Secondary Organic Aerosol Formation from the Oxidation of Decamethylcyclopentasiloxane at Atmospherically Relevant OH Concentrations

Decamethylcyclopentasiloxane (D5, C10H30O5Si5) is measured at ppt levels outdoors and ppb levels indoors. Primarily used in personal care products, its outdoor concentration is correlated to population density. Since understanding the aerosol formation potential of volatile chemical products is critical to understanding particulate matter in urban areas, the secondary organic aerosol yield of D5 was studied under a range of OH concentrations, OH exposures, NOx concentrations, and temperatures. The secondary organic aerosol (SOA) yield from the oxidation of D5 is extremely dependent on the OH 5 concentration, and differing measurements of the SOA yield from the literature are resolved in this study. Here, we compare experimental results from environmental chambers and flow tube reactors. Generally, there are high SOA yields (> 68%) at OH mixing ratios of 5×109 molec cm−3. At atmospherically relevant OH concentrations, the SOA yield is largely <5% and usually ∼1%. This is significantly lower than SOA yields used in emission and particulate matter inventories and demonstrates the necessity of OH concentrations similar to the ambient environment when extrapolating SOA yield data to the outdoor 10

the same generator to create higher concentrations of O 3 . The 254 nm lights photolyze O 3 to form O( 1 D), which reacts with H 2 O to form 2OH. After conditions were changed in the CPOT, no results were collected for at least 2 h. Data were averaged over between 1 and 11 h. D5 was injected through a syringe pump (Harvard Apparatus).
For all experiments, the concentration of D5 was measured with an HP 6890N gas chromatograph with a flame ionization detector (GC-FID) and a DB-5 column. Prior to the beginning of oxidation for the chamber experiments, all contents of the 70 reactor were left to sit for 4 h (2.8 h for Experiment 7) and the initial concentration of D5 was taken as the mean concentration during this time. For the CPOT experiments, the initial concentration of D5 was calculated by measuring the outlet flow with lights off, no water source, and the absence of O 3 . For Experiment 9, the change in D5 was sufficiently small that it was within the uncertainty. For calculating the SOA yield for this experiment, we used the OH exposure calculated from the change in SO 2 concentration to find the change in D5 (7 ppb). 75 To calibrate the GC-FID, a small Teflon bag was filled with 35 ppm of D5 and later diluted to 9 ppm. This bag was sampled using the GC-FID and the concentration was verified with a Fourier transform infrared absorption (FT-IR) spectrometer with a 19 cm path length and absorption cross sections from the Pacific Northwest National Laboratory (PNNL) database. To minimize vapor-wall-loss to the FT-IR enclosure, multiple samples were taken until a consistent spectrum was achieved.
Gas-phase oxidation products were evaluated with a CF 3 O − chemical ionization mass spectrometer (CIMS) equipped with a 80 Varian 1200 triple quadrupole mass analyzer. Concentrations of NO and NO 2 were measured with a Teledyne Nitrogen Oxide Analyzer (Model T200) and O 3 was found with a Horiba Ambient Monitor. Temperature and humidity were determined using a Vaisala HMM211 probe. Aerosol volume was measured by a custom-built scanning mobility particle sizer (SMPS) with a 3081 TSI Differential Mobility Analyzer (DMA) and a TSI 3010 butanol condensation particle counter (CPC). The sheath flow rate was 2.64 Lpm 85 and the aerosol flow rate from the chamber was 0.515 Lpm. A voltage scan from 15 to 9875 V was performed in 240 s every 330 s. Aerosol from the chamber flowed through an x-ray source to provide a known charge distribution, and the size distributions were determined using the data inversion method described by Mai et al. (2018). Experiment 2 required a logarithmic fit to the largest particles present, as described in Charan et al. (2020), which is the source of the higher SOA yield uncertainty than in the other experiments (see Table 1). Conversions to mass concentration were performed by assuming that the aerosol density 90 was 1.1 g cm −3 . This was estimated from data in Wu and Johnston (2017), which measured D5 secondary organic aerosol particles (using information in their Fig. S1a and Table S1).
Uncertainty estimates for all the instruments used in this study were determined as described in Charan et al. (2020). For the chamber experiments, particle-wall-deposition corrections were performed by calculating a diameter-independent first-order exponential fit (β =1-7×10 −4 min −1 ) to the particle volume concentration during the 3 h prior to the onset of oxidation and 95 applying that correction to the rest of the experiment. This method was chosen because it aligns with a diameter-dependent fit as determined using the method in Charan et al. (2018) but is simpler and because, for the chamber experiments, minimal organic aerosol formed and so the particle diameters changed insignificantly throughout the duration of the experiment. For Experiment 7, in which no initial aerosol was present, no aerosol was generated throughout the experiment and so no correction was necessary to determine an apparent SOA yield of 0.

100
For the CPOT experiments, an upper estimate of the wall-deposition-corrected SOA mass was calculated with the inverse of the particle-size-dependent penetration efficiency of the flow-tube component of the reactor (data from Fig. 9d in Huang et al. 2017). Since particles nucleated in the CPOT, the penetration efficiency of only the flow-tube component (and not the static mixer prior to the region of reaction) was used. The penetration efficiency, however, is based on the entire flow tube and nucleated particles may not form immediately at the beginning of the flow-tube component; thus, the wall-deposition correction 105 performed here is an upper bound of the correction. Note that this correction also neglects particle growth throughout the reactor and any particle-particle coagulation.
, where ∆SOA corr is the wall-deposition-corrected change in the aerosol mass concentration and ∆D5 is the mass concentration of reacted D5. Calculations were performed as described by Charan et al. (2020) and with the assumption that a particle, once deposited on the reactor wall, no longer acts as a condensation sink (Trump 110 et al., 2016). Note that since so little aerosol was formed during the chamber experiments, this assumption had a negligible effect on the chamber results. For the CPOT experiments, any deviation from this assumption would have prevented the data from reaching steady-state.
While the vapor-wall-deposition lifetime of D5 to the chamber walls was estimated to be on the order of weeks, the propensity of vapor-wall-deposition of the oxidation products is not extensively investigated in this study. Even at high initial seed 115 surface area concentrations, the SOA yield is still quite small (see Fig. S1). Alton and Browne (2020) estimated that, for their unseeded ∼1 m 3 FEP Teflon chamber, 5% of the ester product of D5 oxidation might partition to the chamber walls during the reaction. The volume of the chamber used in this study is 19 m 3 and seed aerosol is introduced prior to the experiment (except for Experiment 7, which was performed in the absence of seed aerosol). Even if 5% of the oxidation products were lost to the chamber walls, the SOA yields would still be within the reported uncertainty and sufficiently small so as not to affect any 120 conclusions. The CPOT reactor is operated at steady-state and, therefore, any oxidation products that are in equilibrium with the bulk flow (i.e., not lost permanently to the quartz walls) do not need a vapor-wall-deposition correction. For chamber experiments that employed H 2 O 2 , the OH concentration was calculated by fitting the gas-phase D5 concentration to a first-order exponential, fixing the initial point of the fit as the initial D5 concentration (fits had R 2 > 0.75), and using the value for the reaction rate of OH with D5, k = 2.1 ± 0.1 × 10 −12 cm 3 molec −1 s −1 , which was measured using the relative 125 rate method at 297 ± 3 K (Alton and Browne, 2020). Note that other experimental evaluations of the reaction rate of OH with D5 that use the relative rates method vary by less than a factor of 2 (the reasons for this difference are not known), which would not affect the order of magnitude of the OH concentration estimate (Kim and Xu, 2017;Safron et al., 2015;Xiao et al., 2015).
OH is the major loss source in the atmosphere and, we expect, in these experiments: losses to O 3 , NO 3 , and Cl are all negligible (Atkinson, 1991;Alton and Browne, 2020). The ozone concentration did not affect the SOA yield results: Experiments 7 and 130 9, which were performed at substantially different O 3 concentrations, still gave similar results for the SOA yield (0±0.1% and 0.8±0.8% with an upper wall-deposition-corrected bound of 1.4%, respectively). For Experiment 8, in which CH 3 NO 2 served as the OH source, the sharp decrease in the D5 mixing ratio immediately after the commencement of radiation, followed by a gradual decrease in its concentration, indicates that two OH concentrations are relevant for this experiment. Since the D5 concentration is measured every ∼21 min, and the pulse with high OH concentrations occurs within the first 30 min of oxidation, 135 the initial OH concentration is estimated with a two-point first-order exponential fit to the initial concentration and the first data point (12.3 min into radiation). The second OH concentration is estimated with a first-order exponential fit of the second point (33.3 min into radiation) to the end of the experiment.
OH exposure was calculated, for chamber experiments (Experiments 1-8) and experiments from Wu and Johnston (2017), where k OH+D5 = 2.1 × 10 −12 cm 3 molec −1 s −1 (Alton and Browne, 2020). For the CPOT experiments (Experiments 9-19), OH exposure was calculated as where k OH+SO2 = 9 × 10 −13 cm 3 molec −1 s −1 for an identical setup with SO 2 instead of D5 (Janechek et al., 2019). SOA yields and experimental conditions are given in Table 1 with estimated uncertainties. These SOA yields vary from 0 to 79% (114% at the upper bound of the wall-deposition-corrected value), an even wider range than that reported by the literature of 8-50% (Janechek et al., 2019;Wu and Johnston, 2017). Between the experiments performed here and those in the literature, the OH concentrations and OH exposures vary widely.
Determining which of these is the relevant parameter is critical to extrapolating the SOA yield data to the atmosphere: envi-160 ronmentally relevant OH concentrations are on the order of 10 6 molec cm −3 , but since D5 is primarily lost to OH and has a half life of 3.5-7 days, OH exposures on the order of 10 12 molec s cm −3 are also relevant. Due to experimental limitations, in particular an inability to perform experiments for multiple days without diluting the sample and otherwise changing the conditions, these two variables are often correlated.
Nonetheless, differentiating the effects of these two variables is possible. If a chemical process occurs in which the reaction 165 of D5 and OH forms an intermediate or a second-generation product that then either reacts with OH or fragments, then the competition between the two outcomes is moderated by the relative time required for self-reaction or reaction with OH. This means that, as the OH concentration increases, the OH-reaction product will predominate. If this is the chemistry that D5 undergoes, then we would expect the SOA yield to depend solely on the OH concentration and not on the OH exposure. Figure 1a shows the relationship between the measured OH concentrations and SOA yields for the experiments performed 170 here as well as those from the literature. There is very good agreement between the chamber and CPOT experiments for similar OH concentration (shown in purple and orange, respectively). Moreover, the sharp increase in measured Y starting at [OH] ≈ 10 9 molec cm −3 matches the hypothesis that there is a competitive process moderated by OH concentration. measured SOA yields were uniformly higher in experiments that were initiated with ammonium sulfate seed than those that were not (see Fig. 3b). We, therefore, show the seeded and unseeded experiments in Fig. 1 in blue and red, respectively. OH concentration in the experiments reported by Wu and Johnston (2017) were calculated by replacing the precursor with SO 2 , measuring the formation of aerosol, and assuming that all the SO 2 reacts with OH to form H 2 SO 4 and all the sulfuric acid forms aerosol with minimal wall loss (Hall et al., 2013). Because of the uncertainties present for each step of this measurement, it seems reasonable that this [OH] estimate could be too low by at least a factor of 2.
Other instrumental and analysis uncertainties might close the gap between the OH concentrations measured by Wu and 185 Johnston (2017) (2017) is a rectangular bag, regions will exist with differing OH concentrations. If this reactor has slightly higher concentrations in some points or its residence time is overestimated or if the residence time for CPOT is a slight underestimate (we calculated an 190 uncertainty of ∼2%), this could account for the remaining disagreement between the data from the two experimental setups.
Furthermore, differences in the analysis could change the relevant SOA yields calculated. For Experiments 10-19, we measured both the initial and the final D5 concentration and for Experiments 1-8 we continuously measured the concentration.
Wu and Johnston (2017)  Y because they would have assumed less D5 reacted than in actuality. To achieve agreement to experiments performed here, then, the [OH] concentration could be different by less than a factor of 2 because of these confounding variables. We also assumed that the density of the SOA formed was 1.1 g cm −3 and Wu and Johnston (2017) collected the aerosol onto filters and directly measured the mass formed. While we based our density estimate on, crudely, what was found in Wu and Johnston (2017) (particularly, their Experiments 1 and 5), the aerosol density in their experiments could have been as much as 1.6 g 200 cm −3 . Indeed, much secondary organic aerosol has a density of 1.4-1.6 g cm −3 , which would account for a significant portion of the discrepancy (Kostenidou et al., 2007). Since the CPOT experiments were unseeded, seeded experiments increased the measured Y, which could also have led to better agreement. If the SOA yield depends on the OH exposure, instead of the OH concentration, we would expect that the dependence would actually be on the OH exposure normalized to the amount of reacted D5. That is, the number of OH radicals available 220 per reacted D5 molecule, as is shown in Fig. 1b. This figure shows a factor of 10 disagreement between data from Wu and Johnston (2017)  The major difference in Experiments 16-17 and 18-19 is the percent of D5 that reacted by the end of the experiment: 97% for Experiment 16, 98% for Experiment 17, and 100% for Experiments 18-19. Figure 2 shows the fraction reacted compared to the SOA yield for experiments performed in this study and those in the literature. This fit could indicate that there are later generation oxidation products that form large amounts of aerosol and that the gas-phase reaction rate to form the low-volatility later-generation oxidation product is slower than the gas-phase reaction rate to form the first-generation product (Kroll and 230 Seinfeld, 2008). However, if this were the case, Experiments 18-19 (Y=73% and 68%), in which all of the initial D5 reacted throughout the experiment, should show higher SOA yields than Experiments 16-17 (Y=79%), which they do not. Additionally, if later generation oxidation products produced more aerosol, there should be a correlation between Y and the OH exposure normalized to the amount of reacted D5 (Fig. 1b), which is also not accurate. The color axis in Fig. 2  indicate that at high mass loadings, relatively more low-volatility products partition into the particle phase. This could also 240 explain the disagreement in Fig. 1a    OH concentrations were also seeded and showed low SOA yields. For all experiments with [OH]< 10 8 molec cm −3 , the SOA yield is still < 5% and, in general, is closer to ∼1%. We do not expect that either relative humidity or temperature affect the SOA yield sufficiently that these would account for the vastly different measured SOA yields under different OH concentration. Experiments 1-8 were performed at RH levels between 2 and 6%, Experiments 9-19 were between 0 and 30% RH, those by Wu and Johnston (2017) were performed at 27 ○ C and a RH of 8-10%, and the experiments from Janechek et al. (2019) were run at 24 ○ C and an RH of 25% or 45%. At similar values of relative humidity but different OH concentrations (e.g., Experiments 9-12, which all have RH ≤ 6%), the OH 255 concentration matters for determining the SOA yield. For Experiments 3 and 4, the lowest and highest temperatures studied here (17.7 and 27.6 ○ C, respectively), the measured SOA Y varies by < 2%, which is within the uncertainty.
The NO x concentrations also do not seem to affect the SOA yield, as discussed in Appendix A. While the D5 oxidation chemistry may depend on the NO mixing ratio (but not on the NO 2 mixing ratio), this has no effect on the measured SOA yield.

Conclusions
The atmospheric aerosol formation potential of D5 was investigated under a range of OH concentrations and exposures. While secondary organic aerosol (SOA) yields can reach 79% (114% at the upper limit) at OH mixing ratios of ∼ 5×10 9 molec cm −3 , at atmospherically relevant OH concentrations ([OH] ≲ 10 7.5 molec cm −3 ), SOA yields do not exceed 5% and are likely ∼1%.
It is the OH concentration, and not the OH exposure, that affects the SOA yield.

265
This demonstrates the importance of extrapolating to the atmosphere at OH concentrations close to atmospheric levels and of using the appropriate reactor for the chemistry of a precursor to determine the secondary organic aerosol formation: if OH concentration is dominant, environmental chambers may be more useful, but if OH exposure matters, then flow tubes that have high OH mixing ratios may be the best tool.
Despite the relatively low SOA yields of D5 measured here at ambient OH concentrations, silicon has been observed in am-270 bient aerosol and its concentration is likely somewhat population (and not vehicle) dependent (Bzdek et al., 2014;Pennington et al., 2012). Since D5 is so abundant, it could be possible that the silicon present is from D5 or other volatile methyl siloxanes, just in lower concentrations than expected. Another possibility is that silicon in the aerosol-phase comes from polydimethylsiloxanes (Weschler, 1988 This does not imply that the chemistry is independent of NO concentration. Indeed, the concentrations of gas-phase fragments detected by the CIMS at m/z 139, 169, 243, and 317, which likely correspond to oxygenated fragments of D5, depend on 285 the NO concentration but not the NO 2 concentration. Figure A1 shows the signal for these fragments normalized to the reagent ion as a function of the NO concentration at any time. Note that, since some of the methyl nitrite is detected as NO, data from Experiment 8 were not included. Figure A2 shows the NO and NO 2 concentrations in each experiment as a function of time.  Fu et al. (2020) found that the gas-phase rearrangement of methylsiloxanes is dependent on the NO/HO 2 ratio. A comparison of Figs. A1 and A3 shows that the concentration of some gas-phase fragments is dependent on the NO mixing ratio but not 290 on the NO 2 mixing ratio. This is consistent with gas-phase products depending on the NO/HO 2 ratio. Note that at all NO/HO 2 ratios investigated, aerosol formation is still minimal when [OH] is small. Figure A2. For the experiments that included NO x , the NO and NO 2 concentrations as a function of the time since the onset of oxidation.
Experiment 8 is not included, since methyl nitrite was present. The measurement uncertainty is ∼5 ppb, but any organonitrates would also be measured as NO 2 . Figure A3. Dependence of gas-phase D5 oxidation products on the NO 2 concentration in the chamber indicates that oxidation chemistry does not depend on NO 2 (but does depend on NO, see