Formation of highly oxygenated organic molecules from chlorine-atom-initiated oxidation of alpha-pinene

Highly oxygenated organic molecules (HOMs) from atmospheric oxidation of alpha-pinene can irreversibly condense to particles and contribute to secondary organic aerosol (SOA) formation. Recently, the formation of nitryl chloride (ClNO2) from heterogeneous reactions, followed by its subsequent photolysis, is suggested to be an important source of chlorine atoms in many parts of the atmosphere. However, the oxidation of monoterpenes such as alpha-pinene by chlorine atoms has received very little attention, and the ability of this reaction to form HOMs is completely unstudied. Here, chamber experiments were conducted with alpha-pinene and chlorine under lowand high-nitrogen-oxide (NOx , NOx = NO+NO2) conditions. A nitrate-based CI-APi-ToF (chemical ionization–atmospheric pressure interface–time of flight) mass spectrometer was used to measure HOM products. Clear distributions of monomers with 9–10 carbon atoms and dimers with 18– 20 carbon atoms were observed under low-NOx conditions. With increased concentration of NOx within the chamber, the formation of dimers was suppressed due to the reactions of peroxy radicals with NO. We estimated the HOM yields from chlorine-initiated oxidation of alpha-pinene under low-NOx conditions to be around 1.8 %, though with a substantial uncertainty range (0.8 %–4 %) due to lack of suitable calibration methods. Corresponding yields at high NOx could not be determined because of concurrent ozonolysis reactions. Our study demonstrates that also the oxidation of alpha-pinene by chlorine atoms and yield low-volatility organic compounds.


Introduction
Highly oxygenated organic molecules (HOMs) have been identified as key species in the formation of new atmospheric aerosol particles and secondary organic aerosol (SOA; Ehn et al., , 2017Kulmala et al., 2013;Bianchi et al., 2019). Recently, the formation of HOMs in the gas phase was described as an autoxidation process of peroxy radicals (RO 2 ) via multiple intramolecular H atom shifts (Crounse et al., 2013;Jokinen et al., 2014;Mentel et al., 2015;. Oxygen-containing moieties such as carbonyl, carboxylic acid and hydroxyl groups can weaken nearby C−H bonds, making H abstraction and autoxidation competitive with bimolecular RO 2 reactions, e.g. with NO (Crounse et al., 2013;Praske et al., 2018). Until now, all studies on the formation of HOMs have focused on reactions initiated by oxygen-containing oxidants (O 3 and OH; Berndt et al., 2016;Ehn et al., 2014;Jokinen et al., 2015).
Increasing evidence indicates that the chlorine atom (Cl) may also play an important role in transforming atmospheric organics (Tham et al., 2016;Thornton et al., 2010). Chlorine atoms have the greatest reactivity toward volatile organic compounds (VOCs), with rate constants that are, with some exceptions, an order of magnitude higher than those of hydroxyl radicals (OH; Riva et al., 2015). Historically, chlorine atoms were thought to be formed primarily from heterogeneous reaction cycles involving sea salt, and their concentrations estimated to be around 1 %-10 % of that of OH (Thornton et al., 2010). Therefore, the role of chlorine atoms in atmospheric oxidation processes has traditionally been thought to be limited to the marine boundary layer only. In recent years, ClNO 2 , as a significant chlorine atom source, was found in continental regions of America, Canada and Germany, and high concentrations of ClNO 2 were also detected in the urban atmosphere in China (Reyes-Villegas et al., 2018;Tham et al., 2016;Thornton et al., 2010;. The new findings have expanded the potential importance of chlorine atoms from coastal areas to continental urban areas. A recent study also reported that chlorine atoms can be more important than OH radicals for the oxidation of alkanes in the North China Plain (Liu et al., 2017). Therefore, it is desirable to probe the role of chlorine radicals in the degradation of VOCs and related SOA formation.
Emission of biogenic volatile organic compounds (BVOCs) to the atmosphere dominates total hydrocarbon emissions on a global scale, with methane, isoprene and terpenes having the highest source strengths (Guenther et al., 1995). Alpha-pinene is the most abundant monoterpene in the atmosphere, and its oxidation products from ozonolysis and photooxidation contribute to a substantial fraction of SOA mass (Riccobono et al., 2014;Zhang et al., 2018). Many of the regions listed earlier, where ClNO 2 was identified as a chlorine atom source, may also have substantial monoterpene emissions, making Cl a relevant oxidant also for BVOC. Chlorine-atom-initiated reactions of alpha-pinene have also been shown to contribute to the formation of SOA, which implies that low-volatility compounds are efficiently produced also in this process (Cai and Griffin, 2006;Ofner et al., 2013).
Similar to the reaction with OH radicals, the reaction of VOCs with chlorine atoms may proceed either via addition of Cl to unsaturated bonds or via H abstraction.  found that the Cl addition to isoprene can lead to the formation of low-volatility organic compounds. In principle, Cl-initiated reactions could form HOMs in a similar manner to OH-initiated reactions (Berndt et al., 2016), as the initial addition or abstraction step is comparable for both oxidants. In view of the increased understanding of the importance of chlorine atoms in atmospheric chemistry, it is desirable to investigate the formation of HOMs from reactions of common atmospheric VOC with Cl.
Here, laboratory chamber experiments were performed to investigate the ability of the chlorine atom to form HOMs from the oxidation of alpha-pinene. HOMs were characterized using a nitrate-based chemical ionization mass spectrometer, under both low-and high-NO x conditions. The yields of these HOMs were determined under the low-NO x conditions, and the atmospheric implications of this study are discussed. 2 Experiment and method

Experimental set-up
The experiments were conducted in the "COALA" chamber at the University of Helsinki (Peräkylä et al., 2020;Riva et al., 2019b). This is a 2 m 3 Teflon chamber, run as a continuously stirred tank reactor, used with a flow of 45 L min −1 (L min −1 ), resulting in an average residence time of about 45 min. The chamber is surrounded by housing to provide dark conditions. No water vapour was added to the chamber, resulting in an RH < 1 %, and the temperature was the same as the temperature of the room, around 25 • C. The HOM formation targeted in this work is not expected to change markedly as a function of RH, as also indicated in a previous study on both ozone and OH-initiated oxidation of monoterpenes (Li et al., 2019). Although SOA mass yields were not studied in this work, these may be affected by RH (Jonsson et al., 2006(Jonsson et al., , 2008. A general schematic of the chamber facility is shown in Fig. 1. Our experiment was aimed to probe chlorine-atom-initiated formation of HOMs under low-and high-NO x conditions. We use 400 nm LED lights to photolyse chlorine and NO 2 and produce chlorine atoms and NO as follows: The concentration of Cl atoms was varied by changing the amount of 400 nm light. In practice this was done by turning on different numbers of the available lights, with the maximum corresponding to seven. We will refer below to the number of lights that were turned on, although each "light" corresponds to a group of LED strips.

Instrumentation and data analysis
A nitrate-ion-based (NO − 3 -based) chemical ionizationatmospheric pressure interface-time-of-flight (CI-APi-ToF) mass spectrometer was used for measuring HOMs. The instrument has been shown to be sensitive towards this group of compounds, detecting them as adducts with the nitrate ion. Due to a lack of suitable calibration methods, the CI-APi-ToF was not calibrated for HOMs during this study. In order to estimate rough HOM concentrations ([HOM]), we directly use the calibration coefficient C = 1.6 × 10 10 molec. cm −3 utilized by Ehn et al. (2014) to convert the measured HOM ion signals according to the equation below : This value of C is very close to values utilized in several other studies using a CI-APi-ToF (Jokinen et al., , 2015Riva et al., 2019a). Ehn et al. (2014) obtained their calibration factor from a permeation source using a perfluorinated acid and also showed through calculations that there were theoretical limitations for having a much larger or a much smaller value for C. This approach obviously brings large uncertainties to the HOM concentrations, which we estimate to be at least −50 % ± 100 % according to previous calibration results (Jokinen et al., , 2015Riva et al., 2019). More detailed information about the instrument can be found in Jokinen et al. (2012), noting that compared with the CI-APi-ToF used before, the long time-of-flight mass spectrometer used here has a doubled mass-resolving power enabling a more accurate assignment of molecular formulas. Simultaneously, we also used a high-resolution long time-of-flight aerosol mass spectrometer (HR-L-ToF-AMS) to measure bulk aerosol chemical properties (Decarlo et al., 2006). As no seed aerosol particles were added to the chamber, the VOC oxidation products lead to new particle formation and growth to large enough sizes to be measured by the AMS. We also periodically used a filter in front of the AMS inlet to see the influence of the background signal on measured aerosol mass concentration. There were 10 min of filter measurements per hour during our experiments. A PTR-ToF-MS (TOF-8000, Ionicon) was used to measure the concentration of alpha-pinene in the chamber. The instrument background was determined every day for 20 min by guiding the chamber air through a catalytic converter, which removes the VOCs. Then, the background-corrected signals were used to obtain alpha-pinene mixing ratios by using the calibration coefficient determined before the experiments. A description of the used set-up employed for the calibration and zero-air measurements has been introduced earlier (Schallhart et al., 2018). A custom-built DMPS system was used to measure the particle number size distribution from 10 to 400 nm in the chamber. The NO concentration was measured with an ECO-PHYSICS CLD 780 TR instrument with a detection limit of 3 ppt. NO x (= NO+NO 2 ) concentrations were determined by using a Thermo Fisher 42i analyser. O 3 concentration was measured with a Thermo Fisher 49i analyser.

Estimation of chlorine atom concentrations
During steady state in the chamber, average concentrations of the chlorine atom were calculated using the rate coefficients (4.6 ± 1.3) × 10 −10 cm 3 molecule −1 s −1 of Cl atoms with alpha-pinene (Finlayson-Pitts et al., 1999), as follows: where Q in is the flow rate of alpha-pinene continuously injected into the chamber, and Q out is the flow rate that exited the chamber. The term-Q in -related concentration was 13.3 ppb, while the term Q out varied depending on the conditions and is calculated as [AP] / 45 min. During steady state, d[AP]/dt is zero, and then [Cl] concentration is calculated accordingly. As shown in Fig. 2, the concentration of HOMs decreased and alpha-pinene increased as the number of lights switched on changed from seven to four, two and one. We use the variation in alpha-pinene and HOM concentrations during this run to calculate both chlorine atom concentrations and HOM yields. Each change in alpha-pinene concentration was due to the change in Cl atom concentration, and with knowledge of the reaction rate, the concentration of Cl atoms as a function of the number of lights turned on was determined (Fig. 3). The calculated [Cl] concentrations are in the range of (1 − 5) × 10 5 molecules cm −3 , which is within atmospherically relevant concentration ranges (Tham et al., 2016). Raw data from the CI-APi-ToF were recorded in 10 s resolution in HDF format. We used tofTools for data analysis, and detailed protocols of the software have been introduced by Junninen et al. (2010).

HOM molar yield
The change of HOM concentration with time can be described as follows, analogous to Ehn et al. (2014): .
Here, k 1 is the reaction rate coefficient of alpha-pinene with chlorine atoms and γ is the molar yield of HOMs, i.e., the fraction of alpha-pinene + Cl reactions that produced HOMs. k loss is the loss rate of HOMs to the chamber walls and particles, though the latter was negligible in this study due to the low aerosol loadings. We used 300 s as a lifetime of HOMs, i.e. k loss = 1/300 s −1 , in our previous study in the COALA chamber (Riva et al., 2019).    Jokinen et al., 2015;Mentel et al., 2015). The C 20 H 30 O n dimers are most likely formed from reactions of two C 10 H 15 O x radicals, as were many abundant monomers. As noted earlier, Cl oxidation of alkenes may occur via a Cl addition (forming an initial radical containing 16 H atoms and 1 Cl atom) or via an Habstraction reaction (forming a radical with 15 H atoms and no Cl; Fig. 5). The abstraction pathway either leads to HOM formation or the Cl atom is lost during the subsequent reaction in the oxidation processes. With our data, we cannot rule out either of these explanations for this result. Loss of HCl from alpha-pinene products from Cl oxidation has, to our knowledge, only been reported to take place in the aerosol phase (Ofner et al., 2013). Figure 6 shows the variation in several close-shelled HOM products and the peroxy radical C 10 H 15 O 10 measured by NO 3 CI-APi-ToF when we changed the lights from dark conditions to one, two, four and seven lights switched on. Given the low Cl atom concentration, it is expected that no multi-generation oxidation by Cl can take place, and the behaviour of all closed-shell oxidation products should follow similar patterns. As seen in Fig. 6, this was the case both for monomers and dimers. The less steep increase in the radical is also according to expectations, as the formation of RO 2 is linear with the alpha-pinene oxidation rate, but the loss rate (when dominated by RO 2 cross reactions) is proportional to (RO 2 ) 2 . For closed-shell species, the wall loss-driven loss Figure 4. The mass spectra obtained by NO 3 CI-APi-ToF during steady state with one (a), two (b), four (c) and seven (d) lights. All peaks are detected as clusters with NO − 3 . The spectra are plotted as unit mass resolution, with background signals removed, but the peak identifications (labelled in panel d) are based on high-resolution analyses. The spectra correspond to the same four steady-state conditions depicted in Fig. 2. Figure 5. Proposed chemical pathways for chlorine radical oxidation of alpha-pinene and subsequent autoxidation and HOM formation. The upper path shows the chlorine radical addition pathway, while the lower chain shows hydrogen atom abstraction pathway. In both cases, initially a C-centred radical forms (C 10 H 16 Cl or C 10 H 15 ) to which O 2 is added to form an initial peroxy radical. This peroxy radical may then undergo multi-step autoxidation to reach the example molecules C 10 H 16 ClO 10 or C 10 H 15 O 10 before termination. Figure 6. Time series of selected closed-shell HOM monomers, dimers and two RO 2 radicals (C 10 H 15 O 10 and C 10 H 15 O 12 ) detected by NO 3 -CI-APi-ToF as the lights increased from zero to one, two, four and seven. ncps stands for normalized counts per second. rate stays constant throughout the experiment, and therefore they increase linearly with the alpha-pinene oxidation rate while the RO 2 radicals increase as the square root of the oxidation rate. For more detailed discussion on RO 2 dynamics in a steady-state chamber, see Ehn et al. (2014).
In Fig. 7, we plotted time series of the particle number size distribution and the total number concentration together with mass concentrations of particulate chloride and organics as we changed the number of lights. Particle formation was detected even at the lowest Cl atom concentration, as indicated by the increases in aerosol number concentration. An increase in aerosol mass concentration as detected by the AMS only took place at the two highest Cl atom concentrations, when the particles were able to grow into a size range measurable by the AMS. Particulate chloride mass concentrations also increased relatively linearly with the concentration of organics as we increased the number of lights. The Chl / Org ratio was only around 3 %, suggesting that the majority of condensed OVOCs (oxygenated volatile organic compounds) did not contain Cl atoms. However, the exact Chl quantification from organochlorides using the AMS may contain uncertainties (Wang and Ruiz, 2017), and we avoid drawing conclusions that are too far-reaching from this value, keeping also in mind that some fraction of the size distribution was below the lowest detectable size of our AMS. In addition, some part of the chloride signal may also result from adsorption of HCl to particles.

Formation of HOMs at high NO x
Anthropogenic emissions have a significant influence on the formation of SOA, in large part due to the influence of NO x on the atmospheric oxidation chemistry (Lee et al., 2016). In general, the fate of peroxy radicals in chamber experiments can be dominated by reactions with other RO 2 , HO 2 or NO, depending on the exact conditions. In our experi-ments without NO x addition, RO 2 was expected to be the main terminator, as also supported by the high number of detected ROOR dimers. In the atmosphere, all of the three mentioned reaction partners may be relevant at the same time. However, with increased anthropogenic influence, the reaction of RO 2 with NO will often become dominant. Therefore, we added NO x to the chamber, as it allowed for the isolation of the formation pathways leading to HOMs in cases where NO was the main terminator for RO 2 radicals. Figure 8 depicts a HOM mass spectrum at steady state during alpha-pinene oxidation by chlorine radicals in the presence of ∼ 10 ppb NO x , with the maximum seven lights turned on. As anticipated, the dimers above 440 Th were greatly reduced compared to the runs without NO x . As more lights were turned on, both the Cl atoms and NO formation increased, as the 400 nm lights photolyse both Cl 2 and NO 2 . This coupling, together with the fact that the NO 2 photolysis leads to ozone formation, which subsequently can react with alpha-pinene to form HOMs, limits our quantitative analysis of these experiments. However, we conclude that efficient HOM formation took place also under these high-NO x conditions, and thus the autoxidation occurs rapidly enough to still compete with RO 2 termination reactions. The NO x addition also formed an abundance of organonitrate compounds like C 10 H 15 NO 8,9,10,11,12 , as shown in Fig. 8. This family of compounds may also form from H abstraction by the chlorine radical, followed by autoxidation and finally radical termination by NO. We calculated reacted alpha-pinene from the ozone and chlorine atom as follows: The concurrent formation of ozone means that also some alpha-pinene ozonolysis reaction will take place, though oxidation by Cl atoms was still the main loss for alpha-pinene also under these conditions. Figure 9 shows variation in some nitrogen-containing HOMs and variation in alpha-pinene, ozone, NO and NO x , as we changed the lights from dark conditions to one, two, four or seven lights switched on. The concentrations of alphapinene and NO 2 decreased because of the consumption by chlorine radicals and photolysis of NO 2 into NO. Importantly, we did not observe any SOA when we had NO in the chamber. NO may have suppressed the particle formation by suppressing the dimer formation, as these have been shown to be important for initial particle formation (Tröstl et al., 2016).

Estimated HOM production yields
Quantifying the molar yields of HOMs is essential to knowing their potential importance from a specific system. We attempt to estimate the molar yield in the case of Cl oxidation of alpha-pinene in the absence of NO x . The initial  C 10 H 16 concentration is around 13.3 ppb without any UV lights switched on in the chamber. As we changed the lights, alpha-pinene and HOM concentrations varied, as shown in Fig. 3. In addition, we calculated the concentration of Cl radicals as introduced in the "Experiment and method" section. With this information, we can calculate the formation rate of HOM, which in the steady state equals the HOM loss rate [HOM] × k loss . We can also calculate the oxidation rate of alpha-pinene as [α-pinene]×[Cl]×k AP+Cl . The ratio of these two numbers corresponds to the HOM molar yield. We selected the same runs as in Fig. 3, used also for calculating the chlorine radical concentration, and calculated the ratio as a linear fit to these four conditions (Fig. 10). We get a slope of 0.018, meaning a HOM yield of 1.8 %. Considering Figure 9. Time series of (a) selected HOMs measured by NO 3 CI-APi-ToF and (b) NO x , alpha-pinene, ozone and NO as the lights switched on from zero to two, four and seven. Figure 10. HOM loss rate ([HOM] × Kloss) as a function of the alpha-pinene oxidation rate. In steady state the loss rate equals the formation rate, and thus the slope of the points gives the molar yield of HOM from the alpha-pinene + Cl reaction. The data correspond to the conditions with one, two, four and seven lights switched on in the chamber. The slope indicates an average molar yield of HOMs of 1.8 %.
the uncertainty in estimating absolute HOM concentrations, we conservatively estimate that the molar HOM yield from alpha-pinene + Cl is within the range of 0.8 %-4 %. These values are similar to HOM yields reported for alpha-pinene oxidation by ozone and OH (Berndt et al., 2016;Ehn et al., 2014).

Conclusion
We systematically explored the reactions of alpha-pinene with chlorine atoms in a simulation smog chamber under atmospherically relevant conditions. We measured substantial numbers of highly oxidized organic molecules (HOMs) with a NO 3 CI-APi-ToF. With increasing UV lights, and consequently higher chlorine radical concentrations, the concentrations of both HOM and secondary organic aerosol increased. With addition of NO x , HOM monomer formation was still efficient, but the particle formation decreased greatly. We estimated HOM molar yields of around 1.8 % (0.8 %-4 %) from the reaction of alpha-pinene with Cl atoms.
Our study thus indicates that in regions where chlorine atom oxidation is of importance, its possible reactions with monoterpenes can be an important source of HOM and, consequently, SOA.
Data availability. The readers can access the data by contacting Yonghong Wang (yonghong.wang@helisnkifi).