids, alkenes, and water vapor.^5,10–31 The reactions of CIs may produce radicals or low-volatility organic species, which are key components in the formation of secondary organic aerosols.³²

The reactions of O₃ with CIs are potentially important in both laboratory and atmospheric studies. For example, Novelli et al.⁷ have considered that the main loss paths of CIs in their ozonolysis experiments are unimolecular decomposition and reactions with ozone at early reaction times, whereas reactions with organic peroxy radicals, alcohols, aldehydes and organic peroxides become more important at later reaction times. However, to the best of our knowledge, there is no direct measurement for determining the rate of ozone reaction with CIs, while there have been a few theoretical works.^33–36 For CH₂OO, Kjaergaard et al.³³ calculated its reaction rate with O₃ by using the CCSD(T)//B3LYP level of theory and predicted the formation of a cycloaddition intermediate through a significant barrier, yielding small rate coefficients ranging from 4.0 × 10⁻¹⁶ to 1.2 × 10⁻¹⁸ cm³ s⁻¹. Wei et al.³⁴ also investigated the reaction mechanism by using the CCSD(T)//B3LYP level of theory, but they did not report any rate coefficient. Vereecken et al.^35,36 calculated the rate of this reaction by using the CCSD(T)//M06-2X level of theory, and found that CH₂OO and O₃ will firstly form a prereactive complex without a barrier before passing through a submerged chain-addition transition state; they predicted a larger rate coefficient of 4 × 10⁻¹³ cm³ s⁻¹. It appears that it is not easy to accurately estimate the rate coefficient with the present quantum chemistry approaches. Nonetheless, all the theoretical studies predicted CH₂O + 2O₂ as the final products.^33–36

Recently, we have developed a high-resolution mid-infrared quantum cascade laser (QCL) spectrometer, and utilized it to study the spectrum of the ν₄ band of CH₂OO.³⁷ In this work, we used the QCL spectrometer to study the kinetics of the reaction of CH₂OO with O₃. This method has the following advantages: (1) avoids byproduct interferences and baseline drifting by probing narrow spectral lines, (2) high sensitivity due to narrow laser linewidth and long optical path length, (3) long CH₂OO lifetimes (up to 14 ms) due to low required concentration of CH₂OO, and (4) efficient data acquisition using the fast chirped pulse train of the QCL.

Experimental methods

Most of the experimental procedures have been described in our previous work.³⁷ Therefore, only a brief description is provided here. CH₂OO was prepared in a flow cell following the well-established method of CH₂I₂/O₂ photolysis:⁵ CH₂I₂ (1–7 mTorr) mixed with O₂ (30 or 100 Torr) was photolyzed using an unfocused excimer laser beam at 352 nm (laser fluence: (1.3–4.4) × 10¹⁶ photon cm⁻²). We chose this photolysis wavelength (instead of 248 or 308 nm) to minimize the effect of O₃ photolysis (a cross section of O₃ ≈ 1.0 × 10⁻²² cm² at 352 nm, and a photolysis probability of ≤4.4 × 10⁻⁶ at ≤4.4 × 10¹⁶ photon cm⁻²). The O₃ gas was synthesized using a commercial ozone generator, collected by adsorption on silica gel at dry-ice temperature, and further purified by condensation in a stainless steel cylinder at liquid-nitrogen temperature, following the procedures described in our previous work.³⁸ The O₃ concentration was measured via its UV absorption by using an absorption cell with a path length of 5 cm, a deuterium lamp (Hamamatsu, L10671D), and a spectrometer (Ocean Optics, USB2000+UV-VIS-ES), right before the O₃ gas entered the reactor cell. The flow rates of the used gases were controlled using mass flow controllers (Brooks, 5850E).

We used a distributed-feedback quantum cascade laser (Alpes Lasers, CW-DFB-QCL) as a coherent IR source to probe CH₂OO. The laser was driven by an intermittent CW driver (Alpes Lasers), powered by a DC-power supply (HAMEG Instruments, HMP2020). Its frequency coverage was 1279–1290 cm⁻¹ and the practical spectral linewidth was about 0.002 cm⁻¹ with a peak power over 10 mW. The temperature of the laser was controlled by a TEC cooling element inside the laser housing and a temperature controller. The laser was operated in the pulse-mode with a period of 18 μs or 120 μs and 40% duty cycle (a pulse duration of 7 μs or 48 μs). In each laser pulse, the laser frequency was down-chirped (1286.1–1285.5 cm⁻¹ in 48 μs). During the experiment, the IR chirped pulses were repeatedly scanned through the Q branch of the ν₄ fundamental band as shown in Fig. 1. The lifetime of CH₂OO in our experiments was on the millisecond time scale, much longer than the IR pulse period. Thus, for a single photolysis event, we could obtain a full time profile of CH₂OO from probing its transitions with the IR pulse train.

Fig. 1 Examples of transient IR absorption spectra of CH₂OO at different O₃ concentrations at 30 Torr total pressure. The peaks correspond to the Q branch of the ν₄ fundamental band of CH₂OO. The K number is assigned to each (partially) resolved sub-band. The photolysis-probe delay time is 2.4 ms in this example. See Table S1 (ESI†) (expt. 1) for details of the experimental conditions.

Our reactor cell was a glass tube (inner diameter 19 mm) with BaF₂ windows at both ends, which were purged with N₂. The photolysis excimer laser beam (352 nm, repetition rate 1 Hz) was combined with the IR probe beam and introduced into the flow cell by a high reflective BaF₂ mirror at 352 nm (Eksma Optics, custom item). A BaF₂ right-angle prism and a concave aluminium mirror (Edmund Optics, part # 43549) were placed before and after the flow cell, such that the probe IR beam was reflected back-and-forth between the two optics, creating a long overlap with the UV photolysis volume up to 3.9 meters (6 passes though the cell with an effective sample length of 65 cm). After leaving the multi-pass cell, the IR probe beam was guided to a HgCdTe (MCT) detector (Kolmar Technologies, KMPV11-1-J2), and the UV photolysis laser beam was reflected away by another BaF₂ high reflective mirror. Each photolysis pulse was synchronized with the rising edge of one of the IR probe pulses by using a delay generator (SRS, DG535). The calibration of the IR laser wavelength was carried out by measuring a reference gas spectrum (3 Torr N₂O) and an etalon signal (Ge etalon 3′′ in length, free spectral range = 0.0163 cm⁻¹). The IR pulse train signals from the DC outputs of all MCT detectors were acquired using an oscilloscope (LeCroy, HDO4034, 12-bit vertical resolution) at a sampling rate of 1.25 GS s⁻¹. For each [CH₂OO] time profile measurement, we averaged the data for 100 UV laser shots to improve the signal-to-noise ratio.

Finally, we also utilized a UV transient absorption spectrometer to roughly estimate the absolute values of [CH₂OO] under the experimental conditions of this work. A LED (Hamamatsu, LC-L2) with an emission profile centered at 365 nm was used as the probe light source. The probe light was projected into the flow cell by a convex lens, and then it was reflected once by the concave mirror to achieve two passes through the cell. After leaving the flow cell, the remaining probe light was guided to a balanced photodiode detector (Thorlab, PDB450A). The rest of the experimental configurations were the same as those described above. Since the system was not optimized for UV detection, the uncertainty in the absolute values of [CH₂OO] would be a bit larger than those of our previous works.¹²

Results and discussion

Fig. 1 shows the selected transient IR absorption spectra of CH₂OO at representative O₃ concentrations. Transient absorption means the change in the IR absorption intensity with respect to that before the UV photolysis. In this example, the results at a photolysis-probe delay time of 2.4 ms are shown. As mentioned in the Experimental methods, the spectral range corresponds to the Q branch of the ν₄ fundamental band of CH₂OO.³⁷ Note that each of the peaks between 1285.9 and 1285.7 cm⁻¹ corresponds to congested transitions from the same K levels (from K = 3 to 8), while the broad feature peaked at 1285.61 cm⁻¹ is the perturbed transitions of the higher K levels.³⁷ See ref. 37 for detailed spectroscopic assignments and perturbation analysis. Upon increasing the O₃ concentration, the CH₂OO signal decreases accordingly, suggesting that CH₂OO is consumed by its reaction with O₃. The decay trends of these peaks are similar, indicating no rotational dependence.

Fig. 2 shows a few time profiles of [CH₂OO] at various [O₃]. Again, the increasing decay rates at higher [O₃] indicate that CH₂OO reacts with O₃. The following reactions are involved in the formation and decay of CH₂OO in our CH₂I₂/O₂/O₃ photolysis system:¹²

CH₂I₂ + hν (352 nm) → CH₂I + I

CH₂I + O₂ → CH₂OO + I

CH₂OO → products k_first

CH₂OO + O₃ → products k_O3

CH₂OO + I → products k_I

CH₂OO + CH₂OO → products k_self

The pulse width of the photolysis laser was only 20 ns. Under our experimental conditions, very high O₂ concentrations (30 Torr or more) were used, leading to the fast conversion of CH₂I to CH₂OO (<1 μs). Thus, we neglect the formation time of CH₂OO in our kinetic analysis.

Fig. 2 Representative time profiles of [CH₂OO] at various [O₃] at 30 Torr and 298 K. Each data point (symbol) is the integrated intensity of the highest peak (1285.71–1285.74 cm⁻¹) of the Q branch of the CH₂OO ν₄ band (see Fig. 1) probed by each IR pulse. The lines are single-exponential fit to the data. See Table S1 (ESI†) (expt. 11) for details of the experimental conditions.

In our first attempt of data analysis, we described the measured time profile of CH₂OO as:

[CH₂OO](t) = [CH₂OO]₀ exp(−k_efft) (1)

where k_eff is the observed 1st-order rate coefficient of CH₂OO decay, k_O3 is the bimolecular rate coefficient for the reaction of CH₂OO with O₃ and k₀ is the decay rate coefficient of CH₂OO without O₃, which may include the contributions from the first-order loss k_first (mostly wall loss), self-reaction (k_self), and reactions with other radicals (mostly iodine atoms, represented by k_I).

k₀ ≅ k_first + k_I[I] + 2k_self[CH₂OO] (3)

The solid lines in Fig. 2 are the fits to the transient absorption time profiles with eqn (1), yielding k_eff and [CH₂OO]₀. As mentioned in our previous work,³⁷ we also measured the UV absorbance of CH₂OO at 365 nm (where the absolute UV cross section is known)³⁹ under similar experimental conditions, which allows us to roughly estimate the absolute value of [CH₂OO] in the IR experiments.

Fig. 3 plots the value of k_eff as a function of [O₃] at 30 Torr and 298 K, and the solid lines are linear fits to the data points except those at [O₃] = 0. At low [CH₂OO]₀ (≈2.4 × 10¹¹ cm⁻³, Fig. 3(a)), k_eff shows a linear relationship with [O₃], as expected from eqn (2). The slope (6.4 × 10⁻¹⁴ cm³ s⁻¹) should correspond to a measured value of k_O3. To our surprise, at higher [CH₂OO]₀ (for example, Fig. 3(b)), while k_eff is still a linear function of [O₃] for [O₃ ] > 0, this linear line does not go through the data points at [O₃] = 0. Nonetheless, the k_eff values as a function of [O₃] all exhibit a consistent slope for [O₃] > 0. The detailed values are listed in Table S1 (ESI†) in the column of k_O3. The averaged value (with one-sigma error bar) of the slope is (6.72 ± 0.46) × 10⁻¹⁴ cm³ s⁻¹.

Fig. 3 Effective first-order rate coefficient of CH₂OO decay, k_eff, as a function of [O₃] for [CH₂OO]₀ = (a) 2.4 × 10¹¹ cm⁻³ (expt. 11) and (b) 2.1 × 10¹² cm⁻³ (expt. 9). The error bars are 1σ uncertainties from the fitting of the [CH₂OO] time profiles. The red line is a linear fit to k_eff excluding the data at [O₃] = 0 (see text for details). The slope of the red line corresponds to the second-order rate coefficient of the CH₂OO + O₃ reaction.

To investigate the possible source that causes the difference between the intercept value (k_incpt) of k_eff (obtained from the linear fit of k_eff as a function of [O₃] for [O₃] > 0) and the measured k₀ (k_eff at [O₃] = 0), we performed several checking experiments as described below. First, one may wonder whether the O atoms from O₃ photodissociation affect the measurements or not. To check this, we measured k_eff at different 352 nm laser fluences (by a factor of 2.8, which varied [O] by the same factor) and obtained very similar results (see Fig. S1, ESI†), indicating that no O₃ photochemistry was involved. As mentioned in the Experimental methods, the probability of O₃ photolysis is less than 4.4 × 10⁻⁶, thus the upper limit of [O] can be estimated to be (4.4 × 10⁻⁶)(2 × 10¹⁶ cm⁻³) = 9 × 10¹⁰ cm⁻³. The rate of the reaction of CH₂OO with O atom is not known yet. If we assume that every collision between CH₂OO and O atom leads to a reaction (the collision-limit), the rate coefficient would be about 3 × 10⁻¹⁰ cm³ s⁻¹,⁴⁰ and the effect of the O atoms on the CH₂OO decay is still small (<27 s⁻¹ in k_eff).

Because we synthesized and purified the O₃ gas by ourselves, one may worry about its purity. The typical lifetime of our O₃ gas (stored in a stainless steel cylinder at dry-ice temperature) is more than 20 hours. To check the effect of impurities, we deliberately warmed up the O₃ gas to room temperature and waited until all the O₃ molecules had decomposed (∼70 hours, verified using UV absorption). We called this gas “decomposed O₃” which would contain a similar or higher level of impurity, compared to that of our fresh O₃ gas. The results are shown in Fig. S3 (ESI†). The “decomposed O₃” gas does not change the k_eff at all, indicating that the impurity in our O₃ gas has a negligible effect.

As shown in Fig. 3, the difference between k_incpt and k₀ is more significant at higher [CH₂OO]₀. We found that this difference, k_incpt − k₀, is proportional to [CH₂OO]₀ as shown in Fig. S4 (ESI†). This observation gave us an idea that the chemistry may involve I atoms, of which the concentration is proportional to [CH₂OO]₀ in our preparation method. Thus, we propose that the following reactions should also be involved when O₃ is present:

O₃ + I → IO + O₂ k_I+O3

CH₂OO + IO → products (possibly CH₂O + OIO) k_IO

The reaction of I atoms with O₃ has been well studied; the literature value of k_I+O3 is (1.28 ± 0.06) × 10⁻¹² cm³ s⁻¹.⁴¹ The lowest [O₃] in our experiment is at least 1.6 × 10¹⁵ cm⁻³. Thus, I atoms would be quickly converted into IO due to the fast reaction rate (k_I+O3[O₃] > 2000 s⁻¹), which is much faster than the reaction of CH₂OO with O₃ in our experiments. Since [O₃] ≫ [I], the amount of IO is mainly controlled by the initial amount of I atoms, which is proportional to [CH₂OO]₀. Therefore, we may modify eqn (3) to (4) for a better approximation of k_incpt.

k_incpt ≅ k_first + k_I[I] + 2k_self[CH₂OO] + k_IO[IO] (4)

And mass balance gives

[I]₀ = [I] + [IO] ≅ 2[CH₂OO]₀ (5)

If k_IO is larger than k_I, we would have a larger k_incpt when O₃ is present.

We have also determined k_incpt and k₀ at various [CH₂OO]₀ (see Table S1, ESI†). Note that when O₃ is present, the maximum concentration of IO is also controlled by [CH₂OO]₀, [IO]_max ≅ [I]₀ ≅ 2[CH₂OO]₀. Therefore the difference between k_incpt and k₀ would become:

k_incpt − k₀ ≅ (k_IO − k_I) [I]₀ ≅ 2(k_IO − k_I) [CH₂OO]₀. (6)

As a result, the slope (∼1.31 × 10⁻¹⁰ cm³ s⁻¹) in Fig. S4 (ESI†) provides an estimate for 2(k_IO − k_I). Assuming k_I = 9.15 × 10⁻¹¹ cm³ s⁻¹,¹² the value of k_IO is estimated to be on the order of 10⁻¹⁰ cm³ s⁻¹. Examples of kinetic simulation of the time profiles of the involved species can be found in the ESI† (Fig. S5–S8). In the simulation, a value of 1.5 × 10⁻¹⁰ cm³ s⁻¹ of k_IO (k₁₀ in ESI,† which uses different notations for the rate coefficients) gives satisfactory fits to the experimental data.

In brief summary, we have tested a few crucial experimental conditions. The results enable us to exclude the possibility of interferences which originate from the photolysis of O₃ and the impurity in our O₃ gas. In our system, I atoms would be quickly converted into IO when O₃ is present; the side-reaction of CH₂OO + I would be shifted to the reaction of CH₂OO + IO and thus, changes the value of k_incpt. For [O₃] > 1.6 × 10¹⁵ cm⁻³, all the observed data of k_eff are linear with [O₃] with a slope of (6.72 ± 0.46) × 10⁻¹⁴ cm³ s⁻¹, which corresponds to the bimolecular rate coefficient of the reaction of CH₂OO with O₃. Finally, we found that the kinetics at a total pressure of 100 Torr is very similar to that at 30 Torr (Fig. S2, ESI†), indicating a weak pressure dependence in this pressure range.

As described in the Introduction, the mechanism of this reaction and the rate coefficient have been predicted by Kjaergaard et al.³³ and Vereecken et al.^35,36 Our measured rate coefficient is more close to that of Vereecken et al., 4 × 10⁻¹³ cm³ s⁻¹, which has an uncertainty of at least 1 order of magnitude.^35,36 Although O₃ is isoelectronic with CH₂OO, the self-reaction rate of CH₂OO (∼8 × 10⁻¹¹ cm³ s⁻¹)¹² is at least 10 orders of magnitude larger than that of O₃. The reaction rate of CH₂OO and O₃ is somehow in between these two limiting cases. While the zwitterionic character of CH₂OO leads to a pure attractive cycloaddition in its self-reaction,^13,35 the interaction between CH₂OO and O₃ would lead to the formation of a pre-reactive complex instead, which has a barrier for either subsequent chain addition or cycloaddition, according to the predictions by Vereecken et al.^35,36 A further clarification of the reaction mechanism via either theory or experiment is still required.

In our future work we would like to clarify the reaction mechanism via examining the reaction products. Theoretical works have predicted that CH₂O and O₂ are the final products of this reaction. We plan to utilize our QCL spectrometer to probe CH₂O produced from the reactions, while a new QCL with a spectral range dedicated to detecting CH₂O will be required. Vereecken et al.^35,36 further predicted that the oxygen atom in the product CH₂O is purely from O₃, while the two oxygen atoms from CH₂OO are released as O₂. However, the cycloaddition mechanism predicted by Kjaergaard et al.³³ would lead to a 1 : 1 ratio of oxygen atoms of CH₂O originating from CH₂OO or O₃.³⁶ To verify these predictions, our future works will also include isotopic labelling experiments and the IR identification of isotope-substituted CH₂O.

Conclusions

In summary, we investigated the kinetics of CH₂OO reaction with O₃ at 298 K and 30 Torr using a high resolution mid-infrared quantum cascade laser spectrometer. No pressure dependence was observed from 30 to 100 Torr. Different from other reactant molecules like H₂O, SO₂, NO₂, and organic/inorganic acids, ozone would react with I atoms to form IO, an unavoidable reaction in the current kinetic system. Based on our experimental observations, we suggest that CH₂OO should react quickly with IO. Fortunately, the amount of IO is controlled by the initial amount of I atoms under our experimental conditions, where [O₃] ≫ [I]. Thus, the kinetics of the CH₂OO decay is still pseudo-1st-order and the observed k_eff is linear to [O₃] for [O₃] > 0.

The measured rate coefficient of the reaction of CH₂OO with O₃ is (6.72 ± 0.46) × 10⁻¹⁴ cm³ s⁻¹. This value differs from the predicted values in the literature by one order of magnitude or more, suggesting the need of multireference treatment for the quantum chemistry calculations. While the theory has made good predictions or estimations for a number of reaction rates of Criegee intermediates with water, H₂S and some other molecules,²⁴ it still seems tricky to calculate the rate of the reaction of CH₂OO with O₃. We hope the measured results of the present work could benefit future theoretical calculations of this reaction.

This measurement also indicates that the reaction of CH₂OO with O₃ should play a role in laboratory studies of ozonolysis, where the early-time decay of CH₂OO may be controlled by its reactions with O₃ and with the used alkene molecules. The rate coefficients of the reactions of CH₂OO with simple alkenes have been reported by Buras et al.¹⁴ to be 2 × 10⁻¹⁵ to 11 × 10⁻¹⁵ cm³ s⁻¹, which are smaller than that for the O₃ reaction. As a result, if one wishes to produce a higher steady-state concentration of CH₂OO in ozonolysis experiments, an ozone concentration lower than that of alkene may be desirable to slow down the decay of CH₂OO due to its reaction with O₃. For an ozonolysis study with 1 ppm of O₃, the effective decay rate of CH₂OO by O₃ is about 2 s⁻¹, which is slightly larger than the thermal decomposition rate of CH₂OO (0.2 s⁻¹).⁴² On the other hand, the reaction of CH₂OO with atmospheric ozone is relatively slower (∼0.17 s⁻¹ for 100 ppbv O₃) than those with other CH₂OO sinks, such as the reaction with water dimers (>1000 s⁻¹),¹⁷ in the atmosphere.

Conflicts of interest

There are no conflicts of interest to declare.

Acknowledgements

This work was supported by the Ministry of Science and Technology, Taiwan (MOST103-2113-M-001-019-MY3) and Academia Sinica. We thank Mr Wen Chao for his help with the measurements. We also thank Dr Mica Smith and Dr Kaito Takahashi for discussions about the reaction of CH₂OO with IO.

Notes and references

1. R. Criegee, Angew. Chem., Int. Ed. Engl., 1975, 14, 745.
2. C. A. Taatjes, D. E. Shallcross and C. J. Percival, Phys. Chem. Chem. Phys., 2014, 16, 1704.
3. D. L. Osborn and C. A. Taatjes, Int. Rev. Phys. Chem., 2015, 34, 309.
4. Y.-P. Lee, J. Chem. Phys., 2015, 143, 020901.
5. O. Welz, J. D. Savee, D. L. Osborn, S. S. Vasu, C. J. Percival, D. E. Shallcross and C. A. Taatjes, Science, 2012, 335, 204.
6. N. M. Donahue, G. T. Drozd, S. A. Epstein, A. A. Presto and J. H. Kroll, Phys. Chem. Chem. Phys., 2011, 13, 10848.
7. A. Novelli, L. Vereecken, J. Lelieveld and H. Harder, Phys. Chem. Chem. Phys., 2014, 16, 19941.
8. F. Liu, J. M. Beames, A. S. Petit, A. B. McCoy and M. I. Lester, Science, 2014, 345, 1596.
9. N. M. Kidwell, H. Li, X. Wang, J. M. Bowman and M. I. Lester, Nat. Chem., 2016, 8, 509.
10. C. A. Taatjes, O. Welz, A. J. Eskola, J. D. Savee, A. M. Scheer, D. E. Shallcross, B. Rotavera, E. P. F. Lee, J. M. Dyke, D. K. W. Mok, D. L. Osborn and C. J. Percival, Science, 2013, 340, 177.
11. L. Sheps, A. M. Scully and K. Au, Phys. Chem. Chem. Phys., 2014, 16, 26701.
12. W.-L. Ting, C.-H. Chang, Y.-F. Lee, H. Matsui, Y.-P. Lee and J. J.-M. Lin, J. Chem. Phys., 2014, 141, 104308.
13. Y.-T. Su, H.-Y. Lin, R. Putikam, H. Matsui, M. C. Lin and Y.-P. Lee, Nat. Chem., 2014, 6, 477.
14. Z. J. Buras, R. M. I. Elsamra, A. Jalan, J. E. Middaugh and W. H. Green, J. Phys. Chem. A, 2014, 118, 1997.
15. D. Stone, M. Blitz, L. Daubney, N. U. M. Howes and P. Seakins, Phys. Chem. Chem. Phys., 2014, 16, 1139.
16. O. Welz, A. J. Eskola, L. Sheps, B. Rotavera, J. D. Savee, A. M. Scheer, D. L. Osborn, D. Lowe, A. Murray Booth, P. Xiao, M. Anwar, H. Khan, C. J. Percival, D. E. Shallcross and C. A. Taatjes, Angew. Chem., Int. Ed., 2014, 53, 4547.
17. W. Chao, J.-T. Hsieh, C.-H. Chang and J. J.-M. Lin, Science, 2015, 347, 751.
18. H.-L. Huang, W. Chao and J. J.-M. Lin, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 10857.
19. M. C. Smith, C.-H. Chang, W. Chao, L.-C. Lin, K. Takahashi, K. A. Boering and J. J.-M. Lin, J. Phys. Chem. Lett., 2015, 6, 2708.
20. R. Chhantyal-Pun, A. Davey, D. E. Shallcross, C. J. Percival and A. J. Orr-Ewing, Phys. Chem. Chem. Phys., 2015, 17, 3617.
21. E. S. Foreman, K. M. Kapnas and C. Murray, Angew. Chem., 2016, 128, 10575.
22. L.-C. Lin, H.-T. Chang, C.-H. Chang, W. Chao, M. C. Smith, C.-H. Chang, J. Jr-Min Lin and K. Takahashi, Phys. Chem. Chem. Phys., 2016, 18, 4557.
23. M. C. Smith, W. Chao, M. Kumar, J. S. Francisco, K. Takahashi and J. J.-M. Lin, J. Phys. Chem. A, 2017, 121, 938.
24. J. Jr-Min Lin and W. Chao, Chem. Soc. Rev., 2017 10.1039/c7cs00336f.
25. L. Sheps, B. Rotavera, A. J. Eskola, D. L. Osborn, C. A. Taatjes, K. Au, D. E. Shallcross, M. A. H. Khan and C. J. Percival, Phys. Chem. Chem. Phys., 2017, 19, 21970.
26. Z. C. J. Decker, K. Au, L. Vereecken and L. Sheps, Phys. Chem. Chem. Phys., 2017, 19, 8541.
27. Y. Liu, F. Liu, S. Liu, D. Dai, W. Dong and X. Yang, Phys. Chem. Chem. Phys., 2017, 19, 20786.
28. C. A. Taatjes, Annu. Rev. Phys. Chem., 2017, 68, 183.
29. R. L. Caravan, M. A. H. Khan, B. Rotavera, E. Papajak, I. O. Antonov, M.-W. Chen, K. Au, W. Chao, D. L. Osborn, J. J.-M. Lin, C. J. Percival, D. E. Shallcross and C. A. Taatjes, Faraday Discuss., 2017, 200, 313.
30. R. Chhantyal-Pun, M. R. McGillen, J. M. Beames, M. A. H. Khan, C. J. Percival, D. E. Shallcross and A. J. Orr-Ewing, Angew. Chem., Int. Ed., 2017, 56, 9044.
31. R. Chhantyal-Pun, O. Welz, J. D. Savee, A. J. Eskola, E. P. F. Lee, L. Blacker, H. R. Hill, M. Ashcroft, M. A. H. Khan, G. C. Lloyd-Jones, L. Evans, B. Rotavera, H. Huang, D. L. Osborn, D. K. W. Mok, J. M. Dyke, D. E. Shallcross, C. J. Percival, A. J. Orr-Ewing and C. A. Taatjes, J. Phys. Chem. A, 2017, 121, 4.
32. J. H. Kroll and J. H. Seinfeld, Atmos. Environ., 2008, 42, 3593.
33. H. G. Kjaergaard, T. Kurtén, L. B. Nielsen, S. Jørgensen and P. O. Wennberg, J. Phys. Chem. Lett., 2013, 4, 2525.
34. W. Wei, R. Zheng, Y. Pan, Y. Wu, F. Yang and S. Hong, J. Phys. Chem. A, 2014, 118, 1644.
35. L. Vereecken, H. Harder and A. Novelli, Phys. Chem. Chem. Phys., 2014, 16, 4039.
36. L. Vereecken, A. R. Rickard, M. J. Newland and W. J. Bloss, Phys. Chem. Chem. Phys., 2015, 17, 23847.
37. Y.-P. Chang, A. J. Merer, H.-H. Chang, L.-J. Jhang, W. Chao and J. J.-M. Lin, J. Chem. Phys., 2017, 146, 244302.
38. B. Jin, M.-N. Su and J. J.-M. Lin, J. Phys. Chem. A, 2012, 116, 12082.
39. W.-L. Ting, Y.-H. Chen, W. Chao, M. C. Smith and J. J.-M. Lin, Phys. Chem. Chem. Phys., 2014, 16, 10438.
40. J. H. Kroll, J. S. Clarke, N. M. Donahue and J. G. Anderson, J. Phys. Chem. A, 2001, 105, 1554.
41. M. E. Tucceri, T. J. Dillon and J. N. Crowley, Phys. Chem. Chem. Phys., 2005, 7, 1657.
42. T. Berndt, R. Kaethner, J. Voigtländer, F. Stratmann, M. Pfeifle, P. Reichle, M. Sipilä, M. Kulmala and M. Olzmann, Phys. Chem. Chem. Phys., 2015, 17, 19862–19873.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7cp06653h

---