Non-OH chemistry in oxidation flow reactors for the study of atmospheric chemistry systematically examined by modeling

Introduction Conclusions References


Introduction
For decades, environmental chambers have been employed to study atmospheric chemical processes, particularly volatile organic compound (VOC) oxidation processes in the atmosphere (Cocker et al., 2001;Carter et al., 2005;Presto et al., 2005;Wang et al., 2011;Platt et al., 2013), without the interference of some transport processes (e.g., advection and wet deposition).These oxidation processes are the key to secondary organic aerosol (SOA) formation (Odum et al., 1996;Hoffmann et al., 1997;Hallquist et al., 2009) and air pollutant removal (Levy, 1971).Atmospheric simulation chambers usually have volumes on the order of several m 3 and use light sources longer than 300 nm (e.g., sunlight or UV black lights) to generate oxidants (mainly OH).These settings leads to OH concentrations (10 6 -10 8 molecules cm −3 ) that are not much higher than the typical ambient values (10 6 -10 7 molecules cm −3 ; Mao et al., 2009).Relatively low OH concentrations require long residence/simulation times (generally hours) and limit the ability of those setups to reach very high photochemical ages that are atmospherically relevant (George et al., 2007;Kang et al., 2007;Carlton et al., 2009;Seakins, 2010;Wang et al., 2011).Residence times are ultimately limited due to losses of gases and particles to Teflon walls with timescales of tens of minutes to several hours (Cocker et al., 2001;Matsunaga and Ziemann, 2010;Zhang et al., 2014) as well as by the limited volume of the bag relative to the sampling instrumentation (Nguyen et al., 2014).Besides, large sizes and support systems (e.g., clean air generators) make it difficult to use large chambers in field or source studies.
Oxidation flow reactors (OFRs) are an alternative that offers some advantages over environmental chambers, especially for rapid changes of experimental conditions and/or for field experiments.They generally have a smaller size (on the order of 10 L) and typically use low-pressure Hg lamps as light sources for producing OH in large amounts via O 3 and/or H 2 O photolysis.These design choices lead to good portability, short experimental timescales, ability to reach long photochemical ages, and potentially reduced wall losses.
Due to these advantages, OFRs have been employed in many recent field and laboratory studies in atmospheric chemistry, particularly in SOA-related research (George et al., 2007;Kang et al., 2007Kang et al., , 2011;;Smith et al., 2009;Mas-soli et al., 2010;Cubison et al., 2011;Lambe et al., 2011aLambe et al., , b, 2012Lambe et al., , 2013;;Bahreini et al., 2012;Saukko et al., 2012;Wang et al., 2012;Ortega et al., 2013;Li et al., 2013).OFRs are also used in related applied fields, such as scrubbing of pollution from air (Johnson et al., 2014).In contrast to their popularity, the chemistry occurring in OFRs is still incompletely characterized, although the formation and interconversion reactions of most oxidants in OFRs have been well characterized (Sander et al., 2011;Ammann et al., 2015).To our knowledge, there are only three studies of OFR radical oxidation chemistry to date.Ono et al. (2014) focused on the dependence of O 3 destruction on H 2 O concentration.We have recently made progress on the characterization of HO x radical chemistry in OFRs (Li et al., 2015;Peng et al., 2015).We have developed a kinetic model for OFRs, which provides predictions in good agreement with laboratory experiments.This model has also shown that OH exposure (OH exp , i.e., OH concentration integrated over the reactor residence time) increases with H 2 O concentration and UV intensity, and decreases with external OH reactivity (OHR ext = k i c i , i.e., the sum of the products of concentrations of externally introduced OH-consuming species (c i ) and rate constants of their reactions with OH (k i )).The OH exp decrease due to OHR ext was defined as "OH suppression" and can reach 2 orders of magnitude in some cases (Li et al., 2015;Peng et al., 2015).We also showed that relative uncertainties of the outputs of our box model (e.g., OH exp ) due to uncertain kinetic parameters are typically only 20 % (Peng et al., 2015).However, none of these studies directly address the fate of VOCs (including oxygenated VOCs), simply regarded as external OH reactants in the prior studies.
The primary reason for the use of the OFRs studied here is for the study of reactions of species or mixtures of atmospheric relevance with the OH radical.However, other highly reactive species are also present at very elevated concentrations, including the radicals O( 1 D) and O( 3 P), 185 and 254 nm photons, and O 3 .If a substantial fraction of the species of interest reacted with those non-OH reactants, then the chemistry in the OFR would deviate from the OH radical chemistry intended to investigate.The absence of systematic research on VOC fate in OFRs leaves room for some speculation that non-OH or non-tropospheric chemistry can play a major role in OFRs; for example, Johnson et al. (2014) suggested that O( 1 D) and O( 3 P) significantly consumed VOCs.Klems et al. (2015) concluded that photons at 254 nm from Hg-lamp emission played an important role in their OFR experiment, especially for downstream chemistry.Lack of clarity about these types of questions and of clear guidelines about how to apply OFRs to avoid such problems have limited the application of OFRs for years.In this paper, we apply the model in Peng et al. (2015) to systematically investigate whether significant non-tropospheric or non-OH chemistry occurs in OFRs and what experimental conditions make it more important.Considering the enormous complexity of organic radical (particularly organic peroxy) chemistry, we only examine the non-OH fate of stable species in the present work.The fate of organic radicals should be the subject of future studies.The results allow improved OFR operation and experimental design, as well as guidance for the design of future reactors.

Methods
The OFR and the model used here have been described in detail elsewhere (Kang et al., 2007;Li et al., 2015;Peng et al., 2015).Here, we only present a brief introduction for each.Kang et al. (2007) first introduced the PAM flow reactor.Although there were earlier versions of the PAM reactor, the version of cylindrical geometry with a volume of ∼ 13 L has been widely used and is currently in use in many SOA research groups (Massoli et al., 2010;Cubison et al., 2011;Kang et al., 2011;Lambe et al., 2011aLambe et al., , b, 2012Lambe et al., , 2013;;Bahreini et al., 2012;Saukko et al., 2012;Wang et al., 2012;Li et al., 2013;Ortega et al., 2013).The reactor is made of aluminum or of glass and aluminum and equipped with 1-4 low-pressure Hg lamps (model no.82-9304-03, BHK Inc.) located inside the flow tube.The Hg lamps produce UV emissions at 185 and 254 nm, the intensity of which can be rapidly computer-controlled.The operation mode using both 185 and 254 nm emissions is called "OFR185".In this mode, photons at 185 nm dissociate H 2 O and O 2 molecules to produce OH + H and O( 3 P), respectively.Recombination of O( 3 P) with O 2 forms O 3 .UV light at 254 nm then photolyzes O 3 to produce O( 1 D), which reacts with H 2 O and produces additional OH.OFR can also be operated in another mode where photons at 185 nm are filtered by quartz sleeves around the lamps.In this case, only 254 nm UV light is active to generate OH ("OFR254" mode), and injection of externally generated O 3 into the reactor is required for OH production.The amount of injected O 3 plays a critical role in the OFR chemistry (Peng et al., 2015).For this reason this amount (X ppm) is also included in OFR operation mode notation in the form of OFR254-X.For example, OFR254-70 and OFR254-7 denote experiments with 70 and 7 ppm O 3 injected, respectively.We use the PAM as the basic OFR design.Other designs will be specified below if needed.Rapid computer-controlled UV lamp setting allows rapidly scanning UV lamp settings during an experiment and has unique applications to OFR experiments in field studies (Hu et al., 2015;Ortega et al., 2015;Palm et al., 2016).In these experiments, OFRs enable the exploration of a very large range of photochemical age during a short period (∼ 2 h) when ambient conditions often do not significantly change.

Model description
We use the same model as in Peng et al. (2015), a standard chemical-kinetic model under plug-flow conditions.The effect of non-plug-flow residence time distributions (RTDs) was also investigated in that study.Non-plug-flow models result in similar OH exp than plug flow in most cases, except under specific conditions with very high UV, H 2 O, and OHR ext (Peng et al., 2015).Therefore, plug-flow OH exp is used in this study, as a proxy of OH exp estimated from OHR ext decay and to avoid the much increased computational expense for complex RTDs.All O x and HO x reactions available in the JPL Chemical Kinetic Data Evaluation (Sander et al., 2011) are taken into account.Reactions of some external OH reactants (externally introduced reactants destructing OH), such as SO 2 , CO, and NO x , are also included.SO 2 is used as a proxy of other external OH reactants (e.g., VOCs).We believe that this is a realistic approximation in terms of OHR ext decay vs. OH exp for many precursors, as discussed in Peng et al. (2015).
When studying the OFR, we assume a residence time of 180 s and use typical temperature (295 K) and atmospheric pressure (835 mbar) in Boulder, CO, USA.H 2 O mixing ratio (abbr.H 2 O hereafter) ranges from 0.07 to 2.3 % (equivalent to relative humidity, RH, of 2-71 %).According to Li et al. (2015), UV photon fluxes (abbr.UV hereafter) at 185 and 254 nm are estimated to be in the ranges 1.0 × 10 11 -1.0 × 10 14 and 4.2 × 10 13 -8.5 × 10 15 photons cm −2 s −1 , respectively.OHR ext at 0 and between 1 and 1000 s −1 , covering the range of most field and laboratory studies, are investigated.In the explored parameter space, the same threecharacter labels as in Peng et al. (2015) are used to denote typical cases (Table 1).For OFR254, we study OFR254-7 to -70, representing OFR254 experiments with low (Kang et al., 2011;Liu et al., 2015) and high O 3 (Palm et al., 2016), respectively.To model literature OFR studies, we adopt corresponding parameters (reactor volume, H 2 O, residence time, etc.), and estimate parameters that are specified or measured (e.g., UV) as needed.In particular, for some field studies, where long time-series of experimental data (42 days in BEACHON-RoMBAS (Palm et al., 2016), 42 days in SOAS (Hu et al., 2015), and 15 days in CalNex-LA (Ortega et al., 2015)) were recorded, we model all valid data points and present outputs in the form of histograms.Note that the outputs for field studies, i.e., histograms, have a complex dependence on ambient temperature, H 2 O, and OHR ext , as well as UV steps used.The specific histogram shapes for different field campaigns are influenced by both ambient and experimental parameters.

Results and discussions
In the following sections we explore the relative importance of five non-OH pathways (photons at 185 and 254 nm, Table 1.Code of the labels of typical cases.A case label is composed of three characters denoting the water mixing ratio, the photon flux, and the external OH reactivity, respectively.

Water mixing ratio Photon flux
External OH reactivity Options L is low L is low (10 11 photons cm −2 s −1 at 185 nm; 0 (0.07 %) 4.17 × 10 13 photons cm −2 s −1 at 254 nm) M is medium M is medium (10 13 photons cm −2 s −1 at 185 nm; L is low (10 s −1 ) (1 %) 1.45 × 10 15 photons cm −2 s −1 at 254 nm) typically for remote or clean urban air H is high H is high (10 14 photons cm −2 s −1 at 185 nm; H is high (100 s −1 ) (2.3 %) 8.51 × 10 15 photons cm −2 s −1 at 254 nm) typically for polluted urban air "L" is lowest in PAM by Li et al. (2015) V is very high (1000 s −1 ) (7.9 × 10 11 photons cm −2 s −1 at 185 nm; only for laboratory experiments 2.04 × 10 14 photons cm −2 s −1 at 254 nm) Example LH0: low water mixing ratio, high photon flux, no external OH reactivity O( 1 D), O( 3 P), and O 3 ) vs. the OH reaction for species of atmospheric interest, including a variety of typical biogenic and anthropogenic VOCs and a few important inorganic species (e.g., SO 2 and NO 2 ).Because of the huge complexity of VOC oxidation mechanisms, only consumption/oxidation of specific VOCs is investigated.In such an investigation, a large amount of kinetic data is required.We collected the required data (Tables S1-S3 in the Supplement) according to the principles in Sect.S1.Photolysis of SOA is also investigated.

Fractional loss of VOCs to non-OH reactants
As shown in Peng et al. (2015) S5 present the same information in an alternative format that may be useful to evaluate the fate of species not included in our study.They show ratios of rate constants of species with OH to those with non-OH species (X), mathematically equivalent to the X exp / OH exp where corresponding X / (OH + X) is 50 %.X / (OH + X) denotes the fractional importance of X in the sum of the contribution of OH and X to VOC.If a VOC has a higher ratio of rate constant with OH to that with a non-OH reactant (X) than with another reactant (Y), the relative contribution of X, X / (OH + X), should be smaller than that of Y.In these figures, we also show the X exp / OH exp fractional occurrence distribution for OFR254 and OFR185, under different conditions, including key laboratory and field studies, and identify the X exp / OH exp ranges where non-OH contribution to species fate is significant.Unless specifically stated, a dissociation quantum yield of unity is assumed for the photolysis reactions, which results in upper limits for the relative importance of those pathways.In these figures, the relationships of all non-OH reactive species to OH are similar for certain common conditions.We define three types of conditions to help guide experimental design and evaluation in terms of the relative importance of non-OH reactants.Under "riskier conditions" of high/very high OHR ext (≥ 100 s −1 in OFR185 and > 200 s −1 in OFR254-7 to -70) and/or low H 2 O (<0.1 %), non-OH reactions can be significant depending on the species.Conversely, under "safer conditions" with relatively low OHR ext (< 30 s −1 in OFR185 and < 50 s −1 in OFR254) and high H 2 O (> 0.8 % in OFR185 and > 0.5 % in OFR254), and moderate or higher UV (> 1 × 10 12 photons cm −2 s −1 at 185 nm) in OFR185, reaction with OH is dominant .We denote all other conditions as "transition conditions".High H 2 O and zero/low OHR ext lead to strong OH production and no/weak OH suppression, respectively.Thus, OH Fractional importance of the photolysis rate at 185 nm of several species of interest vs. the reaction rate with OH, as a function of the ratio of exposure to 185 nm photons (F185) and OH.F185 exposure (in photons cm −2 ) is the product of 185 nm photon flux (in photons cm −2 s −1 ) and time (in s).F185 exposure / OH exposure is thus in cm s −1 .The modeled frequency distributions of ratios of 185 nm photon exposure to OH exposure under riskier, safer, and transition conditions for OFR185 are also shown.The curves of aromatics and inorganic gases are highlighted by solid dots and upward triangles, respectively.The lower inset shows histograms of model-estimated F185 / OH exposures for three field studies where OFR185 was used to process ambient air.Their ordinate is the fractional occurrence of a given condition (X exp / OH exp ).All histograms are normalized to be of identical total area (i.e., total probability of 1).The upper inset (black and blue markers) shows similar information for source studies of biomass smoke (FLAME-3;Ortega et al., 2013) and an urban tunnel (Tkacik et al., 2014).All curves, markers, and histograms in this figure share the same abscissa.
is more abundant and dominates species consumption under those conditions.In the case of low H 2 O and high OHR ext , OH is generally lower because of less production and more suppression.These conditions increase the relative contribution of non-OH species.UV light intensity is generally less influential on non-OH VOC fate than H 2 O and OHR ext , although OH production is nearly proportional to UV (Peng et al., 2015), because the non-OH reactive species also scale (nearly) proportional to UV.As a result, UV generally has smaller effects on exposure ratios between OH and the non-OH reactants.However, under a UV near the lower bound of the explored range in this study (< 1×10 12 photons cm −2 s −1 at 185 nm) in OFR185, OH production is so small that the effect of OHR ext on OH suppression can be amplified and hence some exposure ratios may be affected.In OFR254 OH is more resilient to suppression even at low UV because of the OH recycling by initially injected O 3 (Peng et al., 2015).Note that we call these conditions "riskier" and "safer" mainly in terms of non-tropospheric VOC fate but not of VOC fate by all non-OH reactants, as some of the non-OH reactant studied in this work may also play a role under some tropospheric conditions (see Sect. 3.1.4 and 3.1.5).In addition to the common features above, individual non-OH reactants have their own features as well as a few exceptions to the above mentioned general observations, which we will detail below.

Reactions with OH vs. photolysis at 185 nm
Under riskier reactor conditions, photolysis at 185 nm of several aromatic compounds, such as toluene, benzene, and pxylene, are estimated to be significant and even dominant vs. the OH reactions (Fig. 1).This results from their aromatic ring, which is not only highly efficient as a chromophore but also relatively resistant to OH attack.
It is not common to perform field studies for SOA at H 2 O as low as 0.1 % or OHR ext ≥ 100 s −1 (Table 1 OFR185 was deployed, i.e., BEACHON-RoMBAS (Palm et al., 2016), SOAS (Hu et al., 2015) and CalNex-LA (Ortega et al., 2015), all these studies are generally under safer conditions (infrequent low H 2 O mixing ratio and ambient OH reactivity estimated to be ∼ 15-25 s −1 ).For instance, none of these field studies fell into the conditions where the fractional importance of photolysis at 185 nm was significant for aromatic species.However, in some source studies using OFRs, e.g., when sampling biomass burning smoke (FLAME-3;Ortega et al., 2013) or air in a traffic tunnel (Tkacik et al., 2014), OHR ext can be very high, reaching values of several 100 s −1 (Table S4).Especially on the smoke study, photolysis of aromatics may have played a role.However, it has long been known that excited aromatic molecules may undergo various deactivation pathways (e.g., vibronic coupling, intersystem crossing, and collisional quenching) without molecular fragmentation (Beddard et al., 1974;Nakashima, 1982;Nakashima and Yoshihara, 1983;Fang and Phillips, 2002), preventing unity quantum yields.Therefore, the photolysis of aromatics at 185 nm in the above mentioned source studies may not be as significant as estimated in Fig. 1.Under riskier conditions, some organic peroxy nitrates and nitrates (e.g., peroxyacetyl nitrate and 2-propyl nitrate in Fig. 1) have an estimated contribution from photolysis at 185 nm to their fate that is comparable to or even larger than that of reaction with OH.Nevertheless, this does not mean that we need to make extra efforts to avoid the photolysis of organic nitrates and peroxy nitrates at 185 nm.Although they have cross sections at 185 nm ∼ 10-100 times smaller than those of aromatics, these organic compounds react remarkably slowly with OH (∼ 2 orders of magnitude slower than reactions of aromatics with OH), so photolysis appears substantial with respect to reaction with OH, but really two slow rates are being compared.Thus, absolute photolyzed amounts of these species are not substantial.For example, only ∼ 10 % of peroxyacyl nitrate is photolyzed by 185 nm photons at the highest OFR185 lamp setting.Even if photolysis of nitrates and peroxy nitrates by low-pressure Hg lamp emission proceeds to a significant extent, it may still not be a problem, as it generally leads to the same products as their ambient photolysis (Renlund and Trott, 1984;Roberts and Fajer, 1989).Nitrate and peroxy nitrate photolysis is actually much more important in the atmosphere than in OFRs for the same photochemical age (see below and Fig. 6).
SO 2 has been used in some studies to calibrate OH exp (Lambe et al., 2015;Li et al., 2015).It does undergo significant photolysis at 185 nm under riskier conditions.However, this photolysis does not lead to an overestimation of OH exp , since the S-bearing product of SO 2 photolysis at 185 nm, SO, converts back to SO 2 very rapidly through its reaction with O 2 .
Oxidation intermediates may also photolyze at 185 nm.However, their photolysis is unlikely to be significant when OFR is operated under safer conditions.To clarify this issue, a detailed discussion about the photolysis of oxidation inter-mediates at 254 nm is required as a premise.We thus discuss oxidation intermediate photolysis at both 185 and 254 nm in Sect.3.1.3.

Reactions with OH vs. photolysis at 254 nm
The photon flux in the reactor at 254 nm is 80-250 times larger than at 185 nm (Li et al., 2015).Although absorption cross sections of all molecules investigated in this study are significantly lower at 254 than 185 nm, the higher photon flux compensates, at least partially, for this effect, so that in OFR185 photolysis of many species at 254 nm is of similar relative importance as photolysis at 185 nm, with potentially important effects at low H 2 O and/or high OHR ext (Figs. 2 and  S2).As for 185 nm, 254 nm photolysis is a concern mainly for aromatic compounds, because of the high light absorptivity and low OH reactivity of aromatic rings as previously discussed.Again, note that this concern may be less serious than shown in Fig. 2 because of possible lower quantum yields.Photolysis of organic nitrates and peroxy nitrates at 254 nm also appears to be important relative to reactions with OH and is still not a concern for the same reasons as photolysis at 185 nm.SO 2 can absorb efficiently at 254 nm, but it is still not a problem for SO 2 -based OH exp calibration, since photons at 254 nm are not sufficiently energetic to dissociate SO 2 molecules.
High UV generally appears to be more problematic than low UV in OFR254 (Fig. S2).This is in contrast to the trend of OFR185.In OFR254, O 3 is the only primary OH source, and a substantial fraction of O 3 can be photolyzed at the highest lamp settings, leading to a substantial reduction of OH production (compared with proportional scaling with UV).OFR254-70 appears to be less prone to riskier conditions than OFR254-7, since higher O 3 favors HO 2 -to-OH recycling, making OH more resilient to suppression (Peng et al., 2015).
Under highly risky conditions (H 2 O < 0.1 % and OHR ext ≥ 100 s −1 for OFR185, and H 2 O < 0.1 % and OHR ext > 200 s −1 for OFR254), some saturated carbonyl compounds (e.g., pyruvic acid, methyl ethyl ketone, and hydroxyacetone) have significant photolysis at 254 nm relative to reactions with OH.This significant relative contribution of photolysis also results from remarkably slow reactions of saturated carbonyl compounds with OH.Although secondary species without C=C double bond, e.g., saturated carbonyls, hydroperoxides, and nitrates, can be photolyzed in OFRs at low H 2 O and/or high OHR ext , their photolysis only proceeds to a ∼ 10-1000 times smaller extent than ambient photolysis at the same photochemical age (Fig. 6).
Unsaturated carbonyls may have much higher absorption cross sections if their C=C bonds are conjugated with carbonyls.However, according to our following analysis, conjugated unsaturated carbonyl compounds do not often cause problems of non-tropospheric photolysis at 254 nm.Car- bonyls have π -π * and n-π * transitions.The former corresponds to high cross section (typically > 10 −18 cm 2 ) and typically occurs around or below 200 nm.The latter is forbidden and thus has weak absorption (cross section on the order of 10 −19 cm 2 or lower) and typically occurs around or above 300 nm (Turro et al., 2009).Conjugation usually does not substantially enhance the absorption of n-π * transition, but it does for π -π * transitions (Turro et al., 2009).As a result, through conjugation, the only reason why cross sections of carbonyls at 254 nm may be elevated above 10 −18 cm 2 is the redshift of the maximum absorption wavelength of their ππ * transitions due to conjugation.According to Woodward's rules (Pretsch et al., 2009) and available cross-section data of α,β-unsaturated carbonyls in Keller-Rudek et al. (2015), a conjugation of at least 3-4 double bonds is required for the excitation at 254 nm to dominantly correspond to π -π * transition.Conjugated oxidation intermediates containing at least 3-4 double bonds including C=C bond(s) are virtually impossible to form from aliphatic hydrocarbon oxidation in OFRs.Nevertheless, such intermediates may form via ringopening pathways of aromatic oxidation (Calvert et al., 2002;Atkinson and Arey, 2003;Strollo and Ziemann, 2013).E,E-2,4-hexadienedial may be regarded as an example of this type of intermediates.Even under assumption of a unity quantum yield, its fraction of photolysis at 254 nm is not much higher than that of aromatic precursors (Fig. 2).Therefore, 254 nm photolysis of conjugated intermediates should not be problematic as long as safer experimental conditions are adopted.
To our knowledge, the only exception that has strong absorption at 254 nm due to conjugation with <2 double bonds are β-diketones, which may be formed in aliphatic hydrocarbon oxidation, particularly that of long-chain alkanes (Ziemann and Atkinson, 2012).The peculiarity of β-diketones is that their enol form may have a highly conjugated ring structure due to very strong resonance (Scheme S1) and hence cross sections on the order of 10 −17 cm 2 at 254 nm (Messaadia et al., 2015).However, even under the assumption of unity quantum yield, the fractional contribution of 254 nm Z. Peng et al.: Non-OH chemistry in oxidation flow reactors for the study of atmospheric chemistry photolysis of acetylacetone (representative of β-diketones) is only slightly larger than for aromatic VOCs (Fig. 2), since its enol form also contains a C=C bond leading to very high reactivity toward OH.Furthermore, we argue that the actual probability that a concrete structural change (in number and type of functional groups, O / C ratio, average C oxidation state, etc.) of β-diketones resulting from photoexcitation at 254 nm may be low.As their excitation at 254 nm corresponds to π -π * transition, their rigid ring structure likely hinders cyclic structural change at the first singlet excited state (S 1 (π ,π * )) while the biradical structure of the first triplet state (T 1 (π ,π * )) may favor H shift between two O atoms, which ends up with the same/similar structure than prior to the H shift (Scheme S1).Also, the excitation of βdiketones at 254 nm may also lead to charge transfer complex formation via direct excitation and/or radiationless transition from a local excited state (Phillips and Smith, 2015), which is likely to result in low quantum yields, as discussed in detail below.
In addition to conjugated species, Phillips andSmith (2014, 2015) reported a new type of highly absorbing species that may be formed from VOC oxidation.Although their studies were conducted in the condensed phase, it is likely that the main conclusions of these studies are generally transferable to the gas-phase conditions, since no long-range interactions, which do not exist in normal gases, were involved in these studies.Phillips andSmith (2014, 2015) investigated the photoabsorption enhancement of multifunctional oxygenated species in SOA and found that the high absorptivity of these species can largely be explained by transitions toward the electronic states of charge transfer complex formed between hydroxyl groups (donor) and spatially neighboring carbonyl groups (acceptor).They also pointed out that charge transfer complexes of this kind have a continuum of states whose energy levels range from that of local excited states (radiative transition wavelength < 300 nm) to very low levels (radiative transition wavelength > 600 nm).The latter are insufficient to cause common photochemical reactions.Relaxation through a continuum of states is usually ultrafast according to Fermi's golden rule (Turro et al., 2009), likely leading to low quantum yields of chemical reactions.The low quantum yields may be seen even from species with only one hydroxyl and one carbonyl: the photolysis of 3-hydroxy-3-methyl-2butanone and 4-hydroxy-2-butanone at wavelengths down to 270 nm has quantum yields around only 0.1 (Bouzidi et al., 2014(Bouzidi et al., , 2015)).Although measurements of photolysis quantum yield for multifunctional species are challenging and rare, it is reasonable to expect even lower quantum yields for larger and/or highly substituted (by hydroxyl and carbonyl) species, since larger species have more degrees of freedom for relaxation of excited molecules, and more and/or larger complex sites generally lead to more efficient relaxation through a continuum of states, in accordance with common photophysical sense (Sharpless and Blough, 2014).Therefore, even though species with a number of hydroxyls and carbonyls are formed in VOC oxidation and can absorb > 1 order of magnitude more efficiently at 254 nm than mono-and difunctional species, they may still have low effective photolysis rates because of low quantum yields.
For this type of species, we estimate an upper limit of the fractional importance of their photolysis at 254 nm.Molar absorption coefficients of charge transfer transitions of organic molecules are usually ∼ 10 3 -1 × 10 4 L mol −1 cm −1 , i.e., cross sections of ∼ 3.9 × 10 −18 -3.9 × 10 −17 cm 2 (Foster, 1969).Based on that, it is reasonable to estimate an upper limit of absorption cross sections of charge transfer transitions of 5 × 10 −17 cm 2 .However, photolysis quantum yields of multifunctional species are unlikely to be larger than that of species with only one carbonyl and one hydroxyl, i.e., ∼ 0.1 (see discussion above).We thus take 0.1 as an upper limit of photolysis quantum yields.Besides, 6×10 −12 cm 3 molecule −1 s −1 can be a conservative estimate of rate constants of multifunctional oxygenated species with OH, as it is roughly an average value for ketones (Atkinson and Arey, 2003), and the enhancement of H-abstraction by hydroxyl groups (Kwok and Atkinson, 1995;Ziemann and Atkinson, 2012) and the fast abstraction of aldehydic H atoms (Atkinson and Arey, 2003) are completely neglected.With the three estimates combined, the estimated maximum fractional contribution from photolysis at 254 nm to the fate of multifunctional species (Fig. 2) is close to that of E,E-2,4-hexadienedial and acetylacetone.
The problem of photolysis of oxidation intermediates at 185 nm is unlikely to be worse than at 254 nm.According to available UV spectra of carbonyl compounds in Keller-Rudek et al. (2015), 185 nm is almost always located within the π -π * transition band, whose maximum cross section is on the order of 10 −17 cm 2 .Even if all types of radiative transitions at normal radiation intensity are considered, an approximate upper limit of absorption cross sections is ∼ 10 −16 cm 2 (Evans et al., 2013).However, UV intensity at 185 nm in the OFR185 mode is ∼ 100 times lower than that at 254 nm (Li et al., 2015).The photolysis rate of oxidation intermediates at 185 nm should thus be generally smaller than at 254 nm.
Therefore, in summary, photolysis of oxidation intermediates are, to our knowledge, conservatively estimated to be of limited importance relative to their reactions with OH, as long as the experimental conditions are in the safer range.Although studies on photolysis quantum yields of oxidation intermediates are very sparse, we reason, based on the existing studies on this topic and common photophysical and photochemical rules, that the photolysis quantum yields of these species may be lower than the values assumed in this study (e.g., 1 for E,E-2,4-hexadienedial and acetylacetone and 0.1 for multifunctional species).As a result, actual rates and relative importance of photolysis might be significantly smaller than the upper limits estimated in our study.
As discussed for photolysis at 185 nm, in all ambient OFR field studies (BEACHON-RoMBAS, SOAS, and CalNex-LA), reactions with OH dominate over photolysis at 254 nm (Fig. 2).The fractional consumption of several anthropogenic aromatic VOCs, such as benzene and naphthalene, in the urban CalNex-LA campaign by 254 nm photolysis is estimated as a few percent under most conditions and at most ∼ 15 %.At the BEACHON-RoMBAS and SOAS forested sites, photolysis at 254 nm should be a negligible contributor to the fate of biogenic VOCs such as isoprene and monoterpenes.
Some laboratory and source studies may have had an appreciable contribution to aromatic species fate from 254 nm at low H 2 O and/or high OHR ext .F254 exp / OH exp in the biomass smoke and urban tunnel source studies (source OHR ext up to ∼ 300 s −1 ; Ortega et al., 2013;Tkacik et al., 2014) and the Kang et al. (2011) laboratory study (H 2 O down to ∼ 0.1 %) can be as high as 10 6 -10 7 cm s −1 .In this range, photolysis of a few aromatic VOCs (e.g., benzene and naphthalene) at 254 nm could account for ∼ 20-80 % of their destruction.
Note that photolysis of oxidation intermediates also needs to be taken into account.If multifunctional species, βdiketones, and extensively conjugated species are photolyzed as shown in Fig. 2, these photolyses would be significant in some previous source and laboratory studies examined here, as they were conducted at relatively low H 2 O and/or high OHR ext .To our knowledge, none of these studies reported a significant photolysis of oxidation intermediates.Klems et al. (2015) attributed large amounts of fragmentation products detected in their OFR experiments with dodecanoic acid to photolysis of peroxy radicals.However, these products may also be at least partially accounted for by photolysis of carbonyls leading to carbon-chain cleavage via Norrish reactions (Laue and Plagens, 2005).The OFR used by Klems et al. (2015) has a different design from the PAM, which is regarded as the base design in this study.Their reactor employs a light source stronger than the PAM's highest lamp setting, with UV at 254 nm estimated to be ∼ 3×10 16 photons cm −2 s −1 (∼ 4 times the value at the highest lamp setting of the PAM OFR) based on the lamp power and the reactor geometry.Such high UV may even result in significant photolysis of saturated carbonyl intermediates, which are very likely formed in the oxidation of long-chain alkane-like dodecanoic acid.

Reactions with OH vs. reactions with O( 1 D) and O( 3 P)
The results for these two radicals are shown in Figs.3-4 (Monks, 2005), except under some riskier conditions.It is believed that the contribution of O( 1 D) to VOC destruction in the atmosphere should be negligible (Calvert et al., 2002), and their relative importance is even lower under most OFR conditions.Reactions with O( 3 P) are small or negligible contributors to VOC consumption except under extreme riskier conditions.Unless at low H 2 O (< 0.1 %) and very high external OH reactivity (∼ 1000 s −1 ), VOC consumption by O( 3 P) cannot be larger than 10 % of that by OH (Figs. 4 and S4).This results from both very low concentrations of O( 3 P) and relatively low reactivity compared to that of OH.Among the species that we examine, biogenic VOC consumption may have some contribution from O( 3 P) under the abovementioned riskier conditions, due to the higher reactivity of double bonds in these species with O( 3 P).For example, α-pinene in the mixture experiments in Kang et al. (2011) may have had a ∼ 5 % contribution from O( 3 P).Similarly to O( 1 D), O( 3 P) exp / OH exp in the troposphere (Calvert et al., 2002) is higher than in the OFR except for riskier conditions.Thus the relative importance of both O( 1 D) and O( 3 P) to OFR chemistry is typically lower than in the troposphere.

Reactions with OH vs. reactions with O 3
Reaction with O 3 is a major, even dominant pathway of the consumption of many biogenic VOCs in the troposphere.However, it is of interest to quantify the relative importance of OH vs. O 3 across OFR experiments (Figs. 5 and S5).This allows comparison with the relative importance in the troposphere, as well as potentially designing experiments where the relative influence of O 3 is minimized or adjusted as desired.
A large amount of O 3 is injected into OFR254, and O 3 concentration in that type of reactor does not change much with UV flux (negligibly under most conditions and up to a factor of ∼ 2 at high H 2 O and UV; Peng et al., 2015).Since OH exp is proportional to UV flux, as UV decreases, OH exp is lowered and the fractional species destruction by O 3 increases.In OFR185, O 3 production is almost linearly dependent on UV, while a significant portion of OH production has a quadratic relationship with UV.Thus OH increases faster with increasing UV than O 3 .Therefore, lower UV in OFR185 also leads to a higher relative importance of O 3 for VOC consumption.
The distribution of O 3exp / OH exp expected for the troposphere was obtained from the GISS ModelE2 climate   model (Schmidt et al., 2014) and is estimated as the ratio of the simulated daily mean concentration of O 3 and OH on a horizontal grid of 2 • in latitude and 2.5 • in longitude for the year 2000.Interestingly, the simulated relative importance of O 3 to OH in the troposphere is higher than when OFRs are operated under safer conditions, and similar to OFRs when they are operated under riskier conditions (Figs. 5 and S5).In those cases, a number of biogenic VOCs can be significantly consumed by O 3 .In particular, some monoterpenes (e.g., α-terpinene) and sesquiterpenes (e.g., βcaryophyllene) have a fractional reaction with O 3 close to 100 % in the troposphere.In contrast to biogenics, reactions with O 3 do not play any role in the consumption of most anthropogenic VOCs, e.g., benzene, toluene, and alkanes.Besides, ozonolysis of saturated oxidation intermediates (e.g., carbonyls and alcohols) is minor or negligible in both OFRs and the atmosphere, since they react with OH at ∼ 10 −13 -10 −11 cm 3 molecule −1 s −1 while their ozonolysis rate constants are < 10 −20 cm 3 molecule −1 s −1 (Atkinson and Arey, 2003).However, unsaturated oxidation intermediates may have larger contributions from O 3 because of C=C bonds.In particular, dihydrofurans, possible intermediates of satu-rated hydrocarbon oxidation (Ziemann and Atkinson, 2012;Aimanant and Ziemann, 2013), may be predominantly oxidized by O 3 in the troposphere.In OFR254, they can still have significant contributions from O 3 even outside riskier conditions.
An experimentalist may be interested in obtaining an O 3exp / OH exp in an OFR close to ambient values, which requires lower H 2 O and higher OHR ext conditions, although care should be taken to avoid other non-tropospheric reactions under those conditions.However, one may want to study OH-dominated chemistry and thus want to avoid significant ozonolysis of VOCs to reduce the complexity of VOC fate.This is analogous to the addition of excess NO to suppress O 3 in some chamber experiments.In this case the OFRs should be operated under opposite conditions, i.e., high H 2 O, high UV, and low OHR ext .This strategy enhancing OH exp is effective for most VOCs, except those with the highest k O 3 /k OH ratios, e.g., α-terpinene and βcaryophyllene.To further decrease the importance of reactions of VOC with O 3 , it is necessary to lower the O 3 concentration.For OFR254, one can inject less O 3 into the reactor and increase the UV lamp setting.The comparison between Figure 6.Ambient photolysis fractions of secondary species in a week (calculated from photolysis rates reported in Hodzic et al., 2015) vs. photolysis fractions of those species in OFR185 and OFR254-70 when reaching the same photochemical age (ambient OH concentration of 1.5 × 10 6 molecules cm −3 assumed) under conditions of 44 % relative humidity (water vapor mixing ratio of 1.4 %) and 25 s −1 initial OHR ext .If the points of a certain species for both OFR185 and OFR254-70 are available, the species name is tagged on the OFR185 point (downward arrow); otherwise it is tagged on the OFR254-70 point (upward arrow).The 1 : 1, 1 : 10, 1 : 100, and 1 : 1000 lines are also shown for comparison.
OFR254-70 and OFR254-7 in Fig. S5 demonstrates this approach.For OFR185, we propose another strategy, i.e., lowering O 2 concentration in the reactor.This decreases O 3 production but affects OH production to a much lesser extent, thanks to the major OH production by H 2 O photolysis.We simulate the OFR185 cases with 2 ‰ O 2 and observe that VOC ozonolysis can be excluded at high H 2 O and high UV (Fig. S5).
Among the literature OFR studies, the field studies employing OFRs in urban and forested areas all operated under O 3exp / OH exp values 100 times lower than in the atmosphere.In these field studies reaction of almost all VOCs with O 3 can be neglected, except for the most reactive biogenics with O 3 , e.g., α-terpinene and β-caryophyllene.The source study in an urban tunnel of Tkacik et al. ( 2014) operated under similar conditions.Some laboratory studies using OFR254 (Kang et al., 2011;Lambe et al., 2011b) as well as the biomass smoke source study (Ortega et al., 2013) operated at O 3exp / OH exp close to tropospheric values, because the injected O 3 plays a key role for OFR254 studies and the biomass smoke experiments were conducted at high OHR ext .Nevertheless, only α-pinene and β-pinene, both biogenics, are significantly consumed by O 3 in these studies.Another OFR254 study, Klems et al. (2015), had O 3exp / OH exp significant lower than tropospheric values, since the initial O 3 in their experiment was only 2 ppm, and the UV light in their experiment was stronger than our lamps' highest setting and further reduced effective O 3 .

Reactions with 1 O 2 and HO 2
Singlet oxygen, 1 O 2 , can be produced in various ways in OFRs (Calvert et al., 2002;Ono et al., 2014) and react with alkenes at appreciable rate constants (∼ 10 −17 -10 −14 cm 3 molecule −1 s −1 ; Huie and Herron, 1973;Eisenberg et al., 1986).We estimate 1 O 2 concentration by the expression proposed by Ono et al. (2014).Only at the lowest H 2 O, the highest lamp setting, and the highest OHR ext in this study (Table 1), may the most reactive alkene (endocyclic conjugated dienes, e.g., cyclopentadiene, α-terpinene, and α-phellandrene) have > 10 % contribution from 1 O 2 to their fate.For all other species and under all other conditions, reactions of VOCs with 1 O 2 are negligible.Thus, this reactant is not discussed further in the present work.
HO 2 is a major radical in the OFR chemistry.However, it is much less reactive than OH toward VOCs.Typically, the rate constants of reactions of HO 2 with alkenes are smaller than 10 −20 cm 3 molecule −1 s −1 at room temperature, and those with almost all saturated VOCs (except aldehydes and ketones) are even smaller (Tsang, 1991;Baulch et al., 1992Baulch et al., , 2005)).Therefore, we briefly discuss reactions of HO 2 with aldehydes and ketones and neglect those with all other VOCs in this study.Ketones react with HO 2 at rate constants on the order of 10 −16 cm 3 molecule −1 s −1 or lower (Gierczak and Ravishankara, 2000;Cours et al., 2007).Therefore, only at low H 2 O, low UV, and high OHR ext , the reaction of acetone with HO 2 may compete with that with OH.The same is likely true for the reactions of acetaldehyde and larger aldehydes with HO 2 , as their rate constants are likely to be around or less than 1 × 10 −14 cm 3 molecule −1 s −1 (da Silva and Bozzelli, 2009).Formaldehyde is the only stable carbonyl compound that may react with HO 2 (rate constant: 7.9 × 10 −14 cm 3 molecule −1 s −1 ; Ammann et al., 2015) at a rate competing with that with OH under conditions that are not low H 2 O, low UV, and high OHR ext .Note that the reaction of formaldehyde with HO 2 is also significant in the atmosphere (Pitts and Finlayson, 1975;Gäb et al., 1985).However, its product, hydroxymethylperoxy radical, dominantly undergoes decomposition via thermal reaction and photolysis (Kumar and Francisco, 2015), compared to the hydroxymethyl hydroperoxide formation pathway via a further reaction with HO 2 (Ziemann and Atkinson, 2012).Even if hydroxymethyl hydroperoxide is produced in appreciable amounts, in the high-OH environment of OFRs, this species can be easily predicted to convert into formic acid (Francisco and Eisfeld, 2009) and eventually CO 2 .All these products have very few interactions with other VOCs and hence should not significantly perturb the reaction system of OFRs.

Overall contribution of non-OH reactants to gas-phase chemistry
In this section we summarize the combined effect of all non-OH reactants to VOC consumption.However, we can no longer use X exp / OH exp to express total non-OH VOC consumption as for individual reactants.Total non-OH VOC consumption is thus discussed case by case.
In the explored range of conditions (i.e., H 2 O, UV, OHR ext , and initial O 3 for OFR254), there are obviously conditions where all non-OH fates of VOCs are negligible.Most simply, the highest H 2 O and UV in this study and a very small non-zero OHR ext result in a VOC consumption nearly 100 % by OH, regardless of the VOC type (Table S6).Lowering UV can make non-OH contribution even smaller for OFR185, but not for OFR254.This difference occurs because OH production is reduced while O 3 roughly remains at the same level in OFR254, leading to enhanced relative contribution from O 3 to the fate of biogenics.At the lowest nonzero UV in the PAM reactor (7.9 × 10 11 photons cm −2 s −1 at 185 nm; 2.0 × 10 14 photons cm −2 s −1 at 254 nm) of Li et al. (2015), the fractional destructions of α-pinene (representative of biogenic VOCs) by O 3 are 21 and 4.2 % in OFR254-70 and OFR254-7, respectively.Other OFR designs may reach lower UV, e.g., the "low UV" case defined in this study (1.0 × 10 11 photons cm −2 s −1 at 185 nm; 4.2 × 10 13 photons cm −2 s −1 at 254 nm).At these UV levels, the fate of α-pinene by O 3 further increases to 44 and 13 % in OFR254-70 and OFR254-7, respectively.
However, non-OH reactants can dominate VOC fate under opposite conditions that lead to low OH production and strong OH suppression.At the lowest H 2 O and UV and the highest OHR ext in this study, > 95 % of α-pinene and ∼ 80 % of toluene are consumed by non-OH reactants in OFR185.In OFR254, almost all α-pinene has non-OH fate while non-OH fate of toluene is still minor or negligible.If OFRs are operated at the low UV setting from the PAM by Li et al. (2015), ∼ 8 times higher the lowest UV in this study, the situation hardly changes, as a very large OH suppression persists.Nevertheless, if UV is increased to the highest level, non-OH fate of α-pinene is lowered to ∼ 60-70 % and that of toluene in OFR185 decreases to 24 %.When H 2 O is at the highest level, non-OH fate is systematically lower than at low H 2 O in all types of OFRs.In particular, the HHV case has non-OH fates of α-pinene only up to ∼ 10 % and negligible non-OH fates of toluene, despite very high OHR ext .
We also summarize VOC fate for key laboratory, source, and field studies examined in the present work in Fig. 7.For each case, the fate of one or a few typical VOCs is investigated.In laboratory studies, Kang et al. (2011) performed experiments with a mixture of α-pinene, m-xylene, and pxylene, and one of experiments by Lambe et al. (2011b) used biogenics (α-pinene and isoprene, respectively) as precursors.In both cases, O 3 plays a role in the fate of biogenics when H 2 O is low (Kang et al., 2011), or OHR ext is high and UV is low (Lambe et al., 2011b), as shown in Fig. 7.The fate of isoprene by O 3 is less significant despite the very high OHR ext , because isoprene, compared to α-pinene, is more reactive with OH and less reactive with O 3 .Besides, O( 3 P) contributes up to a few percent to the fate of biogenics.In the literature experiments performed at a higher H 2 O (Kang et al., 2011) or a higher UV (Lambe et al., 2011b), non-OH fate of both VOCs significantly decreases because of increases in OH exp .About 20 % of p-xylene in the Kang et al. (2011) mixture experiment at very low H 2 O may be destroyed by 254 nm photons, under the assumption of unity quantum yield.Other laboratory study cases with aromatics have lower non-OH fates because of higher H 2 O. N-decane in one of Lambe et al.'s experiments and dodecanoic acid in the Klems et al. (2015) study are consumed ∼ 100 % by OH, as these alkane(-like) species neither react rapidly with O 3 , nor absorb UV efficiently.However, as previously discussed, some carbonyl compounds may be formed and significantly photolyzed at 254 nm in the Klems et al. (2015) experiments, although the huge complexity of intermediates and limited knowledge of reaction mechanisms prevents a quantitative assessment of the fate of carbonyl intermediates by photons at 254 nm.
Source and field studies usually have highly complex precursors.For the urban tunnel study (Tkacik et al., 2014) and the CalNex-LA study in the Los Angeles basin (Ortega et al., 2015), we choose toluene as a representative of  S4. aromatic species, as these are major anthropogenic VOCs and SOA precursors in urban environment (Dzepina et al., 2009;Borbon et al., 2013;Hayes et al., 2015;Jathar et al., 2015).Although alkanes are also major anthropogenic VOCs, their non-OH fate is not quantitatively assessed for the same reason as discussed for dodecanoic acid in the Klems et al. (2015) experiment.For the smoke aging study, FLAME-3 (Ortega et al., 2013), we select benzene and αpinene, which are important VOCs in biomass burning emissions (Warneke et al., 2011).For the BEACHON-RoMBAS and SOAS studies at forested sites, α-pinene and isoprene are chosen, respectively, as they are major emitted biogenic VOCs at those corresponding sites.Both the urban tunnel and FLAME-3 studies have aromatic precursors significantly photolyzed at 185 and 254 nm (assuming quantum yield = 1) under the conditions of high source concentrations (Fig. 7).The toluene fate by UV in the tunnel study is less substantial than that of benzene in FLAME-3, since NO x , the largest fraction of external OH reactant in the tunnel study, is converted into HNO 3 very rapidly (Li et al., 2015) and does not further suppress OH.In the cases of lower OHR ext (e.g., the tunnel experiments with low source concentration and Cal-Nex), toluene is dominantly consumed by OH.It also holds for biogenic VOCs that non-OH fate decreases with decreasing OHR ext due to less OH suppression.The non-OH fate of α-pinene in FLAME-3, dominated by reaction with O 3 , is larger than 20 %, while the non-OH fates of α-pinene in BEACHON-RoMBAS and of isoprene in SOAS are both negligible, since OHR ext in the former study is > 10 times higher than in the latter two studies.

SOA photolysis
Recently, photolysis in the UV range has been found to be a potentially significant sink of some types SOA in the troposphere (Updyke et al., 2012;Lambe et al., 2013;Liu et al., 2013Liu et al., , 2015;;Hodzic et al., 2015;Wong et al., 2015;Romonosky et al., 2016).It is necessary to also investigate SOA photolysis in OFRs, as photons used in OFRs are highly energetic and non-tropospheric.UV extinction due to aerosol optical scattering and in-particle absorption under OFR conditions is generally negligible (Hodzic et al., 2015).For simplicity, we estimate photodegradation ratios of various SOA component surrogates as well as several SOA samples whose absorptivity was measured in previous studies (Updyke et al., 2012;Lambe et al., 2013;Liu et al., 2013;Romonosky et al., 2016) (Fig. 8) under the assumption of unity quantum yield to obtain upper limits of photodegradation ratios, as well as under the assumption of lower (0.1 and 0.01) quantum yields.
Most SOA functional groups are oxygenated (e.g., peroxides, carbonyls, carboxylic acids, alcohols).The absorption cross sections of most of these functional groups are too low at 185 and 254 nm given the OFR residence time and UV light intensity, leading to a small contribution of photolysis to SOA degradation (Fig. 8).For example, glycolaldehyde has a negligible fractional contribution of photolysis except at the highest lamp setting, when only ∼ 5 % of this species photolyzes at each wavelength.Species (e.g., isoprene) with conjugated double bonds as efficient chromophores will not be present in SOA because of their high reactivity with OH and O 3 .Nitrate groups may have a ∼ 30 % contribution from photolysis at the highest UV settings and a negligible contribution at intermediate or low UV settings.
Aromatic rings are more resistant to OH attack and usually strongly absorb UV light.Under our assumptions, the photolysis of some aromatic SOA components (e.g., o-cresol at 185 nm and naphthalene at 254 nm) is already important at medium UV flux.At the highest lamp setting, most aromatics in SOA would be destroyed if the quantum yields are indeed near unity.However, as previously discussed, photolysis quantum yields of aromatics may be significantly lower than 1.This is more probable in the condensed phase (Damschen et al., 1978;Baker et al., 2015) than in the gas phase, as quenching processes in the condensed-phase matrix are usually much more efficient than through gas-phase molecular collisions.It has recently been reported that photolysis quantum yields of aromatics in SOA were low under UVB irradiation (Romonosky et al., 2015).Although the range and relevance of the species investigated in that study are limited, it is reasonable to assume low quantum yield for aromatic photolysis in SOA at 185 and 254 nm.Wong et al. (2015) conducted α-pinene-derived SOA photolysis experiments in a chamber under UVB irradiation (down to 284 nm).They observed at 85 % RH ∼ 30 % SOA photolyzed after > 30 min irradiation and a photolysis quantum yield of ∼ 1 during the first 10 min.However, in OFRs such a high SOA photodegradation percentage would not occur, since the Wong et al. (2015) experiments had a high photon flux (∼ 4×10 15 photons cm −2 s −1 ) and a long irradiation time, and hence a photon flux exposure that is ∼ 5 times that at the highest lamp setting in the OFRs modeled in our work.According to the measurements of Wong et al. (2015), a photolysis fraction of ∼ 6 % would be expected for this type of SOA in our OFRs under the highest UV flux, with lower percentages at lower UV settings.In addition, the approximate unity quantum yield observed in Wong et al. (2015) may be due to (hydro)peroxides in α-pinene-derived SOA, since peroxides have high photolysis quantum yields (Goldstein et al., 2007;Epstein et al., 2012), while other functional groups (i.e., mainly hydroxyl and carbonyl) in oxygenated species in SOA are unlikely to have for reasons discussed below.
Note that a simple addition of absorptivities of different functional groups may not explain SOA absorptivity (Phillips and Smith, 2015).According to the absorption data of SOA samples from Lambe et al. (2013) and Romonosky et al. (2016), real SOA absorbs ∼ 1-3 orders of magnitude more than non-aromatic component surrogate species shown in Fig. 8 at 254 nm.As discussed for multifunctional oxidation intermediates (with carbonyls and hydroxyls), SOA absorption enhancement may be largely due to transitions of charge transfer complexes formed between carbonyls and hydroxyls in multifunctional oxygenated SOA components (Phillips andSmith, 2014, 2015).These complexes between carbonyls and hydroxyls also have continua of states likely leading to ultrafast relaxation and hence low photolysis quantum yields.Charge transfer transitions have been extensively shown in measurements (Alif et al., 1991;Gao and Zepp, 1998;Johannessen and Miller, 2001;O'Sullivan et al., 2005;Zhang et al., 2006;Osburn et al., 2009;Sharpless and Blough, 2014) to have very low quantum yields in the condensed phase.Sharpless and Blough (2014) compiled quantum yields of various products of humic-like matter photolysis down to 280 nm.No quantum yields except those of the product 1 O 2 , which is generally unimportant for OFRs (see Sect. 3.1.6),are higher than 0.01.If the photolysis quantum yields of the SOA samples in Fig. 8 at 254 nm are no more than 0.01, no SOA samples will be photolyzed by 20 % even at the highest OFR lamp setting, and photolysis of most SOA samples at 254 nm will be minor or negligible in OFRs.Thus, to our current knowledge, lack of solid information on quantum yields  Lambe et al., 2013, andRomonosky et al., 2016) are also shown.
of SOA components with multiple carbonyls and hydroxyls at 254 nm prevents a clear assessment of SOA photolysis in OFRs at the medium and high UV.Nonetheless, direct measurements are desirable for this issue and caution should still be exercised for OFR experiments at relatively high UV.
SOA photolysis at 185 nm may be lower compared to that at 254 nm.SOA absorptivity data at 185 nm are not available.According to SOA mass-specific absorption cross-section (MAC) data between 250 and 300 nm in Romonosky et al. (2016), there is a linear relationship between the logarithm of MAC and wavelength for most SOA samples: MAC increases by a factor of ∼ 3 per 50 nm decrease in wavelength.
We thus extrapolate this relationship to 185 nm, where MAC is estimated to be ∼ 3.5 times higher than that at 254 nm.However, the UV flux at 185 nm in our OFR is ∼ 100 times lower than at 254 nm.
Based on the discussion above, the SOA photodegradation ratio of ∼ 30 % in the non-OFR setup of Wong et al. (2015) may be explained.α-Pinene-derived SOA has ∼ 20-50 % weight fraction of peroxides (Docherty et al., 2005;Epstein et al., 2014), which may undergo photolysis in SOA to convert into carbonyls (and hydroxyls) (Epstein et al., 2014).We speculate that after the formation of carbonyls from peroxides, SOA materials cannot proceed significantly further with photolysis as discussed for charge transfer between carbonyl and hydroxyl above.In the experiments of Wong et al. (2015), as well as Epstein et al. (2014), effective photolysis rate constants/quantum yields decreased as SOA photolysis proceeded.Photolysis rates were substantially reduced after a ∼ 30 % mass loss due to photolysis in the Wong et al. (2015) experiments.This mass loss ratio is consistent with the mass percentage of peroxides in α-pinene-derived SOA.Again, we note that, according to the extrapolation from the results from Wong et al. (2015), the mass loss percentage expected in our OFR under the highest UV flux is ∼ 6 % for α-pinene-derived SOA.This value is much lower than that shown in Fig. 8 under the assumption of unity quantum yield (∼ 40 %) because of a substantially decreasing quantum yield in the real photolysis experiments.Therefore, in OFRs, even if (hydro)peroxides in SOA may be photolyzed in appreciable amounts, SOA mass is unlikely to be largely destroyed by photons in OFRs, as (hydro)peroxides may convert into carbonyls and hydroxyls, which may substantially lower subsequent photolysis quantum yields.
According to the discussion above, measurements of quantum yields and/or products of SOA photolysis are highly desirable, especially for the photolysis of SOA containing dominantly carbonyl and hydroxyl groups, as (hydro)peroxides, which are likely to form in OFRs, may convert into hydroxyls and carbonyls.With more data on quantum yields of SOA photolysis, a clearer strategy for including or excluding SOA photolysis in OFRs can be made.
Even though SOA photolysis can be significant in OFRs, it only proceeds to a much lesser extent compared to ambient SOA photolysis.We calculate the numbers of e-fold decay of SOA photolysis in OFR254-70 and the troposphere according to the effective ambient photolysis lifetime of SOA from Romonosky et al. (2016).Under the condition of 44 % RH (H 2 O = 1.4 %) and OHR ext = 25 s −1 (typical of ambient conditions), SOA samples are estimated to undergo ∼ 0.01-0.5 e-fold photolysis timescales (i.e., ∼ 1-35 % OA photolyzed) in OFR254-70 at an equivalent photochemical age of 1 week under the upper limit assumption of unity quantum yields (Table S8).However, in the atmosphere, those samples may proceed with 10 2 -10 4 e-fold decays of photolysis (i.e., virtually complete destruction) at the same photochemical age, if ambient SOA photolysis quantum yields are assumed to be those of H 2 O 2 (unity below 400 nm).Even if the quantum yield of acetone (non-zero below 320 nm; see Romonosky et al., 2016) is taken as a surrogate for SOA, most types of SOA would still be completely or nearly completely photolyzed under ambient conditions.These results demonstrate that ambient SOA photolysis is likely to be much more important than in OFRs.However, they also highlight the need for studies of ambient SOA photolysis quantum yields and photolytic aging, as ambient SOA is unlikely to be completely destroyed by photons within only 1 week.Either their quantum yields are much lower than used in this study or the photolabile groups are destroyed and leave behind others that are not (or less) photolabile during photolytic aging.

Guidelines for OFR experimental design and operation
It is necessary to avoid significant non-tropospheric chemistry in OFRs in order to more accurately simulate tropospheric aging.Only photolysis at 185 and 254 nm are important non-tropospheric pathways in OFRs and reactions with O atoms are generally unimportant.Ozonolysis is also a major VOC sink in the troposphere, and the desirability of including or excluding its effects depends on the goals of each experiment.
In the cases where the exclusion of VOC ozonolysis is desired, there is no dilemma for the experimental design, as the exclusion of both VOC ozonolysis and non-tropospheric VOC consumption requires similar conditions, i.e., safer conditions.As shown above, all examined field studies do not have significant non-tropospheric contribution to VOC fate, while some past laboratory and source studies do because of low H 2 O and/or very high OHR ext in those experiments.It is possible to improve the latter experiments by increasing H 2 O and/or lowering OHR ext .In detail, source humidification and dilution can be feasible measures to increase H 2 O and decrease OHR ext , respectively.For example, increasing RH from 3 to 60 % (H 2 O from ∼ 0.08 to ∼ 1.6 %) lowers the percentage of non-tropospheric consumption of p-xylene in the Kang et al. (2011) mixture experiment from ∼ 20 to 1.5 %.Humidifying the average condition of the BEACHON-RoMBAS (Palm et al., 2016) campaign from H 2 O = 1.6 % (RH = 63 %) to H 2 O = 2.3 % (RH = 92 %) leads to a ∼ 50 % decrease in non-tropospheric photolysis of α-pinene (Table S4).Also, a 5-fold source dilution in FLAME-3 reduces the non-tropospheric fate of benzene from > 60 to ∼ 15 % (Table S4 and "Improved" cases in Fig. 7).Injecting less precursor is a simple way to keep a reasonably low OHR ext in laboratory studies.The comparison between the cases with high and low concentrations in the urban tunnel study (Tkacik et al., 2014) is a good example (Table S4 and Fig. 7).Note that when taking the measures above to limit non-tropospheric VOC fate, one generally reduces the contribution from all non-OH reactants.Increasing H 2 O and source dilution also significantly lower the relative importance of ozonolysis in the fate of α-pinene in the Kang et al. (2011) mixture experiments and the FLAME-3 study (Table S4 and Fig. 7).Although increasing UV may increase OH production, OH reactant destruction, and hence the relative contribution of OH to VOC fate in some cases, one has to be cautious when taking this measure to reduce effective OHR ext , as high UV may cause non-tropospheric photolysis of SOA.
In laboratory experiments, running OFRs under safer conditions ensures a minor contribution of non-tropospheric photolysis, based on the current knowledge of oxidation in- termediate photolysis (Fig. 2).This also reduces the relative contribution of ozonolysis to VOC fate.However, when more information becomes available about photolysis quantum yields of oxidation intermediates (vs. the upper limits assumed here), there may be additional flexibility to include ozonolysis while excluding non-tropospheric VOC consumption.As the precursor composition is usually relatively simple in laboratory experiments, it is sufficient to ensure the insignificance of non-tropospheric consumption of only the precursor(s) and possible intermediates (usually oxygenated species), rather than for a large variety of VOC precursors and intermediates.For example, in the case of quantum yields significantly lower than used in the present work, we may perform OFR254-70 experiments with a large amount of biogenics at medium H 2 O and UV.In this case, a tropospheric O 3exp / OH exp ratio can be obtained without major side effects, because the fractional contribution of photolysis of possible intermediates is still minor due to low quantum yields.However, OFR experiments with some anthropogenic VOCs, such as alkanes, can just be conducted at high H 2 O and low OHR ext to avoid the contribution of all non-OH reactants, since ozonolysis of alkanes is negligible even at a tropospheric O 3exp / OH exp .OFR experiments can be simply conducted under safer conditions to avoid non-tropospheric VOC fate, while riskier conditions can lead to significant non-tropospheric VOC fate, depending on the species under study.The conditions in between, i.e., "transition" conditions, are explicitly discussed above.However, one may want to be able to more quantitatively estimate the relative importance of non-OH reactants under different conditions so that a more detailed experimental planning becomes possible that simultaneously ensures insignificant non-tropospheric VOC fate and specific experimental goals.For this purpose, we provide a series of estimation equations for non-OH reactant exposures (Sect.S3, Table S9, and Fig. S6, as well as Excel file).With these equations, the relative contribution of non-OH reactants under all conditions explored in this study can be easily estimated.In OFR studies where a different OFR design is adopted and/or chemistry beyond the approximations in our model is involved, a new model may need to be established, which can be done in similar manner as Peng et al. (2015), to obtain the relative importance of non-OH VOC fate and then perform experimental design.

Conclusions
We used a kinetic model to investigate non-OH contribution (from 185 and 254 nm photons, O( 1 D), O( 3 P), and O 3 ) to VOC destruction, as well as to SOA photolysis at 185 and 254 nm in OFRs.We assessed the relative significance of the VOC consumption by non-OH reactants to that by OH in OFRs and the troposphere.The only non-tropospheric reaction that can play a major role under OFR conditions is photolysis, especially at 254 nm.The relative importance of photolysis is largest under riskier OFR conditions where OH is low due to low H 2 O and/or high OHR ext .Due to lack of quantum yield data, we estimated upper limits of the relative importance of photolysis for the few most susceptible oxidation intermediates, which are comparable to those from aromatic precursors.Reactions of O atoms are not competitive and are actually of lower relative importance (vs.OH) in OFRs than in the troposphere.VOC ozonolysis is much less important than in the troposphere under typical OFR conditions and of similar importance under riskier OFR conditions.Photolysis of SOA in OFRs could be significant at medium and high UV, but only if corresponding quantum yields are high.If SOA photolysis quantum yields are of the order of 0.01 or lower, as measured for many humic-like substances (Sharpless and Blough, 2014), SOA photolysis in OFRs may be minor or negligible under most conditions.Although the reaction fates may be different, numbers of e-fold decays of photolysis for a given OH exp are at least an order of magnitude lower in the OFRs compared to the troposphere.
We examined some past field, laboratory, and source studies using OFRs.In the field studies of aged urban and forest ambient air, non-OH VOC fate was not important because of relatively high H 2 O and moderate OHR ext .However, some laboratory and source studies were conducted at low H 2 O and/or high OHR ext and have significant non-tropospheric VOC consumption.Humidification and/or dilution are recommended in these cases to reduce the importance of nontropospheric reactants.We proposed different approaches to avoid non-OH VOC consumption, as well as strategies to employ insignificant non-tropospheric photolysis and significant tropospheric ozonolysis simultaneously in laboratory experiments.Our work has implications for the interpretation of past OFR studies and should be useful for designing and conducting future OFR experiments for atmospheric research, as well as in related applied fields.
The need for systematic measurements of photolysis quantum yields, for both VOC and SOA, and both at actinic wavelengths and at 185 and 254 nm, was highlighted in this study.When quantum yield data become available, photolysis of oxidation precursors, oxidation intermediates, and SOA in OFRs can be much better quantified, its relative importance compared to OH oxidation, ambient photolysis, etc. can be better evaluated, and experimental planning might then be able to be less conservative and have more freedom to avoid non-tropospheric photolysis and realize specific experimental objective(s).
The Supplement related to this article is available online at doi:10.5194/acp-16-4283-2016-supplement.

Figure 1 .
Figure 1.Fractional importance of the photolysis rate at 185 nm of several species of interest vs. the reaction rate with OH, as a function of the ratio of exposure to 185 nm photons (F185) and OH.F185 exposure (in photons cm −2 ) is the product of 185 nm photon flux (in photons cm −2 s −1 ) and time (in s).F185 exposure / OH exposure is thus in cm s −1 .The modeled frequency distributions of ratios of 185 nm photon exposure to OH exposure under riskier, safer, and transition conditions for OFR185 are also shown.The curves of aromatics and inorganic gases are highlighted by solid dots and upward triangles, respectively.The lower inset shows histograms of model-estimated F185 / OH exposures for three field studies where OFR185 was used to process ambient air.Their ordinate is the fractional occurrence of a given condition (X exp / OH exp ).All histograms are normalized to be of identical total area (i.e., total probability of 1).The upper inset (black and blue markers) shows similar information for source studies of biomass smoke(FLAME-3;Ortega et al., 2013) and an urban tunnel(Tkacik et al., 2014).All curves, markers, and histograms in this figure share the same abscissa.

Figure 2 .
Figure 2. Same format as Fig.1, but for the fractional importance of the photolysis rate at 254 nm vs. the reaction rate with OH as a function of the ratio of exposure to 254 nm (F254) and OH.The modeled frequency distributions of ratios of 254 nm photon exposure to OH exposure under riskier, safer, and transition conditions for OFR185 and OFR254 (-7 to -70) are also shown.The curves of saturated carbonyl compounds and possible highly absorbing oxidation intermediates are highlighted by downward triangles and squares, respectively.The insets show histograms of model-estimated F254/OH exposures for three field studies where OFR185 was used to process ambient air.In addition to source studies of biomass smoke (FLAME-3) and urban tunnel(Tkacik et al., 2014), F254 exposure / OH exposure ratios in two laboratory studies(Kang et al., 2011;Lambe et al., 2011b) are shown in the upper inset.Colored tags indicate species used in the laboratory experiments.The lower and upper limits of F254 exposure / OH exposure ratios in the experiments with a certain source in a certain study are denoted by tags below and above the markers, respectively.

Figure 3 .
Figure 3. Same format as Fig. 2 but for the ratio of the reaction rate with O( 1 D) vs. OH as a function of the relative exposure to O( 1 D) and OH.A typical value of the relative exposure of O( 1 D) and OH in the troposphere estimated according to Monks (2005) is also shown.

Figure 4 .
Figure 4. Same format as Fig. 3 but for the ratio of the reaction rate with O( 3 P) vs. OH as a function of the relative exposure of O( 3 P) and OH.A typical value of the relative exposure of O( 3 P) and OH in the troposphere from Calvert et al. (2002) is also shown.

Figure 5 .
Figure 5. Same format as Fig. 2 but for the fractional importance of the reaction rate with O 3 vs.OH as a function of the relative exposure of O 3 and OH.The curves of biogenics are highlighted by squares.Also shown are modeled distributions of the relative exposure of O 3 and OH at the Earth's surface (solid line) and throughout the column from the surface to a height with a pressure of 150 hPa (dashed line).The distributions were calculated from the mean daily concentrations of O 3 and OH as simulated by the GISS ModelE2.

Figure 7 .
Figure 7. VOC fate in several representative cases of the laboratory, source, and field studies examined in this work.More details on VOC fate in these studies can be found in TableS4.

Figure 8 .
Figure 8. Percentage of SOA photodegradation at (upper panel) 185 and (lower panel) 254 nm at different UV levels as a function of absorption cross section under the assumptions of quantum yields of 1, 0.1, and 0.01.Absorption cross sections of some SOA component surrogates (black tag) and SOA samples (orange tag; calculated from data in Lambe et al., 2013, and Romonosky et al., 2016) are also shown.
et al.: Non-OH chemistry in oxidation flow reactors for the study of atmospheric chemistry

.1 Common features of all the non-OH reactants
This could be important if CH 4 was used for OH exp calibration under riskier reactor conditions, or if the fate of CH 4 is important to the experiment for other reasons.Other VOCs react much faster with OH.As a result, reactions of VOCs (other than CH 4 ) with O( 1 D) in all laboratory, field, and source studies previously discussed are almost always negligible.We also note that the ratio of O( 1 D) exp / OH exp in the OFR is actually much lower than in the troposphere . The potential impact of O( 1 D) is smaller than for 185 and 254 nm photons, due to the low concentration of O( 1 D) in the reactor.Only for methane may reaction with O( 1 D) be significant, because the reaction of methane with O( 1 D) is close to the collision rate, while CH 4 is the most resistant VOC to H-abstraction by OH.