Trends in stabilisation of Criegee intermediates from alkene ozonolysis †

Criegee Intermediates (CI), formed in the ozonolysis of alkenes, play a central role in tropospheric chemistry as an important source of radicals, with stabilised CI (SCI) able to participate in bimolecular reactions, aﬀecting climate through the formation of inorganic and organic aerosol. However, total SCI yields have only been determined for a few alkene systems, while speciated SCI yields from asymmetrical alkenes are almost entirely unknown. Here we report for the first time a systematic experimental exploration of the stabilisation of CH 2 OO and (CH 3 ) 2 COO CI, formed from ten alkene– ozone systems with a range of different sizes and structures, under atmospherically relevant conditions in the EUPHORE chamber. Experiments in the presence of excess SO 2 (an SCI scavenger) determined total SCI yields from each alkene–ozone system. Comparison of primary carbonyl yields in the presence/absence of SO 2 determined the stabilisation fraction of a given CI. The results show that the stabilisation of a given CI increases as the size of the carbonyl co-product increases. This is interpreted in terms of the nascent population of CI formed following decomposition of the primary ozonide (POZ) having a lower mean energy distribution when formed with a larger carbonyl co-product, as more of the energy from the POZ is taken by the carbonyl. These findings have significant implications for atmospheric modelling of alkene ozonolysis. Higher stabilisation of small CI formed from large alkenes is expected to lead to lower radical yields from CI decomposition, and higher SCI concentrations, increasing the importance of SCI bimolecular reactions.


Introduction
The formation of Criegee intermediates (CI) from gas phase alkene ozonolysis has received attention over the past five decades owing to their role as important non-photolytic sources of radicals (OH, HO 2 and RO 2 )t ot h et r o p o s p h e r e . [1][2][3][4][5] More recently, the potential importance of bimolecular reactions of stabilised CI (SCI) has been the subject of much research (see Vereecken et al. 6 and references within). These reactions contribute significantly to the sulfuric acid budget in certain environments through oxidation of SO 2 , 7 and the acidity of the atmosphere through removal of organic and inorganic acids. 8 Bimolecular reactions of SCI have also been implicated in the formation of aerosol from monoterpene ozonolysis 9 through dimerization, 10,11 oligomerization 12 and reaction with peroxy radicals, 13 contributing to new particle formation 14 and so directly affecting cloud condensation nuclei, 15,16 rainfall, and climate.
The Criegee ozonolysis reaction mechanism 17 proceeds via concerted cycloaddition of the ozone molecule across the CQC double bond to form a chemically activated primary ozonide (POZ), followed by cleavage of the C-C bond and one of the O-O bonds forming a carbonyl molecule and a carbonyl oxide, or 'Criegee intermediate', 18 see Scheme 1. A population of ozonolysis derived CIs in the gas-phase is formed with a broad internal energy distribution. 19 A fraction of CI may be formed 'cold' (although this is not the case for all alkenes 20,21 ), that is, without enough internal energy to undergo prompt decomposition, these are termed stabilised Criegee intermediates (SCI). The remainder are formed chemically activated (CI*). These CI* tend to undergo prompt decomposition on a timescale of nanoseconds. 22 However, they can also be collisionally deactivated to add to the SCI population. The distinction between these two routes to SCI formation has been demonstrated in laboratory experiments performed as a function of pressure. 20,21,[23][24][25] The fraction of each type of CI that is formed will depend on the initial energy distribution of the CI population. 24 Unimolecular decomposition of CI yields a range of radical products, with decomposition of syn-CI (i.e. CI with an alkyl group on the same side of the terminal O atom of the carbonyl oxide moiety), via a vinyl hydroperoxide, producing OH with (near) unit yield. SCI can have sufficiently long lifetimes to undergo bimolecular reactions with H 2 O, RCOOH, and SO 2 , amongst other species, in the atmosphere. Hence the relative yields of CI/SCI from a particular alkene-ozone system determine the effect of that system on atmospheric composition.
Inclusion of alkene ozonolysis chemistry in regional and global chemical transport models is essential to correctly estimate radical concentrations, and the product distribution from the removal of alkenes by reaction with ozone. However, such mechanisms must firstly parameterize SCI yields for a structurally diverse range of alkenes for which no measurements exist, lumping together SCI of different structures and hence reactivities, and secondly lumping together stabilisation of different SCI. This leads to large uncertainties on the SCI yields and hence on the effect of alkene-ozone reactions on atmospheric composition.
Many laboratory studies probing the chemistry of SCI have utilised the facile photolysis of alkyl iodides to yield SCI in the presence of oxygen. 26 This has proved an invaluable resource for the study of SCI chemistry. However, alkene ozonolysis is expected to be the dominant source of CI to the atmosphere, and certain atmospherically relevant questions, such as the fractional CI and SCI yields from alkene ozonolysis, can only be answered by probing the alkene-ozone system.
Total SCI yields have previously been measured from a number of atmospherically relevant alkene-ozone systems including short chain alkenes, [27][28][29][30][31] isoprene, [32][33][34] and monoterpenes. 7,32 These yields are generally measured indirectly, either by the removal of an SCI scavenger (e.g. SO 2 ) or formation of a product from an SCI scavenging reaction (e.g. production of H 2 SO 4 , via the SO 2 + SCI reaction). However, very little experimental information exists on the relative amounts of different SCI formed in nonsymmetrical systems. For non-symmetrical, non-cyclic, alkenes, a pair of CI and carbonyl co-products are formed (Scheme 1). The yield of each SCI depends on: (i) the yield of each CIdetermined by the fragmentation of the POZ (a); (ii) the fraction of each CI that is stabilised (S). The stabilisation of a given CI produced from different alkene-ozone systems might be expected to differ. For example, while the fraction of CH 2 OO stabilised in ethene ozonolysis has been determined to be 0.35-0.54 based on a wide range of experimental studies, the chamber studies of Nguyen et al. 34 suggest that the majority of CH 2 OO formed in isoprene ozonolysis is stabilised.
Here, we present results of a series of alkene ozonolysis experiments carried out at the European PhotoReactor facility (EUPHORE), Valencia, Spain, in which yields of chemically activated and stabilised CH 2 OO and (CH 3 ) 2 COO are determined, for the first time, for a systematic range of alkene-ozone systems under atmospherically relevant conditions. An empirical structureactivity relationship for the stabilisation of these types of Criegee intermediates, based on the size of the carbonyl co-product formed in the decomposition of the primary ozonide is also presented, with the atmospheric implications of the results discussed.

EUPHORE
EUPHORE is a 200 m 3 simulation chamber used for studying reaction mechanisms under atmospheric boundary layer conditions. The chamber is fitted with large horizontal and vertical fans to ensure rapid mixing (three minutes). Further details of the chamber setup and instrumentation are available elsewhere. [35][36][37] Experiments comprised time-resolved measurement of the formation of carbonyl products and the loss of alkene and ozone (and in some experiments SO 2 ). SO 2 and O 3 abundance were measuredu s i n gc o n v e n t i o n a l fluorescence and UV absorption monitors, respectively; alkene and oxygenated volatile organic compound abundance was determined via FTIR spectroscopy and PTR-MS. The precision of the SO 2 and O 3 monitors were 0.25 and 0.47 ppbv respectively (evaluated as 2 standard deviations of the measured value prior to SO 2 or O 3 addition). Experiments were performed in the dark (i.e., with the chamber housing closed; j(NO 2 ) r 10 À6 s À1 ), at atmospheric pressure (ca. 1000 mbar) and temperatures between 297 and 305 K, on timescales of ca. 30-90 minutes. Chamber dilution was monitored via the first order decay of an aliquot of SF 6 , added prior to each experiment. Cyclohexane (ca. 75 ppmv) was added at the beginning of each experiment to act as an OH scavenger, such that SO 2 reaction with OH was calculated to be r1% of the total chemical SO 2 removal in all experiments.

Experimental approach
Experimental procedure comprised addition of SF 6 and cyclohexane, followed by O 3 (ca. 1000 ppbv for the experiments with Scheme 1 Simplified mechanism for the production of CI and primary and secondary carbonyl products in the ozonolysis of 2-methyl propene. alkenes producing CH 2 OO, which generally have reaction rates with ozone B1 Â 10 À17 cm 3 s À1 ; and ca. 500 ppbv for the experiments with alkenes producing (CH 3 ) 2 COO, which generally have reaction rates with ozone B1 Â 10 À16 cm 3 s À1 ) and SO 2 if used (ca. 2000 ppbv). A gap of five minutes was left prior to addition of the alkene, to allow complete mixing. The reaction was then initiated by addition of the alkene (ca. 800 ppbv for the systems producing CH 2 OO, and 400 ppbv for those producing (CH 3 ) 2 COO). The chamber was monitored for 30-90 minutes subsequent to the addition of the alkene depending on the rate of reaction with ozone, the rate of alkene/ozone consumption being dependent on k(alkene + ozone). Roughly 50% of the CH 2 OO producing alkenes were consumed after 60 minutes, while 90% of the faster reacting alkenes were consumed within roughly 25 minutes. The experiments were performed under dry conditions (RH o 1%). A full experiment list is given in Table S1 (ESI †).

Total SCI yields
Alkene ozonolysis experiments were performed in the presence of ca. 2000 ppbv SO 2 , such that the overwhelming majority (Z94%) of the SCI was scavenged and converted to a carbonyl. The total SCI yield, f, was calculated by regressing the loss of ozone (dO 3 ) against the loss of SO 2 (dSO 2 ) (eqn (E1)), both corrected for chamber dilution. In reality the experimentally determined value of f is a minimum value, f min since other loss channels for the SCI (e.g. decomposition) may still be having a small but non-negligible effect (Newland et al. (2015a)), accounted for in eqn (E1) by f.
In addition to the presence of SO 2 as a potential SCI reaction partner in the chamber, there is also H 2 O( o1% RH), HCOOH (of the order of 10 ppbv produced in the ozonolysis reaction), and carbonyls on the order of a few hundred ppbv. For stabilised CH 2 OO, decomposition is slow (0.1 s À1 , IUPAC) and the only other potentially significant loss process under the experimental conditions employed was bimolecular reaction w i t hw a t e rv a p o u ro rH C O O H .B a s e do nt h eI U P A C 38 recommended rate constants, 2000 ppbv of SO 2 is estimated to scavenge 498% of CH 2 OO at RH = 1% and [HCOOH] = 10 ppbv (typical mixing ratio present in the chamber for the experimental conditions). For (CH 3 ) 2 COO the only other important loss process under the experimental conditions employed was unimolecular decomposition. 2000 ppbv of SO 2 is estimated to scavenge 94% of (CH 3 ) 2 COO at 303 K based on the IUPAC (IUPAC) recommended rates of k((CH 3 ) 2 COO + SO 2 ) = 4.2 Â 10 À13 exp (1761/T) cm 3 s À1 and k uni ((CH 3 ) 2 COO) = 1.0 Â 10 7 exp (À3020/T) s À1 . A range of other syn-and anti-CI will be formed as co-products for many of the alkenes studied here. Using syn and anti-CH 3 CHOO as surrogates, under the experimental conditions employed here, 88% of syn-CI will be scavenged by 2000 ppbv SO 2 ,a n d9 7 %o fanti-CI will be scavenged. Elsamra et al. 39 determined k(CH 2 OO + acetone) = 3.0 (AE1.0) Â 10 À13 cm 3 s À1 and k(CH 2 OO + acetaldehyde) = 1.2 (AE0.3) Â 10 À12 cm 3 s À1 ,a t 298 K. Products of the reaction are believed to be the carbonyl + acid (e.g. CH 2 OO + CH 3 CHO -CH 3 COOH + HCHO or HCOOH + CH 3 CHO), formed via decomposition of a secondary ozonide. 40 At 200 ppbv of the carbonyl, this would lead to a sink for CH 2 OO of the order of 1-10 s À1 , o1% in SO 2 scavenger experiments. Other potential SCI sinks, such as self-reaction are negligible under our experimental setup. Total SCI yields were calculated for ten alkene-ozone systems in this work (C 2 -C 10 ). Fig. 1 shows an example plot of DSO 2 vs. DO 3 for an ozonolysis experiment with 2,4-dimethyl-pent-2-ene. Uncertainties are AE2s and represent the combined systematic (estimated measurement uncertainty) and precision components. Fig. S1 (ESI †) shows total SCI plots for each of the ten alkenes studied. Table 1 gives the total SCI yields corrected for additional SCI losses, i.e. f,( s e eT a b l e sS 2a n dS 3 ,E S I † this correction generally increased the yields by B5%) calculated for each alkene, and values previously reported in the literature where available.
For alkenes for which total SCI yields have been measured previously, there is good agreement between the values measured here and those reported in the literature. The ethene-ozone system is the most studied, with reported SCI yields ranging from 0.35-0.54, from a range of different techniques -see Newland et al.; 31 Alam et al. 36   Deng et al. 42 recently reported a total SCI yield of 0.35 for the monoterpene myrcene based on theoretical calculations (82% (CH 3 ) 2 COO, 15% large anti-SCI, 3% large syn-SCI), compared to our measured value of 0.46. It may be expected that our experimental yield is an underestimation due to the very fast unimolecular reaction of the large anti-SCI formed in this system (calculated by Deng et al. 42 to be 7600 s À1 ) which means that probably o50% would be scavenged by 2000 ppbv SO 2 . Based on the yield of anti-SCI predicted by Deng et al. 42 (15%), this may lead to an underestimation of the total SCI yield of B8%. None of the other alkenes studied here have had total SCI yields reported previously to the authors' knowledge. Fig. 2 shows the total SCI yields measured in this study plotted against carbon number of the parent alkene. Previous studies 43 have noted that total SCI yields do not appear to display a strong dependence on alkene size. The systems studied here suggest a weak dependence on the size of the parent alkene, with total SCI yield increasing with alkene size. Propene and 2-methylpropene have total SCI yields of 0.24-0.34, while the larger trisubstituted alkenes, 2,3,4 trimethylpent-2-ene and myrcene have yields of 0.41-0.49, and the largest terminal alkene, 1-heptene has a yield of 0.61 (AE0.03). However, as shown later the total SCI yield is the product of a number of effects and hence any simple relation to alkene size must be treated with caution.

Primary carbonyl yields
For non-symmetrical, non-cyclic, alkenes, a pair of CI and carbonyl co-products are formed in the decomposition of the POZ (Scheme 1). The sum yield of pathways a 1 + a 2 (i.e. the sum yield of the two possible primary carbonyls) should be equal to one. This has been confirmed to be the case in the extensive experimental dataset of Grosjean and Grosjean 44 for systems in which the smaller carbonyl is not also formed from decomposition of the larger CI. Therefore, by determining the yield of just one of the carbonyls (ideally the larger one, as it cannot be formed in CI decomposition) it is possible to determine both the primary yield of the other carbonyl, and the yield of both CI. For most CI, syn and anti conformers can also be formed because of the negligible rotation about the CQO bond of the CI moiety (Vereecken and Francisco, 2012). However, the two CI of focus in this study, CH 2 OO and (CH 3 ) 2 COO, are both symmetrical.
The relative magnitude of pathways a 1 and a 2 is determined by the fragmentation of the POZ. This appears to be determined predominantly by the structure around the double bond at which ozonolysis occurs. 3,45 Af r a c t i o n ,S,o fe a c ho ft h et w oC If o r m e di s stabilised, either due to being formed below the energy threshold for being chemically activated, or via collisional stabilisation. The yield of each SCI is thus given by the product of a i and S i (eqn (E2); Scheme 1). The total yield of SCI from a given alkene is then the sum of the yields of the two specific SCI (eqn (E3)).
The primary carbonyl yields in these experiments were determined by FTIR measurement of the yield of the primary carbonyls relative to the loss of ozone (both corrected for chamber dilution) (eqn (E4)).  41 Carbonyl product yields 2-Methylpropene (C 4 ) 0.24 (AE0.03) 0.17 (AE0.03) 28 DH 2 SO 4 /DO 3 2-Methyl-but-1-ene (C 5 Table 2 shows the primary carbonyl yields determined for each alkene ozonolysis system studied. a 1 is the pathway that leads to the CI being studied here, i.e. CH 2 OO or (CH 3 ) 2 COO. Fig. S2 (ESI †) shows plots of the primary carbonyl yields for all experiments. For alkenes in which HCHO or acetone are expected to be formed in the decomposition of the larger CI, a 1 was determined based on the measured yield of the larger carbonyl. HCHO is expected to be a decomposition product of any CI with a methyl group syn or anti to the carbonyl oxide moiety. 46 For syn-CI this comes via decomposition of a b-oxo-alkoxy radical formed via the vinyl-hydroperoxide mechanism; for anti-CI the methyl radical is formed in decomposition of a bis-oxy radical, formed via 1,3 ring closure of the CI, leading to CH 3 O 2 and ultimately to HCHO. For longer alkyl chains, HCHO is not expected to be formed. For short chain terminal alkenes this is consistent with the extensive database of ozonolysis experiments by Grosjean and Grosjean 44 (and references therein). These show total primary carbonyl yields of unity for straight chain terminal alkenes with the exception of propene (that forms CH 3 CHOO), for which the sum carbonyl yield is well in excess of 1 (1.30). Any alkene that produces (CH 3 ) 2 COO will also have a large secondary HCHO yield from decomposition of this CI.

CI stabilisation
The stabilisation, S, of a given CI is calculated as the ratio of the yield, f, of the SCI to the yield of the CI (eqn (E5)). f CI is calculated as described above, based on the primary carbonyl yields in experiments with no SCI scavenger. f SCI for a specific CI is determined from the difference between the carbonyl yields in experiments with and without an SCI scavenger. The calculated stabilisation of CH 2 OO and (CH 3 ) 2 COO in the alkene systems studied here are given in Table 3.
where carb 21 is the secondary yield of the carbonyl, i.e. from SCI + SO 2 , and carb 11 is the primary yield of the carbonyl. For systems in which the carbonyl of interest (i.e. HCHO/acetone) could also be formed from decomposition of the larger CI, S is calculated by rearranging eqn (E3)-(E6).  where S 1 is the stabilisation of CH 2 OO or (CH 3 ) 2 COO, S 2 is the stabilisation of the other CI, calculated using eqn (E5). Fig. 4 shows the observed relationship between the stabilisation of the CI and the carbon number of the carbonyl coproduct formed in POZ decomposition. There is a clear increase in stabilisation of both CH 2 OO and (CH 3 ) 2 COO with increasing carbon number of the carbonyl co-product. This can be considered in terms of the distribution of the total energy available from decomposition of the POZ. If it is assumed that the total energy liberated in decomposition of the POZ is independent of the size of the alkene, 19 and that the available energy has time to become distributed equally between the non-hydrogen atoms (i.e. C and O), then as the size of the carbonyl co-product increases relative to the CI, so the proportion of the available energy that is taken by the CI will decrease. This would be expected to yield a CI population with a lower mean energy distribution, both increasing the yield of CI that are 'born cold', and also increasing the amount of CI that will be collisionally stabilised (Fig. 5). While theoretical work has shown that the energy distribution between CI and carbonyl fragments in POZ decomposition may be non-statistical, 47,48 it might still be expected that the general relationship will hold. Based on the discussion above, the data presented in Fig. 4 can be fitted by the general relationship given in eqn (E7).

Discussion
where A CI is the number of non-hydrogen atoms (i.e. C and O) in the CI; A tot is the total number of non-hydrogen atoms in the POZ; and F is a factor determined from the total SCI yield of the symmetrical alkene (i.e. ethene for CH 2 OO and 2,3-dimethylbut-2-ene for (CH 3 ) 2 COO). The relationship is plotted in Fig. 4 (red dashed line) for a SCI yield from ethene of 0.43 (lower limit of 0.39, upper limit of 0.43), and from 2,3-dimethyl-but-2-ene of 0.31 (0. 27-0.35). It is seen that the prescribed relationship fits the observed data well. While there are certainly likely to be additional factors influencing the stabilisation of CI, such as specific substituents Table 3 Stabilisation of CH 2 OO/(CH 3 ) 2 COO formed in the ten alkene systems studied. Calculated using either eqn (E5) and (E6) -see text for details. For ethene and 2,3-dimethyl-but-2-ene stabilisation of the CI is equal to the total SCI yield shown in Table 2 Alkene Stab a Symmetrical, therefore total SCI yield ( Table 2). b From model fit to data (Fig. S2, ESI).  and more complex structures, this work shows that the number of carbons in the system is a strong determinant of CI stabilisation, particularly for systems with similar structures. For CI stabilisation from a given alkene, there is expected to be three important effects: (i) collisional stabilisation of the POZ, which will then decompose to yield exclusively stabilised CI -this has been predicted to be significant (65%) for the C 15 sesquiterpene b-caryophyllene 49 but to be insignificant for smaller (oC 15 ) alkenes; 50 (ii) an increased stabilisation of a given CI with increasing parent alkene size, as shown herein; and (iii) an increased stabilisation of larger CI, which have longer lifetimes (due to distributing the initial energy from the POZ decomposition among a greater number of degrees of freedom) allowing greater collisional stabilisation. 20 This effect is particularly noticeable for 1-heptene in this dataset, with stabilisation of the larger CI, hexanal oxide being high (B50%).
Using the relationships shown here it is possible to calculate the stabilisation of a given CI produced from an alkene that also produces either CH 2 OO or (CH 3 ) 2 COO if the total SCI yield and the POZ decomposition branching ratio (a) are known. However, there are still relatively few alkenes for which total SCI yields have been measured. Further useful information to inform structure activity relationships used in atmospheric models would cover: (i) the dependence of a on alkene structure; (ii) trends in the total SCI yields from symmetrical alkenes of increasing size and complexity, to provide values on which to pin the relationships observed herein; (iii) the effect of the size of the CI itself on stabilisation. In addition, information on the relative yields and stabilisation of (E)/(Z) CI is required to fully represent the impact of alkene ozonolysis on atmospheric composition. In addition to further experimental studies, a detailed theoretical study would provide strong corroborating evidence both for the relationship derived in this work, and the further work suggested here.

Atmospheric implications
This work suggests that small CI produced from large alkenes found in the atmosphere, e.g. isoprene, monoterpenes, sesquiterpenes, etc., will predominantly be formed stabilised. For CH 2 OO, the SCI of which reacts almost entirely with water vapour in the boundary layer, this will result in a high yield of the products of the CH 2 OO + H 2 O/(H 2 O) 2 reaction, and low radical (OH and HO 2 ) yields (although it has been suggested that radical yields from CH 2 OO decomposition are low anyway 51,52 ). This finding is particularly important for isoprene, the most abundantly emitted alkene to the atmosphere. 53 Nguyen et al. 34 suggested that the total SCI yield (B0.6) 32-34 measured from isoprene ozonolysis is almost entirely stabilised CH 2 OO (rather than the C 4 -CI). The main products of the reaction of CH 2 OO with water vapour are thought to be hydroxy-methylhydroperoxide (HMHP) and formic acid (HCOOH), 54 b u tr e c e n tf l o wt u b ee x p e r i m e n t s 55 suggest a roughly equal split between HMHP and formaldehyde (HCHO) from the CH 2 OO + (H 2 O) 2 reaction, with direct HCOOH formation o5%. The dominant fate of HMHP in the atmosphere is unclear, with a relatively long lifetime against reaction with OH of about 1 day. 56 OH reaction was shown to yield HCOOH and HCHO in a ratio of 0.88. 56 For (CH 3 ) 2 COO, and other small syn-CI, for which reaction with water vapour is a negligible sink under boundary layer conditions, 55 a higher stabilisation will lead to an increased atmospheric concentration of these SCI. Under typical boundary layer conditions these SCI will participate in bimolecular reactions (e.g. with SO 2 and organic acids), contribute to aerosol formation, or undergo unimolecular decomposition.

Conclusions
Ozonolysis experiments were performed at the EUPHORE atmospheric simulation chamber on a range of alkenes that produce either the CH 2 OO or (CH 3 ) 2 COO Criegee intermediates. Total stabilised Criegee intermediate (SCI) yields were determined from the temporal decay of the SCI scavenger SO 2 .Sp ec ia te dCI yields were determined based on FTIR measurements of primary carbonyl products. Speciated SCI yields were determined from comparison of carbonyl yields in the presence/absence of the SCI scavenger SO 2 . From this information, the stabilisation of CH 2 OO and (CH 3 ) 2 COO from each alkene was determined. The stabilisation is shown to increase with increasing carbon number of the carbonyl co-product formed in the decomposition of the primary ozonide. Stabilisation of CH 2 OO increases from a minimum of B0.4 from ethene, to 0.71 from 1-heptene. Stabilisation of (CH 3 ) 2 COO increases from o0.1 from isobutene, to 0.50 from the monoterpene myrcene. An empirical relationship based on the energy distribution through the molecule on dissociation of the POZ fits the observed data well. This trend has implications for predicted tropospheric concentrations of SCI, with current models generally using SCI yields based on the total yield from the relevant symmetrical alkene-ozone system. From this work it is shown that stabilisation of small CI from many atmospherically relevant alkenes, such as isoprene and monoterpenes, is likely to be considerably higher than currently predicted. This would increase the importance of bimolecular reactions, and reduce radical yields from CI decomposition.

Data availability
Experimental data will be available in the Eurochamp database, www.eurochamp.org, from the H2020 EUROCHAMP2020 project, GA no. 730997.

Conflicts of interest
There are no conflicts to declare. and Fundacion CEAM and the Generalitat Valenciana for the IMAGINA-Prometeo project (PROMETEO2019/110). Fundación CEAM is partly supported by Generalitat Valenciana. Mike Newland and Andrew Rickard also acknowledge support from the Mechanisms for Atmospheric chemistry: GeneratioN, Interpretation and FidelitY -MAGNIFY project, funded by the UK Natural Environment Research Council (NERC, via grant NE/M013448/1). Beth Nelson acknowledges the NERC SPHERES Doctoral Training Partnership (DTP) for her studentship.