NO2 Suppression of Autoxidation–Inhibition of Gas-Phase Highly Oxidized Dimer Product Formation

Atmospheric autoxidation of volatile organic compounds (VOC) leads to prompt formation of highly oxidized multifunctional compounds (HOM) that have been found crucial in forming ambient secondary organic aerosol (SOA). As a radical chain reaction mediated by oxidized peroxy (RO2) and alkoxy (RO) radical intermediates, the formation pathways can be intercepted by suitable reaction partners, preventing the production of the highest oxidized reaction products, and thus the formation of the most condensable material. Commonly, NO is expected to have a detrimental effect on RO2 chemistry, and thus on autoxidation, whereas the influence of NO2 is mostly neglected. Here it is shown by dedicated flow tube experiments, how high concentration of NO2 suppresses cyclohexene ozonolysis initiated autoxidation chain reaction. Importantly, the addition of NO2 ceases covalently bound dimer production, indicating their production involving acylperoxy radical (RC(O)OO•) intermediates. In related experiments NO was also shown to strongly suppress the highly oxidized product formation, but due to possibility for chain propagating reactions (as with RO2 and HO2 too), the suppression is not as absolute as with NO2. Furthermore, it is shown how NOx reactions with oxidized peroxy radicals lead into indistinguishable product compositions, complicating mass spectral assignments in any RO2 + NOx system. The present work was conducted with atmospheric pressure chemical ionization mass spectrometry (CIMS) as the detection method for the highly oxidized end-products and peroxy radical intermediates, under ambient conditions and at short few second reaction times. Specifically, the insight was gained by addition of a large amount of NO2 (and NO) to the oxidation system, upon which acylperoxy radicals reacted in RC(O)O2 + NO2 → RC(O)O2NO2 reaction to form peroxyacylnitrates, consequently shutting down the oxidation sequence.


Charging probabilities and transmission, and their influence on measured concentrations
NO 3 -ionization has been frequently applied in quantifying HOM product signals, with a calibration procedure utilizing SO 2 + OH (+H 2 O) derived H 2 SO 4 to determine the lower limit concentrations of the products observed.
This analysis relies on the assumption that HOM products experience the same collision limit charging in the chemical ionization inlet as does sulfuric acid. In addition, Ehn et al. 1 have further rationalized the measured concentrations by using further empirical constraints and arrived into good agreement with the single-calibrationfactor procedure. On the other hand, the detection efficiency of NO 3charging is known to depend on the available H-bond donating functional groups and on the number of O-atoms in the structure, both increasing the detection sensitivity and thus leading to a positive bias relative to the less oxidized compounds 2, 3 . The sensitivity is further S3 dependent on the detailed settings of the atmospheric pressure interface (APi) altering the ion transmission in the mass range of interest 4 . In considering these issues it becomes clear how difficult it is to adequately account for the compound specific detection sensitivity, and thus this was not attempted here, and instead all the concentrations are given as measured ion signals in ion counts per second (cps), and where relevant (i.e., to be able to compare signals), they are normalized to the total reagent ion current which was c.a. 10 000 to 20 000 cps.

Oxidation pathways
The product peaks detected in cyclohexene ozonolysis experiments without addition of NO x have been collected to Table S1.

S4
Due to cyclohexenes symmetric ring structure with no substituents around the double bond, it produces two aldehyde functionalities containing primary oxidation products as illustrated in Figure 1. In these compounds the aldehydic hydrogens have the lowest C-H bond energies, and thus further oxidation, autoxidation or stepwise, will inevitably lead into further acylperoxy radicals. Also, cyclohexene's carbon chain length enables close to optimal H-shift transition states 6 and facilitate the formation of a pool of secondary acylperoxy radicals. In Scheme S1 further examples for acylperoxy radical formation during the oxidation sequence have been gathered.
Scheme S1 Further potential acylperoxy formation pathways, and interconversions, during cyclohexene autoxidation. Acylperoxy radicals have been marked with red color.

Figure S2
Potential peroxyacylnitrate (PAN) product structures detected in this work. Acylperoxy radicals have been marked with red color.

Suppression reaction kinetics
The suppressing reaction RC(O)O 2 + NO 2 must happen faster than the unimolecular hydrogen shifts progressing the oxidation chain reaction. Previously it has been estimated that H-shift rates of above 0.  Figure 3, a similar analysis was illustrated in graphical format with general assumed reaction rate k(RC(O)O 2 +NO 2 ) between 10 -10 cm 3 s -1 and 10 -13 cm 3 s -1 depicted as a function of [NO 2 ].

NO's enhancing influence
In contrast to RO 2 + NO 2 reaction, RO 2 + NO reaction can also propagate the radical chain by reactive RO generation. In experiments where the NO's influence was investigated, an enhanced C 6 H 8 O 8 product formation was observed after NO was introduced into the gas flow, whereas larger oxidation products were strongly depleted ( Figure S3). The amplification of RO chemistry due to NO could lead to species with C5 and C11 structures, i.e., to compounds which have lost a CO or CO 2 during the oxidation sequence. Whereas a C5 compound was observed to increase with the NO addition (see Figure S3), all the dimer compounds were depleted as a function of [NO x ], and thus this dependence could not be observed for C11 compounds. Due to the cyclic nature of the oxidation chain reaction, yet further NO addition interfered the oxidation at an even earlier stage, and resulted also in reduction of the observed nitrate formation ( Figure S4).  Figure S4 even the observed nitrate signals were plummeting at highest NO or NO 2 additions.

S9
The addition of NO x during various instances of the oxidation sequence is illustrated in Figure S5, where example spectra obtained at several [NO 2 ] and [RO 2 ] combinations are shown. S10 S11 Figure  × 10 14 cm -3 ) at 5.9 second reaction time. Bg stands for a background peak.
With the highest NO x additions, the oxidation suppression was observed even earlier in the oxidation chain, which resulted in decrease of all the product species detected with the utilized nitrate ionization technique (See Figures   S4 and S5 above). In the experiments with very high RO 2 and NO x concentrations, it was impossible to even qualitatively quantify the extent of NO 2 suppression, as removing NO 2 from the gas flow resulted in strong S12 depletion of the reagent ion signals (indicating formation of a very large number of products), and thus rendered the CIMS concentration determination unreliable. However, when a sufficient NO 2 flow was added back to the reacting gas mixture, the HOM formation (except few PAN-HOM) practically ceased, and the reagent ion signals recovered (see Figure 2).  Table 1 and Figure S5]. However, product charged by the reagent ion dimer would be observed with a similar composition {i.e., m/z[C 6 H 9 O 6 NO 2 (*HNO 3 NO 3 -)] = m/z[C 6 H 10 O 6 NO 3 NO 2 (*NO 3 -)]} and seems far more likely, already as the NO 3 is not expected to live long in enough here to perturb the oxidation system. Also, from Figure S5 e it can be seen that when most of the compounds have been depleted due to high [NO 2 ] in the flow, the two peaks that are well resolvable in the spectrum, and have about similar intensity, could correspond to C 6 H 9 O 6 NO 2 charged by NO 3and HNO 3 NO 3 -.
Also in other experiments, the apparently two nitrogen containing compounds were coincident with the highest PAN signals and could be explained with these PAN species clustering with the reagent dimer; similar observations of potential dimer charging were absent from experiments conducted without NO x . Finally, it seems that the nitrate functionality containing products could potentially cluster more efficiently with NO 3 -, and thus potentially also by HNO 3 NO 3 -, as the PAN signals seem to increase somewhat more than the HOM signals decrease, indicating a potential positive bias in their detection.