Organic peroxy radical chemistry in oxidation flow reactors and environmental chambers and their atmospheric relevance

Oxidation flow reactors (OFRs) are a promising complement to environmental chambers for investigating atmospheric oxidation processes and secondary aerosol formation. However, questions have been raised about how representative the chemistry within OFRs is of that in the troposphere. We investigate the fates of organic peroxy radicals (RO2), which play a central role in atmospheric organic chemistry, in OFRs and environmental chambers by chemical kinetic modeling and compare to a variety of ambient conditions to help define a range of atmospherically relevant OFR operating conditions. For most types of RO2, their bimolecular fates in OFRs are mainly RO2+HO2 and RO2+NO, similar to chambers and atmospheric studies. For substituted primary RO2 and acyl RO2, RO2+RO2 can make a significant contribution to the fate of RO2 in OFRs, chambers and the atmosphere, but RO2+RO2 in OFRs is in general somewhat less important than in the atmosphere. At high NO, RO2+NO dominates RO2 fate in OFRs, as in the atmosphere. At a high UV lamp setting in OFRs, RO2+OH can be a major RO2 fate and RO2 isomerization can be negligible for common multifunctional RO2, both of which deviate from common atmospheric conditions. In the OFR254 operation mode (for which OH is generated only from the photolysis of added O3), we cannot identify any conditions that can simultaneously avoid significant organic photolysis at 254 nm and lead to RO2 lifetimes long enough (∼ 10 s) to allow atmospherically relevant RO2 isomerization. In the OFR185 mode (for which OH is generated from reactions initiated by 185 nm photons), high relative humidity, low UV intensity and low precursor concentrations are recommended for the atmospherically relevant gas-phase chemistry of both stable species and RO2. These conditions ensure minor or negligible RO2+OH and a relative importance of RO2 isomerization in RO2 fate in OFRs within ∼ ×2 of that in the atmosphere. Under these conditions, the photochemical age within OFR185 systems can reach a few equivalent days at most, encompassing the typical ages for maximum secondary organic aerosol (SOA) production. A small increase in OFR temperature may allow the relative importance of RO2 isomerization to approach the ambient values. To study the heterogeneous oxidation of SOA formed under atmospherically relevant OFR conditions, a different UV source with higher intensity is needed after the SOA formation stage, which can be done with another reactor in series. Finally, we recommend evaluating the atmospheric relevance of RO2 chemistry by always reporting measured and/or estimated OH, HO2, NO, NO2 and OH reactivity (or at least precursor composition and concentration) in all chamber and flow reactor experiments. An easy-to-use RO2 fate estimator program is included with this paper to facilitate the investigation of this topic in future studies.

isomerization can be negligible for common multifunctional RO2, both of which deviate from common 23 atmospheric conditions. In the OFR254 operation mode (where OH is generated only from photolysis 24 of added O3), we cannot identify any conditions that can simultaneously avoid significant organic 25 photolysis at 254 nm and lead to RO2 lifetimes long enough (~10 s) to allow atmospherically relevant 26 RO2 isomerization. In the OFR185 mode (where OH is generated from reactions initiated by 185 nm 27 photons), high relative humidity, low UV intensity and low precursor concentrations are recommended 28 for atmospherically relevant gas-phase chemistry of both stable species and RO2. These conditions 29 ensure minor or negligible RO2+OH and a relative importance of RO2 isomerization in RO2 fate in OFRs 30 within ~x2 of that in the atmosphere. Under these conditions, the photochemical age within OFR185 31 systems can reach a few equivalent days at most, encompassing the typical ages for maximum 32 secondary organic aerosol (SOA) production. A small increase in OFR temperature may allow the relative 33 importance of RO2 isomerization to approach the ambient values. To study heterogeneous oxidation of 34 SOA formed under atmospherically-relevant OFR conditions, a different UV source with higher intensity 35 is needed after the SOA formation stage, which can be done with another reactor in series. Finally, we 36 recommend evaluating the atmospheric relevance of RO2 chemistry by always reporting measured 37 and/or estimated OH, HO2, NO, NO2 and OH reactivity (or at least precursor composition and 38 concentration) in all chamber and flow reactor experiments. An easy-to-use RO2 fate estimator program 39 is included with this paper to facilitate investigation of this topic in future studies. 40 A single generic RO2 is adopted for modeling purposes, to avoid the huge number of RO2 types 117 that would complicate effective modeling and analysis. In OH-initiated VOC oxidation, RO2 is primarily 118 produced via VOC+OH → R (+H2O) followed by R+O2  RO2, where R is hydrocarbyl or oxygenated 119 hydrocarbyl radical. Since the second step is extremely fast in air (Atkinson and Arey, 2003), the first 120 step controls the RO2 production rate, which depends on OH concentration and OHRext due to VOCs 121 (OHRVOC, see Appendix A for details). OHRVOC also includes the contribution from oxidation 122 intermediates of primary VOCs (e.g. methyl vinyl ketone and pinonic acid). When the information about 123 oxidation intermediates is insufficient to calculate OHRVOC, OHR due to primary VOCs is used instead as 124 an approximant. 125 Table 1 lists all known RO2 loss pathways. Among those, RO2 photolysis, RO2+NO3 and RO2+O3 126 are not included in this study, since they are minor or negligible in OH-dominated atmospheres, 127 chambers and OFRs for the following reasons. 128 - The first-order RO2 photolysis rate constant is of the order of 10 -2 s -1 at the highest lamp setting in 129 OFRs (Kalafut-Pettibone et al., 2013) and of the order of 10 -5 s -1 in the troposphere under the 130 assumption of unity quantum yield (Klems et al., 2015), while RO2 reacts with HO2 at >1 s -1 at the 131 highest lamp setting in OFRs and at ~2x10 -3 s -1 in the troposphere. Note that in this study we assume 132 an average ambient HO2 concentration of 1.5x10 8 molecules cm -3 (Mao et al., 2009;Stone et al., 133 2012) and RO2+HO2 rate constant of 1.5x10 -11 cm 3 molecule -1 s -1 (Orlando and Tyndall, 2012). 134 -When daytime photochemistry is active, NO3 is negligible in the atmosphere. In OFR-iN2O modes, 135 RO2+NO3 is negligible unless at very low H2O and high UV intensity (abbr. UV hereafter), which 136 result in high O3 to oxidize NO2 to NO3 and keep HO2 minimized. However, very low H2O causes 137 serious non-tropospheric organic photolysis (Peng et al., 2016) and thus these conditions are of no 138 experimental interest. 139 -In the atmosphere RO2+O3 is thought to play some role only at night (Orlando and Tyndall, 2012). 140 Similar conditions may exist in some OFR254 cases, if a very large amount of O3 is injected and H2O 141 and UV are kept very low to limit HOx production. These conditions are obviously not OH-142 dominated and not further investigated in this study. 143 Of the RO2 fates considered in this study, RO2+HO2 and RO2+NO and RO2+RO2 have long been 144 known to play a role in the atmosphere (Orlando and Tyndall, 2012). Despite some small dependencies 145 on the type of RO2, recommended general rate constants are available for RO2+HO2 and RO2+NO 146 (Ziemann and Atkinson, 2012; Table 1). We use these recommended values for generic RO2 in this study. 147 However, RO2 self-/cross-reaction rate constants are highly dependent on the specific RO2 types and can 148 vary over a very large range (10 -17 -10 -10 cm 3 molecule -1 s -1 ). Unsubstituted primary, secondary and 149 tertiary RO2 self-react at ~10 -13 , ~10 -15 and ~10 -17 cm 3 molecule -1 s -1 , respectively (Ziemann and Atkinson, 150 2012). Rate constants of cross-reactions between these RO2 types also span this range (Orlando and 151 Tyndall, 2012). Substituted RO2s have higher self-/cross-reaction rate constants (Orlando and Tyndall, 152 2012). RO2+RO2 of highly substituted primary RO2 can be as high as ~10 -11 cm 3 molecule -1 s -1 (Orlando 153 and Tyndall, 2012). Very recently, a few highly oxidized 1,3,5-trimethylbenzene-derived RO2s were 154 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-952 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 24 September 2018 c Author(s) 2018. CC BY 4.0 License.
reported to self-/cross-react at ~10 -10 cm 3 molecule -1 s -1 (Berndt et al., 2018). In the present work, we 155 make a simplification to adapt to the generic RO2 treatment by assuming a single self-/cross-reaction 156 rate constant for generic RO2 in each case. Three levels of RO2+RO2 rate constants, i.e. 1x10 -13 , 1x10 -11 , 157 and 1x10 -10 cm 3 molecule -1 s -1 , are studied in this paper. The first level is referred to as "medium RO2+RO2" 158 as many other RO2 can have self-/cross-reaction rate constants as low as 10 -17 cm 3 molecule -1 s -1 ; the 159 second level is defined as "fast RO2+RO2"; the last level is called "very fast RO2+RO2." No RO2+RO2 rate 160 constant lower than the medium level is investigated in the current work, although there are still a large 161 variety of RO2 whose self-/cross reactions are at lower rate constants, since at the medium level, 162 RO2+RO2 is already negligible in all the environments studied in this work, i.e. OFRs, chambers and the 163 atmosphere (see Section 3.1.1). Since there are only a few very specific examples for very fast RO2+RO2 164 reported to date, we will not systematically explore this category but compare very fast RO2+RO2 as a 165 sensitivity case with the other two types of RO2+RO2 reactions. 166 Acyl RO2 is considered as a separate RO2 type (neither medium nor fast RO2+RO2) in this study 167 since its reaction with NO2 can be a major sink of RO2 in OFR (Peng and Jimenez, 2017). Thermal 168 decomposition lifetimes of the product of RO2+NO2, i.e. acylperoxy nitrates, can be hours at laboratory 169 temperatures (Orlando and Tyndall, 2012; also taken into account in the current work, see Table 1),  170 while OFR residence times are typically minutes. Besides, acyl RO2 react with many RO2 at ~10 -11 cm 3 171 molecule -1 s -1 (Orlando and Tyndall, 2012), similar to that of fast RO2+RO2. We thus assume acyl RO2 172 self-/cross-reaction rate constant to be also 1x10 -11 cm 3 molecule -1 s -1 to facilitate the comparison with 173 fast RO2+RO2 results. 174 In OFRs operated at room temperature, acylperoxy nitrates barely decompose, while peroxy 175 nitrates of non-acyl RO2 do decompose on a timescale of 0.1 s ( Table 1). As a consequence, the 176 production and decomposition of peroxy nitrates of non-acyl RO2 reach a steady state in OFRs, which 177 can be greatly shifted toward the peroxy nitrate side in cases with very high NO2 (Peng and Jimenez, regardless of RO2 type. Therefore, the reaction rate constant of generic RO2 with OH is assigned as 1x10 -184 10 cm 3 molecule -1 s -1 . RO2 isomerization reactivity is highly structure-dependent (Crounse et al., 2013;185 Praske et al., 2018) and rate constant measurements are still scarce, preventing us from assigning a 186 generic RO2 isomerization rate constant. However, for generic RO2, isomerization is generally not a sink 187 but a conversion between two RO2 (both encompassed by the generic one in this study), as RO2 188 isomerization usually generates an oxygenated hydrocarbyl radical, which rapidly recombines with O2 189 and forms another RO2. Therefore, RO2 isomerization is not explicitly taken into account in the modeling, 190 but is considered in the RO2 fate analysis. 191 In summary, 6 pathways are included in the RO2 fate analysis of this study. The need to explore 192 of results challenging. For clarity, we present the results in two steps. In the first step, only well-known 194 RO2 fates (reaction with NO2, HO2, NO and RO2) will be included in the model. In the second step, the 195 results of the first step will be used to guide the modeling and analysis of a more comprehensive set of 196 significant RO2 fates. 197

2.3
Model description 198 The model used in the present work is a standard chemical kinetic box model, implemented in the 199 KinSim 3.4 solver in Igor Pro 7 (WaveMetrics, Lake Oswego, Oregon, USA), and has been described in of RO2 discussed in Section 2.2 are added to the chemical mechanism. A generic slow-reacting VOC 205 (with the same OH rate constant as SO2) is used as the external OH reactant. This slow rate also 206 represents the generation and consumption of latter-generation products that continue to react with 207 OH. The reason for this approximation has been discussed in detail in previous OFR modeling papers 208 (Peng and Jimenez, 2017;Peng et al., 2018). We exclude NOy species, which are explicitly modeled, from 209 the calculation of OHRext; thus OHRext only includes non-NOy OHRext hereafter. As OHRext is dominated 210 by OHRVOC in most OFR experiments, we use OHRext to denote OHRVOC in OFRs (while for ambient and 211 chamber cases OHRVOC is still used to exclude the contribution of CO etc.). The model was estimated to 212 achieve an accuracy of a factor of 2-3 when compared to field OFR experiments; better agreement can 213 generally be obtained for laboratory OFR experiments (Li et al., 2015;Peng et al., 2015). 214 Another key parameter in the model is the HOx recycling ratio (β), defined in this study as the 215 number of HO2 molecule(s) produced per OH molecule destroyed by external OH reactants (Peng et al., 216 2015). This ratio depends on the products of RO2 loss pathways. The main product of RO2+HO2 is usually 217 ROOH (Table 1) In the present work, we model OFR185, OFR254-70, and OFR254-7 (including their -iN2O variants). 225 We specify the same temperature and atmospheric pressure (295 K and 835 mbar, typical values in 226 Boulder, Colorado, USA) as our previous OFR modeling studies (Li et   OFR254-70-iN2O can still make RO2+NO dominate over RO2+HO2 in RO2 fate. OFR and chamber cases 283 span a range of ~0-~100% in relative importance of RO2+NO in RO2 fate (Fig. 2), suggesting that both 284 chambers and OFRs are able to ensure the atmospheric relevance of RO2+NO in RO2 fate. 285 Another important feature that can be easily seen in Fig. 1 is that medium rate RO2+RO2 (and 286 hence also RO2+RO2 slower than 10 -13 cm 3 molecule -1 s -1 ) are of negligible importance in the fate of RO2 287 contribution from RO2+RO2 to their fate. This is already known for ambient RO2 fate (Ziemann and 290 Atkinson, 2012). The reason why this is also true in OFRs is that while OH is much higher than ambient 291 levels, HO2 and NO (high-NO conditions only) are also higher. One can easily verify that steady-state RO2 292 concentrations (see Appendix A for details) would not deviate from ambient levels by orders of 293 magnitude. The reactive fluxes of RO2+RO2 in OFRs are thus not substantially different than in the 294 atmosphere, while RO2+HO2 and RO2+NO (high-NO conditions only) are both faster in OFRs because of 295 higher HO2 and NO. The combined effect is a reduced relative importance of RO2+RO2 in RO2 fate in 296 OFRs compared to the atmosphere. The only exception in OFRs occurs at very high VOC precursor 297 concentrations (OHRext significantly >100 s -1 ) in OFR254 (Fig. S2), where OH levels are not substantially 298 suppressed due to large amounts of O3 (Peng et al., 2015). As a result, RO2 concentration is remarkably 299 increased by strong production and RO2+RO2 relative importance increases roughly quadratically and 300 becomes significant. 301 The generally lower relative importance of RO2+RO2 in OFRs than in the atmosphere is more 302 obvious for the fate of RO2 with fast RO2+RO2 rate constants (Figs. 1b,d and 3). Although OFRs can 303 reasonably reproduce RO2 fates in low-VOC ambient environments (e.g. typical pristine and forested 304 areas; Figs. 1b,d and 3) and low-OHRext chambers, OFR185 cannot achieve relative importance of 305 RO2+RO2 significantly larger than 50%, corresponding to higher-VOC environments (e.g. P1 in Fig. 1)  high-OHRext chamber experiments (e.g. C2 and C5 in Fig. 1; the distribution for C2 is also shown in Fig. 3). 307 In OFR254-70, a relative importance of RO2+RO2 as high as ~90% may be attained (Fig. S3). However, 308 this requires very high OHRext, which leads to medium (and slower) RO2+RO2 showing higher-than-309 ambient relative importance. In reality, fast RO2+RO2 all involve substituted RO2, which almost certainly 310 arise from and coexist with unsubstituted RO2 (with slower self-/cross reactions). Therefore, very high 311 OHRext in OFR254 is not really suitable for attaining dominant RO2+RO2 conditions. In OFR185, a higher 312 OHRext generally also results in a higher RO2+RO2 relative importance because of higher RO2 production 313 2016). It should be noted that although it is difficult to reliably achieve RO2+RO2 with a relative 315 importance larger than 50% in RO2 fate in OFRs, the distributions of RO2+RO2 relative importance in 316 OFRs seems to be within a factor of 2 of those of field/aircraft campaigns (Fig. 3). 317 In the case of very fast RO2+RO2, all features for fast RO2+RO2 discussed above are still present 318 (Fig. S1c,d). The only major difference between the results for fast RO2+RO2 and very fast RO2+RO2 is 319 the significantly higher relative importance of RO2+RO2 in RO2 fate in the latter case, which is expected. 320 In summary, the fast RO2+RO2 is not perfectly reproduced in OFRs in terms of relative importance in RO2 321 fate, but it is significant when this pathway is also important in the atmosphere. 322 The HOx recycling ratio β (see Sect. 2.3) is one of the key factors determining HO2 in the OFR 323 model, yet it is not well constrained. Although we make reasonable assumptions for it in the model 324 input (see Section 2.3 for details), a sensitivity study to explore its effects is also performed here. For 325 RO2 with the fast self-/cross-reaction rate constant, we perform the simulations with the HOx recycling 326 ratios fixed to a number of values from 0 (radical termination) to 2 (radical proliferation) in lieu of those 327 calculated under the assumptions described in Section 2.3. As expected, the contribution of RO2+RO2 328 to RO2 fate increases monotonically between β=2 and β=0 (Fig. S4), as the recycling of the competing 329 reactant HO2 decreases. Nevertheless, the change in the average RO2+RO2 relative importance from β=0 330 to β=2 is generally within a factor of 2. Thus, it still holds that the RO2+RO2 relative importance in OFRs 331 is generally lower than in the atmosphere. Only at β~0 may OFR185 theoretically attain a relative 332 importance of RO2+RO2 of ~70%, as in the P1 case (pristine, but relatively high-VOC, Figure S5). Note 333 that β=0 for all VOC oxidation (including oxidation of intermediates) is extremely unlikely. In OFR254, 334 even if RO2+RO2 may contribute up to ~100% to RO2 fate at very high OHRext at β=0, these conditions 335 still also lead to significant RO2+RO2 in the fate of RO2 that self-/cross-react more slowly, which is not 336 atmospherically relevant. 337

Acyl RO2 338
As described in Section 2.1, the generic acyl RO2 modeled in this study has the same loss 339 pathways as RO2 with the fast self-/cross-reaction rate constant, except for RO2+NO2, which can be a 340 significant acyl RO2 loss pathway in OFRs as well as both chambers and atmosphere. When this reaction 341 is included in the simulations of acyl RO2, it is a minor or negligible loss pathway of RO2 at low N2O, 342 while it can be the dominant fate of acyl RO2 at high N2O (Fig. 4). In general, the RO2+NO2 relative 343 importance increases with initial N2O. This is always true in OFR254-70-iN2O between N2O=0.02% and 344 conditions). In OFR185-iN2O, the relative importance of RO2+RO2 in the sum of the HO2, NO and RO2 355 pathways is reduced (Fig. S6a), compared to that of non-acyl RO2 with the fast RO2+RO2 (Fig. 1b), 356 because RO2+NO2 decrease acyl RO2 concentration. Such a decrease is not significant in OFR254-70-357 iN2O (Fig. S6b, compared to Fig. 1d), since for non-acyl RO2, it is already stored in the form of RO2NO2 358 as RO2 reservoir. In other words, the high initial O3 greatly accelerates NO-to-NO2 oxidation, and shifts 359 the equilibrium RO2+NO2↔RO2NO2 far to the right even for non-acyl RO2. 360 RO2+NO2 is an inevitable sink of most acyl RO2 in high-NOx OFRs. Its contribution to acyl RO2 fate 361 in OFRs is often higher than in urban atmospheres, where the relative amounts of NO  OFRs. In OFR185 cases with medium RO2+RO2, HO2-to-OH ratio around 100 occurs at a combination of 394 low H2O (on the order of 0.1%), low F185 (on the order of 10 11 photons cm -2 s -1 ), and medium OHRext 395 (10-100 s -1 ); and also at medium F185 (~10 12 photons cm -2 s -1 ) combined with very high OHRext (~1000 396 s -1 , Fig. S7). Under both sets of conditions, relatively high external OH reactants suppress OH, whose 397 production is relatively weak, and convert some OH into HO2 through HOx recycling in organic oxidation 398 (e.g. via alkoxy radical chemistry). The reason why such an OH-to-HO2 conversion is needed to attain an 399 ambient-like HO2-to-OH ratio is that OFR185 is unable to achieve this via the internal (mainly assisted 400 by O3) interconversion of HOx. This inability is most evident when F185 (10 13 -10 14 photons cm -2 s -1 ) and 401 H2O (on the order of 1%) are high and OHRext is low (<~10 s -1 ; Fig. S7). Under these conditions, OH 402 production by H2O photolysis is so strong that the HO2-to-OH ratio is lowered to ~1, since OH and H 403 (which recombines with O2 to form HO2) are produced in equal amounts from H2O photolysis. As the 404 RO2+OH rate constant is only roughly 1-order-of-magnitude higher than that for RO2+HO2, slightly lower 405 HO2-to-OH ratios (e.g. ~30) suffice to keep RO2+OH minor in this case. A combination of UV and H2O 406 that are not very high and a moderate OHRext that is able to convert some OH to HO2 and somewhat 407 elevate the HO2-to-OH ratio results in minor relative importance RO2+OH (Figs. S7 and S8). 408 In OFR254-70, it is more difficult to reach an HO2-to-OH ratio of ~100, which can only be realized 409 at a combination of very low H2O and F254 (~0.07% and ~5x10 13 photons cm -2 s -1 , respectively) and very 410 high OHRext (~1000 s -1 ). This is mainly due to high O3 in OFR254-70, which controls the HOx 411 interconversion through HO2+O3OH+2O2 and OH+O3HO2+O2 and makes both OH and HO2 more 412 resilient to changes due to OHRext (Peng et al., 2015). Even without H2O photolysis at 185 nm as a major 413 HO2 source, the HOx interconversion controlled by O3 in OFR254-70 still brings HO2-to-OH ratio to ~1 in 414 the case of minimal external perturbation (see the region at the highest H2O and UV and OHRext=0 in 415 the OFR254-70 part of Fig. S7). This ratio cannot be easily elevated in OFR254-70 because of the 416 resilience of OH to suppression for this mode (Peng et al., 2015). Thus, this ratio is relatively low (<30) 417 under most conditions (Fig. S7), and consequently (and undesirably), RO2+OH is a major RO2 fate in 418 OFR254-70. There is an exception at relatively low H2O and UV with very high OHRext (Fig. S8) Only the results of RO2 with the medium RO2+RO2 are discussed in this subsection. Those of RO2 421 with the fast RO2+RO2 are not shown as they are not qualitatively different. In OFR185, for the fast-self-422 /cross-reacting RO2, RO2+RO2 is relatively important at high OHRext (>~100 s -1 ; Fig. S3), while RO2+OH is 423 a major RO2 fate at low OHRext (generally on the order of 10 s -1 or lower) and relatively high H2O and UV 424 (Fig. S8). These two ranges of conditions are relatively far away from each other, and hence there is no 425 condition under which RO2+RO2 and RO2+OH are both major pathways that compete, which simplifies 426 understanding RO2 fate. However, in OFR254-70, some conditions may lead to both significant RO2+RO2 427 (for the fast-self-/cross-reacting RO2) and RO2+OH (e.g. H2O~0.5%, F254~1x10 15 photons cm -2 s -1 and 428 OHRext~100 s -1 ). Nevertheless, as long as RO2+OH plays a major role, these conditions do not bear much 429 experimental interest and thus do not need to be discussed in detail. 430

RO2 isomerization 431
RO2 isomerization is a first-order reaction. For this type of reactions to occur, RO2 does not need 432 any other species but only a sufficiently long lifetime against all other reactants combined, as most RO2 433 isomerization rate constants are <10 s -1 . Radical (OH, HO2, NO etc.) concentrations in OFRs are much 434 higher than ambient levels and may shorten RO2 lifetimes compared to those in the troposphere. 435 Possibly reduced RO2 lifetimes naturally raise concerns over the potentially diminished importance of 436 RO2 isomerization in OFRs. 437 In this section we examine generic RO2 lifetimes against all reactions (calculated without RO2 438 isomerization taken into account) in OFR (including OFR-iN2O) cases (for the medium RO2+RO2 case) and 439 compare them with the RO2 lifetimes in recent major field/aircraft campaigns in relatively clean 440 environments and a field campaign in an urban area (CalNex-LA), as well as a low-NO chamber 441 experiment (Fig. 6). Indeed, RO2 lifetime in clean ambient cases and in chambers with near-ambient 442 radical levels are generally much longer than those in OFRs. The RO2 lifetime distribution of the explored 443 good and risky cases in OFR254-70 (including OFR254-70-iN2O) barely overlaps with the ambient and 444 chamber cases, while in OFR185 (including OFR185-iN2O), RO2 lifetime can be as long as ~10 s, which is 445 longer than in urban areas and roughly at the lower end of the range of ambient RO2 lifetime in clean 446 environments (Fig. 6). The longest RO2 lifetime in OFR185 occurs at very low F185 (on the order of 10 11 447 photons cm -2 s -1 ) and H2O (~0.1%; Fig. S9), where HOx is low. In OFR254-70, for RO2 to survive for ~10 s, 448 in addition to very low UV and H2O, high OHRext is also needed (Fig. S9) In the discussion about RO2 isomerization above (as in the RO2+OH exploration in Section 3.2.1), 470 we only examine low-NO (or zero-NO for simplicity) conditions with medium RO2+RO2. In high-NO 471 environments, e.g. polluted urban atmospheres with NO of at least ~10 ppb and high-NO OFRs in the 472 iN2O modes, RO2 lifetime is so short that isomerization is no longer a major fate for any but the most 473 rapidly isomerizing multifunctional RO2 discussed above. NO measured in Los Angeles during the 474 CalNex-LA campaign (Ortega et al., 2016) was only ~1 ppb, which would to allow RO2 to survive for a 475 few seconds and isomerize (Fig. 6), even in an urban area. 476 The OFR simulations for the discussions about RO2 isomerization are the same as those 477 conducted to study RO2+OH, i.e. the ones with the medium RO2+RO2 and RO2+OH included. For fast RO2 478 self-/cross-reaction cases, RO2 lifetimes may be significantly shorter than for RO2 with the medium self-479 /cross-reaction rate constant at high OHRext (>~100 s -1 ) in OFR185 (Fig. S3). These high-OHRext conditions 480 are likely to be risky or bad (of little experimental interest)(Peng et al., 2016) and thus do not need to 481 be discussed further in detail. OFR254-70 (a zero-NO mode) does not generate good or risky (of at least 482 some experimental interest in terms of non-tropospheric organic photolysis) conditions also leading to 483 low-NO-atmosphere-relevant RO2 lifetimes (Fig. 6). RO2 with faster self-/cross-reaction rate constants 484 have even shorter lifetimes in OFR254-70 and will not be discussed further. 485

Guidelines for OFR operation 486
In this subsection we discuss OFR operation guidelines for atmospherically relevant RO2 chemistry, 487 with a focus on OFR185 and OFR254 (zero-NO modes). Since RO2+HO2 and RO2+NO both can vary from 488 negligible to dominant RO2 fate in OFRs, chambers and the atmosphere (Figs. 1 and 2), these two 489 pathways are not a concern in OFR atmospheric relevance considerations. Neither is the RO2+RO2 a 490 major concern. Medium or slower RO2+RO2 is minor or negligible in the atmosphere and chambers, as 491 well as in OFRs, as long as high OHRext is avoided in OFR254 (Fig. S2). Fast RO2+RO2 is somewhat less 492 important in OFRs than in the atmosphere (Figs. 1b,d and 3 isomerization. Both the dominance of RO2+NO and the inhibition of RO2 isomerization also occur in the 499 atmosphere and in chambers, so high-NO OFR operation (typically NO>10 ppb) represents these 500 pathways realistically. Some care is, however, required with the RO2+OH and RO2 isomerization 501 pathways at low NO. Since RO2+HO2 in OFRs is always a major RO2 fate at low NO and RO2+RO2 are 502 generally not problematic, RO2+OH and RO2+HO2 can be kept atmospherically relevant as long as HO2-503 to-OH ratio is close to 100 (the ambient average). In addition, RO2 lifetime (calculated without RO2 504 isomerization taken into account) should be at least around 10 s. 505 Practically, OH production should be limited to achieve this goal. Too strong OH production at high 506 H2O and UV can elevate OH and HO2 concentrations, which shortens RO2 lifetime, and decreases the 507 HO2-to-OH ratio to ~1 (see Sect. 3.2.1). OH production is roughly proportional to both H2O and UV (Peng  is reduced without also decreasing UV, F185exp/OHexp and F254exp/OHexp both increase, signifying 514 stronger relative importance of non-tropospheric photolysis. Therefore, reducing UV is strongly 515 preferred as an OH production limitation method, and is effective in making both RO2+OH and RO2 516 isomerization more atmospherically relevant. 517 To further explore the effects of UV reduction on the RO2+OH (Fig. 5) and RO2 isomerization (Fig.  518 6) pathways, we divide our OFR case distributions into higher-UV and lower-UV classes, with the 519 boundary being the mid-level (in logarithmic scale) UV in the explored range. The distributions for 520 lower-UV conditions (solid lines in Figs. 5 and 6) are clearly closer to the ambient cases (i.e. HO2-to-OH 521 ratio closer to 100, smaller RO2+OH relative importance and longer RO2 lifetime). 522 Since OFR254 is unable to achieve conditions with both at least some experimental interest (i.e. 523 with sufficiently low non-tropospheric photolysis) and atmospherically relevant RO2 lifetime, we now 524 discuss preferable conditions for OFR185 only. As F185 close to or lower than 10 12 photons cm -2 s -1 is 525 needed for RO2 lifetime to be around 10 s or longer (Fig. S9), the OH concentration under preferable 526 conditions for atmospherically relevant RO2 chemistry (~10 9 molecules cm -3 or lower) is much lower 527 than the maximum that OFR185 can physically reach (~10 10 -10 11 molecules cm -3 ). Furthermore, lower 528 OH production leads to higher susceptibility to OH suppression by external OH reactants (Peng et al., 529 2015), which can create non-tropospheric photolysis problems (Peng et al., 2016). We thus recommend 530 as high H2O as possible to maintain practically high OH while allowing lower UV to limit the importance 531 of non-tropospheric organic photolysis. 532 The performance of various OFR185 conditions at high H2O (2.3%) is illustrated in Fig. 7  shown. At F185 of ~10 11 -10 12 photons cm -2 s -1 and OHRext around or lower than 10 s -1 , all three criteria 537 are satisfied. Since UV (and hence OH production) is relatively low, a low OHRext (~10 s -1 ) is required to 538 avoid heavy OH suppression and keep conditions good (green area in the bottom panel of Fig. 7). 539 Nevertheless, risky conditions [log(F254exp/OHexp)<7; light red area in the bottom panel of Fig. 7] may 540 also bear some experimental conditions depending on the type of VOC precursors (specifically on their 541 reactivity toward OH and their photolability at 185 and 254 nm, and the same quantities for their 542 oxidation intermediates; Peng et al., 2016; Peng and Jimenez, 2017). Thus, higher OHRext (up to ~100 s -543 1 ) may also be considered in OFR experiments with some precursors (e.g. alkanes). In practice, the 544 preferred conditions may require F185 even lower than that our lowest simulated lamp setting (Li et al., 545 2015). Such a low F185 may be realized e.g. by partially blocking 185 nm photons using non-transparent 546 lamp sleeves with evenly placed holes that allow some 185 nm transmission. 547 Under these preferred conditions, OH concentration in OFR185 is ~10 9 molecules cm -3 , equivalent 548 to a photochemical age of ~1 eq. d for a typical residence time of 180 s. This is much shorter than ages 549 corresponding to the maximal oxidation capacity of OFRs (usually eq. weeks or months; Peng et al.,  (Figs. 5 and 6). During the campaign, relative humidity was high (>60% in most of 554 the period), OHRext was estimated to be relatively low (~15 s -1 ) in this forested area, and UV in the OFR 555 was limited in the case of the maximal SOA formation age (~0.7 eq. d). All these physical conditions 556 were favorable for atmospherically relevant RO2 fate (Figs. 5 and 6). RO2+OH was minor in this case and 557 the relative importance of RO2 isomerization in RO2 fate in the OFR was within a factor of ~2 of that in 558 the atmosphere for all RO2 (regardless of isomerization rate constant) during the BEACHON-RoMBAS 559 campaign (Fig. 6). The effect of UV on the relative importance of RO2 isomerization for this example is 560 also illustrated in Fig. 6. In the sensitivity case with a lower age, a lower UV results in a larger 561 contribution of isomerization to RO2 fate, while the relative importance of RO2 isomerization is lower in 562 a sensitivity case with an age 3 times of that of the maximal SOA formation. In an extreme sensitivity 563 case with the highest UV in the range of this study (with an age of 4 eq. mo), RO2 isomerization becomes 564 minor or negligible for all RO2 except extremely rapidly isomerizing ones. 565 The discussions above indicate that the atmospheric relevance of gas-phase RO2 chemistry in OFRs 566 deteriorates as the photochemical age over the whole residence time (180 s) increases. To reach longer 567 ages, longer residence times (with UV being still low) can be adopted. However, OFR residence times > 568 10 min tend to be limited by the increasing importance of wall losses (Palm et al., 2016). As a result, 569 longer residence times can only increase photochemical age in OFRs up to about a week. This implies 570 that in OFR cases with ages much higher than that of maximal SOA formation (corresponding to the 571 heterogeneous oxidation stage of SOA), the atmospheric relevance of gas-phase RO2 chemistry in the 572 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2018-952 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 24 September 2018 c Author(s) 2018. CC BY 4.0 License. SOA formation stage (before the age of maximal SOA formation) often cannot be ensured. However, 573 under those conditions typically new SOA formation is not observed, and the dominant process 574 affecting OA is heterogeneous oxidation of the pre-existing OA (Palm et al., 2016). If the heterogeneous 575 oxidation of the newly formed SOA is of interest, a two-stage solution may be required. Lower UV can 576 be used in the SOA formation stage to keep the atmospheric relevance of the gas-phase chemistry, while 577 high UV can be used in the heterogeneous aging stage to reach a high equivalent age. The latter 578 approach is viable since heterogeneous oxidation of SOA by OH is slow and particle-phase chemistry is 579 not strongly affected by gas-phase species except OH, when OH is very high (Richards-Henderson et al., increase in the rate constants by a factor of ~5 on average. A 15 K temperature increase in OFRs would 584 lead to RO2 isomerization being accelerated by a factor of ~3, while other major gas-phase radical 585 reactions have weak or no temperature-dependence. As a consequence, the relative importance of RO2 586 isomerization in RO2 fate in OFRs can be elevated and closer to atmospheric values (Fig. 6). Nevertheless, 587 a 15 K increase in temperature may also result in some OA evaporation (Nault et al., 2018). 588 As discussed above, high H2O, low UV and low OHRext are recommended for keeping the 589 atmospheric relevance of RO2 chemistry in OFRs. These three requirements are also part of the 590 requirements for attaining good high-NO conditions in OFR185-iNO (the OFR185 mode with initial NO 591 injection; Peng and Jimenez, 2017). In addition to these three, an initial NO of several tens of ppb is also 592 needed to obtain a good high-NO condition in OFR185-iNO. Under these conditions, RO2+NO dominates 593 over RO2+HO2, and hence RO2+OH; UV is low, the photochemical age is typically ~1 eq. d, and RO2 594 lifetime can be a few seconds. Therefore, these conditions are a good fit for studying the environments 595 in relatively clean urban areas, such as Los Angeles during CalNex-LA (Ortega et al., 2016), where NO is 596 high enough that the dominant bimolecular fate of RO2 is RO2+NO but low enough to maintain RO2 597 lifetimes that allow most common RO2 isomerizations. 598 As RO2 fate in OFRs is a highly complex problem and it can be tricky to find suitable physical 599 conditions to simultaneously achieve experimental goals and keep the atmospheric relevance of the 600 We investigated RO2 chemistry in OFRs with an emphasis on its atmospheric relevance. All 613 potentially major loss pathways of RO2, i.e. reactions of RO2 with HO2, NO and OH, that of acyl RO2 with 614 NO2, self-/cross-reactions of RO2 and RO2 isomerization, were studied and their relative importance in 615 RO2 fate were compared to those in the atmosphere and chamber experiments. OFRs were shown to 616 be able to tune the relative importance of RO2+HO2 vs. RO2+NO by injecting different amounts of N2O. 617 For many RO2 (including all unsubstituted non-acyl RO2 and substituted secondary and tertiary RO2), 618 their self-reactions and the cross-reaction between them are minor or negligible in the atmosphere and 619 chambers. This is also the case in OFR185 (including OFR185-iN2O) and OFR254-iN2O, however those 620 RO2 self-/cross-reactions can be important at high precursor concentrations (OHRext>100 s -1 ) in OFR254. 621 For substituted primary RO2 and acyl RO2, their self-/cross-reactions (including the ones with RO2 whose 622 self-reaction rate constants are slower) can play an important role in RO2 fate in the atmosphere and 623 chambers, and may also be major RO2 loss pathways in OFRs, although they are somewhat less 624 important in OFRs than in the atmosphere. Acylperoxy nitrates are the dominant sink of acyl RO2 at high 625 NOx in OFRs, while only a minor reservoir of acyl RO2 in the atmosphere under most conditions except 626 in urban atmospheres, where acylperoxy nitrate formation can be the dominant acylperoxy loss 627 pathway when most NO is oxidized to NO2. In chambers, most acyl RO2 can be stored in the form of 628 acylperoxy nitrates if NO2 is very high (hundreds of ppb to ppm level). 629 Under typical high-NO conditions, RO2+NO dominates RO2 fate and RO2 lifetime is too short to 630 allow most RO2 isomerizations, regardless of whether in the atmosphere, chambers or OFRs, thus raising 631 no concern over the atmospheric relevance of the OFR RO2 chemistry. However, under low-NO 632 conditions, OFR254 cannot yield any physical conditions leading to sufficiently long RO2 lifetime for its 633 isomerization because of the high radical levels and their resilience to external perturbations in OFR254. 634 In OFR185 with strong OH production (and hence high OH), RO2+OH and RO2 isomerization may strongly 635 deviate from the atmosphere (becoming important and negligible, respectively, for relatively rapidly 636 isomerizing RO2). To attain both atmospherically relevant VOC and RO2 chemistries, OFR185 requires 637 high H2O, low UV and low OHRext, which conditions ensure minor or negligible RO2+OH and a relative 638 importance of RO2 isomerization in RO2 fate in OFRs within x~2 of that in the atmosphere but limit the 639 maximal photochemical age that can be reached to a few eq. days. This age roughly covers SOA 640 formation in ambient air up to its maximum. To reach a much higher age for studying SOA 641 functionalization/fragmentation by heterogeneous oxidation, a sequence of low-UV SOA formation 642 followed by a high UV condition (in the same reactor or in cascade reactors) would be needed. High 643 H2O, low UV and low OHRext in the OFR185-iNO mode can achieve conditions relevant to clean urban 644 atmosphere, i.e. high-NO but not sufficiently high to inhibit common RO2 isomerization. The production rate of a generic RO2 is almost identical to the VOC consumption rate, since the 658 second step of the conversion chain VOCRRO2 is extremely fast. Therefore, the generic RO2 659 production rate, P, can be expressed as follows: 660 corresponding values on this axis. The OFR data points are colored by the logarithm of the exposure ratio 950 between 254 nm photon flux and OH, a measure of badness of OFR conditions in terms of 254 nm organic 951 photolysis. Several typical ambient and chamber cases (see Table 2 for details of these cases) are also 952 shown for comparison. are included in the distributions. Also shown is the relative importance of RO2+NO for several typical 960 ambient and chamber cases (see Table 2 for details of these cases). self/cross reaction rate constant and without RO2+OH and RO2 isomerization considered) for OFR185 965 (including OFR185-iN2O), OFR254-70 (including OFR254-70-iN2O) and a chamber experiment and in the 966 atmosphere (a couple of different environments). The OFR distributions for good and risky conditions (in 967 terms of 254 nm organic photolysis, see Table S1 for the definitions of these conditions) are shown 968 separately. Also shown is the relative importance of RO2+RO2 for several typical chamber cases (see Table  969 2 for details of these cases). The range of the RO2+RO2 relative importance for most high-NO conditions 970 is highlighted in cyan.