Application of chemical derivatization techniques combined with chemical ionization mass spectrometry to detect stabilized Criegee intermediates and peroxy radicals in the gas phase

Short-lived highly reactive atmospheric species, such as organic peroxy radicals (RO2) and stabilized Criegee intermediates (SCIs), play an important role in controlling the oxidative removal and transformation of many natural and anthropogenic trace gases in the atmosphere. Direct speciated measurements of these components are extremely helpful for understanding their atmospheric fate and impact. We describe the development of an online method for measurements of SCIs and RO2 in laboratory experiments using chemical derivatization and spin trapping techniques combined with H3O and NH+4 chemical ionization mass spectrometry (CIMS). Using chemical derivatization agents with low proton affinity, such as electron-poor carbonyls, we scavenge all SCIs produced from a wide range of alkenes without depleting CIMS reagent ions. Comparison between our measurements and results from numeric modeling, using a modified version of the Master Chemical Mechanism, shows that the method can be used for the quantification of SCIs in laboratory experiments with a detection limit of 1.4× 107 molecule cm−3 for an integration time of 30 s with the instrumentation used in this study. We show that spin traps are highly reactive towards atmospheric radicals and form stable adducts with them by studying the gas-phase kinetics of the reaction of spin traps with the hydroxyl radical (OH). We also demonstrate that spin trap adducts with SCIs and RO2 can be simultaneously probed and quantified under laboratory conditions with a detection limit of 1.6× 108 molecule cm−3 for an integration time of 30 s for RO2 species with the instrumentation used in this study. Spin trapping prevents radical secondary reactions and cycling, ensuring that measurements are not biased by chemical interferences, and it can be implemented for detecting RO2 species in laboratory studies and potentially in the ambient atmosphere.

Abstract. Short-lived highly reactive atmospheric species, such as organic peroxy radicals (RO 2 ) and stabilized Criegee intermediates (SCIs), play an important role in controlling the oxidative removal and transformation of many natural and anthropogenic trace gases in the atmosphere. Direct speciated measurements of these components are extremely helpful for understanding their atmospheric fate and impact. We describe the development of an online method for measurements of SCIs and RO 2 in laboratory experiments using chemical derivatization and spin trapping techniques combined with H 3 O + and NH + 4 chemical ionization mass spectrometry (CIMS). Using chemical derivatization agents with low proton affinity, such as electron-poor carbonyls, we scavenge all SCIs produced from a wide range of alkenes without depleting CIMS reagent ions. Comparison between our measurements and results from numeric modeling, using a modified version of the Master Chemical Mechanism, shows that the method can be used for the quantification of SCIs in laboratory experiments with a detection limit of 1.4 × 10 7 molecule cm −3 for an integration time of 30 s with the instrumentation used in this study. We show that spin traps are highly reactive towards atmospheric radicals and form stable adducts with them by studying the gas-phase kinetics of the reaction of spin traps with the hydroxyl radi-cal (OH). We also demonstrate that spin trap adducts with SCIs and RO 2 can be simultaneously probed and quantified under laboratory conditions with a detection limit of 1.6 × 10 8 molecule cm −3 for an integration time of 30 s for RO 2 species with the instrumentation used in this study. Spin trapping prevents radical secondary reactions and cycling, ensuring that measurements are not biased by chemical interferences, and it can be implemented for detecting RO 2 species in laboratory studies and potentially in the ambient atmosphere.
oxy radicals (RO 2 ). In an NO-rich environment, RO 2 radicals predominantly react with NO, while reactions with the hydroperoxy radical (HO 2 ) and potentially other RO 2 become more important at lower NO concentrations, as do unimolecular reactions. The common tendency is to form closed-shell, more oxidized VOCs (OVOCs). OVOCs may have lower volatilities than the parent hydrocarbons and may partition to the particle phase, thereby contributing to secondary organic aerosol (SOA) formation. OH, HO 2 , and RO 2 radicals can form a catalytic reaction cycle, which can lead to the production of tropospheric ozone as a consequence of the shift in the NO/NO 2 ratio to favor the formation of NO 2 . This cycle is terminated by the formation of organic hydroperoxides and nitrates, which can be viewed as reservoirs of the corresponding radicals. Overall, atmospheric radicals -especially their cycling -play an important role in the formation of SOA and tropospheric ozone, as well as in controlling atmospheric oxidation capacity.
Organic peroxy radicals can also be formed via ozonolysis of unsaturated organic compounds. Ozonolysis of alkenes results in the formation of primary ozonides that promptly decompose to a stable carbonyl and a vibrationally excited carbonyl oxide, also known as a Criegee intermediate (CI), some of which are thermally stabilized (SCI). SCIs primarily decompose or react with water vapor  but are also believed to play a role in the oxidation of SO 2 to form H 2 SO 4 in the tropical regions (Khan et al., 2018). syn-SCI can undergo a unimolecular reaction and form a vinyl hydroperoxide, which rapidly decomposes to an OH radical and a vinyl radical. This radical is in resonance with an acetonyl-type radical, which can combine with molecular oxygen to form an RO 2 species (Johnson and Marston, 2008).
Measurements of atmospheric radicals and reactive intermediates such as RO 2 and SCIs are challenging because of their high reactivity towards trace gases and surfaces and rapid cycling, which may lead to potential interferences. Highly sensitive detection systems are required to determine the minute concentrations of these species. Concentrations of the smallest organic peroxy radical, CH 3 O 2 , are typically on the order of 10 8 molecule cm −3 , while concentrations of other RO 2 species can be much lower (Fuchs et al., 2008). As for SCIs, their concentrations are expected to be less than 10 5 molecule cm −3 . With respect to RO 2 species, there are several field-deployable measurement techniques available for nonspeciated measurements of the sum of RO 2 . Matrix isolation electron spin resonance spectroscopy (MIESR) is an established, but rarely used, method for field measurements (Mihelcic et al., 1985). MIESR is an offline technique with a low time resolution (∼ 30 min); however, its main advantage is that it does not require instrument calibration. Besides MIESR, chemical amplification and conversion systems represent another class of instruments for field studies (Edwards et al., 2003;Hornbrook et al., 2011;Cantrell et al., 1984;Wood and Charest, 2014). In these systems, peroxy radicals are not measured directly but are rather converted to other radicals or closedshell molecules (e.g., NO 2 or H 2 SO 4 ). A detection limit of 10 7 molecule cm −3 can be achieved at a temporal resolution of 15 s, but discrimination of different RO 2 species is not possible (Edwards et al., 2003). In addition, secondary chemistry, i.e., additional sources of radical production and destruction, has to be considered, and care needs to be taken to ensure that measurements are not biased by any chemical interferences (Reiner et al., 1997). Finally, laserinduced fluorescence (LIF) was also applied for ambient measurements of RO 2 radicals (Fuchs et al., 2008). This technique is characterized by an excellent detection limit of (2-7) ×10 7 molecule cm −3 for an integration time of 30 s. Similarly to chemical amplifier systems, however, LIF does not allow the differentiation of various RO 2 species. Although it is indirect and converts RO 2 to OH, it does not have an amplification chain. Recently, novel mass spectrometric techniques using different ionization schemes to directly detect individual RO 2 species were developed Berndt et al., 2018Berndt et al., , 2019Nozière and Vereecken, 2019).
As for SCIs, indirect measurement techniques have been widely used. In these techniques, SCIs are chemically converted to other species (e.g., H 2 SO 4 or hydroxymethyl hydroperoxide, HMHP) Sipilä et al., 2014;Neeb et al., 1997). In 2008, the simplest SCI, CH 2 OO, was directly detected for the first time (Taatjes et al., 2008). Later, synchrotron photoionization mass spectrometry was combined with the CI generation technique using diiodoalkane photolysis , which spurred several studies to examine the kinetics of bimolecular and unimolecular SCI reactions Lewis et al., 2015;Chhantyal-Pun et al., 2016). Recently, two new techniques for direct measurements of SCIs using Fourier transform microwave spectroscopy and chemical ionization mass spectrometry (CIMS) were introduced (Womack et al., 2015;Berndt et al., 2017).
Despite the abundance of different analytical methods used for the detection of atmospheric radicals and reactive intermediates, there is still a need for an online, direct, field-deployable technique for measuring these short-lived highly reactive compounds in a speciated way. Free radicals have been conventionally detected by chemical derivatization (CD) techniques such as spin trapping in condensedphase biological and chemical systems (Hawkins and Davies, 2014;Nosaka and Nosaka, 2017). Nonradical spin traps (e.g., nitrone spin traps) are known to react with free radicals to form stable radical adducts that can be detected with electron paramagnetic resonance spectroscopy . In addition, radical spin traps (e.g., nitroxide radicals) are also highly reactive towards radical species such as C-centered radicals and form closed-shell adducts with them (Bagryanskaya and Marque, 2014). However, there are only a few studies in which these techniques were applied for probing atmospheric radicals and intermediates. Watanabe et al. (1982) presented an offline method to quantify hydroxyl radicals using the spin trap α-4-pyridyl-N-tert-butylnitrone α-1-oxide (4-POBN), where condensed-phase stable adducts were detected by electron spin resonance. Recently, Giorio et al. (2017) used the spin trap 5,5-dimethyl-1-pyrroline Noxide (DMPO) to characterize SCIs by detecting gas-phase spin trap adducts with online mass spectrometry.
Here, we explore three types of CD agents, including two spin-trapping agents, and show how they can be used for the detection and quantification of various atmospheric radicals and reactive intermediates (Fig. 1). First, we implement the CD agent hexafluoroacetone (HFA) to characterize a wide range of gas-phase SCIs. HFA is selectively reactive towards SCIs (i.e., it is unreactive towards OH, HO 2 , and RO 2 ), forms stable secondary ozonides with them, and has high vapor pressure and low proton affinity (Fig. 1). Next, we use the radical spin trap (2,2,6,6-tetramethylpiperidin-1yl)oxyl (TEMPO) to demonstrate that spin traps are highly reactive towards radicals in the gas phase -and can therefore effectively scavenge atmospheric radicals -by studying the kinetics of the TEMPO + OH reaction. Finally, we utilize the nonradical spin trap DMPO to simultaneously detect atmospheric gas-phase radicals and intermediates, including SCIs and RO 2 species (Fig. 1). Spin trap adducts and secondary ozonides with CD agents are observed and quantified using H 3 O + and NH + 4 CIMS, which allows for speciated online measurements of stabilized Criegee intermediates and speciated RO 2 radicals formed via ozonolysis of a wide range of parent hydrocarbons. The analytical methods presented here can be used for the quantification of speciated SCIs and RO 2 formed in laboratory experiments, as well as for field measurements.

Ozonolysis experiments with the chemical derivatization agent HFA
The ozonolysis experiments with multiple hydrocarbons including tetramethylethylene (TME), isoprene, pentene, hexene, α-pinene, and limonene were conducted in a flow tube reactor at ambient pressure and temperature (∼ 290 K) and under dry conditions (relative humidity < 2 %). The experimental setup consisted of a flow reactor system with a residence time of ∼ 10 s. The parent hydrocarbon was mixed with ozone and the chemical derivatization agent HFA (C 3 F 6 O) in the flow reactor, leading to SCI formation and scavenging as SCI q HFA adducts. SCIs are known to be highly reactive towards ketones, especially electron-poor ones such as HFA (Horie et al., 1999;Taatjes et al., 2012). HFA has been previously used to effectively scavenge SCIs and prevent their secondary chemistry, thus allowing SCI formation to be probed directly . The other advantage of employing this chemical derivatization agent is its relatively low proton affinity (PA 670.4 kJ mol −1 ; Hunter and Lias, 1998). Since the PA of HFA is lower than that of water, HFA cannot be protonated in H 3 O + CIMS. Hence, one can introduce a significant amount of HFA into the system to make sure that all SCIs are scavenged very rapidly without any concern that the H 3 O + reagent ions will be depleted. The parent hydrocarbon was vaporized from a flask filled with the pure substance by passing a constant flow of zero air regulated via a 0.1-10 cm 3 min −1 mass flow controller (Bronkhorst). The HFA flow was regulated by another mass flow controller (Bronkhorst). Ozone was produced by passing zero air through an ozone generator using a low-pressure mercury ultraviolet lamp. The ozone concentration was measured using an ozone monitor (2B Technologies) ( Table S2 in the Supplement).
A proton-transfer-reaction mass spectrometer (PTR-8000, IONICON Analytik) was used to observe the resulting SCI q HFA adducts as well as the parent hydrocarbons and their oxidation products. This instrument was operated using H 3 O + reagent ions (H 3 O + CIMS) and was directly calibrated to 10 VOCs with different functional groups

Kinetics experiments with the spin trap TEMPO
Highly reactive spin traps are needed for the effective derivatization of radicals and reactive intermediates in the gas phase. A set of experiments in which the reaction rate coefficient between the spin trap TEMPO (C 9 H 18 NO) and OH was measured were conducted in a flow-tube experimental setup at Forschungszentrum Jülich. TEMPO is commonly used to detect carbon-centered radicals in chemical and biological systems (Bagryanskaya and Marque, 2014) and is known to be highly reactive towards OH in the aqueous phase (Samuni et al., 2002). TEMPO was introduced in the flowtube setup using a liquid calibration unit (LCU, IONICON Analytik). The LCU quantitatively evaporates aqueous standards into the gas stream. The TEMPO standard was prepared gravimetrically with an aqueous volume mixing ratio of 485 parts per million (ppm). A known amount (up to 10 µL min −1 ) of this solution was then evaporated into a humidified gas stream of synthetic air (31 SLPM), resulting in a gas-phase TEMPO concentration of up to 4.5 × 10 11 molecule cm −3 . Some of the setup outflow was drawn to a laser photolysis-laser-induced fluorescence (LP-LIF) instrument (Lou et al., 2010), with which the OH reactivity of TEMPO was measured. Laser flash photolysis of ozone was used to produce OH in the experimental setup, while LIF was applied to monitor the time-dependent OH decay. Another part of the outflow was drawn to a CIMS instrument (PTR3, IONICON Analytik) to monitor concentrations of TEMPO and its oxidation products. This instrument was operated in two ionization modes: using Breitenlechner et al., 2017) and NH + 4 q (H 2 O) n , n = 0-2 (as NH + 4 CIMS; Zaytsev et al., 2019) reagent ions. The PTR3 is designed to minimize inlet losses of sampled compounds. It was directly calibrated to 10 VOCs with different functional groups using LCU. Collision-dissociation methods were used to constrain the sensitivities of NH + 4 CIMS to compounds that cannot be calibrated directly (Zaytsev et al., 2019). Sensitivities were calculated in normalized duty-cycle-corrected counts per second per part per billion by volume (ndcps ppb −1 ; the duty-cycle correction was done to the reference m/z 100; ion signals were normalized to the primary ion signal of 10 6 dcps).

Ozonolysis experiments with the spin trap DMPO
An additional set of ozonolysis experiments with several hydrocarbons including TME and α-pinene were conducted in a double flow reactor setup (Fig. 2). The goal of these experiments was to examine how spin traps can be used for the simultaneous detection of stabilized Criegee intermediates and peroxy radicals. The experimental setup consisted of two identical ∼ 2.1 L flow reactors. The parent hydrocarbon was mixed with ozone in the first flow-tube reactor with a residence time of ∼ 28 s. Similar to the previous ozonolysis experiments described in Sect. 2.1, the parent hydrocarbon was vaporized from a flask filled with the pure substance by passing zero air regulated by a mass flow controller, and ozone was generated using a low-pressure mercury ultraviolet lamp. We used an LCU to introduce the spin trap DMPO (C 6 H 11 NO) in the second flow reactor with a residence time of ∼ 23 s. A known amount (up to 10 µL min −1 ) of the DMPO solution was evaporated into a humidified gas stream of synthetic air (5.4-7 SLPM), resulting in a gas-phase DMPO concentration of up to 1.1 × 10 13 molecule cm −3 . The second flow reactor was used to derivatize SCIs and RO 2 species with DMPO while the parent hydrocarbon was still reacting with ozone. Hence, we conducted integrated production measurements of the SCIs and RO 2 species formed in both flow reactors. DMPO represents a class of nonradical spin traps and is widely used to detect oxygen-centered radicals such as OH, HO 2 , and RO 2 in chemical and biological systems Van Der Zee at al., 1996). Recently, DMPO was also employed to detect SCIs in the gas phase (Giorio et al., 2017). The PTR3 was used to detect SCI q DMPO and RO 2 q DMPO adducts, while ozone levels were observed using an ozone monitor (2B Technologies) ( Table S3 in the Supplement).

Kinetic model and quantum-chemical calculations
The Framework for 0-D Atmospheric Modeling v3.1 (F0AM; Wolfe et al., 2016) containing reactions from the Master Chemical Mechanism (MCM v3.3.1) (Jenkin et al., 1997;Saunders et al., 2003) was used to simulate the photooxidation of studied alkenes in the flow reactor system and to compare the modeled concentrations of the products with the measurements. Model calculations were constrained using physical parameters of the experimental setup (pressure and temperature) as well as to observed concentrations of the parent hydrocarbon, ozone, and the chemical derivatization agent.
In order to estimate the proton affinities of the SCI q HFA adducts, we performed geometry optimization and proton affinity calculations with the Gaussian 09 package (Frisch et al., 2009) using the B3LYP functional (Stephens et al., 1994) and TZVP basis sets.
3 Results and discussion 3.1 Detection of speciated stabilized Criegee intermediates using chemical derivatization techniques The primary goal of the first set of experiments was the detection of speciated stabilized Criegee intermediates as adducts with the chemical derivatization agent HFA to prevent secondary reactions within the experimental setup. Starting with (CH 3 ) 2 COO, an SCI produced via the ozonolysis of TME, we tested the formation of SCI q HFA adducts under different experimental conditions (Fig. 3). (CH 3 ) 2 COO q HFA (C 6 H 6 O 3 F 6 q H + , m/z 241.03) can be easily identified in the mass spectrum due to its unique mass defect associated with six F atoms (Fig. S2 in the Supplement). SCI q HFA adducts were observed when TME, ozone, and HFA were present in the experimental setup. Ozonolysis of TME also results in the formation of acetone (C 3 H 6 O q H + , m/z 59.05), which was detected in the presence of TME and ozone and was not affected by HFA addition (Fig. 3). We measured the (CH 3 ) 2 COO q HFA adduct signal as a function of different reactant conditions: initial TME concentrations were in the range of (1.48-1.85) ×10 11 molecule cm −3 ; ozone, (6.77-108.2) ×10 12 molecule cm −3 ; HFA, (1.17-6.13) ×10 15 molecule cm −3 . The measurements are compared to the predictions of the kinetic model in Fig. 4. Concentrations of (CH 3 ) 2 COO species were calculated using the MCM with updated kinetics data from the literature (Newland et al., 2015;Chhantyal-Pun et al., 2016;Long et al., 2018). For more details see the Supplement.
In the presence of HFA, SCIs can react with HFA and form stable adducts: The reaction rate coefficient k 1 was not measured experimentally, and we used the k value for the CH 2 OO + HFA reaction: k 1 = 3 × 10 −11 cm 3 molecule −1 s −1 . It has been suggested that the reaction between HFA and acetone oxide may be slower compared to the CH 2 OO one (Murray et al., 1965;Taatjes et al., 2012), while k (CH 3 ) 2 COO + HFA = 2 × 10 −13 molecule cm −3 s −1 was used in previous studies . However, the concentration of HFA was 2 orders of magnitude higher than the concentrations of other chemical compounds, so, even at lower k values, the reaction with HFA remains the main chemical loss pathway for (CH 3 ) 2 COO (Fig. S3 in the Supplement). Observed concentrations of (CH 3 ) 2 COO q HFA agree to within a factor of 3 with concentrations predicted by the kinetic model (Fig. 4). This discrepancy can be explained by a combination of the following factors. First, a fraction of the (CH 3 ) 2 COO q HFA adducts might be irreversibly deposited on the surfaces inside the experimental setup and the PTR 8000 instrument (Pagonis et al., 2017). In addition, the sensitivity of the observed SCI q HFA adducts depends on the reaction rate constant of the adduct with H 3 O + ion and the degree of fragmentation of the protonated product ions SCI q HFA q H + (Yuan et al., 2017). Since the reaction rate constant of SCI q HFA with H 3 O + ions is unknown, we assumed that all SCI q HFA adducts were ionized via proton  . Correlation plot comparing measured and calculated concentrations of (CH 3 ) 2 COO q HFA. The adducts were detected using H 3 O + CIMS as (CH 3 ) 2 COO q HFA q H + (m/z 241.03). The slope was calculated using weighted least squares (WLS). The 95 % confidence interval was estimated via a Monte Carlo simulation (N = 5000) and is shown using red shading. transfer from hydronium ions and therefore used the sensitivity we obtained from acetone calibration to quantify the detected SCI q HFA species. In addition, we did not take into account possible fragmentation of SCI q HFA q H + ions, which may impede their detection, although a first bond cleavage would likely only break the ozonide ring structure without any loss of mass. Finally, the uncertainty in the kinetic model output is determined by the uncertainty in the SCI yield and the unimolecular and bimolecular reaction rate coefficients. The detection limit for (CH 3 ) 2 COO q HFA adducts was 1.4 × 10 7 molecule cm −3 and was calculated for an integration time of 30 s as 3 standard deviations of the measured background divided by the derived sensitivity.
Besides TME, we also observed the formation of SCI q HFA for a series of precursors including isoprene, pentene, and hexene (Figs. S4-S6 in the Supplement). Proton affinities (PAs) of different CI q HFA adducts were calculated using density-functional theory (DFT) methods (Table 1). A variety of these adducts can be detected using H 3 O + CIMS since their PAs are significantly higher than that of water, which is in agreement with experimental data (Figs. S4-S6). CH 2 OO q HFA cannot be detected because of its low PA (Table 1). We also did not observe SCI q HFA adducts for larger C 10 SCIs produced via ozonolysis of α-pinene and limonene. This can be explained by the lower reactivity of larger SCIs with HFA, potential instability of these secondary ozonides in the gas phase, or their gas-wall partitioning in tubing and inside the PTR-8000 instrument.

Reactivity of spin traps with OH
Spin traps have been shown to be highly reactive towards free radicals and to efficiently form adducts with them in the aqueous phase. However, their reactivity with atmospheric radicals and the stability of the formed adducts in the gas phase remain largely unknown. In order to address these questions, we conducted a set of experiments to estimate the reaction rate between the spin trap TEMPO and the hydroxyl radical by measuring its OH reactivity.  Figure 5. OH reactivity measured as a function of TEMPO concentration. The slope determining the reaction rate coefficient k TEMPO+OH = (9.3 ± 0.9) × 10 −11 cm 3 molecule −1 s −1 was calculated using weighted least squares (WLS). The 95 % confidence interval was estimated via a Monte Carlo simulation (N = 5000) and is shown using blue shading. The intercept (5.1 ± 24) s −1 can be explained by the presence of other OH reactants such as O 3 , NO, NO 2 , and CO.
The OH reactivity of a specific reactant can be calculated as a product of the reactant concentration and its reaction rate with OH (Fuchs et al., 2017): (1) k OH was measured as a function of TEMPO concentration by varying the amount of TEMPO introduced in the experimental setup using the LCU (Fig. 5). The slope of the fitted line in Fig. 5 determines the reaction rate coefficient k OH+TEMPO = (9.3±0.9)×10 −11 cm 3 molecule −1 s −1 . This rate constant is close to the collisional limit of typical radical-molecule reactions in the atmosphere and is 1 order of magnitude greater than the rate constant for the same reaction in the aqueous phase (k aqueous = 7.5×10 −12 cm 3 molecule −1 s −1 ; Samuni et al., 2002). This demonstrates that TEMPO is highly reactive towards OH in the gas phase, emphasizing the applicability of spin trapping for atmospheric measurements. Furthermore, the TEMPO + OH reaction leads to the formation of stable TEMPO q OH adducts that can be detected by H 3 O + CIMS (C 9 H 18 NO q H + , m/z 174.149) and therefore could be used for the quantification of hydroxyl radicals in the atmosphere (Fig. S7 in the Supplement). Further tests are needed to compare the measurement capability of this method (e.g., sensitivity, wall losses, and potential interferences) with that of a well-established technique such as LIF.

Simultaneous detection of SCIs and RO 2 species from the ozonolysis of alkenes using spin-trapping techniques
Next, we implemented spin trapping for the detection of speciated SCIs and RO 2 species formed via the ozonolysis of alkenes, starting with TME. Decomposition of the TME primary ozonide leads to the formation of acetone oxide, (CH 3 ) 2 COO. This SCI can further undergo a unimolecular reaction followed by O 2 addition to form the peroxy radical CH 3 C(=O)CH 2 OO q and OH (Fig. 1). In order to detect SCIs and RO 2 species produced via the ozonolysis of TME, we used a measurement method based on the stabilization of these species using the spin trap DMPO followed by detection via NH + 4 and H 3 O + CIMS. DMPO was shown to form stable secondary ozonides with SCIs in the gas phase (Giorio et al., 2017): DMPO has been shown to be highly reactive towards SCIs (k CI+DMPO ≥ 6×10 −11 cm 3 molecule −1 s −1 ; see the Supplement for more details). In addition, DMPO is known to be highly reactive towards oxygen-centered radicals such as RO 2 and to form stable radical adducts with them ( Fig. 1): We observed the formation of SCI q DMPO and RO 2 q DMPO adducts in both NH + 4 CIMS (e.g., C 9 H 17 NO 3 q NH + 4 , m/z 205.155 and C 9 H 16 NO 4 q NH + 4 , m/z 220.142) and H 3 O + CIMS (e.g., C 9 H 17 NO 3 q H + , m/z 188.128 and C 9 H 16 NO 4 q H + , m/z 203.116) under different experimental conditions (Fig. 6 and Fig. S8 in the Supplement). SCI q DMPO and RO 2 q DMPO were only detected when TME, ozone, and DMPO were present in the experimental setup. Acetone, also formed via the ozonolysis of TME, was observed in the presence of TME and ozone and was not affected by the addition of DMPO ( Fig. 6 and Fig. S8 in the Supplement). OH radicals formed via the decomposition of SCI can, in turn, react with TME, leading to the formation of another RO 2 species, OH−C 6 H 12 OO q . This radical was detected as the C 6 H 13 O 3 q DMPO adduct (C 12 H 24 NO 4 , m/z 264.205; Fig. 6). One of the benefits of NH + 4 CIMS is the possibility of quantifying compounds for which authentic standards are not available, using a voltage scanning procedure based on collision-induced dissociation (Zaytsev et al., 2019). Using this method, DMPO adducts with SCIs and RO 2 were detected at high sensitivities: 2400 ndcps ppbv −1 for SCI q DMPO and 2000 ndcps ppbv −1 for RO 2 q DMPO (Table S1 in the Supplement). Sensitivities were experimentally determined in each ozonolysis experiment and depended on the operational conditions of the PTR3 instrument. Detection limits for SCI q DMPO and RO 2 q DMPO adducts were 3.4×10 7 and 1.6×10 8 molecule cm −3 , respectively. These limits of detection were calculated for an integration time of 30 s as 3 standard deviations of the measured background divided by the derived sensitivity.
In addition, we compared the measured concentrations of RO 2 q DMPO adducts with the concentrations predicted by the kinetic model (Fig. 7). The observed values were an order of magnitude lower than the modeled ones. Similar to experiments described in Sect. 3.1, several factors can contribute to this discrepancy: (1) gas-wall partitioning of RO 2 species and RO 2 q DMPO adducts in the experimental setup and inside the PTR3 instrument; (2) uncertainty in the sensitivity at which RO 2 q DMPO adducts were detected; (3) potential fragmentation of RO 2 q DMPO q NH + 4 product ions; and (4) uncertainty in the reaction rate coefficient k RO 2 +DMPO . In our model, we assumed that the largest fraction of the RO 2 Figure 7. Correlation plot comparing measured and calculated concentrations of CH 3 C(=O)CH 2 OO q DMPO. The adducts were detected using NH + 4 CIMS as CH 3 C(=O)CH 2 OO q DMPO q NH + 4 (m/z 220.142). The slope was calculated using weighted least squares (WLS). The 95 % confidence interval was estimated via a Monte Carlo simulation (N = 5000) and is shown using yellow shading. species was scavenged by DMPO. This assumption is valid if k RO 2 +DMPO is larger than 1 × 10 −12 cm 3 molecule −1 s −1 . Otherwise, other loss channels for peroxy radicals, especially the RO 2 + RO 2 reaction, become more important (Fig. S9 in the Supplement). Additional experiments performed under different conditions and an intercomparison with established methods (i.e., LIF) are needed to further estimate the measurement capability of the proposed analytical method. Finally, we employed spin trapping for the detection of SCIs and organic peroxy radicals formed via the ozonolysis of larger cyclic alkenes, such as α-pinene. Decomposition of the α-pinene primary ozonide yields four different C 10 SCIs, all of which have the same molecular formula C 10 H 16 O 3 (Newland et al., 2018). These SCIs can further isomerize to form the primary peroxy radicals C 10 H 15 O 4 and OH. Autoxidation of C 10 H 15 O 4 −RO 2 species can, in turn, result in the formation of several more oxygenated peroxy radicals C 10 H 15 O x , x = 5-9 (Zhao et al., 2018). Signals of SCI q DMPO (C 10 H 16 O 3 q DMPO) and RO 2 q DMPO (C 10 H 15 O x q DMPO, x = 4-9) adducts were observed in both NH + 4 and H 3 O + CIMS ( Fig. 8 and Figs. S10 and S11 in the Supplement). OH radicals formed via the decomposition of SCI can in turn react with α-pinene, leading to the formation of the OH-derived RO 2 species C 10 H 17 O 3 and the subsequent autoxidation RO 2 species C 10 H 17 O 5 (Berndt et al., 2016). These radicals were detected as the RO 2 q DMPO adducts (Figs. S10 and S11 in the Supplement). This demonstrates that the analytical method allows for the simultaneous detection of a wide range of atmospheric radicals, including those with high oxygen contents (an O : C ratio of up to 0.9) that are formed via the autoxidation pathway, and can be used to study the kinetics of these species in the laboratory.

Conclusions
In summary, we experimentally demonstrated the measurement of speciated, short-lived, highly reactive atmospheric compounds such as Criegee intermediates, organic peroxy radicals, and hydroxyl radicals that are formed during the ozonolysis of alkenes. The analysis was carried out us-ing chemical derivatization techniques, including spin trapping, while the detection of the formed radical adducts and closed-shell secondary ozonides was achieved via H 3 O + and NH + 4 CIMS. The detected adducts and secondary ozonides have unique mass defects and can therefore be clearly separated from other compounds observed in the mass spectrum. Implementing chemical derivatization agents with low proton affinities allows the stabilized Criegee intermediates to be fully scavenged and quantified without depleting CIMS reagent ions. We show that spin traps can be used to effectively scavenge atmospheric radicals and reactive intermediates by demonstrating their high reactivity with radicals in the gas phase, using the TEMPO + OH reaction as an example. Using the spin trap DMPO, SCIs and RO 2 species can be simultaneously detected, while the observed adducts can be quantified without directly calibrating them. Detection limits of spin trap and chemical derivatization agent adducts of 1.4 × 10 7 molecule cm −3 for SCIs and 1.6 × 10 8 molecule cm −3 for RO 2 for an integration time of 30 s were estimated for the instrumentation used here. These limits suggest that the techniques would also work when sampling ambient air. In particular, this method enables any CIMS instrument to detect radicals and SCIs. Since spin traps such as DMPO and TEMPO are reactive towards a plethora of atmospheric radicals and reactive intermediates, including RO 2 , SCIs, and OH, the implementation of such spin traps results in the effective suppression of radical secondary chemistry and thus the elimination of potential chemical interferences. The direct method for speciated SCIs and RO 2 measurements provides a means to study the atmospheric chemistry of these compounds. We stress that the quantification of RO 2 species was done under well-defined laboratory conditions using the collision-induced dissociation (CID) technique, such that the estimated sensitivities are likely unique to the electric fields, pressures, and flows of the NH + 4 CIMS instrument. Further validation of the proposed analytical methods in more complex environments that are closer to the ambient conditions and an intercomparison with established methods (i.e., LIF) are needed.
For future applications of the method in field and laboratory experiments, various modifications of the experimental setup can be implemented to improve its measurement capability. We plan to synthesize and test new chemical derivatization agents that are optimized for gas-phase measurements in terms of their vapor pressures and selective reactivities and by labeling with atomic isotopes to simplify mass spectrometric detection and improve detection limits. With labeled spin traps, the identification of reactive intermediates may be greatly simplified and detection limits may be further improved, as the spin trap can provide a unique signature in the complex mass spectrum and move the observed m/z to a region with very low background.
Data availability. Data used within this work are available upon request. Please email Alexander Zaytsev (zaytsev@g.harvard.edu).
Author contributions. FNK initially conceived the work. AZ and MB developed the laboratory setups, performed the experiments, and analyzed the CIMS data. AZ performed the quantum-chemical calculations and wrote the manuscript. AN and HF provided data and analysis for the LP-LIF experiments. DAK supported the study with PTR-MS measurements. All authors were involved in helpful discussions and contributed to the manuscript.
Competing interests. The authors declare that they have no conflict of interest.
Acknowledgements. Martin Breitenlechner acknowledges support from the Austrian Science Fund. Hendrik Fuchs and Anna Novelli acknowledge support from the European Research Council (ERC). Daniel Knopf acknowledges support from the US National Science Foundation. Jesse Kroll acknowledges support from the US National Science Foundation.
Financial support. This research has been supported by the Harvard Global Institute (A New Strategy for the US and China: Joint Research on Air Pollution and Climate Using Innovative Airborne Instrumentation), the Austrian Science Fund (FWF (grant no. J-3900), the H2020 European Research Council (SARLEP (grant agreement no. 681529)), the National Science Foundation (grant nos. AGS-1446286 and AGS-1638672).
Review statement. This paper was edited by Keding Lu and reviewed by three anonymous referees.