HO x and NO x production in oxidation flow reactors via photolysis of isopropyl nitrite , isopropyl nitrite-d 7 , and 1 , 3-propyl dinitrite at λ = 254 , 350 , and 369 nm

Oxidation flow reactors (OFRs) are an emerging technique for studying the formation and oxidative aging of organic aerosols and other applications. In these flow reactors, hydroxyl radicals (OH), hydroperoxyl radicals (HO2), and nitric oxide (NO) are typically produced in the following ways: photolysis of ozone (O3) at λ= 254 nm, photolysis of H2O at λ= 185 nm, and via reactions of O(1D) with H2O and nitrous oxide (N2O); O(1D) is formed via photolysis of O3 at λ= 254 nm and/or N2O at λ= 185 nm. Here, we adapt a complementary method that uses alkyl nitrite photolysis as a source of OH via its production of HO2 and NO followed by the reaction NO+HO2→NO2+OH. We present experimental and model characterization of the OH exposure and NOx levels generated via photolysis of C3 alkyl nitrites (isopropyl nitrite, perdeuterated isopropyl nitrite, 1,3propyl dinitrite) in the Potential Aerosol Mass (PAM) OFR as a function of photolysis wavelength (λ= 254 to 369 nm) and organic nitrite concentration (0.5 to 20 ppm). We also apply this technique in conjunction with chemical ionization mass spectrometer measurements of multifunctional oxidation products generated following the exposure of α-Pinene to HOx and NOx obtained using both isopropyl nitrite and O3+H2O+N2O as the radical precursors.


Introduction
Hydroxyl (OH) radicals govern the concentrations of most atmospheric organic compounds, including those that lead to secondary organic aerosol (SOA) formation.The relative importance of different primary OH precursors varies in different parts of the atmosphere and may include contributions from O( 1 D)-H 2 O reactions, hydrogen peroxide (H 2 O 2 ), methyl peroxide (CH 3 OOH), nitrous acid (HONO) photolysis, and ozone-alkene reactions.Additionally, ozonehydroperoxy (HO 2 ) reactions and NO-HO 2 reactions recycle HO 2 back to OH (Mao et al., 2009;Lee et al., 2016).For decades, a handful of radical precursors have been used to generate OH radicals in the laboratory to initiate SOA production under controlled conditions.Environmental chambers most commonly photolyze nitrous acid (HONO), methyl nitrite (CH 3 ONO), or hydrogen peroxide (H 2 O 2 ) at λ > 310 nm to mimic SOA production, over experimental timescales of hours to days, simulating up to 2 days of equivalent atmospheric exposure (Atkinson et al., 1981;Matsunaga and Ziemann, 2010;Chhabra et al., 2011;Finewax et al., 2018).
Oxidation flow reactors (OFRs) photolyze H 2 O and O 3 at λ = 185 and 254 nm over experimental timescales of minutes, simulating multiple days of equivalent atmospheric ex-posure (Lambe et al., 2012;Peng et al., 2015).Recent application of O( 1 D) + H 2 O + N 2 O reactions to study NO xdependent SOA formation pathways facilitated characterization of oxidation products generated over a range of low-to high-NO x conditions (Lambe et al., 2017;Peng et al., 2018).Potential limitations of the method include (1) the inability to unambiguously deconvolve contributions from multiple oxidants (O 3 , OH, NO 3 ), which may compete with each other under certain conditions and for specific unsaturated precursors; (2) required use of 254 nm photolysis, which may enhance photolytic losses that compete with OH oxidation, especially for species that are characterized by strong absorption/quantum yield at 254 nm and low-OH reactivity (Peng et al., 2016); (3) optimal high-NO x application at OH exposures corresponding to multiple equivalent days of oxidative aging rather than 1 day or less.
Here, we adapt a complementary method that uses alkyl nitrite photolysis to generate an alkoxy radical (RO•) and NO.In the presence of air, RO• reacts with O 2 to generate a carbonyl product (R'O) and a hydroperoxyl (HO 2 ) radical, and NO and HO 2 subsequently react to generate OH and NO 2 .Using this method, O 3 is not required to generate OH radicals, and insignificant amounts of O 3 or NO 3 are generated as byproducts.We present experimental and model characterization of OH and NO x levels that are generated as a function of photolysis wavelength, and organic nitrite concentration and composition.We furthermore carried out chemical ionization mass spectrometer measurements to compare nitrogen-containing photooxidation products obtained from the reaction of α-Pinene with radicals generated via alkyl nitrite photolysis or the O( 1 D) + H 2 O + N 2 O reaction.

Alkyl nitrite preparation
Figure 1 shows molecular structures of the alkyl nitrites that were used.Isopropyl nitrite (iPrONO; Pfaltz and Bauer, > 95 % purity) was used without additional purification.Perdeuterated isopropyl nitrite (iPrONO-d 7 ) and 1,3-propyl dinitrite [1,3-Pr(ONO) 2 ] were synthesized from the action of HONO on isopropanol-d 8 or 1,3-propanediol, respectively, as described elsewhere (Noyes, 1933;Carrasquillo et al., 2014).Briefly, sodium nitrite (> 99.999 %, Sigma-Aldrich) and alcohol were combined in a 1.1 : 1.0 molar ratio and stirred with a magnetic stirrer inside a round-bottom flask.Sulfuric acid was added dropwise to the flask -thereby generating HONO upon reaction with sodium nitrite -until a 0.5 : 1.0 acid : alcohol molar ratio was achieved.The resulting clear yellow liquid was dried over sodium sulfate, neutralized with excess sodium bicarbonate and then stored in amber vials and refrigerated at 4 • C until use (within 1 week of synthesis in this work).Under these storage conditions, the nominal shelf life of iPrONO and similar organic nitrites is approximately 2 years (Robert Milburn, personal communication, 29 October 2018).
A syringe pump was used to introduce iPrONO, iPrONOd 7 , and 1,3-Pr(ONO) 2 through a 10.2 cm length of 0.0152 cm ID teflon tubing at liquid flow rates ranging from 0.016 to 0.63 µL min −1 .The liquid organic nitrite was evaporated into a 1 L min −1 N 2 carrier gas at the end of the tubing.The flow containing organic nitrite vapor was then mixed with a 7 L min −1 synthetic air carrier gas at the reactor inlet.
The organic nitrite mixing ratio entering the reactor, r RONO , was equal to Q carrier , where Q RONO,g was the volumetric flow rate of organic nitrite vapor (L min −1 ) and Q carrier was the volumetric flow rate of carrier gas (L min −1 ).Q RONO,g was calculated using the ideal gas law as applied by Liu et al. (2015): where Q RONO,l (µL min −1 ) is the volumetric flow of organic nitrite liquid, ρ (g cm −3 ) and MW (g mol −1 ) are the organic nitrite liquid density and molecular weight, R (8.314 J mol −1 K −1 ) is the universal gas constant, T (K) is temperature, P (hPa) is pressure, and 0.01 is a lumped pressure, volume and density unit conversion factor.

Alkyl nitrite photolysis
Alkyl nitrites were photolyzed inside a Potential Aerosol Mass (PAM) oxidation flow reactor (Aerodyne Research, Inc.), which is a horizontal 13.3 L aluminum cylindrical chamber (46 cm long ×22 cm ID) operated in continuous flow mode (Lambe et al., 2017), with 5.1 ± 0.3 L min −1 flow through the reactor unless stated otherwise.The relative humidity (RH) in the reactor was controlled in the range of 31 %-63 % at 21-32 ) was used to regulate current applied to the lamps.The UV irradiance was measured using a photodetector (TOCON-GaP6, sglux GmbH) and was varied by changing the control voltage applied to the ballast between 1.6 and 10 VDC.NO and NO 2 mixing ratios were measured using a NO x analyzer (Model 405 nm, 2B Technologies), which quantified [NO 2 ] (ppb) from the measured absorbance at λ = 405 nm, and [NO] (ppb) by reaction with O 3 to convert to NO 2 .Alkyl nitrites introduced to the reactor with the lamps turned off consistently generated signals in both the NO and NO 2 measurement channels of the NO x analyzer, possibly due to impurities and/or species generated via iPrONO + O 3 reactions inside the analyzer.For example, background NO and NO 2 mixing ratios increased from 0 to 1526 and 0 to 1389 ppb as a function of injected [iPrONO] = 0 to 18.7 ppm with the lamps off (Fig. S2).We attempted to correct [NO] and [NO 2 ] for this apparent alkyl nitrite interference by subtracting background signals measured in the presence of alkyl nitrite with lamps off, to no avail, because background signals (alkyl nitrite present with lamps off) were large compared to signals obtained with alkyl nitrite present with lamps on.Instead, we constrained [NO] and [NO 2 ] using the photochemical model discussed in Sect.2.4.

Actinic flux calibration
To quantify the actinic flux I in the reactor for each lamp type, we measured the rate of NO 2 photolysis as a function of UV irradiance (Fig. S4).Measurements were conducted in the absence of oxygen to avoid O 3 formation.The first-order NO 2 photolysis rate (j NO 2 ) was calculated using Eq.(2): where NO 2,0 and NO 2,τ were the steady-state NO 2 mixing ratios measured at the exit of the reactor with the lamps turned off and on, respectively.The mean NO 2 residence time in the reactor, τ NO 2 , was characterized using 10 s pulsed inputs of NO 2 .To mimic the effect of axial dispersion induced by temperature gradients from the lamps being turned on (Lambe et al., 2011;Huang et al., 2017), residence time distributions were measured in the presence of four lamps centered at λ = 658 nm (F436T5/4P-658; Aerodyne Research, Inc.), where the NO 2 quantum yield is zero (Gardner et al., 1987).NO 2 residence time distributions are shown in Fig. S3, where τ NO 2 ranged from 120 ± 34 s (±1σ ; lamps off) to 98 ± 63 s (±1σ ; lamps on) in a manner that is consistent with previous observations (Lambe et al., 2011;Huang et al., 2017).Assuming τ NO 2 = 98 s, maximum j NO 2 values were 0.12, 0.36, and 0.50 min −1 following photolysis at full lamp power at λ = 254, 350, and 369 nm, respectively.Corresponding I 254 , I 350 , and I 369 values were calculated using a photochemical model implemented in the KinSim chemical kinetics solver (Peng et al., 2015; implemented within Igor Pro 7, WaveMetrics Inc.) that incorporated the following reactions: NO 2 absorption cross sections were averaged across the 254, 350, and 369 nm lamp emission spectra, respectively (Table 1) (Atkinson et al., 2004) and input to the model.Maximum I 254 = 8.6 × 10 16 , I 350 = 6.3 × 10 15 , and I 369 = 6.5 × 10 15 photons cm −2 s −1 were obtained.While I 350 and I 369 values were in agreement with values calculated from lamp manufacturer specifications (I 350 = 5.8 × 10 15 and I 369 = 6.2 × 10 15 photons cm −2 s −1 ) within uncertainties, I 254 obtained from our calibration was ∼ 13 times larger than expected.We hypothesize that this discrepancy was due to the presence of additional minor mercury lines (e.g., λ ∼ 313, 365, 405) that induce NO 2 photolysis and that were not fully accounted for using Eq. ( 2) or the manufacturer spectra (Fig. S1).Thus, we instead assume maximum I 254 = 6.5 × 10 15 photons cm −2 s −1 based on manufacturer specifications.

OH exposure calibration
The OH exposure (OH exp ) obtained from alkyl nitrite photolysis, that is, the product of the OH concentration and mean residence time, was calculated from the addition of between 280 and 420 ppb SO 2 at the reactor inlet.Over the course of these experiments, NO x generated from alkyl nitrite photolysis significantly interfered with the SO 2 mixing ratio measured with an SO 2 analyzer (Model 43i, Thermo Scientific); a representative example is shown in Fig. S5.To circumvent this issue, we measured the initial SO 2 mixing ratio, [SO 2,0 ], prior to alkyl nitrite photolysis, then used an Aerosol Chemical Speciation Monitor (ACSM; Aerodyne Research, Inc.) to measure the concentration of particulate sulfate generated from SO 2 + OH reactions.
To relate the measured [SO 2,0 ] and sulfate to OH exp , we conducted an offline calibration where 493 ppb SO 2 was added to the reactor and OH was generated via O 3 + hν 254 → O( 1 D) + O 2 followed by O( 1 D) + H 2 O → 2OH in the absence of NO x (OFR254 mode).The reactor was operated at the same residence time and humidity that was used in alkyl nitrite experiments, although we note that humidity will not change the response of the ACSM to sulfuric acid aerosols.Because no particulate ammonia was present aside from trace background levels, we assumed an ACSM collection efficiency of unity for the sulfate particles.SO 2 decay and particulate sulfate formation were measured across a range of UV irradiance and [O 3 ], from which a calibration equation relating sulfate to OH exp was obtained (Fig. S6) and applied to alkyl nitrite photolysis experiments.In a separate experiment conducted with 2.2 ppm of iPrONO input to the reactor at I 369 = 6.5×10 15 photons cm −2 s −1 , we verified that the mass of particulate sulfate detected by the ACSM responded linearly to a change in the input mixing ratio of SO 2 between 200 and 473 ppb (Fig. S7).This suggests that the sulfate particles were large enough for efficient transmission through the inlet lens of the ACSM across the range of OH exp used in our experiments.While not applicable in this work, we note that heterogeneous uptake of SO 2 into organic aerosol may bias OH exposure measurements (Ye et al., 2018).

Chemical ionization mass spectrometer (CIMS) measurements
In a separate set of experiments, mass spectra of gas-phase α-Pinene photooxidation products were obtained with an Aerodyne high-resolution time-of-flight chemical ionization mass spectrometer using nitrate as the reagent ion (NO − 3 -HRToF-CIMS, hereafter abbreviated as NO − 3 -CIMS) (Eisele and Tanner, 1993;Ehn et al., 2012).Nitrate (NO − 3 ) and its higher-order clusters (e.g., HNO 3 NO − 3 ) generated from X-ray ionization of HNO 3 were used as the reagent due to their selectivity to highly oxidized organic compounds, including species that contribute to SOA formation (Ehn et al., 2014;Krechmer et al., 2015;Lambe et al., 2017).The NO − 3 -CIMS sampled the reactor output at 10.5 L min −1 .α-Pinene oxidation products were detected as adduct ions of NO − 3 .In these experiments, the reactor was operated with a residence time of approximately 80 s to accommodate the undiluted NO − 3 -CIMS inlet flow requirement.OFR369i(iPrONO) and OFR369-i(iPrONO-d 7 ) were operated using I 369 = 6.5 × 10 15 photons cm −2 s −1 and > 7 ppm alkyl nitrite; in these experiments, α-Pinene was evaporated into the carrier gas by flowing 1 sccm N 2 through a bubbler containing liquid α-Pinene.Assuming the N 2 flow was saturated with α-Pinene vapor, we estimate ∼ 500 ppb α-Pinene was introduced to the OFR based on its vapor pressure at room temperature and known dilution ratio into the main carrier gas.In a separate experiment, OFR254-iN 2 O was operated using I 254 = 3.2 × 10 15 photons cm −2 s −1 and 5 ppm Here, α-Pinene was introduced by flowing 1 sccm of a gas mixture containing 150 ppm α-Pinene in N 2 into the main carrier gas (this gas mixture was unavailable for the iPrONO photolysis experiments); the calculated α-Pinene mixing ratio that was introduced to the OFR was ∼ 16 ppb.

Photochemical model
We used the KinSim OFR photochemical model to calculate concentrations of radical/oxidant species produced (Peng et al., 2015;Peng and Jimenez, 2017).In addition to NO+HO 2 → OH+NO 2 and other reactions included in Peng and Jimenez (2017), the following reactions were added for this study: Model input parameters included pressure, temperature, [H 2 O], [iPrONO], mean residence time, actinic flux, and absorption cross sections and bimolecular rate constants shown in Table 1.We assumed the quantum yield of Reaction (R5) to be 0.50 above 350 nm (Raff and Finlayson-Pitts, 2010).We assumed the quantum yield of Reaction (R6) to be 0.04 above 350 nm (value for t-butyl nitrite) (Calvert and Pitts, 1966) • on ensuing photochemistry may be more significant.This is due to a higher quantum yield of Reaction (R6) at 254 nm, which is estimated to be 0.86 under a vacuum (Calvert and Pitts, 1966).Assuming that all 254 nm photons initiate photolysis, the quantum yield of Reaction (R5) is 0.14.Due to collisional deactivation at 1 atm that prevents i-C 3 H 7 O• decomposition, the quantum yield of Reaction (R5) at λ = 254 nm and 1 atm is expected to be higher than 0.14.Because quantum yield measurements were unavailable at these conditions, we applied an upperlimit quantum yield of 0.50 as applicable at λ > 350 nm and 1 atm (Raff and Finlayson-Pitts, 2010).We calculated a corresponding nominal quantum yield of 0.32 by averaging the lower-and upper-limit values of 0.14 and 0.50, resulting in a quantum yield of 0.68 for Reaction (R6).We assumed that the residence time distribution of iPrONO in the reactor was similar to the residence time distribution of NO 2 .To model iPrONO photolysis at λ = 254 nm, we extended the range of previously measured σ iPrONO values by measuring the gas-phase absorption cross sections of iPrONO (purified via four freeze-pump-thaw cycles prior to measurement) down to λ = 220 nm using a custom-built absorption cell (Raff and Finlayson-Pitts, 2010).Results at λ = 220 to 436 nm are shown in Fig. S1 and are in agreement with previous work (Raff and Finlayson-Pitts, 2010) over the range of overlap at λ = 300 to 450 nm.
To account for uncertainties associated with the assumptions we made for quantum yield values, as well as uncertainties in other kinetic parameters, temperature, residence time, actinic flux, and organic nitrite concentration, we performed Monte Carlo uncertainty propagation (BIPM et al., 2008) as described previously (Peng et al., 2015;Peng and Jimenez, 2017).All uncertain kinetic parameters were assumed to follow lognormal distributions unless stated otherwise below.Uncertainties in rate constants and cross sections newly included in this study were adopted from Burkholder et al. (2015) if available.The relative uncertainty in the rate constant of Reaction (R13) was estimated to be 40 % based on the dispersion of rate constant measurements of published RO 2 + NO reactions.We assumed the random samples of the quantum yields of Reactions (R5) and (R6) at 254 nm and Reaction (R6) at 369 nm followed uniform distributions in the range of [0.50, 0.86], [0.14, 0.50] and [0, 0.20], respectively.We assumed uncertainties of 5 K and 20 s in temperature and residence time (normal distributions assumed) and relative uncertainties of 50 %, 100 %, and 25 % in actinic flux at 369 nm, actinic flux at 254 nm, and organic nitrite concentration.

Results and discussion
We first characterized OH exp and NO x by separately varying the photolysis wavelength (Sect.3.1.1)and input organic nitrite concentration to the reactor (Sect.3.1.2),with the goal of identifying optimal OFR conditions for OH and NO x generation via iPrONO photolysis.Second, we synthesized novel alkyl nitrites and compared their performances to iPrONO (Sect.3.2).Third, we parameterized OH exp and NO 2 production in a set of algebraic equations to guide selection of OFR experimental conditions.Finally, we compared NO − 3 -CIMS spectra of photooxidation products generated from reaction of α-Pinene with radicals produced via alkyl nitrite photolysis and O( 1 D) + H 2 O + N 2 O reactions.

OH exp and NO
x generated from iPrONO photolysis

Effect of photolysis wavelength
Figure 2 shows OH exp , [NO], and [NO 2 ] obtained as a function of actinic flux following photolysis of 1.9 ppm of iPrONO injected into the reactor at λ = 254, 350, or 369 nm.These systems are hereafter designated as OFR254i(iPrONO), OFR350-i(iPrONO), and OFR369-i(iPrONO), respectively; similar nomenclature is adapted for other alkyl nitrites.In these notations, the numbers following "OFR" are the photolysis wavelengths (in nm), and the "i" preceding the parentheses means initial injection of the radical precursor compound noted in the parentheses.Modeled OH exp , NO, and NO 2 values for the OFR254-i(PrONO) and OFR369i(PrONO) modes are shown in Fig. 2 at the same nominal operating conditions.At a fixed photolysis wavelength, OH exp , NO, and NO 2 increased with increasing actinic flux.Measured and modeled OH exp values were in agreement within uncertainties at λ = 369 nm.At λ = 254 nm, model OH exp results were higher than the measurements, perhaps due to uncertainty in assumptions that were necessary to model OFR254-i(iPrONO) (Sect.2.4).Higher NO 2 concentrations were modeled at λ = 254 nm than at λ = 369 nm because more iPrONO was photolyzed and the NO 2 yield was only weakly dependent on the fate of i-C 3 H 7 O•.For example, NO is converted to NO 2 either via reaction with HO 2 obtained via Reaction (R5) or CH 3 O 2 • and CH 3 C(O)O 2 • obtained via Reaction (R6).However, the effect of photolysis wavelength on NO and OH exp was different.Specifically, the highest NO concentration and OH exp were achieved via OFR369-i(iPrONO).OH exp achieved via OFR369-i(iPrONO) was slightly higher than OH exp attained using OFR350-i(iPrONO), likely because photolysis of both iPrONO and NO 2 , the reaction of which with OH suppresses OH exp , is more efficient at λ = 369 nm than at λ = 350 nm (Fig. S1 and Table 1).Further, the NO and OH yields achieved via OFR254-i(iPrONO) were suppressed due to significant (> 73 %) decomposition of iC 3 H 7 O• (Calvert and Pitts, 1966).The dependence of OH, NO, and NO 2 on the quantum yields of Reactions (R5) and (R6) was confirmed by sensitivity analysis of uncertainty propagation inputs and outputs as described in Sect.2.4.OH exp and NO were strongly anticorrelated with the quantum yield of Reaction (R6), whereas the correlation between NO 2 and the quantum yield of Reaction (R6) was negligible.
The products of this decomposition, i.e., CH 3 CHO and CH 3 •, both have adverse effects with regard to our experimental goals: CH 3 CHO is reactive toward OH and can thus suppress OH; the RO 2 • formed through this reaction, CH 3 C(O)O 2 •, consumes NO and generates NO 2 but does not generate OH; CH 3 • rapidly converts to CH 3 O 2 •, which also consumes NO and generates NO 2 but does not directly produce OH.Importantly, Fig. 2 suggests that it is preferable to photolyze alkyl nitrites at λ > 350 because optimal OH exp and NO : NO 2 were attained via OFR369-i(iPrONO).Moreover there is added risk of significant unwanted photolysis of organics via OFR254-i(iPrONO) (Peng et al., 2016).OH destruction from reaction with iPrONO and NO 2 .The model results showed that for [iPrONO] > 5 ppm, the opposite was true and OH exp plateaued or decreased.A maximum OH exp = 7.8×10 10 molecules cm −3 s was achieved via photolysis of 10 ppm iPrONO, with corresponding modeled [NO] and [NO 2 ] values of 148 and 405 ppb respectively.Modeled NO 3 concentrations were negligible in OFR369i(iPrONO) (≤ 1 ppt) because there was no O 3 present and NO 3 production via HNO 3 + OH → NO 3 + H 2 O reactions was insignificant.
We hypothesize that higher OH exp obtained from OFR369-i(iPrONO-d 7 ) relative to OFR369-i(iPrONO) was due to ∼ 2.6 times lower OH reactivity of iPrONO-d 7 relative to iPrONO (Nielsen et al., 1988(Nielsen et al., , 1991) ) and 6 times lower OH reactivity of acetone-d 6 relative to acetone (Raff et al., 2005).This hypothesis is supported by the modeled OH exp attained via OFR369-i(iPrONO-d 7 ), which is in agreement with measured OH exp within uncertainties and is 41 % higher than modeled OH exp attained via OFR369i(iPrONO).Model simulations revealed that this effect was most pronounced near the reactor inlet (e.g., at low residence time), where the local OH concentration was higher than elsewhere in the reactor because NO x was very low, resulting in higher sensitivity of [OH] to the OH reactivity of the specific organic nitrite that was used.On the other hand, OFR369-i(1,3-Pr(ONO) 2 ) was less efficient than OFR369i(iPrONO).In this case, it is possible that higher NO 2 production during 1,3-Pr(ONO) 2 photolysis and/or production of more reactive intermediates (e.g., malonaldehyde) offset any benefit gained from faster OH production via photolysis of both -ONO groups or more efficient photolysis of one -ONO group (Wang and Zu, 2016).

OH exp and NO 2 estimation equations for OFR369-i(iPrONO) and OFR369-i(iPrONO-d 7 )
Previous studies reported empirical OH exp algebraic estimation equations for OFR185 and OFR254 (Li et al., 2015;Peng et al., 2015).These equations parameterize OH exp as a function of readily measured experimental parameters, therefore providing a simpler alternative to detailed photochemical models that aids in experimental planning and analysis.Here, we expand on those studies by deriving OH exp and NO 2 estimation equations for OFR369-i(iPrONO) and OFR369-i(iPrONO-d 7 ).Model results ( 14 where OH exp , I 369 , OHR ext , [iPrONO or iPrONO-d 7 ], and τ are in units of molecules cm −3 s, photons cm −2 s −1 , s −1 , ppm, and s, respectively.Fit coefficients were obtained by fitting Eqs. ( 3) and ( 4) to OH exp model results over the following range of OFR parameters: ([iPrONO/iPrONO-d 7 ]; 0.2-20 ppm), I 369 (1×10 15 -2×10 16 photons cm −2 s −1 ), OHR ext (1-200 s −1 ), and residence time, τ , between 30 and 200 s.We explored 11 logarithmically evenly distributed values in these ranges for each parameter and thus performed simulations for 14641 model cases in total.To determine the functional form of Eqs. ( 3) and ( 4), we used the sum of the logarithms of first-, second-and third-order terms of the four parameters and iteratively removed the terms with very small fit coefficients until further removal of the remaining terms significantly worsened the fit quality.Figure 5a and c compare OH exp estimated from Eqs. (3) and ( 4) and calculated from the model described in Sect.2.4.The mean absolute value of the relative deviation is 29 %, indicating that the estimation equations are typically producing results within the inherent model uncertainties.Care should be taken to not use the equations away from the range in which they were derived, as much larger errors are possible when extrapolating.While several techniques are available to monitor NO 2 , interferences from other nitrogen-containing species are well known and may create issues similar to those shown in Fig. 2f.NO 2 production and loss rates are primarily governed by the alkyl nitrite concentration, actinic flux, and residence time in the OFR.These parameters were experimentally constrained (Sect.2.2.2).Thus, we derived NO 2 estimation equations for OFR369-i(iPrONO) (Eq.5) and OFR369i(iPrONO-d 7 ) (Eq. 6) as a function of [RONO], I 369 , and τ , to all of which NO 2 production is proportional, over the same phase space used to fit Eqs. ( 3) and (4): Figure 5b and d compare NO 2 estimated from Eqs. ( 3) and ( 4) and calculated from the model described in Sect.2.4.The mean absolute value of the relative deviation NO 2 estimated by Eqs. ( 5) and ( 6) and NO 2 computed by the photochemical model is 19 %.The mean model NO:NO 2 fraction is approximately 0.33 (Figs.2-3).
Two additional features are of note in Fig. 6.First, a series of ion signals at m/z = 312, 328, 344, 360, 376, 392, 408 and 340, 356, 372, 388, 402, 420 were observed at higher levels via OFR369-i(iPrONO-d 7 ) relative to OFR369i(iPrONO).These ions are plotted separately in Fig. 6d.The most plausible explanation is the additional contribution of [(NO 3 )C 8 H 10 DNO 8−14 ] − and [(NO 3 )C 10 H 14 DNO 7−14 ] − ions that retain -OD functionality following initial addition of OD (rather than OH) to α-Pinene.There is evidence of other deuterium-containing ions in Fig. 6b that are either less prominent or more difficult to resolve from other ions at the same integer mass.Second, C 10 dinitrates were present in all three spectra, with the highest dinitrate fractions observed in Fig. 6b (0.090) and c (0.081), and the lowest dinitrate fraction observed in Fig. 6a (0.056).Dinitrates are presumably generated from α-Pinene following (1) two OH reactions followed by two RO 2 + NO termination reactions or (2) one NO 3 reaction followed by one RO 2 + NO termination reaction.Previous application of OFR254-iN 2 O could not exclude the contribution of α-Pinene + NO 3 reactions, with NO 3 radicals generated from NO 2 + O 3 and other reactions (Lambe et al., 2017).However, generation of dinitrates via OFR369i(iPrONO-d 7 ), which produced negligible NO 3 , suggests that dinitrates are not an artifact of unwanted α-Pinene + NO 3 reactions.
Overall, Fig. 6 shows that many of the C 7 -C 10 nitrogencontaining compounds observed in Centreville and Hyytiälä were generated via OFR369-i(iPrONO), OFR369-i(iPrONOd 7 ), and OFR254-iN 2 O. Due to the local nature of the ambient terpene emissions at the Centreville and Hyytiälä sites, the associated photochemical age was presumably < 1 day.Thus, while the ambient NO − 3 -CIMS spectra at those sites were more complex and contained contributions from precursors other than α-Pinene, the oxidation state of the ambient terpene-derived organic nitrates was more closely simulated via OFR369-i(iPrONO) or OFR369-i(iPrONO-d 7 ), where the largest C 10 nitrates and dinitrates were C 10 H 15 NO 7 and C 10 H 16 N 2 O 9 (OFR369i(iPrONO); Fig. 6a), and C 10 H 15 NO 8 , C 10 H 15 NO 9 and C 10 H 16 N 2 O 9 (OFR369-i(iPrONO-d 7 ); Fig. 6c).By comparison, C 10 H 15 NO 8 and C 10 H 16 N 2 O 11 were the largest nitrate and dinitrate species generated via OFR254-iN 2 O (Fig. 6b).

Anticipated performance of alternative high-NO
x HO x precursors in OFRs 3.5.1 Methyl nitrite (MeONO) MeONO is commonly used as an OH radical source in environmental chamber studies (Atkinson et al., 1981;Matsunaga and Ziemann, 2010;Chhabra et al., 2011;Finewax et al., 2018).To evaluate its potential use in OFRs, we examined previous measurements in an environmental chamber equipped with blacklights (j NO 2 = 0.27 min −1 , assumed 350 nm wavelength), where photolysis of 10 ppm MeONO generated [OH] ∼ 2 × 10 8 molecules cm −3 for a few minutes (Atkinson et al., 1981).In our OFR, j NO 2 ,max = 0.36 min −1 at λ = 350 nm.Thus, over 98 s exposure time, we anticipate OH exp ≈ 2 × 10 10 molecules cm −3 s would be obtained via photolysis of 10 ppm MeONO in OFRs.This is lower than the OH exp attained via photolysis of 10 ppm iPrONO even after correcting for different j NO 2 values in the different studies.Lower OH exp achieved from MeONO photolysis is presumably due to the higher reactivity of formaldehyde, the primary photolysis product of MeONO, relative to acetone, the primary photolysis product of iPrONO at 369 nm (Raff and Finlayson-Pitts, 2010).Along with less efficient OH production, MeONO must be synthesized, trapped at low temperature, and stored under a vacuum.Thus, there is no advantage to using OFR350-iMeONO (or OFR350-MeONO-d 4 ) in OFRs relative to OFR369-i(iPrONO) or OFR369-i(iPrONOd 7 ).

Nitrous acid (HONO)
HONO is also commonly used as an OH radical source in environmental chamber studies.et al., 1995) that may cause additional OH suppression.For these reasons, we believe that there is no advantage to using HONO as a HO x precursor in OFRs.

Hexafluoroisopropyl nitrite (HFiPrONO)
HFiPrONO has been synthesized from O-nitrosation of hexafluoroisopropanol (Andersen et al., 2003;Shuping et al., 2006).We predict that OFR369-i(HFiPrONO) should attain higher OH exp than OFR369-i(iPrONO) and OFR369i(iPrONO-d 7 ) due to similar photolysis rates (Andersen et al., 2003) and ∼ 200 times lower OH reactivity of HFiPrONO/hexafluoroacetone relative to iPrONO/acetone (Atkinson et al., 1992;Tokuhashi et al., 1999).Simple modeling calculations suggest that application of OFR369i(HFiPrONO) may achieve up to a week of equivalent OH exposure.We made several unsuccessful attempts to synthesize HFiPrONO and other fluorinated alkyl nitrites with a procedure similar to that used by Andersen et al. (2003) and Shuping et al. (2006).The synthesis product was blue (not yellow) in color when trapped or stored in nitrogen, generated negligible OH upon irradiation in the reactor, and evolved into brown vapor in the presence of air or upon warming to room temperature (Fig. S8).(Lambe et al., 2017;Peng et al., 2018).Alkyl nitrite photolysis is an established method that facilitates high-NO x photooxidation studies in modern OFRs.Here, we adapted alkyl nitrite photolysis for new OFR applications by characterizing the photolysis wavelength, nitrite concentration, and nitrite composition that result in optimal HO x and NO x generation capabilities.Based on our results, we recommend photolysis of 5-10 ppm isopropyl nitrite at λ ≈ 365-370 nm photolysis wavelength and I > 10 15 photons cm −2 s −1 .If the user has the resources to synthesize iPrONO-d 7 , better performance is expected relative to iPrONO.Alkyl nitrite photolysis at λ = 254 nm is not recommended.Taken together, OFR254/185-iN 2 O and OFR369-i(iPrONO/iPrONOd 7 ) are complementary methods that provide additional flexibility for NO x -dependent OFR studies.OFR254/185-iN2O generate variable-NO x photooxidation conditions (NO : HO 2 ≈ 0-100) and are suitable for the characterization of multigenerational oxidative aging processes at up to OH exp ≈ (5-10)×10 11 molecules cm −3 s (∼ 5-10 equivalent days).OFR369-i(iPrONO/iPrONO-d7) generate high-NO photooxidation conditions (NO:HO 2 ≈ 10-10 000; NO : NO 2 ≈ 0.2-0.7)with minimal O 3 and NO 3 formation at longer photolysis wavelength than OFR254/185-iN 2 O.We anticipate that alkyl nitrite photolysis is advantageous for the characterization of first-generation, high-NO x photooxidation products of most precursors at up to OH exp ≈ 1 × 10 11 molecules cm −3 s (1 equivalent day), which is comparable to environmental chambers investigating high-NO x conditions.The generation of OD (rather than OH) via OFR369i(iPrONO-d 7 ) may be useful in photooxidation studies of unsaturated precursors due to the shift in the m/z of the addition products, though at the potential expense of generating more complex distributions of oxidation products.Potential disadvantages of the OFR369-i(iPrONO) method are (1) restriction to high-NO photochemical conditions, (2) restriction to OH exp of 1 equivalent day or less, (3) additional complexity involved with integration of the alkyl nitrite source (compared to O 3 +H 2 O+N 2 O), (4) additional cost and complexity to retrofit a specific OFR design with blacklights, and (5) that it acts as an interference that precludes NO x measurements by chemiluminescence detection.Future work will evaluate the ability of each method to mimic polluted atmospheric conditions in specific source regions.

Figure 3
Figure 3 shows measured OH exp and modeled NO x concentrations obtained from photolysis of 0.5 to 20 ppm iPrONO at I 369 ≈ 7×10 15 photons cm −2 s −1 .[NO] and [NO 2 ] increased with increasing [iPrONO], as expected.For [iPrONO] ≤ 5 ppm, OH exp increased with increasing [iPrONO] because the rate of OH production increased faster than the rate of

Figure 3 .
Figure 3. Measured and modeled (a) OH exposure, (b) NO mixing ratio, and (c) NO 2 mixing ratio values obtained using OFR369-i(iPrONO) at I 369 = 7 × 10 15 ph cm −2 s −1 as a function of the iPrONO mixing ratio.Error bars for measurements represent ±50 % uncertainty in OH exp and estimated ±30 % uncertainty in the iPrONO mixing ratio values.
Aerodyne Research, Inc.) were used.Emission spectra obtained from the primary manufacturer (Light Sources, Inc. or LCD Lighting, Inc.) are shown in Fig. S1 in the Supplement.A fluorescent dimming ballast (IZT-2S28-D, Advance Transformer Co.