Evolution of OH reactivity in low-NO volatile organic compound photooxidation 1 investigated by the fully explicit GECKO-A model 2

investigated by the fully explicit GECKO-A model 2 Zhe Peng, Julia Lee-Taylor, Harald Stark, John J. Orlando, Bernard Aumont and Jose L. 3 Jimenez 4 1 Department of Chemistry and Cooperative Institute for Research in Environmental Sciences, 5 University of Colorado, Boulder, Colorado 80309, USA 6 2 Atmospheric Chemistry Observation and Modeling Laboratory, National Center for Atmospheric 7 Research, Boulder, Colorado 80307, USA 8 3 Aerodyne Research Inc., Billerica, Massachusetts 01821, USA 9 4 Laboratoire Inter-Universitaire des Systèmes Atmosphériques (LISA), UMR 7583, Université Paris10 Est Créteil, Université de Paris, CNRS, Institut Pierre Simon Laplace, 94010 Créteil, France 11 12


Introduction
each case (given a typical average ambient OH concentration of 1.5x10 6 molecules cm -3 in the real 122 atmosphere (Mao et al., 2009); see Fig. 1 for the correspondence between equivalent photochemical age 123 and OH exposure (OHexp, i.e., the integral of OH concentration over time)). The simulated OFR in the 124 present work employs the light source parametrization obtained by Li et al. (2015) and Peng et al. (2015).

125
UV at both 185 and 254 nm is used to generate OH, i.e., the "OFR185" mode of operation. The residence 126 time in the OFR is always 3 min.

127
In addition, we simulate illustrative cases of methane oxidation, under ambient and OFR 128 conditions (Table 1 and Section 3.1). Note that these two simulations are performed using the GECKO-

129
A generated mechanism (see Section 2.2) in another chemical-kinetics solver, KinSim (Peng and  (Table 1). To explore the effects of UV sources in OFR (see Section 3.4), two simulations under a typical

134
OFR condition with an additional broad-spectrum UV source (5 and 10000 times the chamber UV source 135 in this study, respectively) are performed for isoprene (Table 1). 1994; Verwer et al., 1996). In mechanism generation, isomer lumping for mechanism reduction purposes 142 is applied to certain products with branching ratios < 1% (here typically N-containing products, which 143 are not relevant for our simulations). It has a negligible impact on the results.

144
The core isoprene scheme in GECKO-A is adopted from Master Chemical Mechanism v3.

148
We tested the effect of solver integration timestep length on output precision. The output species 149 concentrations in all simulations but for isoprene OFR (Table 1) converge well as integration timestep 150 decreases (Fig. S1). In the isoprene OFR test cases, the output values oscillate over a small range (<~5%)

151
for integration timesteps ≤ 0.01 s (Fig. S1). Since this numerical error is smaller than typical rate constant 152 measurement uncertainties (from ~10% to a factor of 2-3; Burkholder et al., 2015), let alone the 153 uncertainties related to the SARs used in GECKO-A, it is deemed acceptable for the relevant simulations 154 in this study. The integration timestep for each simulation in the present work is reported in Table 1.

155
We allow mechanism generation to proceed through to CO2 production in most cases in this study.

156
The only exception is for extremely low-volatility species (saturation vapor pressure < 10 -13 atm) which without gas-wall partitioning, gas-particle partitioning is also disabled to avoid artificial condensation of 162 gases into the particle phase. In environments with very low NO (e.g., remote atmosphere), organic

178
For the current study, we have made several updates to GECKO-A, i.e., i) inclusion of key OFR-179 specific radical reactions, ii) extension of the UV range considered to cover 185 and 254 nm, and iii)

180
updates to the low-NO m-xylene oxidation mechanism, so that GECKO-A is able to simulate OFR 181 chemistry and the entire process of low-NO m-xylene photooxidation (until CO/CO2). We will describe 182 these three updates below.  hydroperoxides and their interconverting peroxy radicals under low-NO conditions. We added two low-

223
NO oxidation reactions to the xylenol branch of the meta-xylene oxidation scheme, Scheme S1. In the 224 51% branch, we allow the unsaturated bicyclic peroxide "MXYLOOH" to react with, sequentially, OH

231
To allow GECKO-A outputs, which are usually highly complex and voluminous, to be explored 232 and visualized in detail on standard (non-UNIX) personal computers, we have developed the GECKO

233
Loader and Plotter based in the data-analyzing and graphic-making package Igor Pro 8.0 (WaveMetrics,

234
Lake Oswego, Oregon, USA). This tool assists on the rapid and detailed analysis of model-chamber/OFR 235 comparison studies.

236
Specifically, the GECKO Loader and Plotter facilitates: i) filtering the (sometimes extremely large 237 and finely-resolved) model results time series to examine specific characteristics, ii) identifying the most 238 abundant and/or influential species in each phase (gas, particle, and wall), iii) selecting species by 239 specific chemical identity (molecular formula, specific formula, and/or functional group identity), iv)  In this section, we will show the evolution of OHRVOC in the photooxidation of different precursors 247 under various conditions. To aid the presentation of this evolution for larger precursors, whose oxidation 248 is more complex, the oxidation of the simplest VOC, i.e., methane, will be first discussed. After

249
presenting the results of individual precursors, we will compare the results between conditions and 250 between precursors to illustrate the general trends. Along with the OHR evolution, OH recycling ratio

251
(β1, defined as number of OH molecules generated from organic reactions per OH consumed by organics)

252
and HOx (= OH + HO2) recycling ratio (β2, defined as number of OH and HO2 molecules generated from 253 organic reactions per OH consumed by organics) will also be discussed, as they are important parameters

273
We also performed a simulation under a typical OFR condition. The OHRVOC peak also appears 274 around 1x10 13 molecules cm -3 s in this case for the same reasons discussed above, but its height is almost 275 twice that of the ambient case (Fig. S3). The OHR of CO in both cases is similar, while that of CH3OH 276 is higher in the ambient case but those of CH3OOH and HCHO are significantly higher in the OFR case.

277
This is because the relative importance of the various reactions involved in CH4 oxidation (Scheme S2)

278
depends on the conditions in each reactor.

279
In the OFR case, OH and HO2 concentrations are ~4 and ~3 orders of magnitude higher than 280 typical ambient values, respectively . The reactions of two intermediates, CH3OOH  decane simulations for OFR conditions is lower than that for ambient conditions.

306
These differences from the methane cases arise because a key assumption of the simple reaction 307 chain model, i.e., slow precursor decay allowing intermediates/products to build up and reach a steady 308 state, no longer holds in decane oxidation. The main first-generation products, i.e., secondary decyl 309 hydroperoxides, react with OH only < x3 more rapidly than does decane, as the significant activation it results in only limited OHRVOC enhancement at peak.

322
The differences between the ambient and OFR cases for decane oxidation are for different reasons 323 than in the case of CH4. In the absence of steady state for the nodes (stable species) in the decane 324 oxidation chains (nodes far downstream insufficiently populated), organic photolysis and RO2 self-and 325 cross-reactions only help move OHR contributors to downstream nodes, but do not significantly change 326 their total concentrations. This is shown by the relatively small differences in the composition of stable

327
OHR contributors between the ambient and OFR cases (Fig. 3). The remarkable difference between these 328 cases is the contribution of RO2 to OHR, which is as high as ~3 s -1 in the OFR case shown in Fig. 3,

329
while estimated to be only up to ~0.1 s -1 in the ambient case, given the RO2 concentration in the 330 simulation.

331
It is known that RO2 + OH can be a significant RO2 loss pathway in OFR, especially when OH 332 and HO2 production is relatively strong (higher relative humidity (RH) and UV). We have previously is still sizable (>30%). However, the evolution of the composition of monofunctional species in this OFR 342 case before the OHRVOC peak equivalent age is similar to that in the ambient case ( Fig. 3), as 343 hydroperoxide production through RO2 + HO2 is still the main loss pathway of the first-generation RO2 344 and RO produced from RO2 + OH can also form ketones, i.e., the main second-generation products. The

345
other main fate of RO, i.e., isomerization, leads to slightly faster production of multifunctional species,

346
since the product of the recombination of the immediate product of this isomerization, i.e., an alkyl 347 radical, with O2 is already a bifunctional RO2. This isomerization also creates a hydroxyl group on the C 348 backbone, resulting in a relatively high share of hydroxyl in the functional groups of the multifunctional 349 species (Fig. 4).

350
Before the OHRVOC peak, as OHexp increases, carbonyls accumulate. They are prone to Norrish- indicates. However, fragmentation of multifunctional species does not appear to be significantly weaker 363 in the OFR case than in the ambient case shown in Fig. 3. This is largely due to fast RO2 + OH. The

364
reactions of acylperoxys with OH lead to direct fragmentation (Orlando and Tyndall, 2012  OHexp (~1x10 12 molecules cm -3 s) are formed via C10 fragmentation and are thus of higher volatility 390 (Fig. 3). The heavy wall partitioning of multifunctional species also significantly slows down their 391 oxidative evolution in both the wall phase and the gas phase relative to the ambient cases (Fig. 4).

392
As OHexp increases and large multifunctional species are formed in increasing amounts from 393 oxidation, their near-complete partitioning to the wall decreases the OHR of decane oxidation 394 intermediates/products by a factor up to 8 around 1x10 12 molecules cm -3 s compared to the chamber cases 395 without gas-wall partitioning (Fig. 2). At higher OHexp (long oxidation times) gas-phase concentrations 396 of partitioning species decline, allowing reverse partitioning back from the wall which then serves as a 397 source rather than a sink. As a result, the ratio of the OHR of oxidation intermediates/products in the 398 chamber case with wall partitioning to that without wall partitioning decreases (Fig. 2).

400
As discussed above, we also compute OH (β1) and HOx (β2) recycling ratios in decane oxidation. Note that these quantities also include OH and HO2 generated as a result of organic photolysis. The 402 differences in these recycling ratios between the simulated cases are relatively small. β1 is close to 0 at 403 OHexp < ~1x10 10 molecules cm -3 s (Fig. 1), as the initial reaction of decane with OH only produces an

416
The HOx recycling ratio (β2) in decane oxidation is similar to β1 before ~1x10 11 molecules cm -3 s 417 for the ambient and chamber cases, as only OH (but not HO2) is recycled at this stage. β2 is a little higher 418 in the OFR cases than in the other cases at this stage because of the HO2 recycling by RO2 + OH.

419
However, at higher OHexp, β2 continues to increase with OHexp to a final value of 1 (Fig. 1). This VOCs are the dominant OHR contributors and many of them recycle HO2 during their oxidation by OH 426 (Fig. 3). Finally, once CO becomes the only remaining OHR contributor, β2 is 1. slightly exceeds that of decane, the OHRVOC peak in m-xylene oxidation occurs at slightly lower OHexp 437 than in decane oxidation (Fig. 1). In the OFR case under the same condition as the decane case shown in 438 Fig. 3, the evolution of OHR of the stable organic species is again similar to that in the ambient case.

439
And OHRVOC is higher in the OFR case again mainly due to OHR from RO2 ( Fig. 1 and S4). Several (Scheme S1) and they are also often unsaturated and prone to further functionalization. Therefore, the 442 degree of functionalization in saturated aliphatic multifunctional species is much higher in m-xylene than 443 in decane oxidation (Fig. 4). Also, as several aromatic-scheme-specific reaction types occur in the early 444 stages of m-xylene oxidation, e.g., endo O2 addition (creating -OO-etc.) and ring-opening (creating -

446
Photolysis again plays a role in species fragmentation and the production of highly oxidized C1 and C2 447 species after the OHRVOC peak (Fig. S4).

448
Wall partitioning also substantially reduces the OHRVOC in the relevant chamber cases of m-449 xylene oxidation (Figs. 1, 2 and S4). The precursor (m-xylene) is a C8 species and even many first-450 generation products of its oxidation are highly oxygenated (Scheme S1) lower-volatility species. The

451
relative reduction of OHR of the intermediates/products also increases with OHexp up to ~1x10 12 452 molecules cm -3 s, as volatile species are oxidized and become more prone to wall partitioning (Fig. 2).

453
At higher OHexp, the wall again serves as an OVOC source (Fig. 2).

454
The evolution of β1 and β2 in m-xylene oxidation is somewhat different than in decane oxidation 455 (Fig. 1). In the ambient cases, they are non-negligible even at OHexp.as low as 1x10 9 molecules cm -3 s

466
HO2 recycling occurs in all simulated cases from the beginning of the oxidation (Fig. 1), since two of the 467 three major channels of m-xylene + OH (i.e., those forming MXYEPOXMUC and xylenol, respectively) 468 produce HO2 as well.

469
As more multifunctional species are formed (particularly through ring-opening) near the OHexp of 470 the peak OHRVOC, HOx recycling is also active, with β1 increasing and β2 remaining high (Fig. 1). There

487
The most salient difference of the OHRVOC evolution in the photooxidation of isoprene from that 488 of the other precursors in this study is the lack of OHRVOC peak in the isoprene cases ( Figs. 1 and S5).

489
The decrease in OHRVOC all along this photooxidation is expected since the reaction of isoprene with

490
OH is very fast (at 1x10 -10 cm 3 molecule -1 s -1 ; Atkinson and Arey, 2003) and all intermediates/products 491 of this photooxidation react with OH more slowly than isoprene. The OHRVOC of the 492 intermediates/products peaks slightly after an OHexp of 1x10 10 molecules cm -3 s ( Fig. 1). At this OHexp,

493
the main type of the first-generation products, oxygenated unsaturated species (e.g., isoprene-derived 494 unsaturated hydroxyl hydroperoxides (ISOPOOH)), are largely produced from isoprene + OH and their 495 loss rates (with rate constant with OH slightly lower than that of isoprene) reach the maxima (Fig. S5).

496
Further oxidation leads to the loss of all C=C bonds in the isoprene C backbone and thus a substantial

509
After OHexp~5x10 10 molecules cm -3 s, the deviation caused by chamber wall partitioning becomes 510 more significant as highly oxidized and lower-volatility multifunctional species (Fig. 4) are formed in 511 significant amounts (Figs. 2 and S5). At very high OHexp, the wall again acts as a source of OVOCs in 512 isoprene oxidation, as in those of the other precursors (Fig. 2). The deviations of OFR cases from the 513 ambient cases are mainly caused by RO2 + OH and lack of organic photolysis. These two effects lead to 514 too much HCHO produced and inefficient production of other C1 and C2 species (Fig. S5).

515
To test whether one of the issues, i.e., lack of organic photolysis in OFR, can be mitigated by Chamber light. Such a strong UV source is obviously not realistic, and, while it does increase both early 522 organic photolysis and the relative contribution of C1 and C2 photoproducts to OHRVOC around 2x10 11 523 molecules cm -3 s (Fig. S6), it increases the deviation of this OFR case from the ambient cases at very 524 high OHexp, where oxidation of C1 and C2 species to CO proceeds much more rapidly than in the 525 atmosphere.

526
Product functionality in isoprene oxidation is more diverse than in decane oxidation (Fig. 4). This 527 is due to both the propensity of the isoprene C=C bonds to addition of various groups, and the active 528 isomerization of isoprene oxidation intermediates (Wennberg et al., 2018). Notably, epoxy groups in 529 species such as isoprene-derived epoxydiol (IEPOX) account for a large fraction of saturated product 530 functionality (Fig. 4), particularly at OHexp on the order of 10 10 molecules cm -3 s. In the gas phase of the 531 chamber cases with wall partitioning, the overwhelming majority of saturated multifunctional organic 532 molecules are IEPOX up to 1x10 11 molecules cm -3 s (Fig. 4), as more highly-oxidized species mostly 533 partition to the wall.

535
(ISOPOOH + OH → IEPOX + OH) leads to the peak of OH recycling around 3x10 10 molecules cm -3 s 536 (Fig. 1). OH recycling is active even at very low OHexp (1x10 9 molecules cm -3 s) because a significant 537 amount of ISOPOOH forms early and can recycle OH through its oxidation, except in the OFR cases 538 with strong water vapor photolysis, where ISOPOOH cannot be efficiently formed from first-generation 539 RO2. HO2 recycling is also active in the entire course of the photooxidation (Fig. 1), because of a number 540 of isomerization and photolysis pathways that form alkoxy radicals and highly oxidized C1 species such 541 as HCOOH, HCHO, and CO at very high OHexp (Fig. S5).

543
To explore some general trends of OHR evolution in VOC photooxidation, simulations are 544 performed for the ambient cases with constant UV for two additional alkanes between methane and 545 decane, i.e., butane and heptane. The results of these simulations are compared to the existing analogous 546 cases in Fig. 5. For all cases, the OHRVOC peak height decreases and the OHexp of the OHRVOC peak shifts 547 towards lower OHexp, as the C number of the precursor alkane increases. This can be explained by the 548 fact that the OH rate constants of these alkanes increase with C number, and suggests a possible general 549 trend between OHR peak location and C number.

550
To explore these trends further, we calculate the OHR per unit starting concentration of C atom 551 (in the precursor) in all ambient cases with constant UV in this study (Fig. 5b). In this study, CO2 is not 552 included initially but produced during the oxidation. Therefore, C atoms in the produced CO2 are taken 553 into account in the calculation of OHR per C atom. For real atmospheric cases, initial CO2 is present but 554 should not be considered in this calculation. Note that OHR per C atom has a unit of cm 3 atom -1 s -1 and 555 represents the average contribution to the rate constant with OH of all considered C atoms. Despite large 556 differences among the reactivities of these precursors, the OHR per C atom in the simulations of all 557 precursors but methane converges near an OHexp of 3x10 11 molecules cm -3 s, and then follows a very 558 similar downward trend (Fig. 5b). This OHexp value is roughly where saturated multifunctional species contribution of aromatics, some of which may artificially persist due to mechanism incompleteness, is 562 excluded (Fig. S4). Also, At OHexp > ~3x10 11 molecules cm -3 s, a C atom in saturated multifunctional 563 species on average has at least 0.3 functional groups in the ambient cases (Fig. 4), and the functional 564 group composition is relatively diverse at this OHexp. Therefore, the convergence value of OHR per C 565 atom of ~2x10 -12 cm 3 atom -1 s -1 at ~3x10 11 molecules cm -3 s can be largely regarded as a relatively

569
Before the convergence, isoprene has the highest OHR per C atom (on the order of 10 -11 cm 3 atom -570 1 s -1 ) among the precursors and intermediates/products (Fig. 5b), because of its conjugated C=C bonds.

571
The OHR per C atom of its first-generation oxidation products is slightly lower and close to that of the 572 oxidation intermediates/products of m-xylene, as the main contributors in both cases are oxygenated 573 monoalkenes. The average OHR per C atom of the studied alkanes increases with C number (Fig. 5b),

574
with the upper limit around 1x10 -12 cm 3 atom -1 s -1 consistent with Kwok and Atkinson (1995), since the 575 less-reactive -CH3 groups (with OHR per C atom of ~1x10 -13 cm 3 atom -1 s -1 ) contribute proportionally 576 less to molecular OHR as C number increases. Conversely, the early-stage products of alkane oxidation

577
(mainly alkyl monohydroperoxides) show higher average OHR per C atom for shorter molecules (Fig.   578 5b), owing to the activating (increasing OHR) contribution of the -OOH group.

579
Following the convergence of OHR per C atom, this quantity in all non-methane ambient cases 580 in this study sees a similar decay (Fig. 5b). This coincides with multifunctional species broken into small 581 highly oxidized C1 and C2 compounds. Although among them there are species with OHR per C atom > 582 5x10 -12 cm 3 atom -1 s -1 (e.g., CH3OOH, CH3CHO, and HCHO), the average OHR per C atom of these C1 583 and C2 species are mainly governed by those reacting more slowly (e.g., HCOOH and particularly CO) 584 and hence reaching higher concentrations amid the fast decay of multifunctional species. The similar fast 585 drop of OHR per C atom after OHexp~1x10 12 molecules cm -3 s for various precursors implies a transition 586 from OHR from saturated multifunctional molecules to OHR from CO before the final oxidation to CO2

589
Integrating OHR per C atom over OHexp allows us to assess the average number of OH molecules 590 consumed by each C atom during the entire course of oxidation. This quantity can also be apportioned 591 to the contributions of different OH reactants (Fig. 6). Due to incomplete oxidation of several species, 592 especially CO, the value of this quantity for an oxidation with all C atoms ending up with CO2 should be 593 higher than those at simulation end (OHexp~4x10 12 molecules cm -3 s). We correct this in Fig. 6 by   594 including additional contribution of CO to make its total contribution 1, since CO, the typical penultimate 595 product, consumes one OH molecule in its final oxidation, but is still present in significant quantities at 596 the end of our simulations. Thus, each C atom reacts with OH ~3 times in the course of the oxidation of 597 isoprene and decane to CO2 (Fig. 6). A simplistic and chemically intuitive explanation for this number 598 is that the average oxidation state (OS C ̅̅̅̅̅̅ ) of both isoprene and decane C atoms is ~-2, and needs to increase with O2, which also lowers the number of OH needed, although the effect is usually small.

607
The surprisingly large contribution of hydroperoxy xylenol (C8H10O3) to OH consumed per C 608 atom in m-xylene oxidation (Fig. 6) is an artifact of mechanism incompleteness. This species may 609 undergo an abstraction of the H atom in its -OOH group by OH. The resulting RO2 may be converted

-
In methane oxidation, the intermediates do not gain dominance in sequence. Instead, they 626 simultaneously increase as the oxidation proceeds, then simultaneously decrease when the methane 627 decay becomes significant. The OHR evolution in methane oxidation is close to the idealized steady-628 state chain model, as the reaction of methane with OH is orders of magnitude slower than those of 629 its oxidation intermediates, which allows the intermediates to reach their steady state.

630
The following discussion refers to the non-methane cases.

-
Where different types of species dominate OHRVOC in sequence, OHRVOC increases after the current 632 dominant type converts to one with a higher average OHR per C atom, and vice versa.

633
Photooxidations of alkanes and aromatics follow the increasing trend from precursor to saturated 634 multifunctional species (via alkyl monohydroperoxides) and from precursor to unsaturated 635 oxygenated species, respectively. The increase in aromatic oxidation is likely to be more significant,

654
In general, the OHR evolution differences resulting from different precursors are larger than those due to  decomposition, may also play a major role in this conversion. This also results in significantly higher

696
(lower) production of HCHO (CH3OOH) in OFR than in the atmosphere at high equivalent ages.

697
With all the key findings in this study presented above, we believe that we have, to some extent, 698 addressed all the three issues for OHR studies raised by Williams and Brune (2015). We largely speciated 699 the likely source of the "missing reactivity", i.e., multifunctional species, by the fully explicit GECKO- has been shown to be similar for most VOC oxidations. This parametrization may be utilized in regional