Direct evidence for a substantive reaction between the Criegee intermediate, CH₂OO, and the water vapour dimer

Abstract

The C1 Criegee intermediate, CH₂OO, reaction with water vapour has been studied. The removal rate constant shows a quadratic dependence on [H₂O], implying reaction with the water dimer, (H₂O)₂. The rate constant, k_CH2OO+(H2O)2 = (4.0 ± 1.2) × 10⁻¹² cm³ molecule⁻¹ s⁻¹, is such that this is the major atmospheric sink for CH₂OO.

---

The Criegee intermediate is the long postulated intermediate formed in the ozonolysis of alkenes.¹ Even though much effort had gone into its direct observation from ozonolysis reactions, it has only recently been directly observed at low pressures via production in the reaction:²

CH₂I + O₂ → CH₂OO + I (R1a)

This reaction is only just exothermic and is the near exclusive channel at low pressure.³ As the total pressure is increased the mechanism for this reaction switches to an association process:

CH₂I + O₂(+M) → CH₂IO₂ (R1b)

where at atmospheric pressure the CH₂OO yield is 0.18.³ Therefore this reaction and analogues using larger organic di-iodides are convenient sources of Criegee intermediates over a wide range of pressures, and in the last few years there has been a plethora of studies^4–7 that have used this type of reaction to determine direct properties of the Criegee intermediate, including the C2 species CH₃CHOO.⁸

Many of these new direct kinetic measurements on Criegee intermediates have determined rate constants significantly higher than older, indirect studies and their importance in atmospheric chemistry has been re-evaluated, in particular its reaction with SO₂ in competition with unimolecular decomposition⁹ and photolysis.^4,10 The importance of these latter processes remains uncertain. An intriguing result of this new work on the Criegee intermediate is its reaction with H₂O:¹¹

CH₂OO + H₂O → Products (HO–CH₂OOH) (R2)

Only upper limits have been placed on this reaction rate constant (cm³ molecule⁻¹ s⁻¹): k₂ < 4 × 10⁻¹⁵ from Welz et al.²via direct detection; <9 × 10⁻¹⁷ from Stone et al.¹²via direct detection of the CH₂O product, and <2 × 10⁻¹⁷ from Ouyang et al.¹³via detection of NO₃. In contrast to these results, end product analysis studies of ethylene ozonolysis, which exclusively generates only the C1 Criegee intermediate, have observed reaction with added water, with implied rate constants of 9 × 10⁻¹⁵ from Suto et al.¹⁴ and 3 × 10⁻¹⁴ cm³ molecule⁻¹ s⁻¹ from Becker et al.¹⁵ Most recently Berndt et al.¹⁶ used ethylene ozonolysis to show that CH₂OO removal has a quadratic dependence on water vapour, which implies that it is water dimer that is reacting. Ozonolysis generates the Criegee intermediate via a highly exothermic reaction while reaction (R1) is only just exothermic, so there is the possibility that the lack of reactivity of CH₂OO with H₂O in direct time-resolved experiments might be a consequence of its method of preparation: “hot” CH₂OO from ozonolysis might intercept H₂O but viaR1 H₂O only encounters “cold” CH₂OO. This effect is known as non-thermal kinetics and it has recently been demonstrated for the reaction of between OH and C₂H₂ in the presence of O₂.¹⁷

In this communication we demonstrate that by generating the Criegee intermediate using reaction (R1) and directly following it in time via UV/Vis spectroscopy, reaction is observed with H₂O vapour that is described by a quadratic dependence on [H₂O]. This observation confirms reaction of the Criegee intermediate with the water dimer and that there is no significant difference in Criegee intermediate chemistry whether the intermediate is generated viareaction (R1) or by ozonolysis.

The experiments were carried out using our newly constructed multiplexing absorption kinetics spectrometer coupled to laser flash photolysis. Full details about the setup will be described in a forthcoming publication. The essential details are as follows: the output from a xenon lamp was multi-passed 14 times through the 1.5 metre reaction cell and configured such that this probe beam was overlapped for the majority of this distance with the 248 nm excimer laser that passed along the length of the reactor. This probe beam was then directed via a fibre optic into a spectrograph (Jobin Yvon CP140-103) where the wavelengths 250–850 nm were simultaneously measured using a CCD image sensor (Hamamatsu S7031, back-thinned FFT-CCD). All the wavelengths were recorded for 1 millisecond intervals for a total of 200 milliseconds and transferred to a PC via a PCI interface board. All these data were processed by the PC using a custom built LabView program before the next photolysis laser pulse; the excimer laser was fired between 1–0.2 Hz. At each wavelength (λ), the 50 points before the excimer laser pulse were averaged and assigned to I₀(λ) (intensity of the probe light), and all these I₀(λ) were compared to all the wavelength time points after the excimer laser fired, I(λ). The program calculated ΔI/I₀ for each wavelength versus time, the time-resolved differential absorption signal for each wavelength.

An example of a spectrum at early time after photolysis is shown in Fig. 1, where it can be seen that the spectrum between 300–400 nm is dominated by the C1 Criegee intermediate. At longer wavelengths absorption by the IO radical is also observable (CH₂I₂ photolysis produces a small amount of CH₂, which reacts with O₂ to produce O(³P)¹⁸ which in turn reacts with the precursor to produce IO¹⁹). The IO is removed from the system much more slowly than CH₂OO. H₂O vapour was added to the system by passing the main gas, N₂ (BOC, OFN), through a bubbler filled with deionized water, where the pressure in the bubbler was measured and could be varied over range 1000–2000 Torr. [O₂] (∼2 × 10¹⁷ molecule cm⁻³) was high enough to ensure R1 was rapid and the total pressure was varied between 50–400 Torr, where N₂ was the main buffer gas. At each pressure the kinetics of the system were recorded without H₂O vapour and then the N₂ flow was switched to the H₂O bubbler, where the pressure can be adjusted. These experiments were carried out at 294 K.

Fig. 1 ΔI/I₀ spectrum of the system at early-times over the wavelength range 300–500 nm. CH₂I₂ was photolysed at 248 nm (energy ∼ 50 mJ per pulse cm⁻²) in the presence of O₂: total pressure (N₂), O₂, CH₂I₂ and H₂O equal 1.52 × 10¹⁸, 1.77 × 10¹⁷, ∼3 × 10¹³ and 2.2 × 10¹⁶ molecule cm⁻³, respectively. The spectrum was recorded 1 milli-second after the photolysis laser. The sharp peaks in the spectrum above 400 nm are due to IO, while the spectrum between 300–460 nm is the C1 Criegee intermediate, CH₂OO. In red and blue the literature spectra of IO and CH₂I₂ (inverted to aid clarity) have been added to guide the eye.

The features of the spectrum in Fig. 1, especially between 350–420 nm, are consistent with the absorption literature spectrum of CH₂OO.^20,21 However, the present experiment records the differential absorption spectrum, ΔI/I₀, and it needs to be corrected for CH₂I₂ photolysis and IO before it can quantitatively be used to compare to the literature. Also, the spectrum is Fig. 1 has been corrected for scattered photons (>850 nm) hitting the CCD camera, see ESI† for further details. Hence this work cannot be compared with absolute cross-sections from any previous study at present. It is at 350 nm where the cross-section value reported by Ting et al.²² is ca. a factor of three times lower than the values reported by Beames et al.⁴ and Sheps.²¹ In our previous study, using a completely different absorption setup on CH₂I₂ photolysis in the presence of O₂ at atmospheric pressure, we mis-assigned the Criegee intermediate spectrum as the CH₂IOO from reaction (R1b).²³ If we now re-assign this spectrum as CH₂OO and divide the cross-sections by 0.18, which our recent measurements have determined as the yield of Criegee intermediate at atmospheric pressure,³ the spectrum is 40% lower at 350 nm than the cross-section value of Ting et al.²²

If the reaction of CH₂OO with water is slow, then self-reaction⁶ should dominate CH₂OO decay. The CH₂OO kinetic traces were analysed using an expression for second-order loss and it was observed that they were always better described by first-order kinetics, even for the traces at the lowest total pressure, see Fig. 2 for example. At this stage it is not clear what is causing the unexpected first order kinetics; a possible explanation is unimolecular decay:^7,9

CH₂OO → products (R3)

It should be emphasised that experiments were always carried out in the absence of water vapour and then in the presence of water vapour, and therefore the difference between the pseudo-first-order decays can be attributed to the presence of water. The reaction with H₂O vapour is slow but it will be pseudo-first-order, and reaction of the Criegee intermediate with water vapour is only significant at the higher total pressures, where more water vapour can be added. Therefore it is reasonable to describe the Criegee intermediate loss as a first-order process:

[CH₂OO] = [CH₂OO]₀ exp(−k_obst) (E1)

where k_obs = k₂′ + k₃, where k₂′ = k₂[H₂O] and k₃ is the first-order rate constant for CH₂OO removal other than H₂O and was typically ∼200 s⁻¹. It is noted that in the study by Sheps²¹ using similar concentrations (∼5 × 10¹¹ molecule cm⁻³) the loss of CH₂OO was also observed to be reasonably described by single exponential behaviour and k_obs was comparable (180 s⁻¹) to this study. Second-order CH₂OO loss rate constants have been measured in studies^20,24 that have used much higher concentrations than used in the present study.

Fig. 2 ΔI/I₀versus time traces for wavelengths that correspond predominantly to the Criegee intermediate (black – 353 nm, red – 350 nm and blue – 344 nm) and IO (green 436 nm). Traces at 341 and 346 nm have been omitted for clarity. CH₂I₂ was photolysed at 248 nm (energy ∼ 50 mJ per pulse cm⁻²) in the presence of O₂: total pressure (N₂), O₂, CH₂I₂ and H₂O equal 1.52 × 10¹⁸, 1.77 × 10¹⁷, ∼3 × 10¹³ and 2.2 × 10¹⁶ molecule cm⁻³, respectively. The Criegee intermediate removal under all conditions is much faster than IO removal. The above data returns k_obs = 221 ± 17 s⁻¹.

The early-time spectrum in Fig. 1 shows that both CH₂OO and IO are present and from Fig. 2 it can be seen that the CH₂OO is removed much faster than IO under all conditions, especially at high water vapour concentrations. The data were analysed at five different wavelengths, 353, 350, 346, 344 and 341 nm, using the equation:

[CH₂OO] = [CH₂OO]₀ exp(−k_obst) + [B]₀ exp(−k_bt) + C (E2)

where [B]₀ exp(−k_bt) takes into account the small but significant contribution to the absorption from IO, k_b is the rate constant for IO loss, and C takes into account CH₂I₂ photolysis, which is significant up to 400 nm. In this analysis all the data at the five wavelengths were fitted simultaneously using eqn (E2), where k_obs was treated as a global parameter and all the other parameters were local. At each total pressure the k_obs was determined in the presence, k₂′ + k₃, and absence of H₂O, k₃. Therefore subtracting k_obs with and without H₂O gives k₂′. As can be seen in Fig. 2, the fits to the data were good and k_obs was defined with errors always less than 10%. The validity of using eqn (E2) is that k_obs and not k_b is significantly changing as [H₂O] is added to the system, and therefore k_obsvs. [H₂O] is a good measure of reaction (R2).

In Fig. 3k₂′ is plotted versus the H₂O vapour concentration, and from this figure it is clear that at the highest concentrations the Criegee intermediate is reacting with water. However, closer inspection of this plot indicates that its dependence on H₂O concentration is better described by a quadratic rather than a linear dependence. The data are better described by a quadratic function based on the value of χ². Also shown in Fig. 3 are linear least squares fits to the data over the full range and [H₂O] < 7.5 × 10¹⁶ molecule cm⁻³, 13 points, where it can be observed that there is a factor is two increase in the slope. These observations, together with visual inspection, highlight the curvature in the data. This observation is in agreement with the recent paper by Berndt et al.¹⁶ where, from ozonolysis of ethylene, the removal of the Criegee intermediate (versus reaction with SO₂) was shown to have a quadratic dependence on [H₂O]. In Fig. 4, k₂′ is plotted versus [(H₂O)₂] and it can be seen that the data are now better described by a linear relationship; good evidence that the Criegee intermediate is reacting predominantly with the dimer. The [(H₂O)₂] was calculated using the parameterisation of Scribano et al.,²⁵ which is the same calculation as used by Berndt et al.¹⁶ Therefore the results from this study can be directly compared to Berndt et al. even though there is an estimated 20% error in the water dimer concentration.

Fig. 3 Bimolecular plot of the removal rate constant, k₂′, in the presence of H₂O vapour, which exhibits distinct upward curvature. The solid line is the least squares fit of a quadratic function to the data, and yields a χ² = 37.5. The dashed lines are linear least squares fits to the data over the full range and [H₂O] < 7.5 × 10¹⁶ molecule cm⁻³, 13 points. The slopes and χ² are 1.5 and 0.8 × 10⁻¹⁵ cm⁻³ molecule⁻¹ s⁻¹ and χ² = 45.3 and 9.2, respectively. The 20% improvement in the fit of the quadratic over the linear function, together with the increases in slope and χ² over the two H₂O ranges, highlights the curvature in the data. The [H₂O] was varied as the total pressure was varied between 50–400 Torr.

Fig. 4 Plot of the removal rate constant, k₂′, in the presence of (H₂O)₂. The plot is reasonably linear and yields a bimolecular rate constant, k₄ = (4.0 ± 1.2) × 10⁻¹² cm³ molecule⁻¹ s⁻¹, 2σ error. X-errors are estimated to be 22% and are propagated into the k₄ determination.

The slope of Fig. 4 is equal to the rate constant for the bimolecular reaction:²⁶

CH₂OO + (H₂O)₂ → HO–CH₂OOH + H₂O (R4)

The rate constant for reaction k₄ is equal to (4.0 ± 1.2) × 10⁻¹² cm³ molecule⁻¹ s⁻¹ and includes uncertainty due to [H₂O] (10%) and [H₂O]₂ (20%). In the ab initio calculations by Ryzhkov et al.²⁶ the channel to HO–CH₂OOH (HMHP) was observed to be the lowest energy for both H₂O (R2) and (H₂O)₂ (R4), with the dimer reacting to a greater extent under atmospheric conditions. It is noted that the present data could have a contribution from reaction (R2), and therefore k₄ should be regarded as an upper limit. However, in the ab initio study by Ryzhkov et al.²⁶ the calculated ratio of rate constants (k₂/k₄) is 3 × 10⁻⁵. At [H₂O] = 2 × 10¹⁷ molecule cm⁻³ the concentration of [H₂O]₂ is 9 × 10¹³ molecule cm⁻³, therefore the contribution from R2 is 0.07, and is only has a minor contribution in the present measurements. In the ESI† we report analysis of the data where both k₂ and k₄ are considered, and it is concluded that k₄ is overestimated by no more than 20%.

The value reported by Berndt et al. was k₄ = 1.01 ± 0.03 × 10⁻¹¹ cm³ molecule⁻¹ s⁻¹, which is about a factor of two larger than our present value. So while both studies are in broad agreement in that reaction (R4) is operating, there is a significant discrepancy in the magnitude of the rate constant. In the present work, the rate constant k₄ is extracted from the change in k_obs on addition of water vapour, where k₂′/k₃ < 10 and hence leads to larger than usual error in the bimolecular rate constant, see Fig. 4, but not as high as a factor of two. [H₂O] was determined from measuring the temperature and pressure of the bubbler and it was assumed that the entire H₂O equilibrium vapour pressure was delivered to the reactor. This is normally a reliable method to estimate the concentration of species introduced via a bubbler; previous work using a bubbler to deliver amines to a kinetic experiment has shown good agreement between calculated concentrations and values measured directly in the cell via UV absorption.²⁷ However, it is acknowledged that there is a potential to overestimate the water vapour concentration. Alternatively, there may be another reason for this discrepancy. The experiments from Berndt et al.¹⁶ used an atmospheric pressure time-of-flight mass spectrometer, where gas was sampled via a small aperture into the low pressure environment of the mass spectrometer. This gas expansion promotes cooling, which promotes dimer formation, and if dimer formation is promoted more rapidly than the reduction in pressure, then Criegee intermediate loss inside the mass spectrometer increases. While this is speculative, there are examples of promoted chemistry inside this type of mass spectrometer.²⁸ At the moment the source of this difference in the rate constant is unclear but it is clear that the Criegee intermediate generated viareaction (R1) or via ozonolysis produces essentially the same chemistry, i.e. there is no non-thermal kinetics.

This brings into question the failure of previous studies to observe any reaction of CH₂OO with H₂O vapour. In the experiments by Welz et al.² the Criegee intermediate was directly monitored and the highest amount of [H₂O] added was 3 × 10¹⁶ molecule cm⁻³ (corresponding to 2 × 10¹² molecule cm⁻³ dimer). This amount of [H₂O] increases the rate constant by no more than 20 s⁻¹, which in the experiments of Welz et al. is too small to observe. In the experiments by Stone et al.¹² CH₂O was used to follow the Criegee kinetics in time and up to [H₂O] = 1.7 × 10¹⁷ molecule cm⁻³ (corresponding to 6 × 10¹³ molecule cm⁻³ dimer) was added to the system. The calculated increase in the Criegee intermediate removal rate constant is between 250–600 s⁻¹ and should be measurable. However, this method relies on CH₂O only coming from characterised CH₂OO and CH₂IO₂ chemistry. If the products of reaction (R4) bring about new chemistry that forms CH₂O then it could mask any reaction with H₂O vapour. This new chemistry would be from radical–radical reactions. Therefore in the experiments of Stone et al.¹² where the radical densities are a few 10¹² molecule cm⁻³ CH₂O could be formed on a timescale not incompatible with this possible explanation. In the experiments of Ouyang et al.¹³ Criegee intermediate kinetics with H₂O were determined in competition with NO₂ by following the NO₃ formed from CH₂OO + NO₂. In these experiments up to 6 × 10¹⁷ molecule cm⁻³ of H₂O (corresponding to 8 × 10¹⁴ molecule cm⁻³ dimer) was added to the system, therefore the Criegee intermediate removal rate constant should have been >3000 s⁻¹, but no removal was observed. However, this method is dependent on Criegee intermediate + NO₂ reacting to make NO₃. The experiments by Ouyang et al.¹³ were not time-resolved; the contents of the reactor flowed into a cavity spectrometer and therefore it is possible that other secondary chemistry was responsible for NO₃ production. In a forthcoming paper it will be demonstrated using the current flash photolysis/UV/Vis absorption setup that NO₃ is not significantly made by reaction of Criegee with NO₂, and the small amount of observed NO₃ is consistent with the iodine chemistry, INO₂ + IONO₂ → NO₃ + NO₂ + I₂. Therefore the lack of change in the NO₃ signal versus added H₂O indicates a lack of reactivity in iodine chemistry and not Criegee intermediate chemistry.

Conclusions

The Criegee intermediate, CH₂OO, has been observed to react in the presence of water vapour. This is the first direct measurement to show that this reaction is occurring and its kinetics implies that the reaction is predominantly with the water dimer, (H₂O)₂, where k₄ = (4.2 ± 1.2) × 10⁻¹² cm³ molecule⁻¹ s⁻¹. This result is in support of the recent indirect measurements by Berndt et al.¹⁶ and indicates that Criegee intermediate chemistry is essentially independent of the method of generation via either ozonolysis or iodo-alkyl radical + O₂. The observed loss contrasts with other previous studies, but we believe the discrepancies can be explained by either the use of relatively low concentrations of water (limiting dimer formation) or via secondary chemistry in more indirect studies monitoring products. The direct observation of Criegee intermediates as used in this study will be less susceptible to such systematic errors.

Using the representative range in (H₂O)₂ concentrations (molecule cm⁻³) reported by Vereecken et al.,²⁹ 8.5 × 10¹³ (mega city) to 5.5 × 10¹⁴ (tropical forest), results in first-order loss rate for C1 Criegee intermediate ranging from 357–2310 s⁻¹. This is significantly greater than first order loss rates with other trace gases. In the atmospheric implications from Vereecken et al.²⁹reaction (R4) was included, using a theoretical estimate of the rate constant, and it was concluded that water vapour was the dominant removal process. This assessment provides a better representation of Criegee chemistry compared to modelling studies that have not included reaction (R4).³⁰ Given the importance of Criegee intermediates, further studies of the reaction with water dimer are required to confirm the fast kinetics reported in this work and to identify the products of the reaction.

Acknowledgements

TRL is grateful to NERC for studentship funding. We acknowledge funding from NERC (NE/K005820/1) and EPSRC (EP/J10871/1).

References

1. R. Criegee, A. Kerckow and H. Zinke, Chem. Ber., 1955, 88, 1878–1888.
2. O. Welz, J. D. Savee, D. L. Osborn, S. S. Vasu, C. J. Percival, D. E. Shallcross and C. A. Taatjes, Science, 2012, 335, 204–207.
3. D. Stone, M. Blitz, L. Daubney, T. Ingham and P. Seakins, Phys. Chem. Chem. Phys., 2013, 15, 19119–19124.
4. J. M. Beames, F. Liu, L. Lu and M. I. Lester, J. Am. Chem. Soc., 2012, 134, 20045–20048.
5. C. A. Taatjes, O. Welz, A. J. Eskola, J. D. Savee, D. L. Osborn, E. P. F. Lee, J. M. Dyke, D. W. K. Mok, D. E. Shallcross and C. J. Percival, Phys. Chem. Chem. Phys., 2012, 14, 10391–10400.
6. Z. J. Buras, R. M. I. Elsamra and W. H. Green, J. Phys. Chem. Lett., 2014, 5, 2224–2228.
7. Y. Liu, K. D. Bayes and S. P. Sander, J. Phys. Chem. A, 2014, 118, 741–747.
8. C. A. Taatjes, D. E. Shallcross and C. J. Percival, Phys. Chem. Chem. Phys., 2014, 16, 1704–1718.
9. A. Novelli, L. Vereecken, J. Lelieveld and H. Harder, Phys. Chem. Chem. Phys., 2014, 16, 19941–19951.
10. F. Liu, J. M. Beames, A. S. Petit, A. B. McCoy and M. I. Lester, Science, 2014, 345, 1596–1598.
11. A. B. Ryzhkov and P. A. Ariya, Chem. Phys. Lett., 2006, 419, 479–485.
12. D. Stone, M. Blitz, L. Daubney, N. U. M. Howes and P. Seakins, Phys. Chem. Chem. Phys., 2014, 16, 1139–1149.
13. B. Ouyang, M. W. McLeod, R. L. Jones and W. J. Bloss, Phys. Chem. Chem. Phys., 2013, 15, 17070–17075.
14. M. Suto, E. R. Manzanares and L. C. Lee, Environ. Sci. Technol., 1985, 19, 815–820.
15. K. H. Becker, J. Bechara and K. J. Brockmann, Atmos. Environ., Part A, 1993, 27, 57–61.
16. T. Berndt, J. Voigtlaender, F. Stratmann, H. Junninen, R. L. Mauldin, III, M. Sipilae, M. Kulmala and H. Herrmann, Phys. Chem. Chem. Phys., 2014, 16, 19130–19136.
17. D. R. Glowacki, J. Lockhart, M. A. Blitz, S. J. Klippenstein, M. J. Pilling, S. H. Robertson and P. W. Seakins, Science, 2012, 337, 1066–1069.
18. M. A. Blitz, C. Kappler, M. J. Pilling and P. W. Seakins, Z. Phys. Chem., 2011, 225, 957–967.
19. T. J. Dillon, M. E. Tucceri, R. Sander and J. N. Crowley, Phys. Chem. Chem. Phys., 2008, 10, 1540–1554.
20. 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.
21. L. Sheps, J. Phys. Chem. Lett., 2013, 4, 4201–4205.
22. W.-L. Ting, Y.-H. Chen, W. Chao, M. C. Smith and J. J.-M. Lin, Phys. Chem. Chem. Phys., 2014, 16, 10438–10443.
23. T. J. Gravestock, M. A. Blitz, W. J. Bloss and D. E. Heard, ChemPhysChem, 2010, 11, 3928–3941.
24. Y.-T. Su, H.-Y. Lin, R. Putikam, H. Matsui, M. C. Lin and Y.-P. Lee, Nat. Chem., 2014, 6, 477–483.
25. Y. Scribano, N. Goldman, R. J. Saykally and C. Leforestier, J. Phys. Chem. A, 2006, 110, 5411–5419.
26. A. B. Ryzhkov and P. A. Ariya, Phys. Chem. Chem. Phys., 2004, 6, 5042–5050.
27. L. Onel, M. Dryden, M. A. Blitz and P. W. Seakins, Environ. Sci. Technol. Lett., 2014, 1, 367–371.
28. M. Ehn, J. A. Thornton, E. Kleist, M. Sipilae, H. Junninen, I. Pullinen, M. Springer, F. Rubach, R. Tillmann, B. Lee, F. Lopez-Hilfiker, S. Andres, I.-H. Acir, M. Rissanen, T. Jokinen, S. Schobesberger, J. Kangasluoma, J. Kontkanen, T. Nieminen, T. Kurten, L. B. Nielsen, S. Jorgensen, H. G. Kjaergaard, M. Canagaratna, M. Dal Maso, T. Berndt, T. Petaejae, A. Wahner, V.-M. Kerminen, M. Kulmala, D. R. Worsnop, J. Wildt and T. F. Mentel, Nature, 2014, 506, 476–479.
29. L. Vereecken, H. Harder and A. Novelli, Phys. Chem. Chem. Phys., 2012, 14, 14682–14695.
30. 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, 4547–4550.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4cp04750h

---